Merge "add vp9_block_error_fp_neon"
diff --git a/README b/README
index 7b44ba7..3adc551 100644
--- a/README
+++ b/README
@@ -85,6 +85,7 @@
x86-win32-vs10
x86-win32-vs11
x86-win32-vs12
+ x86_64-android-gcc
x86_64-darwin9-gcc
x86_64-darwin10-gcc
x86_64-darwin11-gcc
diff --git a/build/make/Android.mk b/build/make/Android.mk
index e971c9d..aa7a8c9 100644
--- a/build/make/Android.mk
+++ b/build/make/Android.mk
@@ -67,6 +67,8 @@
LOCAL_ARM_MODE := arm
else ifeq ($(TARGET_ARCH_ABI),x86)
include $(CONFIG_DIR)libs-x86-android-gcc.mk
+else ifeq ($(TARGET_ARCH_ABI),x86_64)
+ include $(CONFIG_DIR)libs-x86_64-android-gcc.mk
else ifeq ($(TARGET_ARCH_ABI),mips)
include $(CONFIG_DIR)libs-mips-android-gcc.mk
else
diff --git a/build/make/rtcd.pl b/build/make/rtcd.pl
index 6753ee7..7dff028 100755
--- a/build/make/rtcd.pl
+++ b/build/make/rtcd.pl
@@ -319,6 +319,8 @@
print <<EOF;
#if HAVE_DSPR2
+void vpx_dsputil_static_init();
+vpx_dsputil_static_init();
#if CONFIG_VP8
void dsputil_static_init();
dsputil_static_init();
diff --git a/configure b/configure
index 6cac15a..fcab145 100755
--- a/configure
+++ b/configure
@@ -132,6 +132,7 @@
all_platforms="${all_platforms} x86-win32-vs10"
all_platforms="${all_platforms} x86-win32-vs11"
all_platforms="${all_platforms} x86-win32-vs12"
+all_platforms="${all_platforms} x86_64-android-gcc"
all_platforms="${all_platforms} x86_64-darwin9-gcc"
all_platforms="${all_platforms} x86_64-darwin10-gcc"
all_platforms="${all_platforms} x86_64-darwin11-gcc"
diff --git a/test/dct16x16_test.cc b/test/dct16x16_test.cc
index 4db6bf9..b1f7e7e 100644
--- a/test/dct16x16_test.cc
+++ b/test/dct16x16_test.cc
@@ -272,7 +272,7 @@
void fdct16x16_ref(const int16_t *in, tran_low_t *out, int stride,
int /*tx_type*/) {
- vp9_fdct16x16_c(in, out, stride);
+ vpx_fdct16x16_c(in, out, stride);
}
void idct16x16_ref(const tran_low_t *in, uint8_t *dest, int stride,
@@ -822,14 +822,14 @@
INSTANTIATE_TEST_CASE_P(
C, Trans16x16DCT,
::testing::Values(
- make_tuple(&vp9_highbd_fdct16x16_c, &idct16x16_10, 0, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct16x16_c, &idct16x16_12, 0, VPX_BITS_12),
- make_tuple(&vp9_fdct16x16_c, &vp9_idct16x16_256_add_c, 0, VPX_BITS_8)));
+ make_tuple(&vpx_highbd_fdct16x16_c, &idct16x16_10, 0, VPX_BITS_10),
+ make_tuple(&vpx_highbd_fdct16x16_c, &idct16x16_12, 0, VPX_BITS_12),
+ make_tuple(&vpx_fdct16x16_c, &vp9_idct16x16_256_add_c, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
C, Trans16x16DCT,
::testing::Values(
- make_tuple(&vp9_fdct16x16_c, &vp9_idct16x16_256_add_c, 0, VPX_BITS_8)));
+ make_tuple(&vpx_fdct16x16_c, &vp9_idct16x16_256_add_c, 0, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if CONFIG_VP9_HIGHBITDEPTH
@@ -862,7 +862,7 @@
INSTANTIATE_TEST_CASE_P(
NEON, Trans16x16DCT,
::testing::Values(
- make_tuple(&vp9_fdct16x16_c,
+ make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_neon, 0, VPX_BITS_8)));
#endif
@@ -870,7 +870,7 @@
INSTANTIATE_TEST_CASE_P(
SSE2, Trans16x16DCT,
::testing::Values(
- make_tuple(&vp9_fdct16x16_sse2,
+ make_tuple(&vpx_fdct16x16_sse2,
&vp9_idct16x16_256_add_sse2, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
SSE2, Trans16x16HT,
@@ -889,15 +889,15 @@
INSTANTIATE_TEST_CASE_P(
SSE2, Trans16x16DCT,
::testing::Values(
- make_tuple(&vp9_highbd_fdct16x16_sse2,
+ make_tuple(&vpx_highbd_fdct16x16_sse2,
&idct16x16_10, 0, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct16x16_c,
+ make_tuple(&vpx_highbd_fdct16x16_c,
&idct16x16_256_add_10_sse2, 0, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct16x16_sse2,
+ make_tuple(&vpx_highbd_fdct16x16_sse2,
&idct16x16_12, 0, VPX_BITS_12),
- make_tuple(&vp9_highbd_fdct16x16_c,
+ make_tuple(&vpx_highbd_fdct16x16_c,
&idct16x16_256_add_12_sse2, 0, VPX_BITS_12),
- make_tuple(&vp9_fdct16x16_sse2,
+ make_tuple(&vpx_fdct16x16_sse2,
&vp9_idct16x16_256_add_c, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
SSE2, Trans16x16HT,
@@ -926,7 +926,7 @@
INSTANTIATE_TEST_CASE_P(
MSA, Trans16x16DCT,
::testing::Values(
- make_tuple(&vp9_fdct16x16_msa,
+ make_tuple(&vpx_fdct16x16_msa,
&vp9_idct16x16_256_add_msa, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
MSA, Trans16x16HT,
diff --git a/test/dct32x32_test.cc b/test/dct32x32_test.cc
index 25059a5..f406fa2 100644
--- a/test/dct32x32_test.cc
+++ b/test/dct32x32_test.cc
@@ -12,14 +12,15 @@
#include <stdlib.h>
#include <string.h>
-#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/acm_random.h"
#include "test/clear_system_state.h"
#include "test/register_state_check.h"
#include "test/util.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
-#include "./vpx_config.h"
#include "./vp9_rtcd.h"
+#include "./vpx_config.h"
+#include "./vpx_dsp_rtcd.h"
#include "vp9/common/vp9_entropy.h"
#include "vpx/vpx_codec.h"
#include "vpx/vpx_integer.h"
@@ -194,7 +195,7 @@
input_block[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
const int stride = 32;
- vp9_fdct32x32_c(input_block, output_ref_block, stride);
+ vpx_fdct32x32_c(input_block, output_ref_block, stride);
ASM_REGISTER_STATE_CHECK(fwd_txfm_(input_block, output_block, stride));
if (version_ == 0) {
@@ -231,7 +232,7 @@
}
const int stride = 32;
- vp9_fdct32x32_c(input_extreme_block, output_ref_block, stride);
+ vpx_fdct32x32_c(input_extreme_block, output_ref_block, stride);
ASM_REGISTER_STATE_CHECK(
fwd_txfm_(input_extreme_block, output_block, stride));
@@ -314,25 +315,25 @@
INSTANTIATE_TEST_CASE_P(
C, Trans32x32Test,
::testing::Values(
- make_tuple(&vp9_highbd_fdct32x32_c,
+ make_tuple(&vpx_highbd_fdct32x32_c,
&idct32x32_10, 0, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct32x32_rd_c,
+ make_tuple(&vpx_highbd_fdct32x32_rd_c,
&idct32x32_10, 1, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct32x32_c,
+ make_tuple(&vpx_highbd_fdct32x32_c,
&idct32x32_12, 0, VPX_BITS_12),
- make_tuple(&vp9_highbd_fdct32x32_rd_c,
+ make_tuple(&vpx_highbd_fdct32x32_rd_c,
&idct32x32_12, 1, VPX_BITS_12),
- make_tuple(&vp9_fdct32x32_c,
+ make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c, 0, VPX_BITS_8),
- make_tuple(&vp9_fdct32x32_rd_c,
+ make_tuple(&vpx_fdct32x32_rd_c,
&vp9_idct32x32_1024_add_c, 1, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
C, Trans32x32Test,
::testing::Values(
- make_tuple(&vp9_fdct32x32_c,
+ make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c, 0, VPX_BITS_8),
- make_tuple(&vp9_fdct32x32_rd_c,
+ make_tuple(&vpx_fdct32x32_rd_c,
&vp9_idct32x32_1024_add_c, 1, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
@@ -340,9 +341,9 @@
INSTANTIATE_TEST_CASE_P(
NEON, Trans32x32Test,
::testing::Values(
- make_tuple(&vp9_fdct32x32_c,
+ make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_neon, 0, VPX_BITS_8),
- make_tuple(&vp9_fdct32x32_rd_c,
+ make_tuple(&vpx_fdct32x32_rd_c,
&vp9_idct32x32_1024_add_neon, 1, VPX_BITS_8)));
#endif // HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -350,9 +351,9 @@
INSTANTIATE_TEST_CASE_P(
SSE2, Trans32x32Test,
::testing::Values(
- make_tuple(&vp9_fdct32x32_sse2,
+ make_tuple(&vpx_fdct32x32_sse2,
&vp9_idct32x32_1024_add_sse2, 0, VPX_BITS_8),
- make_tuple(&vp9_fdct32x32_rd_sse2,
+ make_tuple(&vpx_fdct32x32_rd_sse2,
&vp9_idct32x32_1024_add_sse2, 1, VPX_BITS_8)));
#endif // HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -360,15 +361,15 @@
INSTANTIATE_TEST_CASE_P(
SSE2, Trans32x32Test,
::testing::Values(
- make_tuple(&vp9_highbd_fdct32x32_sse2, &idct32x32_10, 0, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct32x32_rd_sse2, &idct32x32_10, 1,
+ make_tuple(&vpx_highbd_fdct32x32_sse2, &idct32x32_10, 0, VPX_BITS_10),
+ make_tuple(&vpx_highbd_fdct32x32_rd_sse2, &idct32x32_10, 1,
VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct32x32_sse2, &idct32x32_12, 0, VPX_BITS_12),
- make_tuple(&vp9_highbd_fdct32x32_rd_sse2, &idct32x32_12, 1,
+ make_tuple(&vpx_highbd_fdct32x32_sse2, &idct32x32_12, 0, VPX_BITS_12),
+ make_tuple(&vpx_highbd_fdct32x32_rd_sse2, &idct32x32_12, 1,
VPX_BITS_12),
- make_tuple(&vp9_fdct32x32_sse2, &vp9_idct32x32_1024_add_c, 0,
+ make_tuple(&vpx_fdct32x32_sse2, &vp9_idct32x32_1024_add_c, 0,
VPX_BITS_8),
- make_tuple(&vp9_fdct32x32_rd_sse2, &vp9_idct32x32_1024_add_c, 1,
+ make_tuple(&vpx_fdct32x32_rd_sse2, &vp9_idct32x32_1024_add_c, 1,
VPX_BITS_8)));
#endif // HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -376,9 +377,9 @@
INSTANTIATE_TEST_CASE_P(
AVX2, Trans32x32Test,
::testing::Values(
- make_tuple(&vp9_fdct32x32_avx2,
+ make_tuple(&vpx_fdct32x32_avx2,
&vp9_idct32x32_1024_add_sse2, 0, VPX_BITS_8),
- make_tuple(&vp9_fdct32x32_rd_avx2,
+ make_tuple(&vpx_fdct32x32_rd_avx2,
&vp9_idct32x32_1024_add_sse2, 1, VPX_BITS_8)));
#endif // HAVE_AVX2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -386,9 +387,9 @@
INSTANTIATE_TEST_CASE_P(
MSA, Trans32x32Test,
::testing::Values(
- make_tuple(&vp9_fdct32x32_msa,
+ make_tuple(&vpx_fdct32x32_msa,
&vp9_idct32x32_1024_add_msa, 0, VPX_BITS_8),
- make_tuple(&vp9_fdct32x32_rd_msa,
+ make_tuple(&vpx_fdct32x32_rd_msa,
&vp9_idct32x32_1024_add_msa, 1, VPX_BITS_8)));
#endif // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
} // namespace
diff --git a/test/fdct4x4_test.cc b/test/fdct4x4_test.cc
index b402a48..7573801 100644
--- a/test/fdct4x4_test.cc
+++ b/test/fdct4x4_test.cc
@@ -41,7 +41,7 @@
void fdct4x4_ref(const int16_t *in, tran_low_t *out, int stride,
int tx_type) {
- vp9_fdct4x4_c(in, out, stride);
+ vpx_fdct4x4_c(in, out, stride);
}
void fht4x4_ref(const int16_t *in, tran_low_t *out, int stride, int tx_type) {
@@ -419,14 +419,14 @@
INSTANTIATE_TEST_CASE_P(
C, Trans4x4DCT,
::testing::Values(
- make_tuple(&vp9_highbd_fdct4x4_c, &idct4x4_10, 0, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct4x4_c, &idct4x4_12, 0, VPX_BITS_12),
- make_tuple(&vp9_fdct4x4_c, &vp9_idct4x4_16_add_c, 0, VPX_BITS_8)));
+ make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_10, 0, VPX_BITS_10),
+ make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_12, 0, VPX_BITS_12),
+ make_tuple(&vpx_fdct4x4_c, &vp9_idct4x4_16_add_c, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
C, Trans4x4DCT,
::testing::Values(
- make_tuple(&vp9_fdct4x4_c, &vp9_idct4x4_16_add_c, 0, VPX_BITS_8)));
+ make_tuple(&vpx_fdct4x4_c, &vp9_idct4x4_16_add_c, 0, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if CONFIG_VP9_HIGHBITDEPTH
@@ -473,7 +473,7 @@
INSTANTIATE_TEST_CASE_P(
NEON, Trans4x4DCT,
::testing::Values(
- make_tuple(&vp9_fdct4x4_c,
+ make_tuple(&vpx_fdct4x4_c,
&vp9_idct4x4_16_add_neon, 0, VPX_BITS_8)));
#endif // HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -507,7 +507,7 @@
INSTANTIATE_TEST_CASE_P(
SSE2, Trans4x4DCT,
::testing::Values(
- make_tuple(&vp9_fdct4x4_sse2,
+ make_tuple(&vpx_fdct4x4_sse2,
&vp9_idct4x4_16_add_sse2, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
SSE2, Trans4x4HT,
@@ -522,11 +522,11 @@
INSTANTIATE_TEST_CASE_P(
SSE2, Trans4x4DCT,
::testing::Values(
- make_tuple(&vp9_highbd_fdct4x4_c, &idct4x4_10_sse2, 0, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct4x4_sse2, &idct4x4_10_sse2, 0, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct4x4_c, &idct4x4_12_sse2, 0, VPX_BITS_12),
- make_tuple(&vp9_highbd_fdct4x4_sse2, &idct4x4_12_sse2, 0, VPX_BITS_12),
- make_tuple(&vp9_fdct4x4_sse2, &vp9_idct4x4_16_add_c, 0,
+ make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_10_sse2, 0, VPX_BITS_10),
+ make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_10_sse2, 0, VPX_BITS_10),
+ make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_12_sse2, 0, VPX_BITS_12),
+ make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_12_sse2, 0, VPX_BITS_12),
+ make_tuple(&vpx_fdct4x4_sse2, &vp9_idct4x4_16_add_c, 0,
VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
@@ -542,7 +542,7 @@
INSTANTIATE_TEST_CASE_P(
MSA, Trans4x4DCT,
::testing::Values(
- make_tuple(&vp9_fdct4x4_msa, &vp9_idct4x4_16_add_msa, 0, VPX_BITS_8)));
+ make_tuple(&vpx_fdct4x4_msa, &vp9_idct4x4_16_add_msa, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
MSA, Trans4x4HT,
::testing::Values(
diff --git a/test/fdct8x8_test.cc b/test/fdct8x8_test.cc
index 9fa1ca4..fb7fbb5 100644
--- a/test/fdct8x8_test.cc
+++ b/test/fdct8x8_test.cc
@@ -83,7 +83,7 @@
void fdct8x8_ref(const int16_t *in, tran_low_t *out, int stride, int tx_type) {
- vp9_fdct8x8_c(in, out, stride);
+ vpx_fdct8x8_c(in, out, stride);
}
void fht8x8_ref(const int16_t *in, tran_low_t *out, int stride, int tx_type) {
@@ -658,14 +658,14 @@
INSTANTIATE_TEST_CASE_P(
C, FwdTrans8x8DCT,
::testing::Values(
- make_tuple(&vp9_fdct8x8_c, &vp9_idct8x8_64_add_c, 0, VPX_BITS_8),
- make_tuple(&vp9_highbd_fdct8x8_c, &idct8x8_10, 0, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct8x8_c, &idct8x8_12, 0, VPX_BITS_12)));
+ make_tuple(&vpx_fdct8x8_c, &vp9_idct8x8_64_add_c, 0, VPX_BITS_8),
+ make_tuple(&vpx_highbd_fdct8x8_c, &idct8x8_10, 0, VPX_BITS_10),
+ make_tuple(&vpx_highbd_fdct8x8_c, &idct8x8_12, 0, VPX_BITS_12)));
#else
INSTANTIATE_TEST_CASE_P(
C, FwdTrans8x8DCT,
::testing::Values(
- make_tuple(&vp9_fdct8x8_c, &vp9_idct8x8_64_add_c, 0, VPX_BITS_8)));
+ make_tuple(&vpx_fdct8x8_c, &vp9_idct8x8_64_add_c, 0, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if CONFIG_VP9_HIGHBITDEPTH
@@ -698,7 +698,7 @@
INSTANTIATE_TEST_CASE_P(
NEON, FwdTrans8x8DCT,
::testing::Values(
- make_tuple(&vp9_fdct8x8_neon, &vp9_idct8x8_64_add_neon, 0,
+ make_tuple(&vpx_fdct8x8_neon, &vp9_idct8x8_64_add_neon, 0,
VPX_BITS_8)));
#endif // HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@@ -716,7 +716,7 @@
INSTANTIATE_TEST_CASE_P(
SSE2, FwdTrans8x8DCT,
::testing::Values(
- make_tuple(&vp9_fdct8x8_sse2, &vp9_idct8x8_64_add_sse2, 0,
+ make_tuple(&vpx_fdct8x8_sse2, &vp9_idct8x8_64_add_sse2, 0,
VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
SSE2, FwdTrans8x8HT,
@@ -731,14 +731,14 @@
INSTANTIATE_TEST_CASE_P(
SSE2, FwdTrans8x8DCT,
::testing::Values(
- make_tuple(&vp9_fdct8x8_sse2, &vp9_idct8x8_64_add_c, 0, VPX_BITS_8),
- make_tuple(&vp9_highbd_fdct8x8_c,
+ make_tuple(&vpx_fdct8x8_sse2, &vp9_idct8x8_64_add_c, 0, VPX_BITS_8),
+ make_tuple(&vpx_highbd_fdct8x8_c,
&idct8x8_64_add_10_sse2, 12, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct8x8_sse2,
+ make_tuple(&vpx_highbd_fdct8x8_sse2,
&idct8x8_64_add_10_sse2, 12, VPX_BITS_10),
- make_tuple(&vp9_highbd_fdct8x8_c,
+ make_tuple(&vpx_highbd_fdct8x8_c,
&idct8x8_64_add_12_sse2, 12, VPX_BITS_12),
- make_tuple(&vp9_highbd_fdct8x8_sse2,
+ make_tuple(&vpx_highbd_fdct8x8_sse2,
&idct8x8_64_add_12_sse2, 12, VPX_BITS_12)));
INSTANTIATE_TEST_CASE_P(
@@ -769,7 +769,7 @@
INSTANTIATE_TEST_CASE_P(
SSSE3, FwdTrans8x8DCT,
::testing::Values(
- make_tuple(&vp9_fdct8x8_ssse3, &vp9_idct8x8_64_add_ssse3, 0,
+ make_tuple(&vpx_fdct8x8_ssse3, &vp9_idct8x8_64_add_ssse3, 0,
VPX_BITS_8)));
#endif
@@ -777,7 +777,7 @@
INSTANTIATE_TEST_CASE_P(
MSA, FwdTrans8x8DCT,
::testing::Values(
- make_tuple(&vp9_fdct8x8_msa, &vp9_idct8x8_64_add_msa, 0, VPX_BITS_8)));
+ make_tuple(&vpx_fdct8x8_msa, &vp9_idct8x8_64_add_msa, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
MSA, FwdTrans8x8HT,
::testing::Values(
diff --git a/test/partial_idct_test.cc b/test/partial_idct_test.cc
index 8d19aff..6d77b7c 100644
--- a/test/partial_idct_test.cc
+++ b/test/partial_idct_test.cc
@@ -202,31 +202,31 @@
INSTANTIATE_TEST_CASE_P(
C, PartialIDctTest,
::testing::Values(
- make_tuple(&vp9_fdct32x32_c,
+ make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_34_add_c,
TX_32X32, 34),
- make_tuple(&vp9_fdct32x32_c,
+ make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_1_add_c,
TX_32X32, 1),
- make_tuple(&vp9_fdct16x16_c,
+ make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_10_add_c,
TX_16X16, 10),
- make_tuple(&vp9_fdct16x16_c,
+ make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_1_add_c,
TX_16X16, 1),
- make_tuple(&vp9_fdct8x8_c,
+ make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_12_add_c,
TX_8X8, 12),
- make_tuple(&vp9_fdct8x8_c,
+ make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_1_add_c,
TX_8X8, 1),
- make_tuple(&vp9_fdct4x4_c,
+ make_tuple(&vpx_fdct4x4_c,
&vp9_idct4x4_16_add_c,
&vp9_idct4x4_1_add_c,
TX_4X4, 1)));
@@ -235,27 +235,27 @@
INSTANTIATE_TEST_CASE_P(
NEON, PartialIDctTest,
::testing::Values(
- make_tuple(&vp9_fdct32x32_c,
+ make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_1_add_neon,
TX_32X32, 1),
- make_tuple(&vp9_fdct16x16_c,
+ make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_10_add_neon,
TX_16X16, 10),
- make_tuple(&vp9_fdct16x16_c,
+ make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_1_add_neon,
TX_16X16, 1),
- make_tuple(&vp9_fdct8x8_c,
+ make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_12_add_neon,
TX_8X8, 12),
- make_tuple(&vp9_fdct8x8_c,
+ make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_1_add_neon,
TX_8X8, 1),
- make_tuple(&vp9_fdct4x4_c,
+ make_tuple(&vpx_fdct4x4_c,
&vp9_idct4x4_16_add_c,
&vp9_idct4x4_1_add_neon,
TX_4X4, 1)));
@@ -265,31 +265,31 @@
INSTANTIATE_TEST_CASE_P(
SSE2, PartialIDctTest,
::testing::Values(
- make_tuple(&vp9_fdct32x32_c,
+ make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_34_add_sse2,
TX_32X32, 34),
- make_tuple(&vp9_fdct32x32_c,
+ make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_1_add_sse2,
TX_32X32, 1),
- make_tuple(&vp9_fdct16x16_c,
+ make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_10_add_sse2,
TX_16X16, 10),
- make_tuple(&vp9_fdct16x16_c,
+ make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_1_add_sse2,
TX_16X16, 1),
- make_tuple(&vp9_fdct8x8_c,
+ make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_12_add_sse2,
TX_8X8, 12),
- make_tuple(&vp9_fdct8x8_c,
+ make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_1_add_sse2,
TX_8X8, 1),
- make_tuple(&vp9_fdct4x4_c,
+ make_tuple(&vpx_fdct4x4_c,
&vp9_idct4x4_16_add_c,
&vp9_idct4x4_1_add_sse2,
TX_4X4, 1)));
@@ -300,7 +300,7 @@
INSTANTIATE_TEST_CASE_P(
SSSE3_64, PartialIDctTest,
::testing::Values(
- make_tuple(&vp9_fdct8x8_c,
+ make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_12_add_ssse3,
TX_8X8, 12)));
@@ -310,31 +310,31 @@
INSTANTIATE_TEST_CASE_P(
MSA, PartialIDctTest,
::testing::Values(
- make_tuple(&vp9_fdct32x32_c,
+ make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_34_add_msa,
TX_32X32, 34),
- make_tuple(&vp9_fdct32x32_c,
+ make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_1_add_msa,
TX_32X32, 1),
- make_tuple(&vp9_fdct16x16_c,
+ make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_10_add_msa,
TX_16X16, 10),
- make_tuple(&vp9_fdct16x16_c,
+ make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_1_add_msa,
TX_16X16, 1),
- make_tuple(&vp9_fdct8x8_c,
+ make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_12_add_msa,
TX_8X8, 10),
- make_tuple(&vp9_fdct8x8_c,
+ make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_1_add_msa,
TX_8X8, 1),
- make_tuple(&vp9_fdct4x4_c,
+ make_tuple(&vpx_fdct4x4_c,
&vp9_idct4x4_16_add_c,
&vp9_idct4x4_1_add_msa,
TX_4X4, 1)));
diff --git a/test/test.mk b/test/test.mk
index 8ecc856..6bb08be 100644
--- a/test/test.mk
+++ b/test/test.mk
@@ -164,8 +164,8 @@
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += sad_test.cc
-TEST_INTRA_PRED_SPEED_SRCS-$(CONFIG_VP9_DECODER) := test_intra_pred_speed.cc
-TEST_INTRA_PRED_SPEED_SRCS-$(CONFIG_VP9_DECODER) += ../md5_utils.h ../md5_utils.c
+TEST_INTRA_PRED_SPEED_SRCS-$(CONFIG_VP9) := test_intra_pred_speed.cc
+TEST_INTRA_PRED_SPEED_SRCS-$(CONFIG_VP9) += ../md5_utils.h ../md5_utils.c
endif # CONFIG_SHARED
diff --git a/test/test_intra_pred_speed.cc b/test/test_intra_pred_speed.cc
index c0508a7..5d59e83 100644
--- a/test/test_intra_pred_speed.cc
+++ b/test/test_intra_pred_speed.cc
@@ -7,14 +7,14 @@
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
-// Test and time VP9 intra-predictor functions
+// Test and time VPX intra-predictor functions
#include <stdio.h>
#include <string.h>
#include "third_party/googletest/src/include/gtest/gtest.h"
-#include "./vp9_rtcd.h"
+#include "./vpx_dsp_rtcd.h"
#include "test/acm_random.h"
#include "test/clear_system_state.h"
#include "test/md5_helper.h"
@@ -170,215 +170,215 @@
#define INTRA_PRED_TEST(arch, test_func, dc, dc_left, dc_top, dc_128, v, h, \
d45, d135, d117, d153, d207, d63, tm) \
TEST(arch, test_func) { \
- static const VpxPredFunc vp9_intra_pred[] = { \
+ static const VpxPredFunc vpx_intra_pred[] = { \
dc, dc_left, dc_top, dc_128, v, h, d45, \
d135, d117, d153, d207, d63, tm}; \
- test_func(vp9_intra_pred); \
+ test_func(vpx_intra_pred); \
}
// -----------------------------------------------------------------------------
// 4x4
-INTRA_PRED_TEST(C, TestIntraPred4, vp9_dc_predictor_4x4_c,
- vp9_dc_left_predictor_4x4_c, vp9_dc_top_predictor_4x4_c,
- vp9_dc_128_predictor_4x4_c, vp9_v_predictor_4x4_c,
- vp9_h_predictor_4x4_c, vp9_d45_predictor_4x4_c,
- vp9_d135_predictor_4x4_c, vp9_d117_predictor_4x4_c,
- vp9_d153_predictor_4x4_c, vp9_d207_predictor_4x4_c,
- vp9_d63_predictor_4x4_c, vp9_tm_predictor_4x4_c)
+INTRA_PRED_TEST(C, TestIntraPred4, vpx_dc_predictor_4x4_c,
+ vpx_dc_left_predictor_4x4_c, vpx_dc_top_predictor_4x4_c,
+ vpx_dc_128_predictor_4x4_c, vpx_v_predictor_4x4_c,
+ vpx_h_predictor_4x4_c, vpx_d45_predictor_4x4_c,
+ vpx_d135_predictor_4x4_c, vpx_d117_predictor_4x4_c,
+ vpx_d153_predictor_4x4_c, vpx_d207_predictor_4x4_c,
+ vpx_d63_predictor_4x4_c, vpx_tm_predictor_4x4_c)
#if HAVE_SSE && CONFIG_USE_X86INC
-INTRA_PRED_TEST(SSE, TestIntraPred4, vp9_dc_predictor_4x4_sse,
- vp9_dc_left_predictor_4x4_sse, vp9_dc_top_predictor_4x4_sse,
- vp9_dc_128_predictor_4x4_sse, vp9_v_predictor_4x4_sse, NULL,
- NULL, NULL, NULL, NULL, NULL, NULL, vp9_tm_predictor_4x4_sse)
+INTRA_PRED_TEST(SSE, TestIntraPred4, vpx_dc_predictor_4x4_sse,
+ vpx_dc_left_predictor_4x4_sse, vpx_dc_top_predictor_4x4_sse,
+ vpx_dc_128_predictor_4x4_sse, vpx_v_predictor_4x4_sse, NULL,
+ NULL, NULL, NULL, NULL, NULL, NULL, vpx_tm_predictor_4x4_sse)
#endif // HAVE_SSE && CONFIG_USE_X86INC
#if HAVE_SSSE3 && CONFIG_USE_X86INC
INTRA_PRED_TEST(SSSE3, TestIntraPred4, NULL, NULL, NULL, NULL, NULL,
- vp9_h_predictor_4x4_ssse3, vp9_d45_predictor_4x4_ssse3, NULL,
- NULL, vp9_d153_predictor_4x4_ssse3,
- vp9_d207_predictor_4x4_ssse3, vp9_d63_predictor_4x4_ssse3, NULL)
+ vpx_h_predictor_4x4_ssse3, vpx_d45_predictor_4x4_ssse3, NULL,
+ NULL, vpx_d153_predictor_4x4_ssse3,
+ vpx_d207_predictor_4x4_ssse3, vpx_d63_predictor_4x4_ssse3, NULL)
#endif // HAVE_SSSE3 && CONFIG_USE_X86INC
#if HAVE_DSPR2
-INTRA_PRED_TEST(DSPR2, TestIntraPred4, vp9_dc_predictor_4x4_dspr2, NULL, NULL,
- NULL, NULL, vp9_h_predictor_4x4_dspr2, NULL, NULL, NULL, NULL,
- NULL, NULL, vp9_tm_predictor_4x4_dspr2)
+INTRA_PRED_TEST(DSPR2, TestIntraPred4, vpx_dc_predictor_4x4_dspr2, NULL, NULL,
+ NULL, NULL, vpx_h_predictor_4x4_dspr2, NULL, NULL, NULL, NULL,
+ NULL, NULL, vpx_tm_predictor_4x4_dspr2)
#endif // HAVE_DSPR2
#if HAVE_NEON
-INTRA_PRED_TEST(NEON, TestIntraPred4, vp9_dc_predictor_4x4_neon,
- vp9_dc_left_predictor_4x4_neon, vp9_dc_top_predictor_4x4_neon,
- vp9_dc_128_predictor_4x4_neon, vp9_v_predictor_4x4_neon,
- vp9_h_predictor_4x4_neon, vp9_d45_predictor_4x4_neon,
- vp9_d135_predictor_4x4_neon, NULL, NULL, NULL, NULL,
- vp9_tm_predictor_4x4_neon)
+INTRA_PRED_TEST(NEON, TestIntraPred4, vpx_dc_predictor_4x4_neon,
+ vpx_dc_left_predictor_4x4_neon, vpx_dc_top_predictor_4x4_neon,
+ vpx_dc_128_predictor_4x4_neon, vpx_v_predictor_4x4_neon,
+ vpx_h_predictor_4x4_neon, vpx_d45_predictor_4x4_neon,
+ vpx_d135_predictor_4x4_neon, NULL, NULL, NULL, NULL,
+ vpx_tm_predictor_4x4_neon)
#endif // HAVE_NEON
#if HAVE_MSA
-INTRA_PRED_TEST(MSA, TestIntraPred4, vp9_dc_predictor_4x4_msa,
- vp9_dc_left_predictor_4x4_msa, vp9_dc_top_predictor_4x4_msa,
- vp9_dc_128_predictor_4x4_msa, vp9_v_predictor_4x4_msa,
- vp9_h_predictor_4x4_msa, NULL, NULL, NULL, NULL, NULL,
- NULL, vp9_tm_predictor_4x4_msa)
+INTRA_PRED_TEST(MSA, TestIntraPred4, vpx_dc_predictor_4x4_msa,
+ vpx_dc_left_predictor_4x4_msa, vpx_dc_top_predictor_4x4_msa,
+ vpx_dc_128_predictor_4x4_msa, vpx_v_predictor_4x4_msa,
+ vpx_h_predictor_4x4_msa, NULL, NULL, NULL, NULL, NULL,
+ NULL, vpx_tm_predictor_4x4_msa)
#endif // HAVE_MSA
// -----------------------------------------------------------------------------
// 8x8
-INTRA_PRED_TEST(C, TestIntraPred8, vp9_dc_predictor_8x8_c,
- vp9_dc_left_predictor_8x8_c, vp9_dc_top_predictor_8x8_c,
- vp9_dc_128_predictor_8x8_c, vp9_v_predictor_8x8_c,
- vp9_h_predictor_8x8_c, vp9_d45_predictor_8x8_c,
- vp9_d135_predictor_8x8_c, vp9_d117_predictor_8x8_c,
- vp9_d153_predictor_8x8_c, vp9_d207_predictor_8x8_c,
- vp9_d63_predictor_8x8_c, vp9_tm_predictor_8x8_c)
+INTRA_PRED_TEST(C, TestIntraPred8, vpx_dc_predictor_8x8_c,
+ vpx_dc_left_predictor_8x8_c, vpx_dc_top_predictor_8x8_c,
+ vpx_dc_128_predictor_8x8_c, vpx_v_predictor_8x8_c,
+ vpx_h_predictor_8x8_c, vpx_d45_predictor_8x8_c,
+ vpx_d135_predictor_8x8_c, vpx_d117_predictor_8x8_c,
+ vpx_d153_predictor_8x8_c, vpx_d207_predictor_8x8_c,
+ vpx_d63_predictor_8x8_c, vpx_tm_predictor_8x8_c)
#if HAVE_SSE && CONFIG_USE_X86INC
-INTRA_PRED_TEST(SSE, TestIntraPred8, vp9_dc_predictor_8x8_sse,
- vp9_dc_left_predictor_8x8_sse, vp9_dc_top_predictor_8x8_sse,
- vp9_dc_128_predictor_8x8_sse, vp9_v_predictor_8x8_sse, NULL,
+INTRA_PRED_TEST(SSE, TestIntraPred8, vpx_dc_predictor_8x8_sse,
+ vpx_dc_left_predictor_8x8_sse, vpx_dc_top_predictor_8x8_sse,
+ vpx_dc_128_predictor_8x8_sse, vpx_v_predictor_8x8_sse, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL)
#endif // HAVE_SSE && CONFIG_USE_X86INC
#if HAVE_SSE2 && CONFIG_USE_X86INC
INTRA_PRED_TEST(SSE2, TestIntraPred8, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
- NULL, NULL, NULL, NULL, NULL, vp9_tm_predictor_8x8_sse2)
+ NULL, NULL, NULL, NULL, NULL, vpx_tm_predictor_8x8_sse2)
#endif // HAVE_SSE2 && CONFIG_USE_X86INC
#if HAVE_SSSE3 && CONFIG_USE_X86INC
INTRA_PRED_TEST(SSSE3, TestIntraPred8, NULL, NULL, NULL, NULL, NULL,
- vp9_h_predictor_8x8_ssse3, vp9_d45_predictor_8x8_ssse3, NULL,
- NULL, vp9_d153_predictor_8x8_ssse3,
- vp9_d207_predictor_8x8_ssse3, vp9_d63_predictor_8x8_ssse3, NULL)
+ vpx_h_predictor_8x8_ssse3, vpx_d45_predictor_8x8_ssse3, NULL,
+ NULL, vpx_d153_predictor_8x8_ssse3,
+ vpx_d207_predictor_8x8_ssse3, vpx_d63_predictor_8x8_ssse3, NULL)
#endif // HAVE_SSSE3 && CONFIG_USE_X86INC
#if HAVE_DSPR2
-INTRA_PRED_TEST(DSPR2, TestIntraPred8, vp9_dc_predictor_8x8_dspr2, NULL, NULL,
- NULL, NULL, vp9_h_predictor_8x8_dspr2, NULL, NULL, NULL, NULL,
- NULL, NULL, vp9_tm_predictor_8x8_c)
+INTRA_PRED_TEST(DSPR2, TestIntraPred8, vpx_dc_predictor_8x8_dspr2, NULL, NULL,
+ NULL, NULL, vpx_h_predictor_8x8_dspr2, NULL, NULL, NULL, NULL,
+ NULL, NULL, vpx_tm_predictor_8x8_c)
#endif // HAVE_DSPR2
#if HAVE_NEON
-INTRA_PRED_TEST(NEON, TestIntraPred8, vp9_dc_predictor_8x8_neon,
- vp9_dc_left_predictor_8x8_neon, vp9_dc_top_predictor_8x8_neon,
- vp9_dc_128_predictor_8x8_neon, vp9_v_predictor_8x8_neon,
- vp9_h_predictor_8x8_neon, vp9_d45_predictor_8x8_neon, NULL,
- NULL, NULL, NULL, NULL, vp9_tm_predictor_8x8_neon)
+INTRA_PRED_TEST(NEON, TestIntraPred8, vpx_dc_predictor_8x8_neon,
+ vpx_dc_left_predictor_8x8_neon, vpx_dc_top_predictor_8x8_neon,
+ vpx_dc_128_predictor_8x8_neon, vpx_v_predictor_8x8_neon,
+ vpx_h_predictor_8x8_neon, vpx_d45_predictor_8x8_neon, NULL,
+ NULL, NULL, NULL, NULL, vpx_tm_predictor_8x8_neon)
#endif // HAVE_NEON
#if HAVE_MSA
-INTRA_PRED_TEST(MSA, TestIntraPred8, vp9_dc_predictor_8x8_msa,
- vp9_dc_left_predictor_8x8_msa, vp9_dc_top_predictor_8x8_msa,
- vp9_dc_128_predictor_8x8_msa, vp9_v_predictor_8x8_msa,
- vp9_h_predictor_8x8_msa, NULL, NULL, NULL, NULL, NULL,
- NULL, vp9_tm_predictor_8x8_msa)
+INTRA_PRED_TEST(MSA, TestIntraPred8, vpx_dc_predictor_8x8_msa,
+ vpx_dc_left_predictor_8x8_msa, vpx_dc_top_predictor_8x8_msa,
+ vpx_dc_128_predictor_8x8_msa, vpx_v_predictor_8x8_msa,
+ vpx_h_predictor_8x8_msa, NULL, NULL, NULL, NULL, NULL,
+ NULL, vpx_tm_predictor_8x8_msa)
#endif // HAVE_MSA
// -----------------------------------------------------------------------------
// 16x16
-INTRA_PRED_TEST(C, TestIntraPred16, vp9_dc_predictor_16x16_c,
- vp9_dc_left_predictor_16x16_c, vp9_dc_top_predictor_16x16_c,
- vp9_dc_128_predictor_16x16_c, vp9_v_predictor_16x16_c,
- vp9_h_predictor_16x16_c, vp9_d45_predictor_16x16_c,
- vp9_d135_predictor_16x16_c, vp9_d117_predictor_16x16_c,
- vp9_d153_predictor_16x16_c, vp9_d207_predictor_16x16_c,
- vp9_d63_predictor_16x16_c, vp9_tm_predictor_16x16_c)
+INTRA_PRED_TEST(C, TestIntraPred16, vpx_dc_predictor_16x16_c,
+ vpx_dc_left_predictor_16x16_c, vpx_dc_top_predictor_16x16_c,
+ vpx_dc_128_predictor_16x16_c, vpx_v_predictor_16x16_c,
+ vpx_h_predictor_16x16_c, vpx_d45_predictor_16x16_c,
+ vpx_d135_predictor_16x16_c, vpx_d117_predictor_16x16_c,
+ vpx_d153_predictor_16x16_c, vpx_d207_predictor_16x16_c,
+ vpx_d63_predictor_16x16_c, vpx_tm_predictor_16x16_c)
#if HAVE_SSE2 && CONFIG_USE_X86INC
-INTRA_PRED_TEST(SSE2, TestIntraPred16, vp9_dc_predictor_16x16_sse2,
- vp9_dc_left_predictor_16x16_sse2,
- vp9_dc_top_predictor_16x16_sse2,
- vp9_dc_128_predictor_16x16_sse2, vp9_v_predictor_16x16_sse2,
+INTRA_PRED_TEST(SSE2, TestIntraPred16, vpx_dc_predictor_16x16_sse2,
+ vpx_dc_left_predictor_16x16_sse2,
+ vpx_dc_top_predictor_16x16_sse2,
+ vpx_dc_128_predictor_16x16_sse2, vpx_v_predictor_16x16_sse2,
NULL, NULL, NULL, NULL, NULL, NULL, NULL,
- vp9_tm_predictor_16x16_sse2)
+ vpx_tm_predictor_16x16_sse2)
#endif // HAVE_SSE2 && CONFIG_USE_X86INC
#if HAVE_SSSE3 && CONFIG_USE_X86INC
INTRA_PRED_TEST(SSSE3, TestIntraPred16, NULL, NULL, NULL, NULL, NULL,
- vp9_h_predictor_16x16_ssse3, vp9_d45_predictor_16x16_ssse3,
- NULL, NULL, vp9_d153_predictor_16x16_ssse3,
- vp9_d207_predictor_16x16_ssse3, vp9_d63_predictor_16x16_ssse3,
+ vpx_h_predictor_16x16_ssse3, vpx_d45_predictor_16x16_ssse3,
+ NULL, NULL, vpx_d153_predictor_16x16_ssse3,
+ vpx_d207_predictor_16x16_ssse3, vpx_d63_predictor_16x16_ssse3,
NULL)
#endif // HAVE_SSSE3 && CONFIG_USE_X86INC
#if HAVE_DSPR2
-INTRA_PRED_TEST(DSPR2, TestIntraPred16, vp9_dc_predictor_16x16_dspr2, NULL,
- NULL, NULL, NULL, vp9_h_predictor_16x16_dspr2, NULL, NULL, NULL,
+INTRA_PRED_TEST(DSPR2, TestIntraPred16, vpx_dc_predictor_16x16_dspr2, NULL,
+ NULL, NULL, NULL, vpx_h_predictor_16x16_dspr2, NULL, NULL, NULL,
NULL, NULL, NULL, NULL)
#endif // HAVE_DSPR2
#if HAVE_NEON
-INTRA_PRED_TEST(NEON, TestIntraPred16, vp9_dc_predictor_16x16_neon,
- vp9_dc_left_predictor_16x16_neon,
- vp9_dc_top_predictor_16x16_neon,
- vp9_dc_128_predictor_16x16_neon, vp9_v_predictor_16x16_neon,
- vp9_h_predictor_16x16_neon, vp9_d45_predictor_16x16_neon, NULL,
- NULL, NULL, NULL, NULL, vp9_tm_predictor_16x16_neon)
+INTRA_PRED_TEST(NEON, TestIntraPred16, vpx_dc_predictor_16x16_neon,
+ vpx_dc_left_predictor_16x16_neon,
+ vpx_dc_top_predictor_16x16_neon,
+ vpx_dc_128_predictor_16x16_neon, vpx_v_predictor_16x16_neon,
+ vpx_h_predictor_16x16_neon, vpx_d45_predictor_16x16_neon, NULL,
+ NULL, NULL, NULL, NULL, vpx_tm_predictor_16x16_neon)
#endif // HAVE_NEON
#if HAVE_MSA
-INTRA_PRED_TEST(MSA, TestIntraPred16, vp9_dc_predictor_16x16_msa,
- vp9_dc_left_predictor_16x16_msa, vp9_dc_top_predictor_16x16_msa,
- vp9_dc_128_predictor_16x16_msa, vp9_v_predictor_16x16_msa,
- vp9_h_predictor_16x16_msa, NULL, NULL, NULL, NULL, NULL,
- NULL, vp9_tm_predictor_16x16_msa)
+INTRA_PRED_TEST(MSA, TestIntraPred16, vpx_dc_predictor_16x16_msa,
+ vpx_dc_left_predictor_16x16_msa, vpx_dc_top_predictor_16x16_msa,
+ vpx_dc_128_predictor_16x16_msa, vpx_v_predictor_16x16_msa,
+ vpx_h_predictor_16x16_msa, NULL, NULL, NULL, NULL, NULL,
+ NULL, vpx_tm_predictor_16x16_msa)
#endif // HAVE_MSA
// -----------------------------------------------------------------------------
// 32x32
-INTRA_PRED_TEST(C, TestIntraPred32, vp9_dc_predictor_32x32_c,
- vp9_dc_left_predictor_32x32_c, vp9_dc_top_predictor_32x32_c,
- vp9_dc_128_predictor_32x32_c, vp9_v_predictor_32x32_c,
- vp9_h_predictor_32x32_c, vp9_d45_predictor_32x32_c,
- vp9_d135_predictor_32x32_c, vp9_d117_predictor_32x32_c,
- vp9_d153_predictor_32x32_c, vp9_d207_predictor_32x32_c,
- vp9_d63_predictor_32x32_c, vp9_tm_predictor_32x32_c)
+INTRA_PRED_TEST(C, TestIntraPred32, vpx_dc_predictor_32x32_c,
+ vpx_dc_left_predictor_32x32_c, vpx_dc_top_predictor_32x32_c,
+ vpx_dc_128_predictor_32x32_c, vpx_v_predictor_32x32_c,
+ vpx_h_predictor_32x32_c, vpx_d45_predictor_32x32_c,
+ vpx_d135_predictor_32x32_c, vpx_d117_predictor_32x32_c,
+ vpx_d153_predictor_32x32_c, vpx_d207_predictor_32x32_c,
+ vpx_d63_predictor_32x32_c, vpx_tm_predictor_32x32_c)
#if HAVE_SSE2 && CONFIG_USE_X86INC
#if ARCH_X86_64
-INTRA_PRED_TEST(SSE2, TestIntraPred32, vp9_dc_predictor_32x32_sse2,
- vp9_dc_left_predictor_32x32_sse2,
- vp9_dc_top_predictor_32x32_sse2,
- vp9_dc_128_predictor_32x32_sse2, vp9_v_predictor_32x32_sse2,
+INTRA_PRED_TEST(SSE2, TestIntraPred32, vpx_dc_predictor_32x32_sse2,
+ vpx_dc_left_predictor_32x32_sse2,
+ vpx_dc_top_predictor_32x32_sse2,
+ vpx_dc_128_predictor_32x32_sse2, vpx_v_predictor_32x32_sse2,
NULL, NULL, NULL, NULL, NULL, NULL, NULL,
- vp9_tm_predictor_32x32_sse2)
+ vpx_tm_predictor_32x32_sse2)
#else
-INTRA_PRED_TEST(SSE2, TestIntraPred32, vp9_dc_predictor_32x32_sse2,
- vp9_dc_left_predictor_32x32_sse2,
- vp9_dc_top_predictor_32x32_sse2,
- vp9_dc_128_predictor_32x32_sse2, vp9_v_predictor_32x32_sse2,
+INTRA_PRED_TEST(SSE2, TestIntraPred32, vpx_dc_predictor_32x32_sse2,
+ vpx_dc_left_predictor_32x32_sse2,
+ vpx_dc_top_predictor_32x32_sse2,
+ vpx_dc_128_predictor_32x32_sse2, vpx_v_predictor_32x32_sse2,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL)
#endif // ARCH_X86_64
#endif // HAVE_SSE2 && CONFIG_USE_X86INC
#if HAVE_SSSE3 && CONFIG_USE_X86INC
INTRA_PRED_TEST(SSSE3, TestIntraPred32, NULL, NULL, NULL, NULL, NULL,
- vp9_h_predictor_32x32_ssse3, vp9_d45_predictor_32x32_ssse3,
- NULL, NULL, vp9_d153_predictor_32x32_ssse3,
- vp9_d207_predictor_32x32_ssse3, vp9_d63_predictor_32x32_ssse3,
+ vpx_h_predictor_32x32_ssse3, vpx_d45_predictor_32x32_ssse3,
+ NULL, NULL, vpx_d153_predictor_32x32_ssse3,
+ vpx_d207_predictor_32x32_ssse3, vpx_d63_predictor_32x32_ssse3,
NULL)
#endif // HAVE_SSSE3 && CONFIG_USE_X86INC
#if HAVE_NEON
-INTRA_PRED_TEST(NEON, TestIntraPred32, vp9_dc_predictor_32x32_neon,
- vp9_dc_left_predictor_32x32_neon,
- vp9_dc_top_predictor_32x32_neon,
- vp9_dc_128_predictor_32x32_neon, vp9_v_predictor_32x32_neon,
- vp9_h_predictor_32x32_neon, NULL, NULL, NULL, NULL, NULL, NULL,
- vp9_tm_predictor_32x32_neon)
+INTRA_PRED_TEST(NEON, TestIntraPred32, vpx_dc_predictor_32x32_neon,
+ vpx_dc_left_predictor_32x32_neon,
+ vpx_dc_top_predictor_32x32_neon,
+ vpx_dc_128_predictor_32x32_neon, vpx_v_predictor_32x32_neon,
+ vpx_h_predictor_32x32_neon, NULL, NULL, NULL, NULL, NULL, NULL,
+ vpx_tm_predictor_32x32_neon)
#endif // HAVE_NEON
#if HAVE_MSA
-INTRA_PRED_TEST(MSA, TestIntraPred32, vp9_dc_predictor_32x32_msa,
- vp9_dc_left_predictor_32x32_msa, vp9_dc_top_predictor_32x32_msa,
- vp9_dc_128_predictor_32x32_msa, vp9_v_predictor_32x32_msa,
- vp9_h_predictor_32x32_msa, NULL, NULL, NULL, NULL, NULL,
- NULL, vp9_tm_predictor_32x32_msa)
+INTRA_PRED_TEST(MSA, TestIntraPred32, vpx_dc_predictor_32x32_msa,
+ vpx_dc_left_predictor_32x32_msa, vpx_dc_top_predictor_32x32_msa,
+ vpx_dc_128_predictor_32x32_msa, vpx_v_predictor_32x32_msa,
+ vpx_h_predictor_32x32_msa, NULL, NULL, NULL, NULL, NULL,
+ NULL, vpx_tm_predictor_32x32_msa)
#endif // HAVE_MSA
#include "test/test_libvpx.cc"
diff --git a/test/vp9_intrapred_test.cc b/test/vp9_intrapred_test.cc
index 9533750..c17259a 100644
--- a/test/vp9_intrapred_test.cc
+++ b/test/vp9_intrapred_test.cc
@@ -16,7 +16,7 @@
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vpx_config.h"
-#include "./vp9_rtcd.h"
+#include "./vpx_dsp_rtcd.h"
#include "vp9/common/vp9_blockd.h"
#include "vp9/common/vp9_pred_common.h"
#include "vpx_mem/vpx_mem.h"
@@ -135,165 +135,165 @@
#if ARCH_X86_64
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_8, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_highbd_dc_predictor_32x32_sse2,
- &vp9_highbd_dc_predictor_32x32_c, 32, 8),
- make_tuple(&vp9_highbd_tm_predictor_16x16_sse2,
- &vp9_highbd_tm_predictor_16x16_c, 16, 8),
- make_tuple(&vp9_highbd_tm_predictor_32x32_sse2,
- &vp9_highbd_tm_predictor_32x32_c, 32, 8),
- make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
- &vp9_highbd_dc_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
- &vp9_highbd_dc_predictor_8x8_c, 8, 8),
- make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
- &vp9_highbd_dc_predictor_16x16_c, 16, 8),
- make_tuple(&vp9_highbd_v_predictor_4x4_sse,
- &vp9_highbd_v_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
- &vp9_highbd_v_predictor_8x8_c, 8, 8),
- make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
- &vp9_highbd_v_predictor_16x16_c, 16, 8),
- make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
- &vp9_highbd_v_predictor_32x32_c, 32, 8),
- make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
- &vp9_highbd_tm_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
- &vp9_highbd_tm_predictor_8x8_c, 8, 8)));
+ make_tuple(&vpx_highbd_dc_predictor_32x32_sse2,
+ &vpx_highbd_dc_predictor_32x32_c, 32, 8),
+ make_tuple(&vpx_highbd_tm_predictor_16x16_sse2,
+ &vpx_highbd_tm_predictor_16x16_c, 16, 8),
+ make_tuple(&vpx_highbd_tm_predictor_32x32_sse2,
+ &vpx_highbd_tm_predictor_32x32_c, 32, 8),
+ make_tuple(&vpx_highbd_dc_predictor_4x4_sse,
+ &vpx_highbd_dc_predictor_4x4_c, 4, 8),
+ make_tuple(&vpx_highbd_dc_predictor_8x8_sse2,
+ &vpx_highbd_dc_predictor_8x8_c, 8, 8),
+ make_tuple(&vpx_highbd_dc_predictor_16x16_sse2,
+ &vpx_highbd_dc_predictor_16x16_c, 16, 8),
+ make_tuple(&vpx_highbd_v_predictor_4x4_sse,
+ &vpx_highbd_v_predictor_4x4_c, 4, 8),
+ make_tuple(&vpx_highbd_v_predictor_8x8_sse2,
+ &vpx_highbd_v_predictor_8x8_c, 8, 8),
+ make_tuple(&vpx_highbd_v_predictor_16x16_sse2,
+ &vpx_highbd_v_predictor_16x16_c, 16, 8),
+ make_tuple(&vpx_highbd_v_predictor_32x32_sse2,
+ &vpx_highbd_v_predictor_32x32_c, 32, 8),
+ make_tuple(&vpx_highbd_tm_predictor_4x4_sse,
+ &vpx_highbd_tm_predictor_4x4_c, 4, 8),
+ make_tuple(&vpx_highbd_tm_predictor_8x8_sse2,
+ &vpx_highbd_tm_predictor_8x8_c, 8, 8)));
#else
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_8, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
- &vp9_highbd_dc_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
- &vp9_highbd_dc_predictor_8x8_c, 8, 8),
- make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
- &vp9_highbd_dc_predictor_16x16_c, 16, 8),
- make_tuple(&vp9_highbd_v_predictor_4x4_sse,
- &vp9_highbd_v_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
- &vp9_highbd_v_predictor_8x8_c, 8, 8),
- make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
- &vp9_highbd_v_predictor_16x16_c, 16, 8),
- make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
- &vp9_highbd_v_predictor_32x32_c, 32, 8),
- make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
- &vp9_highbd_tm_predictor_4x4_c, 4, 8),
- make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
- &vp9_highbd_tm_predictor_8x8_c, 8, 8)));
+ make_tuple(&vpx_highbd_dc_predictor_4x4_sse,
+ &vpx_highbd_dc_predictor_4x4_c, 4, 8),
+ make_tuple(&vpx_highbd_dc_predictor_8x8_sse2,
+ &vpx_highbd_dc_predictor_8x8_c, 8, 8),
+ make_tuple(&vpx_highbd_dc_predictor_16x16_sse2,
+ &vpx_highbd_dc_predictor_16x16_c, 16, 8),
+ make_tuple(&vpx_highbd_v_predictor_4x4_sse,
+ &vpx_highbd_v_predictor_4x4_c, 4, 8),
+ make_tuple(&vpx_highbd_v_predictor_8x8_sse2,
+ &vpx_highbd_v_predictor_8x8_c, 8, 8),
+ make_tuple(&vpx_highbd_v_predictor_16x16_sse2,
+ &vpx_highbd_v_predictor_16x16_c, 16, 8),
+ make_tuple(&vpx_highbd_v_predictor_32x32_sse2,
+ &vpx_highbd_v_predictor_32x32_c, 32, 8),
+ make_tuple(&vpx_highbd_tm_predictor_4x4_sse,
+ &vpx_highbd_tm_predictor_4x4_c, 4, 8),
+ make_tuple(&vpx_highbd_tm_predictor_8x8_sse2,
+ &vpx_highbd_tm_predictor_8x8_c, 8, 8)));
#endif // !ARCH_X86_64
#if ARCH_X86_64
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_10, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_highbd_dc_predictor_32x32_sse2,
- &vp9_highbd_dc_predictor_32x32_c, 32,
+ make_tuple(&vpx_highbd_dc_predictor_32x32_sse2,
+ &vpx_highbd_dc_predictor_32x32_c, 32,
10),
- make_tuple(&vp9_highbd_tm_predictor_16x16_sse2,
- &vp9_highbd_tm_predictor_16x16_c, 16,
+ make_tuple(&vpx_highbd_tm_predictor_16x16_sse2,
+ &vpx_highbd_tm_predictor_16x16_c, 16,
10),
- make_tuple(&vp9_highbd_tm_predictor_32x32_sse2,
- &vp9_highbd_tm_predictor_32x32_c, 32,
+ make_tuple(&vpx_highbd_tm_predictor_32x32_sse2,
+ &vpx_highbd_tm_predictor_32x32_c, 32,
10),
- make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
- &vp9_highbd_dc_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
- &vp9_highbd_dc_predictor_8x8_c, 8, 10),
- make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
- &vp9_highbd_dc_predictor_16x16_c, 16,
+ make_tuple(&vpx_highbd_dc_predictor_4x4_sse,
+ &vpx_highbd_dc_predictor_4x4_c, 4, 10),
+ make_tuple(&vpx_highbd_dc_predictor_8x8_sse2,
+ &vpx_highbd_dc_predictor_8x8_c, 8, 10),
+ make_tuple(&vpx_highbd_dc_predictor_16x16_sse2,
+ &vpx_highbd_dc_predictor_16x16_c, 16,
10),
- make_tuple(&vp9_highbd_v_predictor_4x4_sse,
- &vp9_highbd_v_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
- &vp9_highbd_v_predictor_8x8_c, 8, 10),
- make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
- &vp9_highbd_v_predictor_16x16_c, 16,
+ make_tuple(&vpx_highbd_v_predictor_4x4_sse,
+ &vpx_highbd_v_predictor_4x4_c, 4, 10),
+ make_tuple(&vpx_highbd_v_predictor_8x8_sse2,
+ &vpx_highbd_v_predictor_8x8_c, 8, 10),
+ make_tuple(&vpx_highbd_v_predictor_16x16_sse2,
+ &vpx_highbd_v_predictor_16x16_c, 16,
10),
- make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
- &vp9_highbd_v_predictor_32x32_c, 32,
+ make_tuple(&vpx_highbd_v_predictor_32x32_sse2,
+ &vpx_highbd_v_predictor_32x32_c, 32,
10),
- make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
- &vp9_highbd_tm_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
- &vp9_highbd_tm_predictor_8x8_c, 8, 10)));
+ make_tuple(&vpx_highbd_tm_predictor_4x4_sse,
+ &vpx_highbd_tm_predictor_4x4_c, 4, 10),
+ make_tuple(&vpx_highbd_tm_predictor_8x8_sse2,
+ &vpx_highbd_tm_predictor_8x8_c, 8, 10)));
#else
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_10, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
- &vp9_highbd_dc_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
- &vp9_highbd_dc_predictor_8x8_c, 8, 10),
- make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
- &vp9_highbd_dc_predictor_16x16_c, 16,
+ make_tuple(&vpx_highbd_dc_predictor_4x4_sse,
+ &vpx_highbd_dc_predictor_4x4_c, 4, 10),
+ make_tuple(&vpx_highbd_dc_predictor_8x8_sse2,
+ &vpx_highbd_dc_predictor_8x8_c, 8, 10),
+ make_tuple(&vpx_highbd_dc_predictor_16x16_sse2,
+ &vpx_highbd_dc_predictor_16x16_c, 16,
10),
- make_tuple(&vp9_highbd_v_predictor_4x4_sse,
- &vp9_highbd_v_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
- &vp9_highbd_v_predictor_8x8_c, 8, 10),
- make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
- &vp9_highbd_v_predictor_16x16_c, 16, 10),
- make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
- &vp9_highbd_v_predictor_32x32_c, 32, 10),
- make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
- &vp9_highbd_tm_predictor_4x4_c, 4, 10),
- make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
- &vp9_highbd_tm_predictor_8x8_c, 8, 10)));
+ make_tuple(&vpx_highbd_v_predictor_4x4_sse,
+ &vpx_highbd_v_predictor_4x4_c, 4, 10),
+ make_tuple(&vpx_highbd_v_predictor_8x8_sse2,
+ &vpx_highbd_v_predictor_8x8_c, 8, 10),
+ make_tuple(&vpx_highbd_v_predictor_16x16_sse2,
+ &vpx_highbd_v_predictor_16x16_c, 16, 10),
+ make_tuple(&vpx_highbd_v_predictor_32x32_sse2,
+ &vpx_highbd_v_predictor_32x32_c, 32, 10),
+ make_tuple(&vpx_highbd_tm_predictor_4x4_sse,
+ &vpx_highbd_tm_predictor_4x4_c, 4, 10),
+ make_tuple(&vpx_highbd_tm_predictor_8x8_sse2,
+ &vpx_highbd_tm_predictor_8x8_c, 8, 10)));
#endif // !ARCH_X86_64
#if ARCH_X86_64
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_12, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_highbd_dc_predictor_32x32_sse2,
- &vp9_highbd_dc_predictor_32x32_c, 32,
+ make_tuple(&vpx_highbd_dc_predictor_32x32_sse2,
+ &vpx_highbd_dc_predictor_32x32_c, 32,
12),
- make_tuple(&vp9_highbd_tm_predictor_16x16_sse2,
- &vp9_highbd_tm_predictor_16x16_c, 16,
+ make_tuple(&vpx_highbd_tm_predictor_16x16_sse2,
+ &vpx_highbd_tm_predictor_16x16_c, 16,
12),
- make_tuple(&vp9_highbd_tm_predictor_32x32_sse2,
- &vp9_highbd_tm_predictor_32x32_c, 32,
+ make_tuple(&vpx_highbd_tm_predictor_32x32_sse2,
+ &vpx_highbd_tm_predictor_32x32_c, 32,
12),
- make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
- &vp9_highbd_dc_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
- &vp9_highbd_dc_predictor_8x8_c, 8, 12),
- make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
- &vp9_highbd_dc_predictor_16x16_c, 16,
+ make_tuple(&vpx_highbd_dc_predictor_4x4_sse,
+ &vpx_highbd_dc_predictor_4x4_c, 4, 12),
+ make_tuple(&vpx_highbd_dc_predictor_8x8_sse2,
+ &vpx_highbd_dc_predictor_8x8_c, 8, 12),
+ make_tuple(&vpx_highbd_dc_predictor_16x16_sse2,
+ &vpx_highbd_dc_predictor_16x16_c, 16,
12),
- make_tuple(&vp9_highbd_v_predictor_4x4_sse,
- &vp9_highbd_v_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
- &vp9_highbd_v_predictor_8x8_c, 8, 12),
- make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
- &vp9_highbd_v_predictor_16x16_c, 16,
+ make_tuple(&vpx_highbd_v_predictor_4x4_sse,
+ &vpx_highbd_v_predictor_4x4_c, 4, 12),
+ make_tuple(&vpx_highbd_v_predictor_8x8_sse2,
+ &vpx_highbd_v_predictor_8x8_c, 8, 12),
+ make_tuple(&vpx_highbd_v_predictor_16x16_sse2,
+ &vpx_highbd_v_predictor_16x16_c, 16,
12),
- make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
- &vp9_highbd_v_predictor_32x32_c, 32,
+ make_tuple(&vpx_highbd_v_predictor_32x32_sse2,
+ &vpx_highbd_v_predictor_32x32_c, 32,
12),
- make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
- &vp9_highbd_tm_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
- &vp9_highbd_tm_predictor_8x8_c, 8, 12)));
+ make_tuple(&vpx_highbd_tm_predictor_4x4_sse,
+ &vpx_highbd_tm_predictor_4x4_c, 4, 12),
+ make_tuple(&vpx_highbd_tm_predictor_8x8_sse2,
+ &vpx_highbd_tm_predictor_8x8_c, 8, 12)));
#else
INSTANTIATE_TEST_CASE_P(SSE2_TO_C_12, VP9IntraPredTest,
::testing::Values(
- make_tuple(&vp9_highbd_dc_predictor_4x4_sse,
- &vp9_highbd_dc_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_highbd_dc_predictor_8x8_sse2,
- &vp9_highbd_dc_predictor_8x8_c, 8, 12),
- make_tuple(&vp9_highbd_dc_predictor_16x16_sse2,
- &vp9_highbd_dc_predictor_16x16_c, 16,
+ make_tuple(&vpx_highbd_dc_predictor_4x4_sse,
+ &vpx_highbd_dc_predictor_4x4_c, 4, 12),
+ make_tuple(&vpx_highbd_dc_predictor_8x8_sse2,
+ &vpx_highbd_dc_predictor_8x8_c, 8, 12),
+ make_tuple(&vpx_highbd_dc_predictor_16x16_sse2,
+ &vpx_highbd_dc_predictor_16x16_c, 16,
12),
- make_tuple(&vp9_highbd_v_predictor_4x4_sse,
- &vp9_highbd_v_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_highbd_v_predictor_8x8_sse2,
- &vp9_highbd_v_predictor_8x8_c, 8, 12),
- make_tuple(&vp9_highbd_v_predictor_16x16_sse2,
- &vp9_highbd_v_predictor_16x16_c, 16, 12),
- make_tuple(&vp9_highbd_v_predictor_32x32_sse2,
- &vp9_highbd_v_predictor_32x32_c, 32, 12),
- make_tuple(&vp9_highbd_tm_predictor_4x4_sse,
- &vp9_highbd_tm_predictor_4x4_c, 4, 12),
- make_tuple(&vp9_highbd_tm_predictor_8x8_sse2,
- &vp9_highbd_tm_predictor_8x8_c, 8, 12)));
+ make_tuple(&vpx_highbd_v_predictor_4x4_sse,
+ &vpx_highbd_v_predictor_4x4_c, 4, 12),
+ make_tuple(&vpx_highbd_v_predictor_8x8_sse2,
+ &vpx_highbd_v_predictor_8x8_c, 8, 12),
+ make_tuple(&vpx_highbd_v_predictor_16x16_sse2,
+ &vpx_highbd_v_predictor_16x16_c, 16, 12),
+ make_tuple(&vpx_highbd_v_predictor_32x32_sse2,
+ &vpx_highbd_v_predictor_32x32_c, 32, 12),
+ make_tuple(&vpx_highbd_tm_predictor_4x4_sse,
+ &vpx_highbd_tm_predictor_4x4_c, 4, 12),
+ make_tuple(&vpx_highbd_tm_predictor_8x8_sse2,
+ &vpx_highbd_tm_predictor_8x8_c, 8, 12)));
#endif // !ARCH_X86_64
#endif // CONFIG_USE_X86INC
#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vp9/common/mips/dspr2/vp9_itrans16_dspr2.c b/vp9/common/mips/dspr2/vp9_itrans16_dspr2.c
index cccddc7..c4fa2a0 100644
--- a/vp9/common/mips/dspr2/vp9_itrans16_dspr2.c
+++ b/vp9/common/mips/dspr2/vp9_itrans16_dspr2.c
@@ -15,6 +15,7 @@
#include "./vp9_rtcd.h"
#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_blockd.h"
+#include "vp9/common/vp9_idct.h"
#include "vp9/common/mips/dspr2/vp9_common_dspr2.h"
#include "vpx_dsp/txfm_common.h"
#include "vpx_ports/mem.h"
diff --git a/vp9/common/mips/dspr2/vp9_itrans4_dspr2.c b/vp9/common/mips/dspr2/vp9_itrans4_dspr2.c
index 1a1afc5..a0b77f5 100644
--- a/vp9/common/mips/dspr2/vp9_itrans4_dspr2.c
+++ b/vp9/common/mips/dspr2/vp9_itrans4_dspr2.c
@@ -15,6 +15,7 @@
#include "./vp9_rtcd.h"
#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_blockd.h"
+#include "vp9/common/vp9_idct.h"
#include "vp9/common/mips/dspr2/vp9_common_dspr2.h"
#include "vpx_dsp/txfm_common.h"
#include "vpx_ports/mem.h"
diff --git a/vp9/common/vp9_blockd.c b/vp9/common/vp9_blockd.c
index b2bb181..e8334fc 100644
--- a/vp9/common/vp9_blockd.c
+++ b/vp9/common/vp9_blockd.c
@@ -59,16 +59,17 @@
xd->mb_to_right_edge >> (5 + pd->subsampling_x));
const int max_blocks_high = num_4x4_h + (xd->mb_to_bottom_edge >= 0 ? 0 :
xd->mb_to_bottom_edge >> (5 + pd->subsampling_y));
+ const int extra_step = ((num_4x4_w - max_blocks_wide) >> tx_size) * step;
// Keep track of the row and column of the blocks we use so that we know
// if we are in the unrestricted motion border.
for (r = 0; r < max_blocks_high; r += (1 << tx_size)) {
- for (c = 0; c < num_4x4_w; c += (1 << tx_size)) {
- // Skip visiting the sub blocks that are wholly within the UMV.
- if (c < max_blocks_wide)
- visit(plane, i, plane_bsize, tx_size, arg);
+ // Skip visiting the sub blocks that are wholly within the UMV.
+ for (c = 0; c < max_blocks_wide; c += (1 << tx_size)) {
+ visit(plane, i, plane_bsize, tx_size, arg);
i += step;
}
+ i += extra_step;
}
}
diff --git a/vp9/common/vp9_common.h b/vp9/common/vp9_common.h
index 42c3a09..c249ad4 100644
--- a/vp9/common/vp9_common.h
+++ b/vp9/common/vp9_common.h
@@ -44,20 +44,6 @@
return num_values > 0 ? get_msb(num_values) + 1 : 0;
}
-#if CONFIG_VP9_HIGHBITDEPTH
-static INLINE uint16_t clip_pixel_highbd(int val, int bd) {
- switch (bd) {
- case 8:
- default:
- return (uint16_t)clamp(val, 0, 255);
- case 10:
- return (uint16_t)clamp(val, 0, 1023);
- case 12:
- return (uint16_t)clamp(val, 0, 4095);
- }
-}
-#endif // CONFIG_VP9_HIGHBITDEPTH
-
#if CONFIG_DEBUG
#define CHECK_MEM_ERROR(cm, lval, expr) do { \
lval = (expr); \
diff --git a/vp9/common/vp9_reconintra.c b/vp9/common/vp9_reconintra.c
index f969ff1..d87a043 100644
--- a/vp9/common/vp9_reconintra.c
+++ b/vp9/common/vp9_reconintra.c
@@ -9,7 +9,7 @@
*/
#include "./vpx_config.h"
-#include "./vp9_rtcd.h"
+#include "./vpx_dsp_rtcd.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/mem.h"
@@ -50,683 +50,6 @@
NEED_LEFT | NEED_ABOVE, // TM
};
-// This serves as a wrapper function, so that all the prediction functions
-// can be unified and accessed as a pointer array. Note that the boundary
-// above and left are not necessarily used all the time.
-#define intra_pred_sized(type, size) \
- void vp9_##type##_predictor_##size##x##size##_c(uint8_t *dst, \
- ptrdiff_t stride, \
- const uint8_t *above, \
- const uint8_t *left) { \
- type##_predictor(dst, stride, size, above, left); \
- }
-
-#if CONFIG_VP9_HIGHBITDEPTH
-#define intra_pred_highbd_sized(type, size) \
- void vp9_highbd_##type##_predictor_##size##x##size##_c( \
- uint16_t *dst, ptrdiff_t stride, const uint16_t *above, \
- const uint16_t *left, int bd) { \
- highbd_##type##_predictor(dst, stride, size, above, left, bd); \
- }
-
-#define intra_pred_allsizes(type) \
- intra_pred_sized(type, 4) \
- intra_pred_sized(type, 8) \
- intra_pred_sized(type, 16) \
- intra_pred_sized(type, 32) \
- intra_pred_highbd_sized(type, 4) \
- intra_pred_highbd_sized(type, 8) \
- intra_pred_highbd_sized(type, 16) \
- intra_pred_highbd_sized(type, 32)
-
-#define intra_pred_no_4x4(type) \
- intra_pred_sized(type, 8) \
- intra_pred_sized(type, 16) \
- intra_pred_sized(type, 32) \
- intra_pred_highbd_sized(type, 4) \
- intra_pred_highbd_sized(type, 8) \
- intra_pred_highbd_sized(type, 16) \
- intra_pred_highbd_sized(type, 32)
-
-#else
-
-#define intra_pred_allsizes(type) \
- intra_pred_sized(type, 4) \
- intra_pred_sized(type, 8) \
- intra_pred_sized(type, 16) \
- intra_pred_sized(type, 32)
-
-#define intra_pred_no_4x4(type) \
- intra_pred_sized(type, 8) \
- intra_pred_sized(type, 16) \
- intra_pred_sized(type, 32)
-#endif // CONFIG_VP9_HIGHBITDEPTH
-
-#define DST(x, y) dst[(x) + (y) * stride]
-#define AVG3(a, b, c) (((a) + 2 * (b) + (c) + 2) >> 2)
-#define AVG2(a, b) (((a) + (b) + 1) >> 1)
-
-#if CONFIG_VP9_HIGHBITDEPTH
-static INLINE void highbd_d207_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int r, c;
- (void) above;
- (void) bd;
-
- // First column.
- for (r = 0; r < bs - 1; ++r) {
- dst[r * stride] = AVG2(left[r], left[r + 1]);
- }
- dst[(bs - 1) * stride] = left[bs - 1];
- dst++;
-
- // Second column.
- for (r = 0; r < bs - 2; ++r) {
- dst[r * stride] = AVG3(left[r], left[r + 1], left[r + 2]);
- }
- dst[(bs - 2) * stride] = AVG3(left[bs - 2], left[bs - 1], left[bs - 1]);
- dst[(bs - 1) * stride] = left[bs - 1];
- dst++;
-
- // Rest of last row.
- for (c = 0; c < bs - 2; ++c)
- dst[(bs - 1) * stride + c] = left[bs - 1];
-
- for (r = bs - 2; r >= 0; --r) {
- for (c = 0; c < bs - 2; ++c)
- dst[r * stride + c] = dst[(r + 1) * stride + c - 2];
- }
-}
-
-static INLINE void highbd_d63_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int r, c;
- (void) left;
- (void) bd;
- for (r = 0; r < bs; ++r) {
- for (c = 0; c < bs; ++c) {
- dst[c] = r & 1 ? AVG3(above[(r >> 1) + c], above[(r >> 1) + c + 1],
- above[(r >> 1) + c + 2])
- : AVG2(above[(r >> 1) + c], above[(r >> 1) + c + 1]);
- }
- dst += stride;
- }
-}
-
-static INLINE void highbd_d45_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
- const uint16_t *above,
- const uint16_t *left, int bd) {
- int r, c;
- (void) left;
- (void) bd;
- for (r = 0; r < bs; ++r) {
- for (c = 0; c < bs; ++c) {
- dst[c] = r + c + 2 < bs * 2 ? AVG3(above[r + c], above[r + c + 1],
- above[r + c + 2])
- : above[bs * 2 - 1];
- }
- dst += stride;
- }
-}
-
-static INLINE void highbd_d117_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int r, c;
- (void) bd;
-
- // first row
- for (c = 0; c < bs; c++)
- dst[c] = AVG2(above[c - 1], above[c]);
- dst += stride;
-
- // second row
- dst[0] = AVG3(left[0], above[-1], above[0]);
- for (c = 1; c < bs; c++)
- dst[c] = AVG3(above[c - 2], above[c - 1], above[c]);
- dst += stride;
-
- // the rest of first col
- dst[0] = AVG3(above[-1], left[0], left[1]);
- for (r = 3; r < bs; ++r)
- dst[(r - 2) * stride] = AVG3(left[r - 3], left[r - 2], left[r - 1]);
-
- // the rest of the block
- for (r = 2; r < bs; ++r) {
- for (c = 1; c < bs; c++)
- dst[c] = dst[-2 * stride + c - 1];
- dst += stride;
- }
-}
-
-static INLINE void highbd_d135_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int r, c;
- (void) bd;
- dst[0] = AVG3(left[0], above[-1], above[0]);
- for (c = 1; c < bs; c++)
- dst[c] = AVG3(above[c - 2], above[c - 1], above[c]);
-
- dst[stride] = AVG3(above[-1], left[0], left[1]);
- for (r = 2; r < bs; ++r)
- dst[r * stride] = AVG3(left[r - 2], left[r - 1], left[r]);
-
- dst += stride;
- for (r = 1; r < bs; ++r) {
- for (c = 1; c < bs; c++)
- dst[c] = dst[-stride + c - 1];
- dst += stride;
- }
-}
-
-static INLINE void highbd_d153_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int r, c;
- (void) bd;
- dst[0] = AVG2(above[-1], left[0]);
- for (r = 1; r < bs; r++)
- dst[r * stride] = AVG2(left[r - 1], left[r]);
- dst++;
-
- dst[0] = AVG3(left[0], above[-1], above[0]);
- dst[stride] = AVG3(above[-1], left[0], left[1]);
- for (r = 2; r < bs; r++)
- dst[r * stride] = AVG3(left[r - 2], left[r - 1], left[r]);
- dst++;
-
- for (c = 0; c < bs - 2; c++)
- dst[c] = AVG3(above[c - 1], above[c], above[c + 1]);
- dst += stride;
-
- for (r = 1; r < bs; ++r) {
- for (c = 0; c < bs - 2; c++)
- dst[c] = dst[-stride + c - 2];
- dst += stride;
- }
-}
-
-static INLINE void highbd_v_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int r;
- (void) left;
- (void) bd;
- for (r = 0; r < bs; r++) {
- memcpy(dst, above, bs * sizeof(uint16_t));
- dst += stride;
- }
-}
-
-static INLINE void highbd_h_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int r;
- (void) above;
- (void) bd;
- for (r = 0; r < bs; r++) {
- vpx_memset16(dst, left[r], bs);
- dst += stride;
- }
-}
-
-static INLINE void highbd_tm_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int r, c;
- int ytop_left = above[-1];
- (void) bd;
-
- for (r = 0; r < bs; r++) {
- for (c = 0; c < bs; c++)
- dst[c] = clip_pixel_highbd(left[r] + above[c] - ytop_left, bd);
- dst += stride;
- }
-}
-
-static INLINE void highbd_dc_128_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int r;
- (void) above;
- (void) left;
-
- for (r = 0; r < bs; r++) {
- vpx_memset16(dst, 128 << (bd - 8), bs);
- dst += stride;
- }
-}
-
-static INLINE void highbd_dc_left_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int i, r, expected_dc, sum = 0;
- (void) above;
- (void) bd;
-
- for (i = 0; i < bs; i++)
- sum += left[i];
- expected_dc = (sum + (bs >> 1)) / bs;
-
- for (r = 0; r < bs; r++) {
- vpx_memset16(dst, expected_dc, bs);
- dst += stride;
- }
-}
-
-static INLINE void highbd_dc_top_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int i, r, expected_dc, sum = 0;
- (void) left;
- (void) bd;
-
- for (i = 0; i < bs; i++)
- sum += above[i];
- expected_dc = (sum + (bs >> 1)) / bs;
-
- for (r = 0; r < bs; r++) {
- vpx_memset16(dst, expected_dc, bs);
- dst += stride;
- }
-}
-
-static INLINE void highbd_dc_predictor(uint16_t *dst, ptrdiff_t stride,
- int bs, const uint16_t *above,
- const uint16_t *left, int bd) {
- int i, r, expected_dc, sum = 0;
- const int count = 2 * bs;
- (void) bd;
-
- for (i = 0; i < bs; i++) {
- sum += above[i];
- sum += left[i];
- }
-
- expected_dc = (sum + (count >> 1)) / count;
-
- for (r = 0; r < bs; r++) {
- vpx_memset16(dst, expected_dc, bs);
- dst += stride;
- }
-}
-#endif // CONFIG_VP9_HIGHBITDEPTH
-
-void vp9_d207_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
- const uint8_t *above, const uint8_t *left) {
- const int I = left[0];
- const int J = left[1];
- const int K = left[2];
- const int L = left[3];
- (void)above;
- DST(0, 0) = AVG2(I, J);
- DST(2, 0) = DST(0, 1) = AVG2(J, K);
- DST(2, 1) = DST(0, 2) = AVG2(K, L);
- DST(1, 0) = AVG3(I, J, K);
- DST(3, 0) = DST(1, 1) = AVG3(J, K, L);
- DST(3, 1) = DST(1, 2) = AVG3(K, L, L);
- DST(3, 2) = DST(2, 2) =
- DST(0, 3) = DST(1, 3) = DST(2, 3) = DST(3, 3) = L;
-}
-
-static INLINE void d207_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- int r, c;
- (void) above;
- // first column
- for (r = 0; r < bs - 1; ++r)
- dst[r * stride] = AVG2(left[r], left[r + 1]);
- dst[(bs - 1) * stride] = left[bs - 1];
- dst++;
-
- // second column
- for (r = 0; r < bs - 2; ++r)
- dst[r * stride] = AVG3(left[r], left[r + 1], left[r + 2]);
- dst[(bs - 2) * stride] = AVG3(left[bs - 2], left[bs - 1], left[bs - 1]);
- dst[(bs - 1) * stride] = left[bs - 1];
- dst++;
-
- // rest of last row
- for (c = 0; c < bs - 2; ++c)
- dst[(bs - 1) * stride + c] = left[bs - 1];
-
- for (r = bs - 2; r >= 0; --r)
- for (c = 0; c < bs - 2; ++c)
- dst[r * stride + c] = dst[(r + 1) * stride + c - 2];
-}
-intra_pred_no_4x4(d207)
-
-void vp9_d63_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
- const uint8_t *above, const uint8_t *left) {
- const int A = above[0];
- const int B = above[1];
- const int C = above[2];
- const int D = above[3];
- const int E = above[4];
- const int F = above[5];
- const int G = above[6];
- (void)left;
- DST(0, 0) = AVG2(A, B);
- DST(1, 0) = DST(0, 2) = AVG2(B, C);
- DST(2, 0) = DST(1, 2) = AVG2(C, D);
- DST(3, 0) = DST(2, 2) = AVG2(D, E);
- DST(3, 2) = AVG2(E, F); // differs from vp8
-
- DST(0, 1) = AVG3(A, B, C);
- DST(1, 1) = DST(0, 3) = AVG3(B, C, D);
- DST(2, 1) = DST(1, 3) = AVG3(C, D, E);
- DST(3, 1) = DST(2, 3) = AVG3(D, E, F);
- DST(3, 3) = AVG3(E, F, G); // differs from vp8
-}
-
-static INLINE void d63_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- int r, c;
- int size;
- (void)left;
- for (c = 0; c < bs; ++c) {
- dst[c] = AVG2(above[c], above[c + 1]);
- dst[stride + c] = AVG3(above[c], above[c + 1], above[c + 2]);
- }
- for (r = 2, size = bs - 2; r < bs; r += 2, --size) {
- memcpy(dst + (r + 0) * stride, dst + (r >> 1), size);
- memset(dst + (r + 0) * stride + size, above[bs - 1], bs - size);
- memcpy(dst + (r + 1) * stride, dst + stride + (r >> 1), size);
- memset(dst + (r + 1) * stride + size, above[bs - 1], bs - size);
- }
-}
-intra_pred_no_4x4(d63)
-
-void vp9_d45_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
- const uint8_t *above, const uint8_t *left) {
- const int A = above[0];
- const int B = above[1];
- const int C = above[2];
- const int D = above[3];
- const int E = above[4];
- const int F = above[5];
- const int G = above[6];
- const int H = above[7];
- (void)stride;
- (void)left;
- DST(0, 0) = AVG3(A, B, C);
- DST(1, 0) = DST(0, 1) = AVG3(B, C, D);
- DST(2, 0) = DST(1, 1) = DST(0, 2) = AVG3(C, D, E);
- DST(3, 0) = DST(2, 1) = DST(1, 2) = DST(0, 3) = AVG3(D, E, F);
- DST(3, 1) = DST(2, 2) = DST(1, 3) = AVG3(E, F, G);
- DST(3, 2) = DST(2, 3) = AVG3(F, G, H);
- DST(3, 3) = H; // differs from vp8
-}
-
-static INLINE void d45_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- const uint8_t above_right = above[bs - 1];
- const uint8_t *const dst_row0 = dst;
- int x, size;
- (void)left;
-
- for (x = 0; x < bs - 1; ++x) {
- dst[x] = AVG3(above[x], above[x + 1], above[x + 2]);
- }
- dst[bs - 1] = above_right;
- dst += stride;
- for (x = 1, size = bs - 2; x < bs; ++x, --size) {
- memcpy(dst, dst_row0 + x, size);
- memset(dst + size, above_right, x + 1);
- dst += stride;
- }
-}
-intra_pred_no_4x4(d45)
-
-void vp9_d117_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
- const uint8_t *above, const uint8_t *left) {
- const int I = left[0];
- const int J = left[1];
- const int K = left[2];
- const int X = above[-1];
- const int A = above[0];
- const int B = above[1];
- const int C = above[2];
- const int D = above[3];
- DST(0, 0) = DST(1, 2) = AVG2(X, A);
- DST(1, 0) = DST(2, 2) = AVG2(A, B);
- DST(2, 0) = DST(3, 2) = AVG2(B, C);
- DST(3, 0) = AVG2(C, D);
-
- DST(0, 3) = AVG3(K, J, I);
- DST(0, 2) = AVG3(J, I, X);
- DST(0, 1) = DST(1, 3) = AVG3(I, X, A);
- DST(1, 1) = DST(2, 3) = AVG3(X, A, B);
- DST(2, 1) = DST(3, 3) = AVG3(A, B, C);
- DST(3, 1) = AVG3(B, C, D);
-}
-
-static INLINE void d117_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- int r, c;
-
- // first row
- for (c = 0; c < bs; c++)
- dst[c] = AVG2(above[c - 1], above[c]);
- dst += stride;
-
- // second row
- dst[0] = AVG3(left[0], above[-1], above[0]);
- for (c = 1; c < bs; c++)
- dst[c] = AVG3(above[c - 2], above[c - 1], above[c]);
- dst += stride;
-
- // the rest of first col
- dst[0] = AVG3(above[-1], left[0], left[1]);
- for (r = 3; r < bs; ++r)
- dst[(r - 2) * stride] = AVG3(left[r - 3], left[r - 2], left[r - 1]);
-
- // the rest of the block
- for (r = 2; r < bs; ++r) {
- for (c = 1; c < bs; c++)
- dst[c] = dst[-2 * stride + c - 1];
- dst += stride;
- }
-}
-intra_pred_no_4x4(d117)
-
-void vp9_d135_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
- const uint8_t *above, const uint8_t *left) {
- const int I = left[0];
- const int J = left[1];
- const int K = left[2];
- const int L = left[3];
- const int X = above[-1];
- const int A = above[0];
- const int B = above[1];
- const int C = above[2];
- const int D = above[3];
- (void)stride;
- DST(0, 3) = AVG3(J, K, L);
- DST(1, 3) = DST(0, 2) = AVG3(I, J, K);
- DST(2, 3) = DST(1, 2) = DST(0, 1) = AVG3(X, I, J);
- DST(3, 3) = DST(2, 2) = DST(1, 1) = DST(0, 0) = AVG3(A, X, I);
- DST(3, 2) = DST(2, 1) = DST(1, 0) = AVG3(B, A, X);
- DST(3, 1) = DST(2, 0) = AVG3(C, B, A);
- DST(3, 0) = AVG3(D, C, B);
-}
-
-static INLINE void d135_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- int r, c;
- dst[0] = AVG3(left[0], above[-1], above[0]);
- for (c = 1; c < bs; c++)
- dst[c] = AVG3(above[c - 2], above[c - 1], above[c]);
-
- dst[stride] = AVG3(above[-1], left[0], left[1]);
- for (r = 2; r < bs; ++r)
- dst[r * stride] = AVG3(left[r - 2], left[r - 1], left[r]);
-
- dst += stride;
- for (r = 1; r < bs; ++r) {
- for (c = 1; c < bs; c++)
- dst[c] = dst[-stride + c - 1];
- dst += stride;
- }
-}
-intra_pred_no_4x4(d135)
-
-void vp9_d153_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
- const uint8_t *above, const uint8_t *left) {
- const int I = left[0];
- const int J = left[1];
- const int K = left[2];
- const int L = left[3];
- const int X = above[-1];
- const int A = above[0];
- const int B = above[1];
- const int C = above[2];
-
- DST(0, 0) = DST(2, 1) = AVG2(I, X);
- DST(0, 1) = DST(2, 2) = AVG2(J, I);
- DST(0, 2) = DST(2, 3) = AVG2(K, J);
- DST(0, 3) = AVG2(L, K);
-
- DST(3, 0) = AVG3(A, B, C);
- DST(2, 0) = AVG3(X, A, B);
- DST(1, 0) = DST(3, 1) = AVG3(I, X, A);
- DST(1, 1) = DST(3, 2) = AVG3(J, I, X);
- DST(1, 2) = DST(3, 3) = AVG3(K, J, I);
- DST(1, 3) = AVG3(L, K, J);
-}
-
-static INLINE void d153_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- int r, c;
- dst[0] = AVG2(above[-1], left[0]);
- for (r = 1; r < bs; r++)
- dst[r * stride] = AVG2(left[r - 1], left[r]);
- dst++;
-
- dst[0] = AVG3(left[0], above[-1], above[0]);
- dst[stride] = AVG3(above[-1], left[0], left[1]);
- for (r = 2; r < bs; r++)
- dst[r * stride] = AVG3(left[r - 2], left[r - 1], left[r]);
- dst++;
-
- for (c = 0; c < bs - 2; c++)
- dst[c] = AVG3(above[c - 1], above[c], above[c + 1]);
- dst += stride;
-
- for (r = 1; r < bs; ++r) {
- for (c = 0; c < bs - 2; c++)
- dst[c] = dst[-stride + c - 2];
- dst += stride;
- }
-}
-intra_pred_no_4x4(d153)
-
-static INLINE void v_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- int r;
- (void) left;
-
- for (r = 0; r < bs; r++) {
- memcpy(dst, above, bs);
- dst += stride;
- }
-}
-intra_pred_allsizes(v)
-
-static INLINE void h_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- int r;
- (void) above;
-
- for (r = 0; r < bs; r++) {
- memset(dst, left[r], bs);
- dst += stride;
- }
-}
-intra_pred_allsizes(h)
-
-static INLINE void tm_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- int r, c;
- int ytop_left = above[-1];
-
- for (r = 0; r < bs; r++) {
- for (c = 0; c < bs; c++)
- dst[c] = clip_pixel(left[r] + above[c] - ytop_left);
- dst += stride;
- }
-}
-intra_pred_allsizes(tm)
-
-static INLINE void dc_128_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- int r;
- (void) above;
- (void) left;
-
- for (r = 0; r < bs; r++) {
- memset(dst, 128, bs);
- dst += stride;
- }
-}
-intra_pred_allsizes(dc_128)
-
-static INLINE void dc_left_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above,
- const uint8_t *left) {
- int i, r, expected_dc, sum = 0;
- (void) above;
-
- for (i = 0; i < bs; i++)
- sum += left[i];
- expected_dc = (sum + (bs >> 1)) / bs;
-
- for (r = 0; r < bs; r++) {
- memset(dst, expected_dc, bs);
- dst += stride;
- }
-}
-intra_pred_allsizes(dc_left)
-
-static INLINE void dc_top_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- int i, r, expected_dc, sum = 0;
- (void) left;
-
- for (i = 0; i < bs; i++)
- sum += above[i];
- expected_dc = (sum + (bs >> 1)) / bs;
-
- for (r = 0; r < bs; r++) {
- memset(dst, expected_dc, bs);
- dst += stride;
- }
-}
-intra_pred_allsizes(dc_top)
-
-static INLINE void dc_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
- const uint8_t *above, const uint8_t *left) {
- int i, r, expected_dc, sum = 0;
- const int count = 2 * bs;
-
- for (i = 0; i < bs; i++) {
- sum += above[i];
- sum += left[i];
- }
-
- expected_dc = (sum + (count >> 1)) / count;
-
- for (r = 0; r < bs; r++) {
- memset(dst, expected_dc, bs);
- dst += stride;
- }
-}
-intra_pred_allsizes(dc)
-#undef intra_pred_allsizes
-
typedef void (*intra_pred_fn)(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left);
@@ -743,10 +66,10 @@
static void vp9_init_intra_predictors_internal(void) {
#define INIT_ALL_SIZES(p, type) \
- p[TX_4X4] = vp9_##type##_predictor_4x4; \
- p[TX_8X8] = vp9_##type##_predictor_8x8; \
- p[TX_16X16] = vp9_##type##_predictor_16x16; \
- p[TX_32X32] = vp9_##type##_predictor_32x32
+ p[TX_4X4] = vpx_##type##_predictor_4x4; \
+ p[TX_8X8] = vpx_##type##_predictor_8x8; \
+ p[TX_16X16] = vpx_##type##_predictor_16x16; \
+ p[TX_32X32] = vpx_##type##_predictor_32x32
INIT_ALL_SIZES(pred[V_PRED], v);
INIT_ALL_SIZES(pred[H_PRED], h);
diff --git a/vp9/common/vp9_rtcd_defs.pl b/vp9/common/vp9_rtcd_defs.pl
index be445f8..ee99d7a 100644
--- a/vp9/common/vp9_rtcd_defs.pl
+++ b/vp9/common/vp9_rtcd_defs.pl
@@ -61,165 +61,6 @@
}
#
-# RECON
-#
-add_proto qw/void vp9_d207_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d207_predictor_4x4/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d45_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d45_predictor_4x4 neon/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d63_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d63_predictor_4x4/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_h_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_h_predictor_4x4 neon dspr2 msa/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d117_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d117_predictor_4x4/;
-
-add_proto qw/void vp9_d135_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d135_predictor_4x4 neon/;
-
-add_proto qw/void vp9_d153_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d153_predictor_4x4/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_v_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_v_predictor_4x4 neon msa/, "$sse_x86inc";
-
-add_proto qw/void vp9_tm_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_tm_predictor_4x4 neon dspr2 msa/, "$sse_x86inc";
-
-add_proto qw/void vp9_dc_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_predictor_4x4 dspr2 msa neon/, "$sse_x86inc";
-
-add_proto qw/void vp9_dc_top_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_top_predictor_4x4 msa neon/, "$sse_x86inc";
-
-add_proto qw/void vp9_dc_left_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_left_predictor_4x4 msa neon/, "$sse_x86inc";
-
-add_proto qw/void vp9_dc_128_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_128_predictor_4x4 msa neon/, "$sse_x86inc";
-
-add_proto qw/void vp9_d207_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d207_predictor_8x8/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d45_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d45_predictor_8x8 neon/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d63_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d63_predictor_8x8/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_h_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_h_predictor_8x8 neon dspr2 msa/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d117_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d117_predictor_8x8/;
-
-add_proto qw/void vp9_d135_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d135_predictor_8x8/;
-
-add_proto qw/void vp9_d153_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d153_predictor_8x8/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_v_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_v_predictor_8x8 neon msa/, "$sse_x86inc";
-
-add_proto qw/void vp9_tm_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_tm_predictor_8x8 neon dspr2 msa/, "$sse2_x86inc";
-
-add_proto qw/void vp9_dc_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_predictor_8x8 dspr2 neon msa/, "$sse_x86inc";
-
-add_proto qw/void vp9_dc_top_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_top_predictor_8x8 neon msa/, "$sse_x86inc";
-
-add_proto qw/void vp9_dc_left_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_left_predictor_8x8 neon msa/, "$sse_x86inc";
-
-add_proto qw/void vp9_dc_128_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_128_predictor_8x8 neon msa/, "$sse_x86inc";
-
-add_proto qw/void vp9_d207_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d207_predictor_16x16/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d45_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d45_predictor_16x16 neon/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d63_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d63_predictor_16x16/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_h_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_h_predictor_16x16 neon dspr2 msa/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d117_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d117_predictor_16x16/;
-
-add_proto qw/void vp9_d135_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d135_predictor_16x16/;
-
-add_proto qw/void vp9_d153_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d153_predictor_16x16/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_v_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_v_predictor_16x16 neon msa/, "$sse2_x86inc";
-
-add_proto qw/void vp9_tm_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_tm_predictor_16x16 neon msa/, "$sse2_x86inc";
-
-add_proto qw/void vp9_dc_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_predictor_16x16 dspr2 neon msa/, "$sse2_x86inc";
-
-add_proto qw/void vp9_dc_top_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_top_predictor_16x16 neon msa/, "$sse2_x86inc";
-
-add_proto qw/void vp9_dc_left_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_left_predictor_16x16 neon msa/, "$sse2_x86inc";
-
-add_proto qw/void vp9_dc_128_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_128_predictor_16x16 neon msa/, "$sse2_x86inc";
-
-add_proto qw/void vp9_d207_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d207_predictor_32x32/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d45_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d45_predictor_32x32/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d63_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d63_predictor_32x32/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_h_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_h_predictor_32x32 neon msa/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_d117_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d117_predictor_32x32/;
-
-add_proto qw/void vp9_d135_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d135_predictor_32x32/;
-
-add_proto qw/void vp9_d153_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_d153_predictor_32x32/, "$ssse3_x86inc";
-
-add_proto qw/void vp9_v_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_v_predictor_32x32 neon msa/, "$sse2_x86inc";
-
-add_proto qw/void vp9_tm_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_tm_predictor_32x32 neon msa/, "$sse2_x86_64_x86inc";
-
-add_proto qw/void vp9_dc_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_predictor_32x32 msa neon/, "$sse2_x86inc";
-
-add_proto qw/void vp9_dc_top_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_top_predictor_32x32 msa neon/, "$sse2_x86inc";
-
-add_proto qw/void vp9_dc_left_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_left_predictor_32x32 msa neon/, "$sse2_x86inc";
-
-add_proto qw/void vp9_dc_128_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_128_predictor_32x32 msa neon/, "$sse2_x86inc";
-
-#
# post proc
#
if (vpx_config("CONFIG_VP9_POSTPROC") eq "yes") {
@@ -439,165 +280,6 @@
# High bitdepth functions
if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
#
- # Intra prediction
- #
- add_proto qw/void vp9_highbd_d207_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d207_predictor_4x4/;
-
- add_proto qw/void vp9_highbd_d45_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d45_predictor_4x4/;
-
- add_proto qw/void vp9_highbd_d63_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d63_predictor_4x4/;
-
- add_proto qw/void vp9_highbd_h_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_h_predictor_4x4/;
-
- add_proto qw/void vp9_highbd_d117_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d117_predictor_4x4/;
-
- add_proto qw/void vp9_highbd_d135_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d135_predictor_4x4/;
-
- add_proto qw/void vp9_highbd_d153_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d153_predictor_4x4/;
-
- add_proto qw/void vp9_highbd_v_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_v_predictor_4x4/, "$sse_x86inc";
-
- add_proto qw/void vp9_highbd_tm_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_tm_predictor_4x4/, "$sse_x86inc";
-
- add_proto qw/void vp9_highbd_dc_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_predictor_4x4/, "$sse_x86inc";
-
- add_proto qw/void vp9_highbd_dc_top_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_top_predictor_4x4/;
-
- add_proto qw/void vp9_highbd_dc_left_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_left_predictor_4x4/;
-
- add_proto qw/void vp9_highbd_dc_128_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_128_predictor_4x4/;
-
- add_proto qw/void vp9_highbd_d207_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d207_predictor_8x8/;
-
- add_proto qw/void vp9_highbd_d45_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d45_predictor_8x8/;
-
- add_proto qw/void vp9_highbd_d63_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d63_predictor_8x8/;
-
- add_proto qw/void vp9_highbd_h_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_h_predictor_8x8/;
-
- add_proto qw/void vp9_highbd_d117_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d117_predictor_8x8/;
-
- add_proto qw/void vp9_highbd_d135_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d135_predictor_8x8/;
-
- add_proto qw/void vp9_highbd_d153_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d153_predictor_8x8/;
-
- add_proto qw/void vp9_highbd_v_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_v_predictor_8x8/, "$sse2_x86inc";
-
- add_proto qw/void vp9_highbd_tm_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_tm_predictor_8x8/, "$sse2_x86inc";
-
- add_proto qw/void vp9_highbd_dc_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_predictor_8x8/, "$sse2_x86inc";;
-
- add_proto qw/void vp9_highbd_dc_top_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_top_predictor_8x8/;
-
- add_proto qw/void vp9_highbd_dc_left_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_left_predictor_8x8/;
-
- add_proto qw/void vp9_highbd_dc_128_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_128_predictor_8x8/;
-
- add_proto qw/void vp9_highbd_d207_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d207_predictor_16x16/;
-
- add_proto qw/void vp9_highbd_d45_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d45_predictor_16x16/;
-
- add_proto qw/void vp9_highbd_d63_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d63_predictor_16x16/;
-
- add_proto qw/void vp9_highbd_h_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_h_predictor_16x16/;
-
- add_proto qw/void vp9_highbd_d117_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d117_predictor_16x16/;
-
- add_proto qw/void vp9_highbd_d135_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d135_predictor_16x16/;
-
- add_proto qw/void vp9_highbd_d153_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d153_predictor_16x16/;
-
- add_proto qw/void vp9_highbd_v_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_v_predictor_16x16/, "$sse2_x86inc";
-
- add_proto qw/void vp9_highbd_tm_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_tm_predictor_16x16/, "$sse2_x86_64_x86inc";
-
- add_proto qw/void vp9_highbd_dc_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_predictor_16x16/, "$sse2_x86inc";
-
- add_proto qw/void vp9_highbd_dc_top_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_top_predictor_16x16/;
-
- add_proto qw/void vp9_highbd_dc_left_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_left_predictor_16x16/;
-
- add_proto qw/void vp9_highbd_dc_128_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_128_predictor_16x16/;
-
- add_proto qw/void vp9_highbd_d207_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d207_predictor_32x32/;
-
- add_proto qw/void vp9_highbd_d45_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d45_predictor_32x32/;
-
- add_proto qw/void vp9_highbd_d63_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d63_predictor_32x32/;
-
- add_proto qw/void vp9_highbd_h_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_h_predictor_32x32/;
-
- add_proto qw/void vp9_highbd_d117_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d117_predictor_32x32/;
-
- add_proto qw/void vp9_highbd_d135_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d135_predictor_32x32/;
-
- add_proto qw/void vp9_highbd_d153_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_d153_predictor_32x32/;
-
- add_proto qw/void vp9_highbd_v_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_v_predictor_32x32/, "$sse2_x86inc";
-
- add_proto qw/void vp9_highbd_tm_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_tm_predictor_32x32/, "$sse2_x86_64_x86inc";
-
- add_proto qw/void vp9_highbd_dc_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_predictor_32x32/, "$sse2_x86_64_x86inc";
-
- add_proto qw/void vp9_highbd_dc_top_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_top_predictor_32x32/;
-
- add_proto qw/void vp9_highbd_dc_left_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_left_predictor_32x32/;
-
- add_proto qw/void vp9_highbd_dc_128_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
- specialize qw/vp9_highbd_dc_128_predictor_32x32/;
-
- #
# Sub Pixel Filters
#
add_proto qw/void vp9_highbd_convolve_copy/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
@@ -825,24 +507,6 @@
add_proto qw/void vp9_fwht4x4/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/vp9_fwht4x4/, "$mmx_x86inc";
-
- add_proto qw/void vp9_fdct4x4_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct4x4_1 sse2/;
-
- add_proto qw/void vp9_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct8x8_1 sse2/;
-
- add_proto qw/void vp9_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct16x16_1 sse2/;
-
- add_proto qw/void vp9_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct32x32_1 sse2/;
-
- add_proto qw/void vp9_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct32x32 sse2/;
-
- add_proto qw/void vp9_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct32x32_rd sse2/;
} else {
add_proto qw/void vp9_fht4x4/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
specialize qw/vp9_fht4x4 sse2 msa/;
@@ -855,24 +519,6 @@
add_proto qw/void vp9_fwht4x4/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/vp9_fwht4x4 msa/, "$mmx_x86inc";
-
- add_proto qw/void vp9_fdct4x4_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct4x4_1 sse2/;
-
- add_proto qw/void vp9_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct8x8_1 sse2 neon msa/;
-
- add_proto qw/void vp9_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct16x16_1 sse2 msa/;
-
- add_proto qw/void vp9_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct32x32_1 sse2 msa/;
-
- add_proto qw/void vp9_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct32x32 sse2 avx2 msa/;
-
- add_proto qw/void vp9_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct32x32_rd sse2 avx2 msa/;
}
#
@@ -926,21 +572,6 @@
add_proto qw/void vp9_highbd_fwht4x4/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/vp9_highbd_fwht4x4/;
- add_proto qw/void vp9_highbd_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_highbd_fdct8x8_1/;
-
- add_proto qw/void vp9_highbd_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_highbd_fdct16x16_1/;
-
- add_proto qw/void vp9_highbd_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_highbd_fdct32x32_1/;
-
- add_proto qw/void vp9_highbd_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_highbd_fdct32x32 sse2/;
-
- add_proto qw/void vp9_highbd_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_highbd_fdct32x32_rd sse2/;
-
add_proto qw/void vp9_highbd_temporal_filter_apply/, "uint8_t *frame1, unsigned int stride, uint8_t *frame2, unsigned int block_width, unsigned int block_height, int strength, int filter_weight, unsigned int *accumulator, uint16_t *count";
specialize qw/vp9_highbd_temporal_filter_apply/;
diff --git a/vp9/encoder/arm/neon/vp9_dct_neon.c b/vp9/encoder/arm/neon/vp9_dct_neon.c
index 941ad20..11e8773 100644
--- a/vp9/encoder/arm/neon/vp9_dct_neon.c
+++ b/vp9/encoder/arm/neon/vp9_dct_neon.c
@@ -9,30 +9,14 @@
*/
#include <arm_neon.h>
+
#include "./vp9_rtcd.h"
-#include "./vpx_dsp_rtcd.h"
#include "./vpx_config.h"
+#include "./vpx_dsp_rtcd.h"
#include "vp9/common/vp9_blockd.h"
#include "vpx_dsp/txfm_common.h"
-void vp9_fdct8x8_1_neon(const int16_t *input, int16_t *output, int stride) {
- int r;
- int16x8_t sum = vld1q_s16(&input[0]);
- for (r = 1; r < 8; ++r) {
- const int16x8_t input_00 = vld1q_s16(&input[r * stride]);
- sum = vaddq_s16(sum, input_00);
- }
- {
- const int32x4_t a = vpaddlq_s16(sum);
- const int64x2_t b = vpaddlq_s32(a);
- const int32x2_t c = vadd_s32(vreinterpret_s32_s64(vget_low_s64(b)),
- vreinterpret_s32_s64(vget_high_s64(b)));
- output[0] = vget_lane_s16(vreinterpret_s16_s32(c), 0);
- output[1] = 0;
- }
-}
-
void vp9_fdct8x8_quant_neon(const int16_t *input, int stride,
int16_t* coeff_ptr, intptr_t n_coeffs,
int skip_block, const int16_t* zbin_ptr,
@@ -45,7 +29,7 @@
int16_t temp_buffer[64];
(void)coeff_ptr;
- vp9_fdct8x8_neon(input, temp_buffer, stride);
+ vpx_fdct8x8_neon(input, temp_buffer, stride);
vp9_quantize_fp_neon(temp_buffer, n_coeffs, skip_block, zbin_ptr, round_ptr,
quant_ptr, quant_shift_ptr, qcoeff_ptr, dqcoeff_ptr,
dequant_ptr, eob_ptr, scan_ptr, iscan_ptr);
diff --git a/vp9/encoder/mips/msa/vp9_fdct16x16_msa.c b/vp9/encoder/mips/msa/vp9_fdct16x16_msa.c
index de02956..6dabb58 100644
--- a/vp9/encoder/mips/msa/vp9_fdct16x16_msa.c
+++ b/vp9/encoder/mips/msa/vp9_fdct16x16_msa.c
@@ -10,20 +10,10 @@
#include <assert.h>
-#include "./vp9_rtcd.h"
+#include "vp9/common/vp9_enums.h"
#include "vp9/encoder/mips/msa/vp9_fdct_msa.h"
#include "vpx_dsp/mips/fwd_txfm_msa.h"
-void vp9_fdct16x16_1_msa(const int16_t *input, int16_t *out, int32_t stride) {
- out[1] = 0;
-
- out[0] = LD_HADD(input, stride);
- out[0] += LD_HADD(input + 8, stride);
- out[0] += LD_HADD(input + 16 * 8, stride);
- out[0] += LD_HADD(input + 16 * 8 + 8, stride);
- out[0] >>= 1;
-}
-
static void fadst16_cols_step1_msa(const int16_t *input, int32_t stride,
const int32_t *const0, int16_t *int_buf) {
v8i16 r0, r1, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, r13, r14, r15;
diff --git a/vp9/encoder/mips/msa/vp9_fdct4x4_msa.c b/vp9/encoder/mips/msa/vp9_fdct4x4_msa.c
index ce47457..574016f 100644
--- a/vp9/encoder/mips/msa/vp9_fdct4x4_msa.c
+++ b/vp9/encoder/mips/msa/vp9_fdct4x4_msa.c
@@ -10,7 +10,7 @@
#include <assert.h>
-#include "./vp9_rtcd.h"
+#include "vp9/common/vp9_enums.h"
#include "vp9/encoder/mips/msa/vp9_fdct_msa.h"
void vp9_fwht4x4_msa(const int16_t *input, int16_t *output,
diff --git a/vp9/encoder/mips/msa/vp9_fdct8x8_msa.c b/vp9/encoder/mips/msa/vp9_fdct8x8_msa.c
index 253b5e8..7c3c635 100644
--- a/vp9/encoder/mips/msa/vp9_fdct8x8_msa.c
+++ b/vp9/encoder/mips/msa/vp9_fdct8x8_msa.c
@@ -10,14 +10,9 @@
#include <assert.h>
-#include "./vp9_rtcd.h"
+#include "vp9/common/vp9_enums.h"
#include "vp9/encoder/mips/msa/vp9_fdct_msa.h"
-void vp9_fdct8x8_1_msa(const int16_t *input, int16_t *out, int32_t stride) {
- out[0] = LD_HADD(input, stride);
- out[1] = 0;
-}
-
void vp9_fht8x8_msa(const int16_t *input, int16_t *output, int32_t stride,
int32_t tx_type) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
diff --git a/vp9/encoder/mips/msa/vp9_fdct_msa.h b/vp9/encoder/mips/msa/vp9_fdct_msa.h
index 425ba9d..d7d40cb 100644
--- a/vp9/encoder/mips/msa/vp9_fdct_msa.h
+++ b/vp9/encoder/mips/msa/vp9_fdct_msa.h
@@ -81,37 +81,6 @@
out5 = -out5; \
}
-#define LD_HADD(psrc, stride) ({ \
- v8i16 in0_m, in1_m, in2_m, in3_m, in4_m, in5_m, in6_m, in7_m; \
- v4i32 vec_w_m; \
- \
- LD_SH4((psrc), stride, in0_m, in1_m, in2_m, in3_m); \
- ADD2(in0_m, in1_m, in2_m, in3_m, in0_m, in2_m); \
- LD_SH4(((psrc) + 4 * stride), stride, in4_m, in5_m, in6_m, in7_m); \
- ADD4(in4_m, in5_m, in6_m, in7_m, in0_m, in2_m, in4_m, in6_m, \
- in4_m, in6_m, in0_m, in4_m); \
- in0_m += in4_m; \
- \
- vec_w_m = __msa_hadd_s_w(in0_m, in0_m); \
- HADD_SW_S32(vec_w_m); \
-})
-
-#define FDCT_POSTPROC_2V_NEG_H(vec0, vec1) { \
- v8i16 tp0_m, tp1_m; \
- v8i16 one_m = __msa_ldi_h(1); \
- \
- tp0_m = __msa_clti_s_h(vec0, 0); \
- tp1_m = __msa_clti_s_h(vec1, 0); \
- vec0 += 1; \
- vec1 += 1; \
- tp0_m = one_m & tp0_m; \
- tp1_m = one_m & tp1_m; \
- vec0 += tp0_m; \
- vec1 += tp1_m; \
- vec0 >>= 2; \
- vec1 >>= 2; \
-}
-
#define VP9_FADST4(in0, in1, in2, in3, out0, out1, out2, out3) { \
v4i32 s0_m, s1_m, s2_m, s3_m, constant_m; \
v4i32 in0_r_m, in1_r_m, in2_r_m, in3_r_m; \
@@ -145,67 +114,4 @@
PCKEV_H4_SH(in0_r_m, in0_r_m, in1_r_m, in1_r_m, s2_m, s2_m, \
s3_m, s3_m, out0, out1, out2, out3); \
}
-
-#define FDCT32_POSTPROC_NEG_W(vec) { \
- v4i32 temp_m; \
- v4i32 one_m = __msa_ldi_w(1); \
- \
- temp_m = __msa_clti_s_w(vec, 0); \
- vec += 1; \
- temp_m = one_m & temp_m; \
- vec += temp_m; \
- vec >>= 2; \
-}
-
-#define FDCT32_POSTPROC_2V_POS_H(vec0, vec1) { \
- v8i16 tp0_m, tp1_m; \
- v8i16 one = __msa_ldi_h(1); \
- \
- tp0_m = __msa_clei_s_h(vec0, 0); \
- tp1_m = __msa_clei_s_h(vec1, 0); \
- tp0_m = (v8i16)__msa_xori_b((v16u8)tp0_m, 255); \
- tp1_m = (v8i16)__msa_xori_b((v16u8)tp1_m, 255); \
- vec0 += 1; \
- vec1 += 1; \
- tp0_m = one & tp0_m; \
- tp1_m = one & tp1_m; \
- vec0 += tp0_m; \
- vec1 += tp1_m; \
- vec0 >>= 2; \
- vec1 >>= 2; \
-}
-
-#define DOTP_CONST_PAIR_W(reg0_left, reg1_left, reg0_right, \
- reg1_right, const0, const1, \
- out0, out1, out2, out3) { \
- v4i32 s0_m, s1_m, s2_m, s3_m, s4_m, s5_m, s6_m, s7_m; \
- v2i64 tp0_m, tp1_m, tp2_m, tp3_m; \
- v4i32 k0_m = __msa_fill_w((int32_t) const0); \
- \
- s0_m = __msa_fill_w((int32_t) const1); \
- k0_m = __msa_ilvev_w(s0_m, k0_m); \
- \
- ILVRL_W2_SW(-reg1_left, reg0_left, s1_m, s0_m); \
- ILVRL_W2_SW(reg0_left, reg1_left, s3_m, s2_m); \
- ILVRL_W2_SW(-reg1_right, reg0_right, s5_m, s4_m); \
- ILVRL_W2_SW(reg0_right, reg1_right, s7_m, s6_m); \
- \
- DOTP_SW2_SD(s0_m, s1_m, k0_m, k0_m, tp0_m, tp1_m); \
- DOTP_SW2_SD(s4_m, s5_m, k0_m, k0_m, tp2_m, tp3_m); \
- tp0_m = __msa_srari_d(tp0_m, DCT_CONST_BITS); \
- tp1_m = __msa_srari_d(tp1_m, DCT_CONST_BITS); \
- tp2_m = __msa_srari_d(tp2_m, DCT_CONST_BITS); \
- tp3_m = __msa_srari_d(tp3_m, DCT_CONST_BITS); \
- out0 = __msa_pckev_w((v4i32)tp0_m, (v4i32)tp1_m); \
- out1 = __msa_pckev_w((v4i32)tp2_m, (v4i32)tp3_m); \
- \
- DOTP_SW2_SD(s2_m, s3_m, k0_m, k0_m, tp0_m, tp1_m); \
- DOTP_SW2_SD(s6_m, s7_m, k0_m, k0_m, tp2_m, tp3_m); \
- tp0_m = __msa_srari_d(tp0_m, DCT_CONST_BITS); \
- tp1_m = __msa_srari_d(tp1_m, DCT_CONST_BITS); \
- tp2_m = __msa_srari_d(tp2_m, DCT_CONST_BITS); \
- tp3_m = __msa_srari_d(tp3_m, DCT_CONST_BITS); \
- out2 = __msa_pckev_w((v4i32)tp0_m, (v4i32)tp1_m); \
- out3 = __msa_pckev_w((v4i32)tp2_m, (v4i32)tp3_m); \
-}
#endif /* VP9_ENCODER_MIPS_MSA_VP9_FDCT_MSA_H_ */
diff --git a/vp9/encoder/vp9_context_tree.c b/vp9/encoder/vp9_context_tree.c
index fbff9d7..e87cccb 100644
--- a/vp9/encoder/vp9_context_tree.c
+++ b/vp9/encoder/vp9_context_tree.c
@@ -69,10 +69,13 @@
alloc_mode_context(cm, num_4x4_blk/2, &tree->horizontal[0]);
alloc_mode_context(cm, num_4x4_blk/2, &tree->vertical[0]);
- /* TODO(Jbb): for 4x8 and 8x4 these allocated values are not used.
- * Figure out a better way to do this. */
- alloc_mode_context(cm, num_4x4_blk/2, &tree->horizontal[1]);
- alloc_mode_context(cm, num_4x4_blk/2, &tree->vertical[1]);
+ if (num_4x4_blk > 4) {
+ alloc_mode_context(cm, num_4x4_blk/2, &tree->horizontal[1]);
+ alloc_mode_context(cm, num_4x4_blk/2, &tree->vertical[1]);
+ } else {
+ memset(&tree->horizontal[1], 0, sizeof(tree->horizontal[1]));
+ memset(&tree->vertical[1], 0, sizeof(tree->vertical[1]));
+ }
}
static void free_tree_contexts(PC_TREE *tree) {
diff --git a/vp9/encoder/vp9_dct.c b/vp9/encoder/vp9_dct.c
index 5dcbd1f..8d0d10f 100644
--- a/vp9/encoder/vp9_dct.c
+++ b/vp9/encoder/vp9_dct.c
@@ -11,16 +11,15 @@
#include <assert.h>
#include <math.h>
-#include "./vpx_config.h"
#include "./vp9_rtcd.h"
+#include "./vpx_config.h"
#include "./vpx_dsp_rtcd.h"
#include "vp9/common/vp9_blockd.h"
#include "vp9/common/vp9_idct.h"
#include "vp9/common/vp9_systemdependent.h"
-#include "vp9/encoder/vp9_dct.h"
-#include "vpx_ports/mem.h"
#include "vpx_dsp/fwd_txfm.h"
+#include "vpx_ports/mem.h"
static void fdct4(const tran_low_t *input, tran_low_t *output) {
tran_high_t step[4];
@@ -529,21 +528,10 @@
{ fadst16, fadst16 } // ADST_ADST = 3
};
-void vp9_fdct4x4_1_c(const int16_t *input, tran_low_t *output, int stride) {
- int r, c;
- tran_low_t sum = 0;
- for (r = 0; r < 4; ++r)
- for (c = 0; c < 4; ++c)
- sum += input[r * stride + c];
-
- output[0] = sum << 1;
- output[1] = 0;
-}
-
void vp9_fht4x4_c(const int16_t *input, tran_low_t *output,
int stride, int tx_type) {
if (tx_type == DCT_DCT) {
- vp9_fdct4x4_c(input, output, stride);
+ vpx_fdct4x4_c(input, output, stride);
} else {
tran_low_t out[4 * 4];
int i, j;
@@ -572,17 +560,6 @@
}
}
-void vp9_fdct8x8_1_c(const int16_t *input, tran_low_t *output, int stride) {
- int r, c;
- tran_low_t sum = 0;
- for (r = 0; r < 8; ++r)
- for (c = 0; c < 8; ++c)
- sum += input[r * stride + c];
-
- output[0] = sum;
- output[1] = 0;
-}
-
void vp9_fdct8x8_quant_c(const int16_t *input, int stride,
tran_low_t *coeff_ptr, intptr_t n_coeffs,
int skip_block,
@@ -695,21 +672,10 @@
*eob_ptr = eob + 1;
}
-void vp9_fdct16x16_1_c(const int16_t *input, tran_low_t *output, int stride) {
- int r, c;
- tran_low_t sum = 0;
- for (r = 0; r < 16; ++r)
- for (c = 0; c < 16; ++c)
- sum += input[r * stride + c];
-
- output[0] = sum >> 1;
- output[1] = 0;
-}
-
void vp9_fht8x8_c(const int16_t *input, tran_low_t *output,
int stride, int tx_type) {
if (tx_type == DCT_DCT) {
- vp9_fdct8x8_c(input, output, stride);
+ vpx_fdct8x8_c(input, output, stride);
} else {
tran_low_t out[64];
int i, j;
@@ -795,7 +761,7 @@
void vp9_fht16x16_c(const int16_t *input, tran_low_t *output,
int stride, int tx_type) {
if (tx_type == DCT_DCT) {
- vp9_fdct16x16_c(input, output, stride);
+ vpx_fdct16x16_c(input, output, stride);
} else {
tran_low_t out[256];
int i, j;
@@ -822,426 +788,12 @@
}
}
-static INLINE tran_high_t dct_32_round(tran_high_t input) {
- tran_high_t rv = ROUND_POWER_OF_TWO(input, DCT_CONST_BITS);
- // TODO(debargha, peter.derivaz): Find new bounds for this assert,
- // and make the bounds consts.
- // assert(-131072 <= rv && rv <= 131071);
- return rv;
-}
-
-static INLINE tran_high_t half_round_shift(tran_high_t input) {
- tran_high_t rv = (input + 1 + (input < 0)) >> 2;
- return rv;
-}
-
-void vp9_fdct32(const tran_high_t *input, tran_high_t *output, int round) {
- tran_high_t step[32];
- // Stage 1
- step[0] = input[0] + input[(32 - 1)];
- step[1] = input[1] + input[(32 - 2)];
- step[2] = input[2] + input[(32 - 3)];
- step[3] = input[3] + input[(32 - 4)];
- step[4] = input[4] + input[(32 - 5)];
- step[5] = input[5] + input[(32 - 6)];
- step[6] = input[6] + input[(32 - 7)];
- step[7] = input[7] + input[(32 - 8)];
- step[8] = input[8] + input[(32 - 9)];
- step[9] = input[9] + input[(32 - 10)];
- step[10] = input[10] + input[(32 - 11)];
- step[11] = input[11] + input[(32 - 12)];
- step[12] = input[12] + input[(32 - 13)];
- step[13] = input[13] + input[(32 - 14)];
- step[14] = input[14] + input[(32 - 15)];
- step[15] = input[15] + input[(32 - 16)];
- step[16] = -input[16] + input[(32 - 17)];
- step[17] = -input[17] + input[(32 - 18)];
- step[18] = -input[18] + input[(32 - 19)];
- step[19] = -input[19] + input[(32 - 20)];
- step[20] = -input[20] + input[(32 - 21)];
- step[21] = -input[21] + input[(32 - 22)];
- step[22] = -input[22] + input[(32 - 23)];
- step[23] = -input[23] + input[(32 - 24)];
- step[24] = -input[24] + input[(32 - 25)];
- step[25] = -input[25] + input[(32 - 26)];
- step[26] = -input[26] + input[(32 - 27)];
- step[27] = -input[27] + input[(32 - 28)];
- step[28] = -input[28] + input[(32 - 29)];
- step[29] = -input[29] + input[(32 - 30)];
- step[30] = -input[30] + input[(32 - 31)];
- step[31] = -input[31] + input[(32 - 32)];
-
- // Stage 2
- output[0] = step[0] + step[16 - 1];
- output[1] = step[1] + step[16 - 2];
- output[2] = step[2] + step[16 - 3];
- output[3] = step[3] + step[16 - 4];
- output[4] = step[4] + step[16 - 5];
- output[5] = step[5] + step[16 - 6];
- output[6] = step[6] + step[16 - 7];
- output[7] = step[7] + step[16 - 8];
- output[8] = -step[8] + step[16 - 9];
- output[9] = -step[9] + step[16 - 10];
- output[10] = -step[10] + step[16 - 11];
- output[11] = -step[11] + step[16 - 12];
- output[12] = -step[12] + step[16 - 13];
- output[13] = -step[13] + step[16 - 14];
- output[14] = -step[14] + step[16 - 15];
- output[15] = -step[15] + step[16 - 16];
-
- output[16] = step[16];
- output[17] = step[17];
- output[18] = step[18];
- output[19] = step[19];
-
- output[20] = dct_32_round((-step[20] + step[27]) * cospi_16_64);
- output[21] = dct_32_round((-step[21] + step[26]) * cospi_16_64);
- output[22] = dct_32_round((-step[22] + step[25]) * cospi_16_64);
- output[23] = dct_32_round((-step[23] + step[24]) * cospi_16_64);
-
- output[24] = dct_32_round((step[24] + step[23]) * cospi_16_64);
- output[25] = dct_32_round((step[25] + step[22]) * cospi_16_64);
- output[26] = dct_32_round((step[26] + step[21]) * cospi_16_64);
- output[27] = dct_32_round((step[27] + step[20]) * cospi_16_64);
-
- output[28] = step[28];
- output[29] = step[29];
- output[30] = step[30];
- output[31] = step[31];
-
- // dump the magnitude by 4, hence the intermediate values are within
- // the range of 16 bits.
- if (round) {
- output[0] = half_round_shift(output[0]);
- output[1] = half_round_shift(output[1]);
- output[2] = half_round_shift(output[2]);
- output[3] = half_round_shift(output[3]);
- output[4] = half_round_shift(output[4]);
- output[5] = half_round_shift(output[5]);
- output[6] = half_round_shift(output[6]);
- output[7] = half_round_shift(output[7]);
- output[8] = half_round_shift(output[8]);
- output[9] = half_round_shift(output[9]);
- output[10] = half_round_shift(output[10]);
- output[11] = half_round_shift(output[11]);
- output[12] = half_round_shift(output[12]);
- output[13] = half_round_shift(output[13]);
- output[14] = half_round_shift(output[14]);
- output[15] = half_round_shift(output[15]);
-
- output[16] = half_round_shift(output[16]);
- output[17] = half_round_shift(output[17]);
- output[18] = half_round_shift(output[18]);
- output[19] = half_round_shift(output[19]);
- output[20] = half_round_shift(output[20]);
- output[21] = half_round_shift(output[21]);
- output[22] = half_round_shift(output[22]);
- output[23] = half_round_shift(output[23]);
- output[24] = half_round_shift(output[24]);
- output[25] = half_round_shift(output[25]);
- output[26] = half_round_shift(output[26]);
- output[27] = half_round_shift(output[27]);
- output[28] = half_round_shift(output[28]);
- output[29] = half_round_shift(output[29]);
- output[30] = half_round_shift(output[30]);
- output[31] = half_round_shift(output[31]);
- }
-
- // Stage 3
- step[0] = output[0] + output[(8 - 1)];
- step[1] = output[1] + output[(8 - 2)];
- step[2] = output[2] + output[(8 - 3)];
- step[3] = output[3] + output[(8 - 4)];
- step[4] = -output[4] + output[(8 - 5)];
- step[5] = -output[5] + output[(8 - 6)];
- step[6] = -output[6] + output[(8 - 7)];
- step[7] = -output[7] + output[(8 - 8)];
- step[8] = output[8];
- step[9] = output[9];
- step[10] = dct_32_round((-output[10] + output[13]) * cospi_16_64);
- step[11] = dct_32_round((-output[11] + output[12]) * cospi_16_64);
- step[12] = dct_32_round((output[12] + output[11]) * cospi_16_64);
- step[13] = dct_32_round((output[13] + output[10]) * cospi_16_64);
- step[14] = output[14];
- step[15] = output[15];
-
- step[16] = output[16] + output[23];
- step[17] = output[17] + output[22];
- step[18] = output[18] + output[21];
- step[19] = output[19] + output[20];
- step[20] = -output[20] + output[19];
- step[21] = -output[21] + output[18];
- step[22] = -output[22] + output[17];
- step[23] = -output[23] + output[16];
- step[24] = -output[24] + output[31];
- step[25] = -output[25] + output[30];
- step[26] = -output[26] + output[29];
- step[27] = -output[27] + output[28];
- step[28] = output[28] + output[27];
- step[29] = output[29] + output[26];
- step[30] = output[30] + output[25];
- step[31] = output[31] + output[24];
-
- // Stage 4
- output[0] = step[0] + step[3];
- output[1] = step[1] + step[2];
- output[2] = -step[2] + step[1];
- output[3] = -step[3] + step[0];
- output[4] = step[4];
- output[5] = dct_32_round((-step[5] + step[6]) * cospi_16_64);
- output[6] = dct_32_round((step[6] + step[5]) * cospi_16_64);
- output[7] = step[7];
- output[8] = step[8] + step[11];
- output[9] = step[9] + step[10];
- output[10] = -step[10] + step[9];
- output[11] = -step[11] + step[8];
- output[12] = -step[12] + step[15];
- output[13] = -step[13] + step[14];
- output[14] = step[14] + step[13];
- output[15] = step[15] + step[12];
-
- output[16] = step[16];
- output[17] = step[17];
- output[18] = dct_32_round(step[18] * -cospi_8_64 + step[29] * cospi_24_64);
- output[19] = dct_32_round(step[19] * -cospi_8_64 + step[28] * cospi_24_64);
- output[20] = dct_32_round(step[20] * -cospi_24_64 + step[27] * -cospi_8_64);
- output[21] = dct_32_round(step[21] * -cospi_24_64 + step[26] * -cospi_8_64);
- output[22] = step[22];
- output[23] = step[23];
- output[24] = step[24];
- output[25] = step[25];
- output[26] = dct_32_round(step[26] * cospi_24_64 + step[21] * -cospi_8_64);
- output[27] = dct_32_round(step[27] * cospi_24_64 + step[20] * -cospi_8_64);
- output[28] = dct_32_round(step[28] * cospi_8_64 + step[19] * cospi_24_64);
- output[29] = dct_32_round(step[29] * cospi_8_64 + step[18] * cospi_24_64);
- output[30] = step[30];
- output[31] = step[31];
-
- // Stage 5
- step[0] = dct_32_round((output[0] + output[1]) * cospi_16_64);
- step[1] = dct_32_round((-output[1] + output[0]) * cospi_16_64);
- step[2] = dct_32_round(output[2] * cospi_24_64 + output[3] * cospi_8_64);
- step[3] = dct_32_round(output[3] * cospi_24_64 - output[2] * cospi_8_64);
- step[4] = output[4] + output[5];
- step[5] = -output[5] + output[4];
- step[6] = -output[6] + output[7];
- step[7] = output[7] + output[6];
- step[8] = output[8];
- step[9] = dct_32_round(output[9] * -cospi_8_64 + output[14] * cospi_24_64);
- step[10] = dct_32_round(output[10] * -cospi_24_64 + output[13] * -cospi_8_64);
- step[11] = output[11];
- step[12] = output[12];
- step[13] = dct_32_round(output[13] * cospi_24_64 + output[10] * -cospi_8_64);
- step[14] = dct_32_round(output[14] * cospi_8_64 + output[9] * cospi_24_64);
- step[15] = output[15];
-
- step[16] = output[16] + output[19];
- step[17] = output[17] + output[18];
- step[18] = -output[18] + output[17];
- step[19] = -output[19] + output[16];
- step[20] = -output[20] + output[23];
- step[21] = -output[21] + output[22];
- step[22] = output[22] + output[21];
- step[23] = output[23] + output[20];
- step[24] = output[24] + output[27];
- step[25] = output[25] + output[26];
- step[26] = -output[26] + output[25];
- step[27] = -output[27] + output[24];
- step[28] = -output[28] + output[31];
- step[29] = -output[29] + output[30];
- step[30] = output[30] + output[29];
- step[31] = output[31] + output[28];
-
- // Stage 6
- output[0] = step[0];
- output[1] = step[1];
- output[2] = step[2];
- output[3] = step[3];
- output[4] = dct_32_round(step[4] * cospi_28_64 + step[7] * cospi_4_64);
- output[5] = dct_32_round(step[5] * cospi_12_64 + step[6] * cospi_20_64);
- output[6] = dct_32_round(step[6] * cospi_12_64 + step[5] * -cospi_20_64);
- output[7] = dct_32_round(step[7] * cospi_28_64 + step[4] * -cospi_4_64);
- output[8] = step[8] + step[9];
- output[9] = -step[9] + step[8];
- output[10] = -step[10] + step[11];
- output[11] = step[11] + step[10];
- output[12] = step[12] + step[13];
- output[13] = -step[13] + step[12];
- output[14] = -step[14] + step[15];
- output[15] = step[15] + step[14];
-
- output[16] = step[16];
- output[17] = dct_32_round(step[17] * -cospi_4_64 + step[30] * cospi_28_64);
- output[18] = dct_32_round(step[18] * -cospi_28_64 + step[29] * -cospi_4_64);
- output[19] = step[19];
- output[20] = step[20];
- output[21] = dct_32_round(step[21] * -cospi_20_64 + step[26] * cospi_12_64);
- output[22] = dct_32_round(step[22] * -cospi_12_64 + step[25] * -cospi_20_64);
- output[23] = step[23];
- output[24] = step[24];
- output[25] = dct_32_round(step[25] * cospi_12_64 + step[22] * -cospi_20_64);
- output[26] = dct_32_round(step[26] * cospi_20_64 + step[21] * cospi_12_64);
- output[27] = step[27];
- output[28] = step[28];
- output[29] = dct_32_round(step[29] * cospi_28_64 + step[18] * -cospi_4_64);
- output[30] = dct_32_round(step[30] * cospi_4_64 + step[17] * cospi_28_64);
- output[31] = step[31];
-
- // Stage 7
- step[0] = output[0];
- step[1] = output[1];
- step[2] = output[2];
- step[3] = output[3];
- step[4] = output[4];
- step[5] = output[5];
- step[6] = output[6];
- step[7] = output[7];
- step[8] = dct_32_round(output[8] * cospi_30_64 + output[15] * cospi_2_64);
- step[9] = dct_32_round(output[9] * cospi_14_64 + output[14] * cospi_18_64);
- step[10] = dct_32_round(output[10] * cospi_22_64 + output[13] * cospi_10_64);
- step[11] = dct_32_round(output[11] * cospi_6_64 + output[12] * cospi_26_64);
- step[12] = dct_32_round(output[12] * cospi_6_64 + output[11] * -cospi_26_64);
- step[13] = dct_32_round(output[13] * cospi_22_64 + output[10] * -cospi_10_64);
- step[14] = dct_32_round(output[14] * cospi_14_64 + output[9] * -cospi_18_64);
- step[15] = dct_32_round(output[15] * cospi_30_64 + output[8] * -cospi_2_64);
-
- step[16] = output[16] + output[17];
- step[17] = -output[17] + output[16];
- step[18] = -output[18] + output[19];
- step[19] = output[19] + output[18];
- step[20] = output[20] + output[21];
- step[21] = -output[21] + output[20];
- step[22] = -output[22] + output[23];
- step[23] = output[23] + output[22];
- step[24] = output[24] + output[25];
- step[25] = -output[25] + output[24];
- step[26] = -output[26] + output[27];
- step[27] = output[27] + output[26];
- step[28] = output[28] + output[29];
- step[29] = -output[29] + output[28];
- step[30] = -output[30] + output[31];
- step[31] = output[31] + output[30];
-
- // Final stage --- outputs indices are bit-reversed.
- output[0] = step[0];
- output[16] = step[1];
- output[8] = step[2];
- output[24] = step[3];
- output[4] = step[4];
- output[20] = step[5];
- output[12] = step[6];
- output[28] = step[7];
- output[2] = step[8];
- output[18] = step[9];
- output[10] = step[10];
- output[26] = step[11];
- output[6] = step[12];
- output[22] = step[13];
- output[14] = step[14];
- output[30] = step[15];
-
- output[1] = dct_32_round(step[16] * cospi_31_64 + step[31] * cospi_1_64);
- output[17] = dct_32_round(step[17] * cospi_15_64 + step[30] * cospi_17_64);
- output[9] = dct_32_round(step[18] * cospi_23_64 + step[29] * cospi_9_64);
- output[25] = dct_32_round(step[19] * cospi_7_64 + step[28] * cospi_25_64);
- output[5] = dct_32_round(step[20] * cospi_27_64 + step[27] * cospi_5_64);
- output[21] = dct_32_round(step[21] * cospi_11_64 + step[26] * cospi_21_64);
- output[13] = dct_32_round(step[22] * cospi_19_64 + step[25] * cospi_13_64);
- output[29] = dct_32_round(step[23] * cospi_3_64 + step[24] * cospi_29_64);
- output[3] = dct_32_round(step[24] * cospi_3_64 + step[23] * -cospi_29_64);
- output[19] = dct_32_round(step[25] * cospi_19_64 + step[22] * -cospi_13_64);
- output[11] = dct_32_round(step[26] * cospi_11_64 + step[21] * -cospi_21_64);
- output[27] = dct_32_round(step[27] * cospi_27_64 + step[20] * -cospi_5_64);
- output[7] = dct_32_round(step[28] * cospi_7_64 + step[19] * -cospi_25_64);
- output[23] = dct_32_round(step[29] * cospi_23_64 + step[18] * -cospi_9_64);
- output[15] = dct_32_round(step[30] * cospi_15_64 + step[17] * -cospi_17_64);
- output[31] = dct_32_round(step[31] * cospi_31_64 + step[16] * -cospi_1_64);
-}
-
-void vp9_fdct32x32_1_c(const int16_t *input, tran_low_t *output, int stride) {
- int r, c;
- tran_low_t sum = 0;
- for (r = 0; r < 32; ++r)
- for (c = 0; c < 32; ++c)
- sum += input[r * stride + c];
-
- output[0] = sum >> 3;
- output[1] = 0;
-}
-
-void vp9_fdct32x32_c(const int16_t *input, tran_low_t *out, int stride) {
- int i, j;
- tran_high_t output[32 * 32];
-
- // Columns
- for (i = 0; i < 32; ++i) {
- tran_high_t temp_in[32], temp_out[32];
- for (j = 0; j < 32; ++j)
- temp_in[j] = input[j * stride + i] * 4;
- vp9_fdct32(temp_in, temp_out, 0);
- for (j = 0; j < 32; ++j)
- output[j * 32 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2;
- }
-
- // Rows
- for (i = 0; i < 32; ++i) {
- tran_high_t temp_in[32], temp_out[32];
- for (j = 0; j < 32; ++j)
- temp_in[j] = output[j + i * 32];
- vp9_fdct32(temp_in, temp_out, 0);
- for (j = 0; j < 32; ++j)
- out[j + i * 32] =
- (tran_low_t)((temp_out[j] + 1 + (temp_out[j] < 0)) >> 2);
- }
-}
-
-// Note that although we use dct_32_round in dct32 computation flow,
-// this 2d fdct32x32 for rate-distortion optimization loop is operating
-// within 16 bits precision.
-void vp9_fdct32x32_rd_c(const int16_t *input, tran_low_t *out, int stride) {
- int i, j;
- tran_high_t output[32 * 32];
-
- // Columns
- for (i = 0; i < 32; ++i) {
- tran_high_t temp_in[32], temp_out[32];
- for (j = 0; j < 32; ++j)
- temp_in[j] = input[j * stride + i] * 4;
- vp9_fdct32(temp_in, temp_out, 0);
- for (j = 0; j < 32; ++j)
- // TODO(cd): see quality impact of only doing
- // output[j * 32 + i] = (temp_out[j] + 1) >> 2;
- // PS: also change code in vp9/encoder/x86/vp9_dct_sse2.c
- output[j * 32 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2;
- }
-
- // Rows
- for (i = 0; i < 32; ++i) {
- tran_high_t temp_in[32], temp_out[32];
- for (j = 0; j < 32; ++j)
- temp_in[j] = output[j + i * 32];
- vp9_fdct32(temp_in, temp_out, 1);
- for (j = 0; j < 32; ++j)
- out[j + i * 32] = (tran_low_t)temp_out[j];
- }
-}
-
#if CONFIG_VP9_HIGHBITDEPTH
void vp9_highbd_fht4x4_c(const int16_t *input, tran_low_t *output,
int stride, int tx_type) {
vp9_fht4x4_c(input, output, stride, tx_type);
}
-void vp9_highbd_fdct8x8_1_c(const int16_t *input, tran_low_t *final_output,
- int stride) {
- vp9_fdct8x8_1_c(input, final_output, stride);
-}
-
-void vp9_highbd_fdct16x16_1_c(const int16_t *input, tran_low_t *output,
- int stride) {
- vp9_fdct16x16_1_c(input, output, stride);
-}
-
void vp9_highbd_fht8x8_c(const int16_t *input, tran_low_t *output,
int stride, int tx_type) {
vp9_fht8x8_c(input, output, stride, tx_type);
@@ -1256,18 +808,4 @@
int stride, int tx_type) {
vp9_fht16x16_c(input, output, stride, tx_type);
}
-
-void vp9_highbd_fdct32x32_1_c(const int16_t *input, tran_low_t *out,
- int stride) {
- vp9_fdct32x32_1_c(input, out, stride);
-}
-
-void vp9_highbd_fdct32x32_c(const int16_t *input, tran_low_t *out, int stride) {
- vp9_fdct32x32_c(input, out, stride);
-}
-
-void vp9_highbd_fdct32x32_rd_c(const int16_t *input, tran_low_t *out,
- int stride) {
- vp9_fdct32x32_rd_c(input, out, stride);
-}
#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vp9/encoder/vp9_dct.h b/vp9/encoder/vp9_dct.h
deleted file mode 100644
index 6ce7e96..0000000
--- a/vp9/encoder/vp9_dct.h
+++ /dev/null
@@ -1,24 +0,0 @@
-/*
- * Copyright (c) 2014 The WebM project authors. All Rights Reserved.
- *
- * Use of this source code is governed by a BSD-style license
- * that can be found in the LICENSE file in the root of the source
- * tree. An additional intellectual property rights grant can be found
- * in the file PATENTS. All contributing project authors may
- * be found in the AUTHORS file in the root of the source tree.
- */
-
-#ifndef VP9_ENCODER_VP9_DCT_H_
-#define VP9_ENCODER_VP9_DCT_H_
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-void vp9_fdct32(const tran_high_t *input, tran_high_t *output, int round);
-
-#ifdef __cplusplus
-} // extern "C"
-#endif
-
-#endif // VP9_ENCODER_VP9_DCT_H_
diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index 658a11c..6199db4 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -3885,13 +3885,13 @@
#if CONFIG_VP9_HIGHBITDEPTH
if (cm->use_highbitdepth)
- x->fwd_txm4x4 = xd->lossless ? vp9_highbd_fwht4x4 : vp9_highbd_fdct4x4;
+ x->fwd_txm4x4 = xd->lossless ? vp9_highbd_fwht4x4 : vpx_highbd_fdct4x4;
else
- x->fwd_txm4x4 = xd->lossless ? vp9_fwht4x4 : vp9_fdct4x4;
+ x->fwd_txm4x4 = xd->lossless ? vp9_fwht4x4 : vpx_fdct4x4;
x->highbd_itxm_add = xd->lossless ? vp9_highbd_iwht4x4_add :
vp9_highbd_idct4x4_add;
#else
- x->fwd_txm4x4 = xd->lossless ? vp9_fwht4x4 : vp9_fdct4x4;
+ x->fwd_txm4x4 = xd->lossless ? vp9_fwht4x4 : vpx_fdct4x4;
#endif // CONFIG_VP9_HIGHBITDEPTH
x->itxm_add = xd->lossless ? vp9_iwht4x4_add : vp9_idct4x4_add;
diff --git a/vp9/encoder/vp9_encodemb.c b/vp9/encoder/vp9_encodemb.c
index 4e4b2a4..4cf5bbf 100644
--- a/vp9/encoder/vp9_encodemb.c
+++ b/vp9/encoder/vp9_encodemb.c
@@ -308,18 +308,18 @@
const int16_t *src, tran_low_t *dst,
int src_stride) {
if (rd_transform)
- vp9_fdct32x32_rd(src, dst, src_stride);
+ vpx_fdct32x32_rd(src, dst, src_stride);
else
- vp9_fdct32x32(src, dst, src_stride);
+ vpx_fdct32x32(src, dst, src_stride);
}
#if CONFIG_VP9_HIGHBITDEPTH
static INLINE void highbd_fdct32x32(int rd_transform, const int16_t *src,
tran_low_t *dst, int src_stride) {
if (rd_transform)
- vp9_highbd_fdct32x32_rd(src, dst, src_stride);
+ vpx_highbd_fdct32x32_rd(src, dst, src_stride);
else
- vp9_highbd_fdct32x32(src, dst, src_stride);
+ vpx_highbd_fdct32x32(src, dst, src_stride);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
@@ -351,14 +351,14 @@
scan_order->iscan);
break;
case TX_16X16:
- vp9_highbd_fdct16x16(src_diff, coeff, diff_stride);
+ vpx_highbd_fdct16x16(src_diff, coeff, diff_stride);
vp9_highbd_quantize_fp(coeff, 256, x->skip_block, p->zbin, p->round_fp,
p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
pd->dequant, eob,
scan_order->scan, scan_order->iscan);
break;
case TX_8X8:
- vp9_highbd_fdct8x8(src_diff, coeff, diff_stride);
+ vpx_highbd_fdct8x8(src_diff, coeff, diff_stride);
vp9_highbd_quantize_fp(coeff, 64, x->skip_block, p->zbin, p->round_fp,
p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
pd->dequant, eob,
@@ -387,7 +387,7 @@
scan_order->iscan);
break;
case TX_16X16:
- vp9_fdct16x16(src_diff, coeff, diff_stride);
+ vpx_fdct16x16(src_diff, coeff, diff_stride);
vp9_quantize_fp(coeff, 256, x->skip_block, p->zbin, p->round_fp,
p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
pd->dequant, eob,
@@ -433,19 +433,19 @@
if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
switch (tx_size) {
case TX_32X32:
- vp9_highbd_fdct32x32_1(src_diff, coeff, diff_stride);
+ vpx_highbd_fdct32x32_1(src_diff, coeff, diff_stride);
vp9_highbd_quantize_dc_32x32(coeff, x->skip_block, p->round,
p->quant_fp[0], qcoeff, dqcoeff,
pd->dequant[0], eob);
break;
case TX_16X16:
- vp9_highbd_fdct16x16_1(src_diff, coeff, diff_stride);
+ vpx_highbd_fdct16x16_1(src_diff, coeff, diff_stride);
vp9_highbd_quantize_dc(coeff, 256, x->skip_block, p->round,
p->quant_fp[0], qcoeff, dqcoeff,
pd->dequant[0], eob);
break;
case TX_8X8:
- vp9_highbd_fdct8x8_1(src_diff, coeff, diff_stride);
+ vpx_highbd_fdct8x8_1(src_diff, coeff, diff_stride);
vp9_highbd_quantize_dc(coeff, 64, x->skip_block, p->round,
p->quant_fp[0], qcoeff, dqcoeff,
pd->dequant[0], eob);
@@ -465,19 +465,19 @@
switch (tx_size) {
case TX_32X32:
- vp9_fdct32x32_1(src_diff, coeff, diff_stride);
+ vpx_fdct32x32_1(src_diff, coeff, diff_stride);
vp9_quantize_dc_32x32(coeff, x->skip_block, p->round,
p->quant_fp[0], qcoeff, dqcoeff,
pd->dequant[0], eob);
break;
case TX_16X16:
- vp9_fdct16x16_1(src_diff, coeff, diff_stride);
+ vpx_fdct16x16_1(src_diff, coeff, diff_stride);
vp9_quantize_dc(coeff, 256, x->skip_block, p->round,
p->quant_fp[0], qcoeff, dqcoeff,
pd->dequant[0], eob);
break;
case TX_8X8:
- vp9_fdct8x8_1(src_diff, coeff, diff_stride);
+ vpx_fdct8x8_1(src_diff, coeff, diff_stride);
vp9_quantize_dc(coeff, 64, x->skip_block, p->round,
p->quant_fp[0], qcoeff, dqcoeff,
pd->dequant[0], eob);
@@ -521,14 +521,14 @@
scan_order->scan, scan_order->iscan);
break;
case TX_16X16:
- vp9_highbd_fdct16x16(src_diff, coeff, diff_stride);
+ vpx_highbd_fdct16x16(src_diff, coeff, diff_stride);
vp9_highbd_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
p->quant, p->quant_shift, qcoeff, dqcoeff,
pd->dequant, eob,
scan_order->scan, scan_order->iscan);
break;
case TX_8X8:
- vp9_highbd_fdct8x8(src_diff, coeff, diff_stride);
+ vpx_highbd_fdct8x8(src_diff, coeff, diff_stride);
vp9_highbd_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
p->quant, p->quant_shift, qcoeff, dqcoeff,
pd->dequant, eob,
@@ -557,14 +557,14 @@
scan_order->iscan);
break;
case TX_16X16:
- vp9_fdct16x16(src_diff, coeff, diff_stride);
+ vpx_fdct16x16(src_diff, coeff, diff_stride);
vp9_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
p->quant, p->quant_shift, qcoeff, dqcoeff,
pd->dequant, eob,
scan_order->scan, scan_order->iscan);
break;
case TX_8X8:
- vp9_fdct8x8(src_diff, coeff, diff_stride);
+ vpx_fdct8x8(src_diff, coeff, diff_stride);
vp9_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
p->quant, p->quant_shift, qcoeff, dqcoeff,
pd->dequant, eob,
@@ -829,7 +829,7 @@
vpx_highbd_subtract_block(16, 16, src_diff, diff_stride,
src, src_stride, dst, dst_stride, xd->bd);
if (tx_type == DCT_DCT)
- vp9_highbd_fdct16x16(src_diff, coeff, diff_stride);
+ vpx_highbd_fdct16x16(src_diff, coeff, diff_stride);
else
vp9_highbd_fht16x16(src_diff, coeff, diff_stride, tx_type);
vp9_highbd_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
@@ -847,7 +847,7 @@
vpx_highbd_subtract_block(8, 8, src_diff, diff_stride,
src, src_stride, dst, dst_stride, xd->bd);
if (tx_type == DCT_DCT)
- vp9_highbd_fdct8x8(src_diff, coeff, diff_stride);
+ vpx_highbd_fdct8x8(src_diff, coeff, diff_stride);
else
vp9_highbd_fht8x8(src_diff, coeff, diff_stride, tx_type);
vp9_highbd_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
diff --git a/vp9/encoder/vp9_encoder.c b/vp9/encoder/vp9_encoder.c
index 7a41a7e..53e747c 100644
--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -4306,13 +4306,13 @@
#if CONFIG_VP9_HIGHBITDEPTH
if (cpi->oxcf.use_highbitdepth)
cpi->td.mb.fwd_txm4x4 = lossless ?
- vp9_highbd_fwht4x4 : vp9_highbd_fdct4x4;
+ vp9_highbd_fwht4x4 : vpx_highbd_fdct4x4;
else
- cpi->td.mb.fwd_txm4x4 = lossless ? vp9_fwht4x4 : vp9_fdct4x4;
+ cpi->td.mb.fwd_txm4x4 = lossless ? vp9_fwht4x4 : vpx_fdct4x4;
cpi->td.mb.highbd_itxm_add = lossless ? vp9_highbd_iwht4x4_add :
vp9_highbd_idct4x4_add;
#else
- cpi->td.mb.fwd_txm4x4 = lossless ? vp9_fwht4x4 : vp9_fdct4x4;
+ cpi->td.mb.fwd_txm4x4 = lossless ? vp9_fwht4x4 : vpx_fdct4x4;
#endif // CONFIG_VP9_HIGHBITDEPTH
cpi->td.mb.itxm_add = lossless ? vp9_iwht4x4_add : vp9_idct4x4_add;
vp9_first_pass(cpi, source);
diff --git a/vp9/encoder/vp9_pickmode.c b/vp9/encoder/vp9_pickmode.c
index 6d09dbe..34bf8a6 100644
--- a/vp9/encoder/vp9_pickmode.c
+++ b/vp9/encoder/vp9_pickmode.c
@@ -611,7 +611,7 @@
switch (tx_size) {
case TX_32X32:
- vp9_fdct32x32_rd(src_diff, coeff, diff_stride);
+ vpx_fdct32x32_rd(src_diff, coeff, diff_stride);
vp9_quantize_fp_32x32(coeff, 1024, x->skip_block, p->zbin,
p->round_fp, p->quant_fp, p->quant_shift,
qcoeff, dqcoeff, pd->dequant, eob,
diff --git a/vp9/encoder/vp9_psnrhvs.c b/vp9/encoder/vp9_psnrhvs.c
index a91a4ee..5104b9a 100644
--- a/vp9/encoder/vp9_psnrhvs.c
+++ b/vp9/encoder/vp9_psnrhvs.c
@@ -26,7 +26,7 @@
void od_bin_fdct8x8(tran_low_t *y, int ystride, const int16_t *x, int xstride) {
(void) xstride;
- vp9_fdct8x8(x, y, ystride);
+ vpx_fdct8x8(x, y, ystride);
}
/* Normalized inverse quantization matrix for 8x8 DCT at the point of
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 8ae6783..2a16ce7 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -846,7 +846,7 @@
const TX_TYPE tx_type = get_tx_type_4x4(PLANE_TYPE_Y, xd, block);
const scan_order *so = &vp9_scan_orders[TX_4X4][tx_type];
if (tx_type == DCT_DCT)
- vp9_highbd_fdct4x4(src_diff, coeff, 8);
+ vpx_highbd_fdct4x4(src_diff, coeff, 8);
else
vp9_highbd_fht4x4(src_diff, coeff, 8, tx_type);
vp9_regular_quantize_b_4x4(x, 0, block, so->scan, so->iscan);
diff --git a/vp9/encoder/vp9_tokenize.c b/vp9/encoder/vp9_tokenize.c
index a1c076a..85cb2fc 100644
--- a/vp9/encoder/vp9_tokenize.c
+++ b/vp9/encoder/vp9_tokenize.c
@@ -568,23 +568,24 @@
}
struct is_skippable_args {
- MACROBLOCK *x;
+ uint16_t *eobs;
int *skippable;
};
static void is_skippable(int plane, int block,
BLOCK_SIZE plane_bsize, TX_SIZE tx_size,
void *argv) {
struct is_skippable_args *args = argv;
+ (void)plane;
(void)plane_bsize;
(void)tx_size;
- args->skippable[0] &= (!args->x->plane[plane].eobs[block]);
+ args->skippable[0] &= (!args->eobs[block]);
}
// TODO(yaowu): rewrite and optimize this function to remove the usage of
// vp9_foreach_transform_block() and simplify is_skippable().
int vp9_is_skippable_in_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane) {
int result = 1;
- struct is_skippable_args args = {x, &result};
+ struct is_skippable_args args = {x->plane[plane].eobs, &result};
vp9_foreach_transformed_block_in_plane(&x->e_mbd, bsize, plane, is_skippable,
&args);
return result;
@@ -595,14 +596,15 @@
void *argv) {
struct is_skippable_args *args = argv;
int eobs = (tx_size == TX_4X4) ? 3 : 10;
+ (void) plane;
(void) plane_bsize;
- *(args->skippable) |= (args->x->plane[plane].eobs[block] > eobs);
+ *(args->skippable) |= (args->eobs[block] > eobs);
}
int vp9_has_high_freq_in_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane) {
int result = 0;
- struct is_skippable_args args = {x, &result};
+ struct is_skippable_args args = {x->plane[plane].eobs, &result};
vp9_foreach_transformed_block_in_plane(&x->e_mbd, bsize, plane,
has_high_freq_coeff, &args);
return result;
diff --git a/vp9/encoder/x86/vp9_dct_sse2.c b/vp9/encoder/x86/vp9_dct_sse2.c
index f263c84..fa37b6f 100644
--- a/vp9/encoder/x86/vp9_dct_sse2.c
+++ b/vp9/encoder/x86/vp9_dct_sse2.c
@@ -18,35 +18,6 @@
#include "vpx_dsp/x86/txfm_common_sse2.h"
#include "vpx_ports/mem.h"
-void vp9_fdct4x4_1_sse2(const int16_t *input, tran_low_t *output, int stride) {
- __m128i in0, in1;
- __m128i tmp;
- const __m128i zero = _mm_setzero_si128();
- in0 = _mm_loadl_epi64((const __m128i *)(input + 0 * stride));
- in1 = _mm_loadl_epi64((const __m128i *)(input + 1 * stride));
- in1 = _mm_unpacklo_epi64(in1, _mm_loadl_epi64((const __m128i *)
- (input + 2 * stride)));
- in0 = _mm_unpacklo_epi64(in0, _mm_loadl_epi64((const __m128i *)
- (input + 3 * stride)));
-
- tmp = _mm_add_epi16(in0, in1);
- in0 = _mm_unpacklo_epi16(zero, tmp);
- in1 = _mm_unpackhi_epi16(zero, tmp);
- in0 = _mm_srai_epi32(in0, 16);
- in1 = _mm_srai_epi32(in1, 16);
-
- tmp = _mm_add_epi32(in0, in1);
- in0 = _mm_unpacklo_epi32(tmp, zero);
- in1 = _mm_unpackhi_epi32(tmp, zero);
-
- tmp = _mm_add_epi32(in0, in1);
- in0 = _mm_srli_si128(tmp, 8);
-
- in1 = _mm_add_epi32(tmp, in0);
- in0 = _mm_slli_epi32(in1, 1);
- store_output(&in0, output);
-}
-
static INLINE void load_buffer_4x4(const int16_t *input, __m128i *in,
int stride) {
const __m128i k__nonzero_bias_a = _mm_setr_epi16(0, 1, 1, 1, 1, 1, 1, 1);
@@ -186,7 +157,7 @@
switch (tx_type) {
case DCT_DCT:
- vp9_fdct4x4_sse2(input, output, stride);
+ vpx_fdct4x4_sse2(input, output, stride);
break;
case ADST_DCT:
load_buffer_4x4(input, in, stride);
@@ -212,46 +183,6 @@
}
}
-void vp9_fdct8x8_1_sse2(const int16_t *input, tran_low_t *output, int stride) {
- __m128i in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
- __m128i in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
- __m128i in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
- __m128i in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
- __m128i u0, u1, sum;
-
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
-
- in0 = _mm_load_si128((const __m128i *)(input + 4 * stride));
- in1 = _mm_load_si128((const __m128i *)(input + 5 * stride));
- in2 = _mm_load_si128((const __m128i *)(input + 6 * stride));
- in3 = _mm_load_si128((const __m128i *)(input + 7 * stride));
-
- sum = _mm_add_epi16(u0, u1);
-
- in0 = _mm_add_epi16(in0, in1);
- in2 = _mm_add_epi16(in2, in3);
- sum = _mm_add_epi16(sum, in0);
-
- u0 = _mm_setzero_si128();
- sum = _mm_add_epi16(sum, in2);
-
- in0 = _mm_unpacklo_epi16(u0, sum);
- in1 = _mm_unpackhi_epi16(u0, sum);
- in0 = _mm_srai_epi32(in0, 16);
- in1 = _mm_srai_epi32(in1, 16);
-
- sum = _mm_add_epi32(in0, in1);
- in0 = _mm_unpacklo_epi32(sum, u0);
- in1 = _mm_unpackhi_epi32(sum, u0);
-
- sum = _mm_add_epi32(in0, in1);
- in0 = _mm_srli_si128(sum, 8);
-
- in1 = _mm_add_epi32(sum, in0);
- store_output(&in1, output);
-}
-
void vp9_fdct8x8_quant_sse2(const int16_t *input, int stride,
int16_t* coeff_ptr, intptr_t n_coeffs,
int skip_block, const int16_t* zbin_ptr,
@@ -1210,7 +1141,7 @@
switch (tx_type) {
case DCT_DCT:
- vp9_fdct8x8_sse2(input, output, stride);
+ vpx_fdct8x8_sse2(input, output, stride);
break;
case ADST_DCT:
load_buffer_8x8(input, in, stride);
@@ -1239,75 +1170,6 @@
}
}
-void vp9_fdct16x16_1_sse2(const int16_t *input, tran_low_t *output,
- int stride) {
- __m128i in0, in1, in2, in3;
- __m128i u0, u1;
- __m128i sum = _mm_setzero_si128();
- int i;
-
- for (i = 0; i < 2; ++i) {
- input += 8 * i;
- in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
- in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
- in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
- in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
-
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
- sum = _mm_add_epi16(sum, u0);
-
- in0 = _mm_load_si128((const __m128i *)(input + 4 * stride));
- in1 = _mm_load_si128((const __m128i *)(input + 5 * stride));
- in2 = _mm_load_si128((const __m128i *)(input + 6 * stride));
- in3 = _mm_load_si128((const __m128i *)(input + 7 * stride));
-
- sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
- sum = _mm_add_epi16(sum, u0);
-
- in0 = _mm_load_si128((const __m128i *)(input + 8 * stride));
- in1 = _mm_load_si128((const __m128i *)(input + 9 * stride));
- in2 = _mm_load_si128((const __m128i *)(input + 10 * stride));
- in3 = _mm_load_si128((const __m128i *)(input + 11 * stride));
-
- sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
- sum = _mm_add_epi16(sum, u0);
-
- in0 = _mm_load_si128((const __m128i *)(input + 12 * stride));
- in1 = _mm_load_si128((const __m128i *)(input + 13 * stride));
- in2 = _mm_load_si128((const __m128i *)(input + 14 * stride));
- in3 = _mm_load_si128((const __m128i *)(input + 15 * stride));
-
- sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
- sum = _mm_add_epi16(sum, u0);
-
- sum = _mm_add_epi16(sum, u1);
- }
-
- u0 = _mm_setzero_si128();
- in0 = _mm_unpacklo_epi16(u0, sum);
- in1 = _mm_unpackhi_epi16(u0, sum);
- in0 = _mm_srai_epi32(in0, 16);
- in1 = _mm_srai_epi32(in1, 16);
-
- sum = _mm_add_epi32(in0, in1);
- in0 = _mm_unpacklo_epi32(sum, u0);
- in1 = _mm_unpackhi_epi32(sum, u0);
-
- sum = _mm_add_epi32(in0, in1);
- in0 = _mm_srli_si128(sum, 8);
-
- in1 = _mm_add_epi32(sum, in0);
- in1 = _mm_srai_epi32(in1, 1);
- store_output(&in1, output);
-}
-
static INLINE void load_buffer_16x16(const int16_t* input, __m128i *in0,
__m128i *in1, int stride) {
// load first 8 columns
@@ -2166,7 +2028,7 @@
switch (tx_type) {
case DCT_DCT:
- vp9_fdct16x16_sse2(input, output, stride);
+ vpx_fdct16x16_sse2(input, output, stride);
break;
case ADST_DCT:
load_buffer_16x16(input, in0, in1, stride);
@@ -2194,119 +2056,3 @@
break;
}
}
-
-void vp9_fdct32x32_1_sse2(const int16_t *input, tran_low_t *output,
- int stride) {
- __m128i in0, in1, in2, in3;
- __m128i u0, u1;
- __m128i sum = _mm_setzero_si128();
- int i;
-
- for (i = 0; i < 8; ++i) {
- in0 = _mm_load_si128((const __m128i *)(input + 0));
- in1 = _mm_load_si128((const __m128i *)(input + 8));
- in2 = _mm_load_si128((const __m128i *)(input + 16));
- in3 = _mm_load_si128((const __m128i *)(input + 24));
-
- input += stride;
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
- sum = _mm_add_epi16(sum, u0);
-
- in0 = _mm_load_si128((const __m128i *)(input + 0));
- in1 = _mm_load_si128((const __m128i *)(input + 8));
- in2 = _mm_load_si128((const __m128i *)(input + 16));
- in3 = _mm_load_si128((const __m128i *)(input + 24));
-
- input += stride;
- sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
- sum = _mm_add_epi16(sum, u0);
-
- in0 = _mm_load_si128((const __m128i *)(input + 0));
- in1 = _mm_load_si128((const __m128i *)(input + 8));
- in2 = _mm_load_si128((const __m128i *)(input + 16));
- in3 = _mm_load_si128((const __m128i *)(input + 24));
-
- input += stride;
- sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
- sum = _mm_add_epi16(sum, u0);
-
- in0 = _mm_load_si128((const __m128i *)(input + 0));
- in1 = _mm_load_si128((const __m128i *)(input + 8));
- in2 = _mm_load_si128((const __m128i *)(input + 16));
- in3 = _mm_load_si128((const __m128i *)(input + 24));
-
- input += stride;
- sum = _mm_add_epi16(sum, u1);
- u0 = _mm_add_epi16(in0, in1);
- u1 = _mm_add_epi16(in2, in3);
- sum = _mm_add_epi16(sum, u0);
-
- sum = _mm_add_epi16(sum, u1);
- }
-
- u0 = _mm_setzero_si128();
- in0 = _mm_unpacklo_epi16(u0, sum);
- in1 = _mm_unpackhi_epi16(u0, sum);
- in0 = _mm_srai_epi32(in0, 16);
- in1 = _mm_srai_epi32(in1, 16);
-
- sum = _mm_add_epi32(in0, in1);
- in0 = _mm_unpacklo_epi32(sum, u0);
- in1 = _mm_unpackhi_epi32(sum, u0);
-
- sum = _mm_add_epi32(in0, in1);
- in0 = _mm_srli_si128(sum, 8);
-
- in1 = _mm_add_epi32(sum, in0);
- in1 = _mm_srai_epi32(in1, 3);
- store_output(&in1, output);
-}
-
-/*
- * The DCTnxn functions are defined using the macros below. The main code for
- * them is in separate files (vp9/encoder/x86/vp9_dct_sse2_impl.h &
- * vp9/encoder/x86/vp9_dct32x32_sse2_impl.h) which are used by both the 8 bit code
- * and the high bit depth code.
- */
-
-#define DCT_HIGH_BIT_DEPTH 0
-
-#define FDCT32x32_2D vp9_fdct32x32_rd_sse2
-#define FDCT32x32_HIGH_PRECISION 0
-#include "vp9/encoder/x86/vp9_dct32x32_sse2_impl.h"
-#undef FDCT32x32_2D
-#undef FDCT32x32_HIGH_PRECISION
-
-#define FDCT32x32_2D vp9_fdct32x32_sse2
-#define FDCT32x32_HIGH_PRECISION 1
-#include "vp9/encoder/x86/vp9_dct32x32_sse2_impl.h" // NOLINT
-#undef FDCT32x32_2D
-#undef FDCT32x32_HIGH_PRECISION
-
-#undef DCT_HIGH_BIT_DEPTH
-
-
-#if CONFIG_VP9_HIGHBITDEPTH
-
-#define DCT_HIGH_BIT_DEPTH 1
-
-#define FDCT32x32_2D vp9_highbd_fdct32x32_rd_sse2
-#define FDCT32x32_HIGH_PRECISION 0
-#include "vp9/encoder/x86/vp9_dct32x32_sse2_impl.h" // NOLINT
-#undef FDCT32x32_2D
-#undef FDCT32x32_HIGH_PRECISION
-
-#define FDCT32x32_2D vp9_highbd_fdct32x32_sse2
-#define FDCT32x32_HIGH_PRECISION 1
-#include "vp9/encoder/x86/vp9_dct32x32_sse2_impl.h" // NOLINT
-#undef FDCT32x32_2D
-#undef FDCT32x32_HIGH_PRECISION
-
-#undef DCT_HIGH_BIT_DEPTH
-
-#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vp9/vp9_common.mk b/vp9/vp9_common.mk
index 50c11b9..78ea63f 100644
--- a/vp9/vp9_common.mk
+++ b/vp9/vp9_common.mk
@@ -84,16 +84,11 @@
ifeq ($(CONFIG_USE_X86INC),yes)
VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_copy_sse2.asm
-VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_intrapred_sse2.asm
-VP9_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/vp9_intrapred_ssse3.asm
endif
ifeq ($(CONFIG_VP9_HIGHBITDEPTH),yes)
VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_high_subpixel_8t_sse2.asm
VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_high_subpixel_bilinear_sse2.asm
-ifeq ($(CONFIG_USE_X86INC),yes)
-VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_high_intrapred_sse2.asm
-endif
endif
# common (c)
@@ -108,9 +103,6 @@
VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_convolve8_dspr2.c
VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_convolve8_horiz_dspr2.c
VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_convolve8_vert_dspr2.c
-VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_intrapred4_dspr2.c
-VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_intrapred8_dspr2.c
-VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_intrapred16_dspr2.c
ifneq ($(CONFIG_VP9_HIGHBITDEPTH),yes)
VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_itrans4_dspr2.c
@@ -135,7 +127,6 @@
VP9_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/vp9_idct16x16_msa.c
VP9_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/vp9_idct32x32_msa.c
VP9_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/vp9_idct_msa.h
-VP9_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/vp9_intra_predict_msa.c
ifeq ($(CONFIG_VP9_POSTPROC),yes)
VP9_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/vp9_mfqe_msa.c
@@ -174,7 +165,6 @@
VP9_COMMON_SRCS-yes += common/arm/neon/vp9_idct4x4_add_neon_asm$(ASM)
VP9_COMMON_SRCS-yes += common/arm/neon/vp9_idct8x8_1_add_neon_asm$(ASM)
VP9_COMMON_SRCS-yes += common/arm/neon/vp9_idct8x8_add_neon_asm$(ASM)
-VP9_COMMON_SRCS-yes += common/arm/neon/vp9_reconintra_neon_asm$(ASM)
else
ifeq ($(HAVE_NEON), yes)
VP9_COMMON_SRCS-yes += common/arm/neon/vp9_convolve8_avg_neon.c
@@ -194,6 +184,4 @@
endif # HAVE_NEON
endif # HAVE_NEON_ASM
-VP9_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/vp9_reconintra_neon.c
-
$(eval $(call rtcd_h_template,vp9_rtcd,vp9/common/vp9_rtcd_defs.pl))
diff --git a/vp9/vp9cx.mk b/vp9/vp9cx.mk
index 1e28454..186ce11 100644
--- a/vp9/vp9cx.mk
+++ b/vp9/vp9cx.mk
@@ -24,7 +24,6 @@
VP9_CX_SRCS-yes += encoder/vp9_cost.h
VP9_CX_SRCS-yes += encoder/vp9_cost.c
VP9_CX_SRCS-yes += encoder/vp9_dct.c
-VP9_CX_SRCS-yes += encoder/vp9_dct.h
VP9_CX_SRCS-$(CONFIG_VP9_TEMPORAL_DENOISING) += encoder/vp9_denoiser.c
VP9_CX_SRCS-$(CONFIG_VP9_TEMPORAL_DENOISING) += encoder/vp9_denoiser.h
VP9_CX_SRCS-yes += encoder/vp9_encodeframe.c
@@ -118,14 +117,11 @@
VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_dct_sse2.c
VP9_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/vp9_dct_ssse3.c
-VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_dct32x32_sse2_impl.h
ifeq ($(CONFIG_VP9_TEMPORAL_DENOISING),yes)
VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_denoiser_sse2.c
endif
-VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_dct32x32_avx2_impl.h
-VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_dct_avx2.c
VP9_CX_SRCS-$(HAVE_AVX2) += encoder/x86/vp9_error_intrin_avx2.c
ifneq ($(CONFIG_VP9_HIGHBITDEPTH),yes)
@@ -140,7 +136,6 @@
VP9_CX_SRCS-$(HAVE_MSA) += encoder/mips/msa/vp9_fdct4x4_msa.c
VP9_CX_SRCS-$(HAVE_MSA) += encoder/mips/msa/vp9_fdct8x8_msa.c
VP9_CX_SRCS-$(HAVE_MSA) += encoder/mips/msa/vp9_fdct16x16_msa.c
-VP9_CX_SRCS-$(HAVE_MSA) += encoder/mips/msa/vp9_fdct32x32_msa.c
VP9_CX_SRCS-$(HAVE_MSA) += encoder/mips/msa/vp9_fdct_msa.h
VP9_CX_SRCS-$(HAVE_MSA) += encoder/mips/msa/vp9_temporal_filter_msa.c
diff --git a/vpx_dsp/arm/fwd_txfm_neon.c b/vpx_dsp/arm/fwd_txfm_neon.c
index 6258b61..79afc91 100644
--- a/vpx_dsp/arm/fwd_txfm_neon.c
+++ b/vpx_dsp/arm/fwd_txfm_neon.c
@@ -13,7 +13,7 @@
#include "./vpx_config.h"
#include "vpx_dsp/txfm_common.h"
-void vp9_fdct8x8_neon(const int16_t *input, int16_t *final_output, int stride) {
+void vpx_fdct8x8_neon(const int16_t *input, int16_t *final_output, int stride) {
int i;
// stage 1
int16x8_t input_0 = vshlq_n_s16(vld1q_s16(&input[0 * stride]), 2);
@@ -201,3 +201,20 @@
vst1q_s16(&final_output[7 * 8], input_7);
}
}
+
+void vpx_fdct8x8_1_neon(const int16_t *input, int16_t *output, int stride) {
+ int r;
+ int16x8_t sum = vld1q_s16(&input[0]);
+ for (r = 1; r < 8; ++r) {
+ const int16x8_t input_00 = vld1q_s16(&input[r * stride]);
+ sum = vaddq_s16(sum, input_00);
+ }
+ {
+ const int32x4_t a = vpaddlq_s16(sum);
+ const int64x2_t b = vpaddlq_s32(a);
+ const int32x2_t c = vadd_s32(vreinterpret_s32_s64(vget_low_s64(b)),
+ vreinterpret_s32_s64(vget_high_s64(b)));
+ output[0] = vget_lane_s16(vreinterpret_s16_s32(c), 0);
+ output[1] = 0;
+ }
+}
diff --git a/vp9/common/arm/neon/vp9_reconintra_neon.c b/vpx_dsp/arm/intrapred_neon.c
similarity index 92%
rename from vp9/common/arm/neon/vp9_reconintra_neon.c
rename to vpx_dsp/arm/intrapred_neon.c
index 92706bf..0a37610 100644
--- a/vp9/common/arm/neon/vp9_reconintra_neon.c
+++ b/vpx_dsp/arm/intrapred_neon.c
@@ -10,8 +10,8 @@
#include <arm_neon.h>
-#include "./vp9_rtcd.h"
#include "./vpx_config.h"
+#include "./vpx_dsp_rtcd.h"
#include "vpx/vpx_integer.h"
//------------------------------------------------------------------------------
@@ -59,24 +59,24 @@
}
}
-void vp9_dc_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
dc_4x4(dst, stride, above, left, 1, 1);
}
-void vp9_dc_left_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_left_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
dc_4x4(dst, stride, NULL, left, 0, 1);
}
-void vp9_dc_top_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_top_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
(void)left;
dc_4x4(dst, stride, above, NULL, 1, 0);
}
-void vp9_dc_128_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_128_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
(void)left;
@@ -130,24 +130,24 @@
}
}
-void vp9_dc_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
dc_8x8(dst, stride, above, left, 1, 1);
}
-void vp9_dc_left_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_left_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
dc_8x8(dst, stride, NULL, left, 0, 1);
}
-void vp9_dc_top_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_top_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
(void)left;
dc_8x8(dst, stride, above, NULL, 1, 0);
}
-void vp9_dc_128_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_128_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
(void)left;
@@ -203,26 +203,26 @@
}
}
-void vp9_dc_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
dc_16x16(dst, stride, above, left, 1, 1);
}
-void vp9_dc_left_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_left_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above,
const uint8_t *left) {
(void)above;
dc_16x16(dst, stride, NULL, left, 0, 1);
}
-void vp9_dc_top_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_top_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above,
const uint8_t *left) {
(void)left;
dc_16x16(dst, stride, above, NULL, 1, 0);
}
-void vp9_dc_128_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_128_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above,
const uint8_t *left) {
(void)above;
@@ -286,26 +286,26 @@
}
}
-void vp9_dc_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
dc_32x32(dst, stride, above, left, 1, 1);
}
-void vp9_dc_left_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_left_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above,
const uint8_t *left) {
(void)above;
dc_32x32(dst, stride, NULL, left, 0, 1);
}
-void vp9_dc_top_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_top_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above,
const uint8_t *left) {
(void)left;
dc_32x32(dst, stride, above, NULL, 1, 0);
}
-void vp9_dc_128_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_128_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above,
const uint8_t *left) {
(void)above;
@@ -315,7 +315,7 @@
// -----------------------------------------------------------------------------
-void vp9_d45_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_d45_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
const uint64x1_t A0 = vreinterpret_u64_u8(vld1_u8(above)); // top row
const uint64x1_t A1 = vshr_n_u64(A0, 8);
@@ -338,7 +338,7 @@
dst[3 * stride + 3] = above[7];
}
-void vp9_d45_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_d45_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
static const uint8_t shuffle1[8] = { 1, 2, 3, 4, 5, 6, 7, 7 };
static const uint8_t shuffle2[8] = { 2, 3, 4, 5, 6, 7, 7, 7 };
@@ -358,7 +358,7 @@
vst1_u8(dst + i * stride, row);
}
-void vp9_d45_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_d45_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
const uint8x16_t A0 = vld1q_u8(above); // top row
const uint8x16_t above_right = vld1q_dup_u8(above + 15);
@@ -377,7 +377,7 @@
// -----------------------------------------------------------------------------
-void vp9_d135_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_d135_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
const uint8x8_t XABCD_u8 = vld1_u8(above - 1);
const uint64x1_t XABCD = vreinterpret_u64_u8(XABCD_u8);
@@ -407,7 +407,7 @@
#if !HAVE_NEON_ASM
-void vp9_v_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_v_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int i;
uint32x2_t d0u32 = vdup_n_u32(0);
@@ -418,7 +418,7 @@
vst1_lane_u32((uint32_t *)dst, d0u32, 0);
}
-void vp9_v_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_v_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int i;
uint8x8_t d0u8 = vdup_n_u8(0);
@@ -429,7 +429,7 @@
vst1_u8(dst, d0u8);
}
-void vp9_v_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_v_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int i;
uint8x16_t q0u8 = vdupq_n_u8(0);
@@ -440,7 +440,7 @@
vst1q_u8(dst, q0u8);
}
-void vp9_v_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_v_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int i;
uint8x16_t q0u8 = vdupq_n_u8(0);
@@ -455,7 +455,7 @@
}
}
-void vp9_h_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_h_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
uint8x8_t d0u8 = vdup_n_u8(0);
uint32x2_t d1u32 = vdup_n_u32(0);
@@ -476,7 +476,7 @@
vst1_lane_u32((uint32_t *)dst, vreinterpret_u32_u8(d0u8), 0);
}
-void vp9_h_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_h_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
uint8x8_t d0u8 = vdup_n_u8(0);
uint64x1_t d1u64 = vdup_n_u64(0);
@@ -509,7 +509,7 @@
vst1_u8(dst, d0u8);
}
-void vp9_h_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_h_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int j;
uint8x8_t d2u8 = vdup_n_u8(0);
@@ -547,7 +547,7 @@
}
}
-void vp9_h_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_h_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int j, k;
uint8x8_t d2u8 = vdup_n_u8(0);
@@ -595,7 +595,7 @@
}
}
-void vp9_tm_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_tm_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int i;
uint16x8_t q1u16, q3u16;
@@ -615,7 +615,7 @@
}
}
-void vp9_tm_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_tm_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int j;
uint16x8_t q0u16, q3u16, q10u16;
@@ -657,7 +657,7 @@
}
}
-void vp9_tm_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_tm_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int j, k;
uint16x8_t q0u16, q2u16, q3u16, q8u16, q10u16;
@@ -720,7 +720,7 @@
}
}
-void vp9_tm_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
+void vpx_tm_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int j, k;
uint16x8_t q0u16, q3u16, q8u16, q9u16, q10u16, q11u16;
diff --git a/vp9/common/arm/neon/vp9_reconintra_neon_asm.asm b/vpx_dsp/arm/intrapred_neon_asm.asm
similarity index 88%
rename from vp9/common/arm/neon/vp9_reconintra_neon_asm.asm
rename to vpx_dsp/arm/intrapred_neon_asm.asm
index 14f574a..115790d 100644
--- a/vp9/common/arm/neon/vp9_reconintra_neon_asm.asm
+++ b/vpx_dsp/arm/intrapred_neon_asm.asm
@@ -8,25 +8,25 @@
; be found in the AUTHORS file in the root of the source tree.
;
- EXPORT |vp9_v_predictor_4x4_neon|
- EXPORT |vp9_v_predictor_8x8_neon|
- EXPORT |vp9_v_predictor_16x16_neon|
- EXPORT |vp9_v_predictor_32x32_neon|
- EXPORT |vp9_h_predictor_4x4_neon|
- EXPORT |vp9_h_predictor_8x8_neon|
- EXPORT |vp9_h_predictor_16x16_neon|
- EXPORT |vp9_h_predictor_32x32_neon|
- EXPORT |vp9_tm_predictor_4x4_neon|
- EXPORT |vp9_tm_predictor_8x8_neon|
- EXPORT |vp9_tm_predictor_16x16_neon|
- EXPORT |vp9_tm_predictor_32x32_neon|
+ EXPORT |vpx_v_predictor_4x4_neon|
+ EXPORT |vpx_v_predictor_8x8_neon|
+ EXPORT |vpx_v_predictor_16x16_neon|
+ EXPORT |vpx_v_predictor_32x32_neon|
+ EXPORT |vpx_h_predictor_4x4_neon|
+ EXPORT |vpx_h_predictor_8x8_neon|
+ EXPORT |vpx_h_predictor_16x16_neon|
+ EXPORT |vpx_h_predictor_32x32_neon|
+ EXPORT |vpx_tm_predictor_4x4_neon|
+ EXPORT |vpx_tm_predictor_8x8_neon|
+ EXPORT |vpx_tm_predictor_16x16_neon|
+ EXPORT |vpx_tm_predictor_32x32_neon|
ARM
REQUIRE8
PRESERVE8
AREA ||.text||, CODE, READONLY, ALIGN=2
-;void vp9_v_predictor_4x4_neon(uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_v_predictor_4x4_neon(uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -34,16 +34,16 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_v_predictor_4x4_neon| PROC
+|vpx_v_predictor_4x4_neon| PROC
vld1.32 {d0[0]}, [r2]
vst1.32 {d0[0]}, [r0], r1
vst1.32 {d0[0]}, [r0], r1
vst1.32 {d0[0]}, [r0], r1
vst1.32 {d0[0]}, [r0], r1
bx lr
- ENDP ; |vp9_v_predictor_4x4_neon|
+ ENDP ; |vpx_v_predictor_4x4_neon|
-;void vp9_v_predictor_8x8_neon(uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_v_predictor_8x8_neon(uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -51,7 +51,7 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_v_predictor_8x8_neon| PROC
+|vpx_v_predictor_8x8_neon| PROC
vld1.8 {d0}, [r2]
vst1.8 {d0}, [r0], r1
vst1.8 {d0}, [r0], r1
@@ -62,9 +62,9 @@
vst1.8 {d0}, [r0], r1
vst1.8 {d0}, [r0], r1
bx lr
- ENDP ; |vp9_v_predictor_8x8_neon|
+ ENDP ; |vpx_v_predictor_8x8_neon|
-;void vp9_v_predictor_16x16_neon(uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_v_predictor_16x16_neon(uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -72,7 +72,7 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_v_predictor_16x16_neon| PROC
+|vpx_v_predictor_16x16_neon| PROC
vld1.8 {q0}, [r2]
vst1.8 {q0}, [r0], r1
vst1.8 {q0}, [r0], r1
@@ -91,9 +91,9 @@
vst1.8 {q0}, [r0], r1
vst1.8 {q0}, [r0], r1
bx lr
- ENDP ; |vp9_v_predictor_16x16_neon|
+ ENDP ; |vpx_v_predictor_16x16_neon|
-;void vp9_v_predictor_32x32_neon(uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_v_predictor_32x32_neon(uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -101,7 +101,7 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_v_predictor_32x32_neon| PROC
+|vpx_v_predictor_32x32_neon| PROC
vld1.8 {q0, q1}, [r2]
mov r2, #2
loop_v
@@ -124,9 +124,9 @@
subs r2, r2, #1
bgt loop_v
bx lr
- ENDP ; |vp9_v_predictor_32x32_neon|
+ ENDP ; |vpx_v_predictor_32x32_neon|
-;void vp9_h_predictor_4x4_neon(uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_h_predictor_4x4_neon(uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -134,7 +134,7 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_h_predictor_4x4_neon| PROC
+|vpx_h_predictor_4x4_neon| PROC
vld1.32 {d1[0]}, [r3]
vdup.8 d0, d1[0]
vst1.32 {d0[0]}, [r0], r1
@@ -145,9 +145,9 @@
vdup.8 d0, d1[3]
vst1.32 {d0[0]}, [r0], r1
bx lr
- ENDP ; |vp9_h_predictor_4x4_neon|
+ ENDP ; |vpx_h_predictor_4x4_neon|
-;void vp9_h_predictor_8x8_neon(uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_h_predictor_8x8_neon(uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -155,7 +155,7 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_h_predictor_8x8_neon| PROC
+|vpx_h_predictor_8x8_neon| PROC
vld1.64 {d1}, [r3]
vdup.8 d0, d1[0]
vst1.64 {d0}, [r0], r1
@@ -174,9 +174,9 @@
vdup.8 d0, d1[7]
vst1.64 {d0}, [r0], r1
bx lr
- ENDP ; |vp9_h_predictor_8x8_neon|
+ ENDP ; |vpx_h_predictor_8x8_neon|
-;void vp9_h_predictor_16x16_neon(uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_h_predictor_16x16_neon(uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -184,7 +184,7 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_h_predictor_16x16_neon| PROC
+|vpx_h_predictor_16x16_neon| PROC
vld1.8 {q1}, [r3]
vdup.8 q0, d2[0]
vst1.8 {q0}, [r0], r1
@@ -219,9 +219,9 @@
vdup.8 q0, d3[7]
vst1.8 {q0}, [r0], r1
bx lr
- ENDP ; |vp9_h_predictor_16x16_neon|
+ ENDP ; |vpx_h_predictor_16x16_neon|
-;void vp9_h_predictor_32x32_neon(uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_h_predictor_32x32_neon(uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -229,7 +229,7 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_h_predictor_32x32_neon| PROC
+|vpx_h_predictor_32x32_neon| PROC
sub r1, r1, #16
mov r2, #2
loop_h
@@ -285,9 +285,9 @@
subs r2, r2, #1
bgt loop_h
bx lr
- ENDP ; |vp9_h_predictor_32x32_neon|
+ ENDP ; |vpx_h_predictor_32x32_neon|
-;void vp9_tm_predictor_4x4_neon (uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_tm_predictor_4x4_neon (uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -295,7 +295,7 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_tm_predictor_4x4_neon| PROC
+|vpx_tm_predictor_4x4_neon| PROC
; Load ytop_left = above[-1];
sub r12, r2, #1
vld1.u8 {d0[]}, [r12]
@@ -331,9 +331,9 @@
vst1.32 {d0[0]}, [r0], r1
vst1.32 {d1[0]}, [r0], r1
bx lr
- ENDP ; |vp9_tm_predictor_4x4_neon|
+ ENDP ; |vpx_tm_predictor_4x4_neon|
-;void vp9_tm_predictor_8x8_neon (uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_tm_predictor_8x8_neon (uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -341,7 +341,7 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_tm_predictor_8x8_neon| PROC
+|vpx_tm_predictor_8x8_neon| PROC
; Load ytop_left = above[-1];
sub r12, r2, #1
vld1.8 {d0[]}, [r12]
@@ -403,9 +403,9 @@
vst1.64 {d3}, [r0], r1
bx lr
- ENDP ; |vp9_tm_predictor_8x8_neon|
+ ENDP ; |vpx_tm_predictor_8x8_neon|
-;void vp9_tm_predictor_16x16_neon (uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_tm_predictor_16x16_neon (uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -413,7 +413,7 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_tm_predictor_16x16_neon| PROC
+|vpx_tm_predictor_16x16_neon| PROC
; Load ytop_left = above[-1];
sub r12, r2, #1
vld1.8 {d0[]}, [r12]
@@ -496,9 +496,9 @@
bgt loop_16x16_neon
bx lr
- ENDP ; |vp9_tm_predictor_16x16_neon|
+ ENDP ; |vpx_tm_predictor_16x16_neon|
-;void vp9_tm_predictor_32x32_neon (uint8_t *dst, ptrdiff_t y_stride,
+;void vpx_tm_predictor_32x32_neon (uint8_t *dst, ptrdiff_t y_stride,
; const uint8_t *above,
; const uint8_t *left)
; r0 uint8_t *dst
@@ -506,7 +506,7 @@
; r2 const uint8_t *above
; r3 const uint8_t *left
-|vp9_tm_predictor_32x32_neon| PROC
+|vpx_tm_predictor_32x32_neon| PROC
; Load ytop_left = above[-1];
sub r12, r2, #1
vld1.8 {d0[]}, [r12]
@@ -625,6 +625,6 @@
bgt loop_32x32_neon
bx lr
- ENDP ; |vp9_tm_predictor_32x32_neon|
+ ENDP ; |vpx_tm_predictor_32x32_neon|
END
diff --git a/vpx_dsp/fwd_txfm.c b/vpx_dsp/fwd_txfm.c
index bdb55ee..c82e1c1 100644
--- a/vpx_dsp/fwd_txfm.c
+++ b/vpx_dsp/fwd_txfm.c
@@ -10,7 +10,7 @@
#include "vpx_dsp/fwd_txfm.h"
-void vp9_fdct4x4_c(const int16_t *input, tran_low_t *output, int stride) {
+void vpx_fdct4x4_c(const int16_t *input, tran_low_t *output, int stride) {
// The 2D transform is done with two passes which are actually pretty
// similar. In the first one, we transform the columns and transpose
// the results. In the second one, we transform the rows. To achieve that,
@@ -77,7 +77,18 @@
}
}
-void vp9_fdct8x8_c(const int16_t *input, tran_low_t *final_output, int stride) {
+void vpx_fdct4x4_1_c(const int16_t *input, tran_low_t *output, int stride) {
+ int r, c;
+ tran_low_t sum = 0;
+ for (r = 0; r < 4; ++r)
+ for (c = 0; c < 4; ++c)
+ sum += input[r * stride + c];
+
+ output[0] = sum << 1;
+ output[1] = 0;
+}
+
+void vpx_fdct8x8_c(const int16_t *input, tran_low_t *final_output, int stride) {
int i, j;
tran_low_t intermediate[64];
int pass;
@@ -163,7 +174,18 @@
}
}
-void vp9_fdct16x16_c(const int16_t *input, tran_low_t *output, int stride) {
+void vpx_fdct8x8_1_c(const int16_t *input, tran_low_t *output, int stride) {
+ int r, c;
+ tran_low_t sum = 0;
+ for (r = 0; r < 8; ++r)
+ for (c = 0; c < 8; ++c)
+ sum += input[r * stride + c];
+
+ output[0] = sum;
+ output[1] = 0;
+}
+
+void vpx_fdct16x16_c(const int16_t *input, tran_low_t *output, int stride) {
// The 2D transform is done with two passes which are actually pretty
// similar. In the first one, we transform the columns and transpose
// the results. In the second one, we transform the rows. To achieve that,
@@ -343,19 +365,458 @@
}
}
+void vpx_fdct16x16_1_c(const int16_t *input, tran_low_t *output, int stride) {
+ int r, c;
+ tran_low_t sum = 0;
+ for (r = 0; r < 16; ++r)
+ for (c = 0; c < 16; ++c)
+ sum += input[r * stride + c];
+
+ output[0] = sum >> 1;
+ output[1] = 0;
+}
+
+static INLINE tran_high_t dct_32_round(tran_high_t input) {
+ tran_high_t rv = ROUND_POWER_OF_TWO(input, DCT_CONST_BITS);
+ // TODO(debargha, peter.derivaz): Find new bounds for this assert,
+ // and make the bounds consts.
+ // assert(-131072 <= rv && rv <= 131071);
+ return rv;
+}
+
+static INLINE tran_high_t half_round_shift(tran_high_t input) {
+ tran_high_t rv = (input + 1 + (input < 0)) >> 2;
+ return rv;
+}
+
+void vpx_fdct32(const tran_high_t *input, tran_high_t *output, int round) {
+ tran_high_t step[32];
+ // Stage 1
+ step[0] = input[0] + input[(32 - 1)];
+ step[1] = input[1] + input[(32 - 2)];
+ step[2] = input[2] + input[(32 - 3)];
+ step[3] = input[3] + input[(32 - 4)];
+ step[4] = input[4] + input[(32 - 5)];
+ step[5] = input[5] + input[(32 - 6)];
+ step[6] = input[6] + input[(32 - 7)];
+ step[7] = input[7] + input[(32 - 8)];
+ step[8] = input[8] + input[(32 - 9)];
+ step[9] = input[9] + input[(32 - 10)];
+ step[10] = input[10] + input[(32 - 11)];
+ step[11] = input[11] + input[(32 - 12)];
+ step[12] = input[12] + input[(32 - 13)];
+ step[13] = input[13] + input[(32 - 14)];
+ step[14] = input[14] + input[(32 - 15)];
+ step[15] = input[15] + input[(32 - 16)];
+ step[16] = -input[16] + input[(32 - 17)];
+ step[17] = -input[17] + input[(32 - 18)];
+ step[18] = -input[18] + input[(32 - 19)];
+ step[19] = -input[19] + input[(32 - 20)];
+ step[20] = -input[20] + input[(32 - 21)];
+ step[21] = -input[21] + input[(32 - 22)];
+ step[22] = -input[22] + input[(32 - 23)];
+ step[23] = -input[23] + input[(32 - 24)];
+ step[24] = -input[24] + input[(32 - 25)];
+ step[25] = -input[25] + input[(32 - 26)];
+ step[26] = -input[26] + input[(32 - 27)];
+ step[27] = -input[27] + input[(32 - 28)];
+ step[28] = -input[28] + input[(32 - 29)];
+ step[29] = -input[29] + input[(32 - 30)];
+ step[30] = -input[30] + input[(32 - 31)];
+ step[31] = -input[31] + input[(32 - 32)];
+
+ // Stage 2
+ output[0] = step[0] + step[16 - 1];
+ output[1] = step[1] + step[16 - 2];
+ output[2] = step[2] + step[16 - 3];
+ output[3] = step[3] + step[16 - 4];
+ output[4] = step[4] + step[16 - 5];
+ output[5] = step[5] + step[16 - 6];
+ output[6] = step[6] + step[16 - 7];
+ output[7] = step[7] + step[16 - 8];
+ output[8] = -step[8] + step[16 - 9];
+ output[9] = -step[9] + step[16 - 10];
+ output[10] = -step[10] + step[16 - 11];
+ output[11] = -step[11] + step[16 - 12];
+ output[12] = -step[12] + step[16 - 13];
+ output[13] = -step[13] + step[16 - 14];
+ output[14] = -step[14] + step[16 - 15];
+ output[15] = -step[15] + step[16 - 16];
+
+ output[16] = step[16];
+ output[17] = step[17];
+ output[18] = step[18];
+ output[19] = step[19];
+
+ output[20] = dct_32_round((-step[20] + step[27]) * cospi_16_64);
+ output[21] = dct_32_round((-step[21] + step[26]) * cospi_16_64);
+ output[22] = dct_32_round((-step[22] + step[25]) * cospi_16_64);
+ output[23] = dct_32_round((-step[23] + step[24]) * cospi_16_64);
+
+ output[24] = dct_32_round((step[24] + step[23]) * cospi_16_64);
+ output[25] = dct_32_round((step[25] + step[22]) * cospi_16_64);
+ output[26] = dct_32_round((step[26] + step[21]) * cospi_16_64);
+ output[27] = dct_32_round((step[27] + step[20]) * cospi_16_64);
+
+ output[28] = step[28];
+ output[29] = step[29];
+ output[30] = step[30];
+ output[31] = step[31];
+
+ // dump the magnitude by 4, hence the intermediate values are within
+ // the range of 16 bits.
+ if (round) {
+ output[0] = half_round_shift(output[0]);
+ output[1] = half_round_shift(output[1]);
+ output[2] = half_round_shift(output[2]);
+ output[3] = half_round_shift(output[3]);
+ output[4] = half_round_shift(output[4]);
+ output[5] = half_round_shift(output[5]);
+ output[6] = half_round_shift(output[6]);
+ output[7] = half_round_shift(output[7]);
+ output[8] = half_round_shift(output[8]);
+ output[9] = half_round_shift(output[9]);
+ output[10] = half_round_shift(output[10]);
+ output[11] = half_round_shift(output[11]);
+ output[12] = half_round_shift(output[12]);
+ output[13] = half_round_shift(output[13]);
+ output[14] = half_round_shift(output[14]);
+ output[15] = half_round_shift(output[15]);
+
+ output[16] = half_round_shift(output[16]);
+ output[17] = half_round_shift(output[17]);
+ output[18] = half_round_shift(output[18]);
+ output[19] = half_round_shift(output[19]);
+ output[20] = half_round_shift(output[20]);
+ output[21] = half_round_shift(output[21]);
+ output[22] = half_round_shift(output[22]);
+ output[23] = half_round_shift(output[23]);
+ output[24] = half_round_shift(output[24]);
+ output[25] = half_round_shift(output[25]);
+ output[26] = half_round_shift(output[26]);
+ output[27] = half_round_shift(output[27]);
+ output[28] = half_round_shift(output[28]);
+ output[29] = half_round_shift(output[29]);
+ output[30] = half_round_shift(output[30]);
+ output[31] = half_round_shift(output[31]);
+ }
+
+ // Stage 3
+ step[0] = output[0] + output[(8 - 1)];
+ step[1] = output[1] + output[(8 - 2)];
+ step[2] = output[2] + output[(8 - 3)];
+ step[3] = output[3] + output[(8 - 4)];
+ step[4] = -output[4] + output[(8 - 5)];
+ step[5] = -output[5] + output[(8 - 6)];
+ step[6] = -output[6] + output[(8 - 7)];
+ step[7] = -output[7] + output[(8 - 8)];
+ step[8] = output[8];
+ step[9] = output[9];
+ step[10] = dct_32_round((-output[10] + output[13]) * cospi_16_64);
+ step[11] = dct_32_round((-output[11] + output[12]) * cospi_16_64);
+ step[12] = dct_32_round((output[12] + output[11]) * cospi_16_64);
+ step[13] = dct_32_round((output[13] + output[10]) * cospi_16_64);
+ step[14] = output[14];
+ step[15] = output[15];
+
+ step[16] = output[16] + output[23];
+ step[17] = output[17] + output[22];
+ step[18] = output[18] + output[21];
+ step[19] = output[19] + output[20];
+ step[20] = -output[20] + output[19];
+ step[21] = -output[21] + output[18];
+ step[22] = -output[22] + output[17];
+ step[23] = -output[23] + output[16];
+ step[24] = -output[24] + output[31];
+ step[25] = -output[25] + output[30];
+ step[26] = -output[26] + output[29];
+ step[27] = -output[27] + output[28];
+ step[28] = output[28] + output[27];
+ step[29] = output[29] + output[26];
+ step[30] = output[30] + output[25];
+ step[31] = output[31] + output[24];
+
+ // Stage 4
+ output[0] = step[0] + step[3];
+ output[1] = step[1] + step[2];
+ output[2] = -step[2] + step[1];
+ output[3] = -step[3] + step[0];
+ output[4] = step[4];
+ output[5] = dct_32_round((-step[5] + step[6]) * cospi_16_64);
+ output[6] = dct_32_round((step[6] + step[5]) * cospi_16_64);
+ output[7] = step[7];
+ output[8] = step[8] + step[11];
+ output[9] = step[9] + step[10];
+ output[10] = -step[10] + step[9];
+ output[11] = -step[11] + step[8];
+ output[12] = -step[12] + step[15];
+ output[13] = -step[13] + step[14];
+ output[14] = step[14] + step[13];
+ output[15] = step[15] + step[12];
+
+ output[16] = step[16];
+ output[17] = step[17];
+ output[18] = dct_32_round(step[18] * -cospi_8_64 + step[29] * cospi_24_64);
+ output[19] = dct_32_round(step[19] * -cospi_8_64 + step[28] * cospi_24_64);
+ output[20] = dct_32_round(step[20] * -cospi_24_64 + step[27] * -cospi_8_64);
+ output[21] = dct_32_round(step[21] * -cospi_24_64 + step[26] * -cospi_8_64);
+ output[22] = step[22];
+ output[23] = step[23];
+ output[24] = step[24];
+ output[25] = step[25];
+ output[26] = dct_32_round(step[26] * cospi_24_64 + step[21] * -cospi_8_64);
+ output[27] = dct_32_round(step[27] * cospi_24_64 + step[20] * -cospi_8_64);
+ output[28] = dct_32_round(step[28] * cospi_8_64 + step[19] * cospi_24_64);
+ output[29] = dct_32_round(step[29] * cospi_8_64 + step[18] * cospi_24_64);
+ output[30] = step[30];
+ output[31] = step[31];
+
+ // Stage 5
+ step[0] = dct_32_round((output[0] + output[1]) * cospi_16_64);
+ step[1] = dct_32_round((-output[1] + output[0]) * cospi_16_64);
+ step[2] = dct_32_round(output[2] * cospi_24_64 + output[3] * cospi_8_64);
+ step[3] = dct_32_round(output[3] * cospi_24_64 - output[2] * cospi_8_64);
+ step[4] = output[4] + output[5];
+ step[5] = -output[5] + output[4];
+ step[6] = -output[6] + output[7];
+ step[7] = output[7] + output[6];
+ step[8] = output[8];
+ step[9] = dct_32_round(output[9] * -cospi_8_64 + output[14] * cospi_24_64);
+ step[10] = dct_32_round(output[10] * -cospi_24_64 + output[13] * -cospi_8_64);
+ step[11] = output[11];
+ step[12] = output[12];
+ step[13] = dct_32_round(output[13] * cospi_24_64 + output[10] * -cospi_8_64);
+ step[14] = dct_32_round(output[14] * cospi_8_64 + output[9] * cospi_24_64);
+ step[15] = output[15];
+
+ step[16] = output[16] + output[19];
+ step[17] = output[17] + output[18];
+ step[18] = -output[18] + output[17];
+ step[19] = -output[19] + output[16];
+ step[20] = -output[20] + output[23];
+ step[21] = -output[21] + output[22];
+ step[22] = output[22] + output[21];
+ step[23] = output[23] + output[20];
+ step[24] = output[24] + output[27];
+ step[25] = output[25] + output[26];
+ step[26] = -output[26] + output[25];
+ step[27] = -output[27] + output[24];
+ step[28] = -output[28] + output[31];
+ step[29] = -output[29] + output[30];
+ step[30] = output[30] + output[29];
+ step[31] = output[31] + output[28];
+
+ // Stage 6
+ output[0] = step[0];
+ output[1] = step[1];
+ output[2] = step[2];
+ output[3] = step[3];
+ output[4] = dct_32_round(step[4] * cospi_28_64 + step[7] * cospi_4_64);
+ output[5] = dct_32_round(step[5] * cospi_12_64 + step[6] * cospi_20_64);
+ output[6] = dct_32_round(step[6] * cospi_12_64 + step[5] * -cospi_20_64);
+ output[7] = dct_32_round(step[7] * cospi_28_64 + step[4] * -cospi_4_64);
+ output[8] = step[8] + step[9];
+ output[9] = -step[9] + step[8];
+ output[10] = -step[10] + step[11];
+ output[11] = step[11] + step[10];
+ output[12] = step[12] + step[13];
+ output[13] = -step[13] + step[12];
+ output[14] = -step[14] + step[15];
+ output[15] = step[15] + step[14];
+
+ output[16] = step[16];
+ output[17] = dct_32_round(step[17] * -cospi_4_64 + step[30] * cospi_28_64);
+ output[18] = dct_32_round(step[18] * -cospi_28_64 + step[29] * -cospi_4_64);
+ output[19] = step[19];
+ output[20] = step[20];
+ output[21] = dct_32_round(step[21] * -cospi_20_64 + step[26] * cospi_12_64);
+ output[22] = dct_32_round(step[22] * -cospi_12_64 + step[25] * -cospi_20_64);
+ output[23] = step[23];
+ output[24] = step[24];
+ output[25] = dct_32_round(step[25] * cospi_12_64 + step[22] * -cospi_20_64);
+ output[26] = dct_32_round(step[26] * cospi_20_64 + step[21] * cospi_12_64);
+ output[27] = step[27];
+ output[28] = step[28];
+ output[29] = dct_32_round(step[29] * cospi_28_64 + step[18] * -cospi_4_64);
+ output[30] = dct_32_round(step[30] * cospi_4_64 + step[17] * cospi_28_64);
+ output[31] = step[31];
+
+ // Stage 7
+ step[0] = output[0];
+ step[1] = output[1];
+ step[2] = output[2];
+ step[3] = output[3];
+ step[4] = output[4];
+ step[5] = output[5];
+ step[6] = output[6];
+ step[7] = output[7];
+ step[8] = dct_32_round(output[8] * cospi_30_64 + output[15] * cospi_2_64);
+ step[9] = dct_32_round(output[9] * cospi_14_64 + output[14] * cospi_18_64);
+ step[10] = dct_32_round(output[10] * cospi_22_64 + output[13] * cospi_10_64);
+ step[11] = dct_32_round(output[11] * cospi_6_64 + output[12] * cospi_26_64);
+ step[12] = dct_32_round(output[12] * cospi_6_64 + output[11] * -cospi_26_64);
+ step[13] = dct_32_round(output[13] * cospi_22_64 + output[10] * -cospi_10_64);
+ step[14] = dct_32_round(output[14] * cospi_14_64 + output[9] * -cospi_18_64);
+ step[15] = dct_32_round(output[15] * cospi_30_64 + output[8] * -cospi_2_64);
+
+ step[16] = output[16] + output[17];
+ step[17] = -output[17] + output[16];
+ step[18] = -output[18] + output[19];
+ step[19] = output[19] + output[18];
+ step[20] = output[20] + output[21];
+ step[21] = -output[21] + output[20];
+ step[22] = -output[22] + output[23];
+ step[23] = output[23] + output[22];
+ step[24] = output[24] + output[25];
+ step[25] = -output[25] + output[24];
+ step[26] = -output[26] + output[27];
+ step[27] = output[27] + output[26];
+ step[28] = output[28] + output[29];
+ step[29] = -output[29] + output[28];
+ step[30] = -output[30] + output[31];
+ step[31] = output[31] + output[30];
+
+ // Final stage --- outputs indices are bit-reversed.
+ output[0] = step[0];
+ output[16] = step[1];
+ output[8] = step[2];
+ output[24] = step[3];
+ output[4] = step[4];
+ output[20] = step[5];
+ output[12] = step[6];
+ output[28] = step[7];
+ output[2] = step[8];
+ output[18] = step[9];
+ output[10] = step[10];
+ output[26] = step[11];
+ output[6] = step[12];
+ output[22] = step[13];
+ output[14] = step[14];
+ output[30] = step[15];
+
+ output[1] = dct_32_round(step[16] * cospi_31_64 + step[31] * cospi_1_64);
+ output[17] = dct_32_round(step[17] * cospi_15_64 + step[30] * cospi_17_64);
+ output[9] = dct_32_round(step[18] * cospi_23_64 + step[29] * cospi_9_64);
+ output[25] = dct_32_round(step[19] * cospi_7_64 + step[28] * cospi_25_64);
+ output[5] = dct_32_round(step[20] * cospi_27_64 + step[27] * cospi_5_64);
+ output[21] = dct_32_round(step[21] * cospi_11_64 + step[26] * cospi_21_64);
+ output[13] = dct_32_round(step[22] * cospi_19_64 + step[25] * cospi_13_64);
+ output[29] = dct_32_round(step[23] * cospi_3_64 + step[24] * cospi_29_64);
+ output[3] = dct_32_round(step[24] * cospi_3_64 + step[23] * -cospi_29_64);
+ output[19] = dct_32_round(step[25] * cospi_19_64 + step[22] * -cospi_13_64);
+ output[11] = dct_32_round(step[26] * cospi_11_64 + step[21] * -cospi_21_64);
+ output[27] = dct_32_round(step[27] * cospi_27_64 + step[20] * -cospi_5_64);
+ output[7] = dct_32_round(step[28] * cospi_7_64 + step[19] * -cospi_25_64);
+ output[23] = dct_32_round(step[29] * cospi_23_64 + step[18] * -cospi_9_64);
+ output[15] = dct_32_round(step[30] * cospi_15_64 + step[17] * -cospi_17_64);
+ output[31] = dct_32_round(step[31] * cospi_31_64 + step[16] * -cospi_1_64);
+}
+
+void vpx_fdct32x32_c(const int16_t *input, tran_low_t *out, int stride) {
+ int i, j;
+ tran_high_t output[32 * 32];
+
+ // Columns
+ for (i = 0; i < 32; ++i) {
+ tran_high_t temp_in[32], temp_out[32];
+ for (j = 0; j < 32; ++j)
+ temp_in[j] = input[j * stride + i] * 4;
+ vpx_fdct32(temp_in, temp_out, 0);
+ for (j = 0; j < 32; ++j)
+ output[j * 32 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2;
+ }
+
+ // Rows
+ for (i = 0; i < 32; ++i) {
+ tran_high_t temp_in[32], temp_out[32];
+ for (j = 0; j < 32; ++j)
+ temp_in[j] = output[j + i * 32];
+ vpx_fdct32(temp_in, temp_out, 0);
+ for (j = 0; j < 32; ++j)
+ out[j + i * 32] =
+ (tran_low_t)((temp_out[j] + 1 + (temp_out[j] < 0)) >> 2);
+ }
+}
+
+// Note that although we use dct_32_round in dct32 computation flow,
+// this 2d fdct32x32 for rate-distortion optimization loop is operating
+// within 16 bits precision.
+void vpx_fdct32x32_rd_c(const int16_t *input, tran_low_t *out, int stride) {
+ int i, j;
+ tran_high_t output[32 * 32];
+
+ // Columns
+ for (i = 0; i < 32; ++i) {
+ tran_high_t temp_in[32], temp_out[32];
+ for (j = 0; j < 32; ++j)
+ temp_in[j] = input[j * stride + i] * 4;
+ vpx_fdct32(temp_in, temp_out, 0);
+ for (j = 0; j < 32; ++j)
+ // TODO(cd): see quality impact of only doing
+ // output[j * 32 + i] = (temp_out[j] + 1) >> 2;
+ // PS: also change code in vp9/encoder/x86/vp9_dct_sse2.c
+ output[j * 32 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2;
+ }
+
+ // Rows
+ for (i = 0; i < 32; ++i) {
+ tran_high_t temp_in[32], temp_out[32];
+ for (j = 0; j < 32; ++j)
+ temp_in[j] = output[j + i * 32];
+ vpx_fdct32(temp_in, temp_out, 1);
+ for (j = 0; j < 32; ++j)
+ out[j + i * 32] = (tran_low_t)temp_out[j];
+ }
+}
+
+void vpx_fdct32x32_1_c(const int16_t *input, tran_low_t *output, int stride) {
+ int r, c;
+ tran_low_t sum = 0;
+ for (r = 0; r < 32; ++r)
+ for (c = 0; c < 32; ++c)
+ sum += input[r * stride + c];
+
+ output[0] = sum >> 3;
+ output[1] = 0;
+}
+
#if CONFIG_VP9_HIGHBITDEPTH
-void vp9_highbd_fdct4x4_c(const int16_t *input, tran_low_t *output,
+void vpx_highbd_fdct4x4_c(const int16_t *input, tran_low_t *output,
int stride) {
- vp9_fdct4x4_c(input, output, stride);
+ vpx_fdct4x4_c(input, output, stride);
}
-void vp9_highbd_fdct8x8_c(const int16_t *input, tran_low_t *final_output,
+void vpx_highbd_fdct8x8_c(const int16_t *input, tran_low_t *final_output,
int stride) {
- vp9_fdct8x8_c(input, final_output, stride);
+ vpx_fdct8x8_c(input, final_output, stride);
}
-void vp9_highbd_fdct16x16_c(const int16_t *input, tran_low_t *output,
+void vpx_highbd_fdct8x8_1_c(const int16_t *input, tran_low_t *final_output,
int stride) {
- vp9_fdct16x16_c(input, output, stride);
+ vpx_fdct8x8_1_c(input, final_output, stride);
+}
+
+void vpx_highbd_fdct16x16_c(const int16_t *input, tran_low_t *output,
+ int stride) {
+ vpx_fdct16x16_c(input, output, stride);
+}
+
+void vpx_highbd_fdct16x16_1_c(const int16_t *input, tran_low_t *output,
+ int stride) {
+ vpx_fdct16x16_1_c(input, output, stride);
+}
+
+void vpx_highbd_fdct32x32_c(const int16_t *input, tran_low_t *out, int stride) {
+ vpx_fdct32x32_c(input, out, stride);
+}
+
+void vpx_highbd_fdct32x32_rd_c(const int16_t *input, tran_low_t *out,
+ int stride) {
+ vpx_fdct32x32_rd_c(input, out, stride);
+}
+
+void vpx_highbd_fdct32x32_1_c(const int16_t *input, tran_low_t *out,
+ int stride) {
+ vpx_fdct32x32_1_c(input, out, stride);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vpx_dsp/fwd_txfm.h b/vpx_dsp/fwd_txfm.h
index 509fe7f..29e139c 100644
--- a/vpx_dsp/fwd_txfm.h
+++ b/vpx_dsp/fwd_txfm.h
@@ -8,6 +8,9 @@
* be found in the AUTHORS file in the root of the source tree.
*/
+#ifndef VPX_DSP_FWD_TXFM_H_
+#define VPX_DSP_FWD_TXFM_H_
+
#include "vpx_dsp/txfm_common.h"
static INLINE tran_high_t fdct_round_shift(tran_high_t input) {
@@ -17,3 +20,6 @@
// assert(INT16_MIN <= rv && rv <= INT16_MAX);
return rv;
}
+
+void vpx_fdct32(const tran_high_t *input, tran_high_t *output, int round);
+#endif // VPX_DSP_FWD_TXFM_H_
diff --git a/vpx_dsp/intrapred.c b/vpx_dsp/intrapred.c
new file mode 100644
index 0000000..9ba0f64
--- /dev/null
+++ b/vpx_dsp/intrapred.c
@@ -0,0 +1,692 @@
+/*
+ * Copyright (c) 2015 The WebM project authors. All Rights Reserved.
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "./vpx_config.h"
+#include "./vpx_dsp_rtcd.h"
+
+#include "vpx_dsp/vpx_dsp_common.h"
+#include "vpx_mem/vpx_mem.h"
+
+#define DST(x, y) dst[(x) + (y) * stride]
+#define AVG3(a, b, c) (((a) + 2 * (b) + (c) + 2) >> 2)
+#define AVG2(a, b) (((a) + (b) + 1) >> 1)
+
+static INLINE void d207_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ int r, c;
+ (void) above;
+ // first column
+ for (r = 0; r < bs - 1; ++r)
+ dst[r * stride] = AVG2(left[r], left[r + 1]);
+ dst[(bs - 1) * stride] = left[bs - 1];
+ dst++;
+
+ // second column
+ for (r = 0; r < bs - 2; ++r)
+ dst[r * stride] = AVG3(left[r], left[r + 1], left[r + 2]);
+ dst[(bs - 2) * stride] = AVG3(left[bs - 2], left[bs - 1], left[bs - 1]);
+ dst[(bs - 1) * stride] = left[bs - 1];
+ dst++;
+
+ // rest of last row
+ for (c = 0; c < bs - 2; ++c)
+ dst[(bs - 1) * stride + c] = left[bs - 1];
+
+ for (r = bs - 2; r >= 0; --r)
+ for (c = 0; c < bs - 2; ++c)
+ dst[r * stride + c] = dst[(r + 1) * stride + c - 2];
+}
+
+static INLINE void d63_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ int r, c;
+ int size;
+ (void)left;
+ for (c = 0; c < bs; ++c) {
+ dst[c] = AVG2(above[c], above[c + 1]);
+ dst[stride + c] = AVG3(above[c], above[c + 1], above[c + 2]);
+ }
+ for (r = 2, size = bs - 2; r < bs; r += 2, --size) {
+ memcpy(dst + (r + 0) * stride, dst + (r >> 1), size);
+ memset(dst + (r + 0) * stride + size, above[bs - 1], bs - size);
+ memcpy(dst + (r + 1) * stride, dst + stride + (r >> 1), size);
+ memset(dst + (r + 1) * stride + size, above[bs - 1], bs - size);
+ }
+}
+
+static INLINE void d45_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ const uint8_t above_right = above[bs - 1];
+ const uint8_t *const dst_row0 = dst;
+ int x, size;
+ (void)left;
+
+ for (x = 0; x < bs - 1; ++x) {
+ dst[x] = AVG3(above[x], above[x + 1], above[x + 2]);
+ }
+ dst[bs - 1] = above_right;
+ dst += stride;
+ for (x = 1, size = bs - 2; x < bs; ++x, --size) {
+ memcpy(dst, dst_row0 + x, size);
+ memset(dst + size, above_right, x + 1);
+ dst += stride;
+ }
+}
+
+static INLINE void d117_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ int r, c;
+
+ // first row
+ for (c = 0; c < bs; c++)
+ dst[c] = AVG2(above[c - 1], above[c]);
+ dst += stride;
+
+ // second row
+ dst[0] = AVG3(left[0], above[-1], above[0]);
+ for (c = 1; c < bs; c++)
+ dst[c] = AVG3(above[c - 2], above[c - 1], above[c]);
+ dst += stride;
+
+ // the rest of first col
+ dst[0] = AVG3(above[-1], left[0], left[1]);
+ for (r = 3; r < bs; ++r)
+ dst[(r - 2) * stride] = AVG3(left[r - 3], left[r - 2], left[r - 1]);
+
+ // the rest of the block
+ for (r = 2; r < bs; ++r) {
+ for (c = 1; c < bs; c++)
+ dst[c] = dst[-2 * stride + c - 1];
+ dst += stride;
+ }
+}
+
+static INLINE void d135_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ int r, c;
+ dst[0] = AVG3(left[0], above[-1], above[0]);
+ for (c = 1; c < bs; c++)
+ dst[c] = AVG3(above[c - 2], above[c - 1], above[c]);
+
+ dst[stride] = AVG3(above[-1], left[0], left[1]);
+ for (r = 2; r < bs; ++r)
+ dst[r * stride] = AVG3(left[r - 2], left[r - 1], left[r]);
+
+ dst += stride;
+ for (r = 1; r < bs; ++r) {
+ for (c = 1; c < bs; c++)
+ dst[c] = dst[-stride + c - 1];
+ dst += stride;
+ }
+}
+
+static INLINE void d153_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ int r, c;
+ dst[0] = AVG2(above[-1], left[0]);
+ for (r = 1; r < bs; r++)
+ dst[r * stride] = AVG2(left[r - 1], left[r]);
+ dst++;
+
+ dst[0] = AVG3(left[0], above[-1], above[0]);
+ dst[stride] = AVG3(above[-1], left[0], left[1]);
+ for (r = 2; r < bs; r++)
+ dst[r * stride] = AVG3(left[r - 2], left[r - 1], left[r]);
+ dst++;
+
+ for (c = 0; c < bs - 2; c++)
+ dst[c] = AVG3(above[c - 1], above[c], above[c + 1]);
+ dst += stride;
+
+ for (r = 1; r < bs; ++r) {
+ for (c = 0; c < bs - 2; c++)
+ dst[c] = dst[-stride + c - 2];
+ dst += stride;
+ }
+}
+
+static INLINE void v_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ int r;
+ (void) left;
+
+ for (r = 0; r < bs; r++) {
+ memcpy(dst, above, bs);
+ dst += stride;
+ }
+}
+
+static INLINE void h_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ int r;
+ (void) above;
+
+ for (r = 0; r < bs; r++) {
+ memset(dst, left[r], bs);
+ dst += stride;
+ }
+}
+
+static INLINE void tm_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ int r, c;
+ int ytop_left = above[-1];
+
+ for (r = 0; r < bs; r++) {
+ for (c = 0; c < bs; c++)
+ dst[c] = clip_pixel(left[r] + above[c] - ytop_left);
+ dst += stride;
+ }
+}
+
+static INLINE void dc_128_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ int r;
+ (void) above;
+ (void) left;
+
+ for (r = 0; r < bs; r++) {
+ memset(dst, 128, bs);
+ dst += stride;
+ }
+}
+
+static INLINE void dc_left_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above,
+ const uint8_t *left) {
+ int i, r, expected_dc, sum = 0;
+ (void) above;
+
+ for (i = 0; i < bs; i++)
+ sum += left[i];
+ expected_dc = (sum + (bs >> 1)) / bs;
+
+ for (r = 0; r < bs; r++) {
+ memset(dst, expected_dc, bs);
+ dst += stride;
+ }
+}
+
+static INLINE void dc_top_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ int i, r, expected_dc, sum = 0;
+ (void) left;
+
+ for (i = 0; i < bs; i++)
+ sum += above[i];
+ expected_dc = (sum + (bs >> 1)) / bs;
+
+ for (r = 0; r < bs; r++) {
+ memset(dst, expected_dc, bs);
+ dst += stride;
+ }
+}
+
+static INLINE void dc_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
+ const uint8_t *above, const uint8_t *left) {
+ int i, r, expected_dc, sum = 0;
+ const int count = 2 * bs;
+
+ for (i = 0; i < bs; i++) {
+ sum += above[i];
+ sum += left[i];
+ }
+
+ expected_dc = (sum + (count >> 1)) / count;
+
+ for (r = 0; r < bs; r++) {
+ memset(dst, expected_dc, bs);
+ dst += stride;
+ }
+}
+
+void vpx_d207_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {
+ const int I = left[0];
+ const int J = left[1];
+ const int K = left[2];
+ const int L = left[3];
+ (void)above;
+ DST(0, 0) = AVG2(I, J);
+ DST(2, 0) = DST(0, 1) = AVG2(J, K);
+ DST(2, 1) = DST(0, 2) = AVG2(K, L);
+ DST(1, 0) = AVG3(I, J, K);
+ DST(3, 0) = DST(1, 1) = AVG3(J, K, L);
+ DST(3, 1) = DST(1, 2) = AVG3(K, L, L);
+ DST(3, 2) = DST(2, 2) =
+ DST(0, 3) = DST(1, 3) = DST(2, 3) = DST(3, 3) = L;
+}
+
+void vpx_d63_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {
+ const int A = above[0];
+ const int B = above[1];
+ const int C = above[2];
+ const int D = above[3];
+ const int E = above[4];
+ const int F = above[5];
+ const int G = above[6];
+ (void)left;
+ DST(0, 0) = AVG2(A, B);
+ DST(1, 0) = DST(0, 2) = AVG2(B, C);
+ DST(2, 0) = DST(1, 2) = AVG2(C, D);
+ DST(3, 0) = DST(2, 2) = AVG2(D, E);
+ DST(3, 2) = AVG2(E, F); // differs from vp8
+
+ DST(0, 1) = AVG3(A, B, C);
+ DST(1, 1) = DST(0, 3) = AVG3(B, C, D);
+ DST(2, 1) = DST(1, 3) = AVG3(C, D, E);
+ DST(3, 1) = DST(2, 3) = AVG3(D, E, F);
+ DST(3, 3) = AVG3(E, F, G); // differs from vp8
+}
+
+void vpx_d45_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {
+ const int A = above[0];
+ const int B = above[1];
+ const int C = above[2];
+ const int D = above[3];
+ const int E = above[4];
+ const int F = above[5];
+ const int G = above[6];
+ const int H = above[7];
+ (void)stride;
+ (void)left;
+ DST(0, 0) = AVG3(A, B, C);
+ DST(1, 0) = DST(0, 1) = AVG3(B, C, D);
+ DST(2, 0) = DST(1, 1) = DST(0, 2) = AVG3(C, D, E);
+ DST(3, 0) = DST(2, 1) = DST(1, 2) = DST(0, 3) = AVG3(D, E, F);
+ DST(3, 1) = DST(2, 2) = DST(1, 3) = AVG3(E, F, G);
+ DST(3, 2) = DST(2, 3) = AVG3(F, G, H);
+ DST(3, 3) = H; // differs from vp8
+}
+
+void vpx_d117_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {
+ const int I = left[0];
+ const int J = left[1];
+ const int K = left[2];
+ const int X = above[-1];
+ const int A = above[0];
+ const int B = above[1];
+ const int C = above[2];
+ const int D = above[3];
+ DST(0, 0) = DST(1, 2) = AVG2(X, A);
+ DST(1, 0) = DST(2, 2) = AVG2(A, B);
+ DST(2, 0) = DST(3, 2) = AVG2(B, C);
+ DST(3, 0) = AVG2(C, D);
+
+ DST(0, 3) = AVG3(K, J, I);
+ DST(0, 2) = AVG3(J, I, X);
+ DST(0, 1) = DST(1, 3) = AVG3(I, X, A);
+ DST(1, 1) = DST(2, 3) = AVG3(X, A, B);
+ DST(2, 1) = DST(3, 3) = AVG3(A, B, C);
+ DST(3, 1) = AVG3(B, C, D);
+}
+
+void vpx_d135_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {
+ const int I = left[0];
+ const int J = left[1];
+ const int K = left[2];
+ const int L = left[3];
+ const int X = above[-1];
+ const int A = above[0];
+ const int B = above[1];
+ const int C = above[2];
+ const int D = above[3];
+ (void)stride;
+ DST(0, 3) = AVG3(J, K, L);
+ DST(1, 3) = DST(0, 2) = AVG3(I, J, K);
+ DST(2, 3) = DST(1, 2) = DST(0, 1) = AVG3(X, I, J);
+ DST(3, 3) = DST(2, 2) = DST(1, 1) = DST(0, 0) = AVG3(A, X, I);
+ DST(3, 2) = DST(2, 1) = DST(1, 0) = AVG3(B, A, X);
+ DST(3, 1) = DST(2, 0) = AVG3(C, B, A);
+ DST(3, 0) = AVG3(D, C, B);
+}
+
+void vpx_d153_predictor_4x4_c(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {
+ const int I = left[0];
+ const int J = left[1];
+ const int K = left[2];
+ const int L = left[3];
+ const int X = above[-1];
+ const int A = above[0];
+ const int B = above[1];
+ const int C = above[2];
+
+ DST(0, 0) = DST(2, 1) = AVG2(I, X);
+ DST(0, 1) = DST(2, 2) = AVG2(J, I);
+ DST(0, 2) = DST(2, 3) = AVG2(K, J);
+ DST(0, 3) = AVG2(L, K);
+
+ DST(3, 0) = AVG3(A, B, C);
+ DST(2, 0) = AVG3(X, A, B);
+ DST(1, 0) = DST(3, 1) = AVG3(I, X, A);
+ DST(1, 1) = DST(3, 2) = AVG3(J, I, X);
+ DST(1, 2) = DST(3, 3) = AVG3(K, J, I);
+ DST(1, 3) = AVG3(L, K, J);
+}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+static INLINE void highbd_d207_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int r, c;
+ (void) above;
+ (void) bd;
+
+ // First column.
+ for (r = 0; r < bs - 1; ++r) {
+ dst[r * stride] = AVG2(left[r], left[r + 1]);
+ }
+ dst[(bs - 1) * stride] = left[bs - 1];
+ dst++;
+
+ // Second column.
+ for (r = 0; r < bs - 2; ++r) {
+ dst[r * stride] = AVG3(left[r], left[r + 1], left[r + 2]);
+ }
+ dst[(bs - 2) * stride] = AVG3(left[bs - 2], left[bs - 1], left[bs - 1]);
+ dst[(bs - 1) * stride] = left[bs - 1];
+ dst++;
+
+ // Rest of last row.
+ for (c = 0; c < bs - 2; ++c)
+ dst[(bs - 1) * stride + c] = left[bs - 1];
+
+ for (r = bs - 2; r >= 0; --r) {
+ for (c = 0; c < bs - 2; ++c)
+ dst[r * stride + c] = dst[(r + 1) * stride + c - 2];
+ }
+}
+
+static INLINE void highbd_d63_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int r, c;
+ (void) left;
+ (void) bd;
+ for (r = 0; r < bs; ++r) {
+ for (c = 0; c < bs; ++c) {
+ dst[c] = r & 1 ? AVG3(above[(r >> 1) + c], above[(r >> 1) + c + 1],
+ above[(r >> 1) + c + 2])
+ : AVG2(above[(r >> 1) + c], above[(r >> 1) + c + 1]);
+ }
+ dst += stride;
+ }
+}
+
+static INLINE void highbd_d45_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
+ const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int r, c;
+ (void) left;
+ (void) bd;
+ for (r = 0; r < bs; ++r) {
+ for (c = 0; c < bs; ++c) {
+ dst[c] = r + c + 2 < bs * 2 ? AVG3(above[r + c], above[r + c + 1],
+ above[r + c + 2])
+ : above[bs * 2 - 1];
+ }
+ dst += stride;
+ }
+}
+
+static INLINE void highbd_d117_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int r, c;
+ (void) bd;
+
+ // first row
+ for (c = 0; c < bs; c++)
+ dst[c] = AVG2(above[c - 1], above[c]);
+ dst += stride;
+
+ // second row
+ dst[0] = AVG3(left[0], above[-1], above[0]);
+ for (c = 1; c < bs; c++)
+ dst[c] = AVG3(above[c - 2], above[c - 1], above[c]);
+ dst += stride;
+
+ // the rest of first col
+ dst[0] = AVG3(above[-1], left[0], left[1]);
+ for (r = 3; r < bs; ++r)
+ dst[(r - 2) * stride] = AVG3(left[r - 3], left[r - 2], left[r - 1]);
+
+ // the rest of the block
+ for (r = 2; r < bs; ++r) {
+ for (c = 1; c < bs; c++)
+ dst[c] = dst[-2 * stride + c - 1];
+ dst += stride;
+ }
+}
+
+static INLINE void highbd_d135_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int r, c;
+ (void) bd;
+ dst[0] = AVG3(left[0], above[-1], above[0]);
+ for (c = 1; c < bs; c++)
+ dst[c] = AVG3(above[c - 2], above[c - 1], above[c]);
+
+ dst[stride] = AVG3(above[-1], left[0], left[1]);
+ for (r = 2; r < bs; ++r)
+ dst[r * stride] = AVG3(left[r - 2], left[r - 1], left[r]);
+
+ dst += stride;
+ for (r = 1; r < bs; ++r) {
+ for (c = 1; c < bs; c++)
+ dst[c] = dst[-stride + c - 1];
+ dst += stride;
+ }
+}
+
+static INLINE void highbd_d153_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int r, c;
+ (void) bd;
+ dst[0] = AVG2(above[-1], left[0]);
+ for (r = 1; r < bs; r++)
+ dst[r * stride] = AVG2(left[r - 1], left[r]);
+ dst++;
+
+ dst[0] = AVG3(left[0], above[-1], above[0]);
+ dst[stride] = AVG3(above[-1], left[0], left[1]);
+ for (r = 2; r < bs; r++)
+ dst[r * stride] = AVG3(left[r - 2], left[r - 1], left[r]);
+ dst++;
+
+ for (c = 0; c < bs - 2; c++)
+ dst[c] = AVG3(above[c - 1], above[c], above[c + 1]);
+ dst += stride;
+
+ for (r = 1; r < bs; ++r) {
+ for (c = 0; c < bs - 2; c++)
+ dst[c] = dst[-stride + c - 2];
+ dst += stride;
+ }
+}
+
+static INLINE void highbd_v_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int r;
+ (void) left;
+ (void) bd;
+ for (r = 0; r < bs; r++) {
+ memcpy(dst, above, bs * sizeof(uint16_t));
+ dst += stride;
+ }
+}
+
+static INLINE void highbd_h_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int r;
+ (void) above;
+ (void) bd;
+ for (r = 0; r < bs; r++) {
+ vpx_memset16(dst, left[r], bs);
+ dst += stride;
+ }
+}
+
+static INLINE void highbd_tm_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int r, c;
+ int ytop_left = above[-1];
+ (void) bd;
+
+ for (r = 0; r < bs; r++) {
+ for (c = 0; c < bs; c++)
+ dst[c] = clip_pixel_highbd(left[r] + above[c] - ytop_left, bd);
+ dst += stride;
+ }
+}
+
+static INLINE void highbd_dc_128_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int r;
+ (void) above;
+ (void) left;
+
+ for (r = 0; r < bs; r++) {
+ vpx_memset16(dst, 128 << (bd - 8), bs);
+ dst += stride;
+ }
+}
+
+static INLINE void highbd_dc_left_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int i, r, expected_dc, sum = 0;
+ (void) above;
+ (void) bd;
+
+ for (i = 0; i < bs; i++)
+ sum += left[i];
+ expected_dc = (sum + (bs >> 1)) / bs;
+
+ for (r = 0; r < bs; r++) {
+ vpx_memset16(dst, expected_dc, bs);
+ dst += stride;
+ }
+}
+
+static INLINE void highbd_dc_top_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int i, r, expected_dc, sum = 0;
+ (void) left;
+ (void) bd;
+
+ for (i = 0; i < bs; i++)
+ sum += above[i];
+ expected_dc = (sum + (bs >> 1)) / bs;
+
+ for (r = 0; r < bs; r++) {
+ vpx_memset16(dst, expected_dc, bs);
+ dst += stride;
+ }
+}
+
+static INLINE void highbd_dc_predictor(uint16_t *dst, ptrdiff_t stride,
+ int bs, const uint16_t *above,
+ const uint16_t *left, int bd) {
+ int i, r, expected_dc, sum = 0;
+ const int count = 2 * bs;
+ (void) bd;
+
+ for (i = 0; i < bs; i++) {
+ sum += above[i];
+ sum += left[i];
+ }
+
+ expected_dc = (sum + (count >> 1)) / count;
+
+ for (r = 0; r < bs; r++) {
+ vpx_memset16(dst, expected_dc, bs);
+ dst += stride;
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
+// This serves as a wrapper function, so that all the prediction functions
+// can be unified and accessed as a pointer array. Note that the boundary
+// above and left are not necessarily used all the time.
+#define intra_pred_sized(type, size) \
+ void vpx_##type##_predictor_##size##x##size##_c(uint8_t *dst, \
+ ptrdiff_t stride, \
+ const uint8_t *above, \
+ const uint8_t *left) { \
+ type##_predictor(dst, stride, size, above, left); \
+ }
+
+#if CONFIG_VP9_HIGHBITDEPTH
+#define intra_pred_highbd_sized(type, size) \
+ void vpx_highbd_##type##_predictor_##size##x##size##_c( \
+ uint16_t *dst, ptrdiff_t stride, const uint16_t *above, \
+ const uint16_t *left, int bd) { \
+ highbd_##type##_predictor(dst, stride, size, above, left, bd); \
+ }
+
+#define intra_pred_allsizes(type) \
+ intra_pred_sized(type, 4) \
+ intra_pred_sized(type, 8) \
+ intra_pred_sized(type, 16) \
+ intra_pred_sized(type, 32) \
+ intra_pred_highbd_sized(type, 4) \
+ intra_pred_highbd_sized(type, 8) \
+ intra_pred_highbd_sized(type, 16) \
+ intra_pred_highbd_sized(type, 32)
+
+#define intra_pred_no_4x4(type) \
+ intra_pred_sized(type, 8) \
+ intra_pred_sized(type, 16) \
+ intra_pred_sized(type, 32) \
+ intra_pred_highbd_sized(type, 4) \
+ intra_pred_highbd_sized(type, 8) \
+ intra_pred_highbd_sized(type, 16) \
+ intra_pred_highbd_sized(type, 32)
+
+#else
+#define intra_pred_allsizes(type) \
+ intra_pred_sized(type, 4) \
+ intra_pred_sized(type, 8) \
+ intra_pred_sized(type, 16) \
+ intra_pred_sized(type, 32)
+
+#define intra_pred_no_4x4(type) \
+ intra_pred_sized(type, 8) \
+ intra_pred_sized(type, 16) \
+ intra_pred_sized(type, 32)
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
+intra_pred_no_4x4(d207)
+intra_pred_no_4x4(d63)
+intra_pred_no_4x4(d45)
+intra_pred_no_4x4(d117)
+intra_pred_no_4x4(d135)
+intra_pred_no_4x4(d153)
+intra_pred_allsizes(v)
+intra_pred_allsizes(h)
+intra_pred_allsizes(tm)
+intra_pred_allsizes(dc_128)
+intra_pred_allsizes(dc_left)
+intra_pred_allsizes(dc_top)
+intra_pred_allsizes(dc)
+#undef intra_pred_allsizes
diff --git a/vpx_dsp/mips/common_dspr2.h b/vpx_dsp/mips/common_dspr2.h
index 8278101..7a10bf1 100644
--- a/vpx_dsp/mips/common_dspr2.h
+++ b/vpx_dsp/mips/common_dspr2.h
@@ -21,6 +21,8 @@
#if HAVE_DSPR2
#define CROP_WIDTH 512
+extern uint8_t *vpx_ff_cropTbl; // From "vpx_dsp/mips/intrapred4_dspr2.c"
+
static INLINE void prefetch_load(const unsigned char *src) {
__asm__ __volatile__ (
"pref 0, 0(%[src]) \n\t"
diff --git a/vp9/encoder/mips/msa/vp9_fdct32x32_msa.c b/vpx_dsp/mips/fwd_dct32x32_msa.c
similarity index 98%
rename from vp9/encoder/mips/msa/vp9_fdct32x32_msa.c
rename to vpx_dsp/mips/fwd_dct32x32_msa.c
index 17a3f94..2115a34 100644
--- a/vp9/encoder/mips/msa/vp9_fdct32x32_msa.c
+++ b/vpx_dsp/mips/fwd_dct32x32_msa.c
@@ -8,8 +8,7 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-#include "./vp9_rtcd.h"
-#include "vp9/encoder/mips/msa/vp9_fdct_msa.h"
+#include "vpx_dsp/mips/fwd_txfm_msa.h"
static void fdct8x32_1d_column_load_butterfly(const int16_t *input,
int32_t src_stride,
@@ -676,7 +675,7 @@
fdct8x32_1d_row_transpose_store(tmp_buf, output);
}
-void vp9_fdct32x32_msa(const int16_t *input, int16_t *output,
+void vpx_fdct32x32_msa(const int16_t *input, int16_t *output,
int32_t src_stride) {
int32_t i;
DECLARE_ALIGNED(32, int16_t, tmp_buf_big[1024]);
@@ -697,28 +696,6 @@
}
}
-void vp9_fdct32x32_1_msa(const int16_t *input, int16_t *out, int32_t stride) {
- out[1] = 0;
-
- out[0] = LD_HADD(input, stride);
- out[0] += LD_HADD(input + 8, stride);
- out[0] += LD_HADD(input + 16, stride);
- out[0] += LD_HADD(input + 24, stride);
- out[0] += LD_HADD(input + 32 * 8, stride);
- out[0] += LD_HADD(input + 32 * 8 + 8, stride);
- out[0] += LD_HADD(input + 32 * 8 + 16, stride);
- out[0] += LD_HADD(input + 32 * 8 + 24, stride);
- out[0] += LD_HADD(input + 32 * 16, stride);
- out[0] += LD_HADD(input + 32 * 16 + 8, stride);
- out[0] += LD_HADD(input + 32 * 16 + 16, stride);
- out[0] += LD_HADD(input + 32 * 16 + 24, stride);
- out[0] += LD_HADD(input + 32 * 24, stride);
- out[0] += LD_HADD(input + 32 * 24 + 8, stride);
- out[0] += LD_HADD(input + 32 * 24 + 16, stride);
- out[0] += LD_HADD(input + 32 * 24 + 24, stride);
- out[0] >>= 3;
-}
-
static void fdct8x32_1d_row_even_rd(int16_t *temp, int16_t *out) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v8i16 in8, in9, in10, in11, in12, in13, in14, in15;
@@ -936,7 +913,7 @@
fdct8x32_1d_row_transpose_store(tmp_buf, output);
}
-void vp9_fdct32x32_rd_msa(const int16_t *input, int16_t *out,
+void vpx_fdct32x32_rd_msa(const int16_t *input, int16_t *out,
int32_t src_stride) {
int32_t i;
DECLARE_ALIGNED(32, int16_t, tmp_buf_big[1024]);
@@ -954,3 +931,25 @@
out + (8 * i * 32));
}
}
+
+void vpx_fdct32x32_1_msa(const int16_t *input, int16_t *out, int32_t stride) {
+ out[1] = 0;
+
+ out[0] = LD_HADD(input, stride);
+ out[0] += LD_HADD(input + 8, stride);
+ out[0] += LD_HADD(input + 16, stride);
+ out[0] += LD_HADD(input + 24, stride);
+ out[0] += LD_HADD(input + 32 * 8, stride);
+ out[0] += LD_HADD(input + 32 * 8 + 8, stride);
+ out[0] += LD_HADD(input + 32 * 8 + 16, stride);
+ out[0] += LD_HADD(input + 32 * 8 + 24, stride);
+ out[0] += LD_HADD(input + 32 * 16, stride);
+ out[0] += LD_HADD(input + 32 * 16 + 8, stride);
+ out[0] += LD_HADD(input + 32 * 16 + 16, stride);
+ out[0] += LD_HADD(input + 32 * 16 + 24, stride);
+ out[0] += LD_HADD(input + 32 * 24, stride);
+ out[0] += LD_HADD(input + 32 * 24 + 8, stride);
+ out[0] += LD_HADD(input + 32 * 24 + 16, stride);
+ out[0] += LD_HADD(input + 32 * 24 + 24, stride);
+ out[0] >>= 3;
+}
diff --git a/vpx_dsp/mips/fwd_txfm_msa.c b/vpx_dsp/mips/fwd_txfm_msa.c
index 8a7e7b6..f66dd5f 100644
--- a/vpx_dsp/mips/fwd_txfm_msa.c
+++ b/vpx_dsp/mips/fwd_txfm_msa.c
@@ -166,7 +166,7 @@
ST_SH8(tmp4, in4, tmp5, in5, tmp6, in6, tmp7, in7, output + 8, 16);
}
-void vp9_fdct4x4_msa(const int16_t *input, int16_t *output,
+void vpx_fdct4x4_msa(const int16_t *input, int16_t *output,
int32_t src_stride) {
v8i16 in0, in1, in2, in3;
@@ -196,7 +196,7 @@
ST_SH2(in0, in2, output, 8);
}
-void vp9_fdct8x8_msa(const int16_t *input, int16_t *output,
+void vpx_fdct8x8_msa(const int16_t *input, int16_t *output,
int32_t src_stride) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
@@ -215,7 +215,12 @@
ST_SH8(in0, in1, in2, in3, in4, in5, in6, in7, output, 8);
}
-void vp9_fdct16x16_msa(const int16_t *input, int16_t *output,
+void vpx_fdct8x8_1_msa(const int16_t *input, int16_t *out, int32_t stride) {
+ out[0] = LD_HADD(input, stride);
+ out[1] = 0;
+}
+
+void vpx_fdct16x16_msa(const int16_t *input, int16_t *output,
int32_t src_stride) {
int32_t i;
DECLARE_ALIGNED(32, int16_t, tmp_buf[16 * 16]);
@@ -230,3 +235,13 @@
fdct16x8_1d_row((&tmp_buf[0] + (128 * i)), (output + (128 * i)));
}
}
+
+void vpx_fdct16x16_1_msa(const int16_t *input, int16_t *out, int32_t stride) {
+ out[1] = 0;
+
+ out[0] = LD_HADD(input, stride);
+ out[0] += LD_HADD(input + 8, stride);
+ out[0] += LD_HADD(input + 16 * 8, stride);
+ out[0] += LD_HADD(input + 16 * 8 + 8, stride);
+ out[0] >>= 1;
+}
diff --git a/vpx_dsp/mips/fwd_txfm_msa.h b/vpx_dsp/mips/fwd_txfm_msa.h
index d4c68ec..d1e160e 100644
--- a/vpx_dsp/mips/fwd_txfm_msa.h
+++ b/vpx_dsp/mips/fwd_txfm_msa.h
@@ -14,6 +14,21 @@
#include "vpx_dsp/mips/txfm_macros_msa.h"
#include "vpx_dsp/txfm_common.h"
+#define LD_HADD(psrc, stride) ({ \
+ v8i16 in0_m, in1_m, in2_m, in3_m, in4_m, in5_m, in6_m, in7_m; \
+ v4i32 vec_w_m; \
+ \
+ LD_SH4((psrc), stride, in0_m, in1_m, in2_m, in3_m); \
+ ADD2(in0_m, in1_m, in2_m, in3_m, in0_m, in2_m); \
+ LD_SH4(((psrc) + 4 * stride), stride, in4_m, in5_m, in6_m, in7_m); \
+ ADD4(in4_m, in5_m, in6_m, in7_m, in0_m, in2_m, in4_m, in6_m, \
+ in4_m, in6_m, in0_m, in4_m); \
+ in0_m += in4_m; \
+ \
+ vec_w_m = __msa_hadd_s_w(in0_m, in0_m); \
+ HADD_SW_S32(vec_w_m); \
+})
+
#define VP9_FDCT4(in0, in1, in2, in3, out0, out1, out2, out3) { \
v8i16 cnst0_m, cnst1_m, cnst2_m, cnst3_m; \
v8i16 vec0_m, vec1_m, vec2_m, vec3_m; \
@@ -273,6 +288,85 @@
out3 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m); \
}
+#define FDCT_POSTPROC_2V_NEG_H(vec0, vec1) { \
+ v8i16 tp0_m, tp1_m; \
+ v8i16 one_m = __msa_ldi_h(1); \
+ \
+ tp0_m = __msa_clti_s_h(vec0, 0); \
+ tp1_m = __msa_clti_s_h(vec1, 0); \
+ vec0 += 1; \
+ vec1 += 1; \
+ tp0_m = one_m & tp0_m; \
+ tp1_m = one_m & tp1_m; \
+ vec0 += tp0_m; \
+ vec1 += tp1_m; \
+ vec0 >>= 2; \
+ vec1 >>= 2; \
+}
+
+#define FDCT32_POSTPROC_NEG_W(vec) { \
+ v4i32 temp_m; \
+ v4i32 one_m = __msa_ldi_w(1); \
+ \
+ temp_m = __msa_clti_s_w(vec, 0); \
+ vec += 1; \
+ temp_m = one_m & temp_m; \
+ vec += temp_m; \
+ vec >>= 2; \
+}
+
+#define FDCT32_POSTPROC_2V_POS_H(vec0, vec1) { \
+ v8i16 tp0_m, tp1_m; \
+ v8i16 one = __msa_ldi_h(1); \
+ \
+ tp0_m = __msa_clei_s_h(vec0, 0); \
+ tp1_m = __msa_clei_s_h(vec1, 0); \
+ tp0_m = (v8i16)__msa_xori_b((v16u8)tp0_m, 255); \
+ tp1_m = (v8i16)__msa_xori_b((v16u8)tp1_m, 255); \
+ vec0 += 1; \
+ vec1 += 1; \
+ tp0_m = one & tp0_m; \
+ tp1_m = one & tp1_m; \
+ vec0 += tp0_m; \
+ vec1 += tp1_m; \
+ vec0 >>= 2; \
+ vec1 >>= 2; \
+}
+
+#define DOTP_CONST_PAIR_W(reg0_left, reg1_left, reg0_right, \
+ reg1_right, const0, const1, \
+ out0, out1, out2, out3) { \
+ v4i32 s0_m, s1_m, s2_m, s3_m, s4_m, s5_m, s6_m, s7_m; \
+ v2i64 tp0_m, tp1_m, tp2_m, tp3_m; \
+ v4i32 k0_m = __msa_fill_w((int32_t) const0); \
+ \
+ s0_m = __msa_fill_w((int32_t) const1); \
+ k0_m = __msa_ilvev_w(s0_m, k0_m); \
+ \
+ ILVRL_W2_SW(-reg1_left, reg0_left, s1_m, s0_m); \
+ ILVRL_W2_SW(reg0_left, reg1_left, s3_m, s2_m); \
+ ILVRL_W2_SW(-reg1_right, reg0_right, s5_m, s4_m); \
+ ILVRL_W2_SW(reg0_right, reg1_right, s7_m, s6_m); \
+ \
+ DOTP_SW2_SD(s0_m, s1_m, k0_m, k0_m, tp0_m, tp1_m); \
+ DOTP_SW2_SD(s4_m, s5_m, k0_m, k0_m, tp2_m, tp3_m); \
+ tp0_m = __msa_srari_d(tp0_m, DCT_CONST_BITS); \
+ tp1_m = __msa_srari_d(tp1_m, DCT_CONST_BITS); \
+ tp2_m = __msa_srari_d(tp2_m, DCT_CONST_BITS); \
+ tp3_m = __msa_srari_d(tp3_m, DCT_CONST_BITS); \
+ out0 = __msa_pckev_w((v4i32)tp0_m, (v4i32)tp1_m); \
+ out1 = __msa_pckev_w((v4i32)tp2_m, (v4i32)tp3_m); \
+ \
+ DOTP_SW2_SD(s2_m, s3_m, k0_m, k0_m, tp0_m, tp1_m); \
+ DOTP_SW2_SD(s6_m, s7_m, k0_m, k0_m, tp2_m, tp3_m); \
+ tp0_m = __msa_srari_d(tp0_m, DCT_CONST_BITS); \
+ tp1_m = __msa_srari_d(tp1_m, DCT_CONST_BITS); \
+ tp2_m = __msa_srari_d(tp2_m, DCT_CONST_BITS); \
+ tp3_m = __msa_srari_d(tp3_m, DCT_CONST_BITS); \
+ out2 = __msa_pckev_w((v4i32)tp0_m, (v4i32)tp1_m); \
+ out3 = __msa_pckev_w((v4i32)tp2_m, (v4i32)tp3_m); \
+}
+
void fdct8x16_1d_column(const int16_t *input, int16_t *tmp_ptr,
int32_t src_stride);
void fdct16x8_1d_row(int16_t *input, int16_t *output);
diff --git a/vp9/common/mips/dspr2/vp9_intrapred16_dspr2.c b/vpx_dsp/mips/intrapred16_dspr2.c
similarity index 97%
rename from vp9/common/mips/dspr2/vp9_intrapred16_dspr2.c
rename to vpx_dsp/mips/intrapred16_dspr2.c
index b0dc496..11444c7 100644
--- a/vp9/common/mips/dspr2/vp9_intrapred16_dspr2.c
+++ b/vpx_dsp/mips/intrapred16_dspr2.c
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2013 The WebM project authors. All Rights Reserved.
+ * Copyright (c) 2015 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
@@ -7,14 +7,11 @@
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
-#include <stdlib.h>
-#include "./vp9_rtcd.h"
-#include "vp9/common/vp9_common.h"
-#include "vp9/common/mips/dspr2/vp9_common_dspr2.h"
+#include "vpx_dsp/mips/common_dspr2.h"
#if HAVE_DSPR2
-void vp9_h_predictor_16x16_dspr2(uint8_t *dst, ptrdiff_t stride,
+void vpx_h_predictor_16x16_dspr2(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int32_t tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8;
int32_t tmp9, tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16;
@@ -161,7 +158,7 @@
);
}
-void vp9_dc_predictor_16x16_dspr2(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_predictor_16x16_dspr2(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int32_t expected_dc;
int32_t average;
diff --git a/vp9/common/mips/dspr2/vp9_intrapred4_dspr2.c b/vpx_dsp/mips/intrapred4_dspr2.c
similarity index 93%
rename from vp9/common/mips/dspr2/vp9_intrapred4_dspr2.c
rename to vpx_dsp/mips/intrapred4_dspr2.c
index a53c623..3872f24 100644
--- a/vp9/common/mips/dspr2/vp9_intrapred4_dspr2.c
+++ b/vpx_dsp/mips/intrapred4_dspr2.c
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2013 The WebM project authors. All Rights Reserved.
+ * Copyright (c) 2015 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
@@ -7,14 +7,27 @@
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
-#include <stdlib.h>
-#include "./vp9_rtcd.h"
-#include "vp9/common/vp9_common.h"
-#include "vp9/common/mips/dspr2/vp9_common_dspr2.h"
+#include "vpx_dsp/mips/common_dspr2.h"
#if HAVE_DSPR2
-void vp9_h_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
+uint8_t vpx_ff_cropTbl_a[256 + 2 * CROP_WIDTH];
+uint8_t *vpx_ff_cropTbl;
+
+void vpx_dsputil_static_init(void) {
+ int i;
+
+ for (i = 0; i < 256; i++) vpx_ff_cropTbl_a[i + CROP_WIDTH] = i;
+
+ for (i = 0; i < CROP_WIDTH; i++) {
+ vpx_ff_cropTbl_a[i] = 0;
+ vpx_ff_cropTbl_a[i + CROP_WIDTH + 256] = 255;
+ }
+
+ vpx_ff_cropTbl = &vpx_ff_cropTbl_a[CROP_WIDTH];
+}
+
+void vpx_h_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int32_t tmp1, tmp2, tmp3, tmp4;
@@ -41,7 +54,7 @@
);
}
-void vp9_dc_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int32_t expected_dc;
int32_t average;
@@ -83,7 +96,7 @@
);
}
-void vp9_tm_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
+void vpx_tm_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int32_t abovel, abover;
int32_t left0, left1, left2, left3;
@@ -91,7 +104,7 @@
int32_t resl;
int32_t resr;
int32_t top_left;
- uint8_t *cm = vp9_ff_cropTbl;
+ uint8_t *cm = vpx_ff_cropTbl;
__asm__ __volatile__ (
"ulw %[resl], (%[above]) \n\t"
diff --git a/vp9/common/mips/dspr2/vp9_intrapred8_dspr2.c b/vpx_dsp/mips/intrapred8_dspr2.c
similarity index 98%
rename from vp9/common/mips/dspr2/vp9_intrapred8_dspr2.c
rename to vpx_dsp/mips/intrapred8_dspr2.c
index 40d93ae..196ff5a 100644
--- a/vp9/common/mips/dspr2/vp9_intrapred8_dspr2.c
+++ b/vpx_dsp/mips/intrapred8_dspr2.c
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2013 The WebM project authors. All Rights Reserved.
+ * Copyright (c) 2015 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
@@ -7,14 +7,11 @@
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
-#include <stdlib.h>
-#include "./vp9_rtcd.h"
-#include "vp9/common/vp9_common.h"
-#include "vp9/common/mips/dspr2/vp9_common_dspr2.h"
+#include "vpx_dsp/mips/common_dspr2.h"
#if HAVE_DSPR2
-void vp9_h_predictor_8x8_dspr2(uint8_t *dst, ptrdiff_t stride,
+void vpx_h_predictor_8x8_dspr2(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int32_t tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8;
@@ -70,7 +67,7 @@
);
}
-void vp9_dc_predictor_8x8_dspr2(uint8_t *dst, ptrdiff_t stride,
+void vpx_dc_predictor_8x8_dspr2(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int32_t expected_dc;
int32_t average;
@@ -153,7 +150,7 @@
);
}
-void vp9_tm_predictor_8x8_dspr2(uint8_t *dst, ptrdiff_t stride,
+void vpx_tm_predictor_8x8_dspr2(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
int32_t abovel, abover;
int32_t abovel_1, abover_1;
@@ -161,7 +158,7 @@
int32_t res0, res1, res2, res3;
int32_t reshw;
int32_t top_left;
- uint8_t *cm = vp9_ff_cropTbl;
+ uint8_t *cm = vpx_ff_cropTbl;
__asm__ __volatile__ (
"ulw %[reshw], (%[above]) \n\t"
diff --git a/vp9/common/mips/msa/vp9_intra_predict_msa.c b/vpx_dsp/mips/intrapred_msa.c
similarity index 91%
rename from vp9/common/mips/msa/vp9_intra_predict_msa.c
rename to vpx_dsp/mips/intrapred_msa.c
index abf2704..f6fbe40 100644
--- a/vp9/common/mips/msa/vp9_intra_predict_msa.c
+++ b/vpx_dsp/mips/intrapred_msa.c
@@ -8,7 +8,7 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-#include "./vp9_rtcd.h"
+#include "./vpx_dsp_rtcd.h"
#include "vpx_dsp/mips/macros_msa.h"
#define IPRED_SUBS_UH2_UH(in0, in1, out0, out1) { \
@@ -550,125 +550,125 @@
}
}
-void vp9_v_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_v_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)left;
intra_predict_vert_4x4_msa(above, dst, y_stride);
}
-void vp9_v_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_v_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)left;
intra_predict_vert_8x8_msa(above, dst, y_stride);
}
-void vp9_v_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_v_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)left;
intra_predict_vert_16x16_msa(above, dst, y_stride);
}
-void vp9_v_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_v_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)left;
intra_predict_vert_32x32_msa(above, dst, y_stride);
}
-void vp9_h_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_h_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
intra_predict_horiz_4x4_msa(left, dst, y_stride);
}
-void vp9_h_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_h_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
intra_predict_horiz_8x8_msa(left, dst, y_stride);
}
-void vp9_h_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_h_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
intra_predict_horiz_16x16_msa(left, dst, y_stride);
}
-void vp9_h_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_h_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
intra_predict_horiz_32x32_msa(left, dst, y_stride);
}
-void vp9_dc_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
intra_predict_dc_4x4_msa(above, left, dst, y_stride);
}
-void vp9_dc_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
intra_predict_dc_8x8_msa(above, left, dst, y_stride);
}
-void vp9_dc_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
intra_predict_dc_16x16_msa(above, left, dst, y_stride);
}
-void vp9_dc_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
intra_predict_dc_32x32_msa(above, left, dst, y_stride);
}
-void vp9_dc_top_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_top_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)left;
intra_predict_dc_tl_4x4_msa(above, dst, y_stride);
}
-void vp9_dc_top_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_top_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)left;
intra_predict_dc_tl_8x8_msa(above, dst, y_stride);
}
-void vp9_dc_top_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_top_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)left;
intra_predict_dc_tl_16x16_msa(above, dst, y_stride);
}
-void vp9_dc_top_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_top_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)left;
intra_predict_dc_tl_32x32_msa(above, dst, y_stride);
}
-void vp9_dc_left_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_left_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
intra_predict_dc_tl_4x4_msa(left, dst, y_stride);
}
-void vp9_dc_left_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_left_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
intra_predict_dc_tl_8x8_msa(left, dst, y_stride);
}
-void vp9_dc_left_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_left_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above,
const uint8_t *left) {
(void)above;
@@ -676,7 +676,7 @@
intra_predict_dc_tl_16x16_msa(left, dst, y_stride);
}
-void vp9_dc_left_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_left_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above,
const uint8_t *left) {
(void)above;
@@ -684,7 +684,7 @@
intra_predict_dc_tl_32x32_msa(left, dst, y_stride);
}
-void vp9_dc_128_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_128_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
(void)left;
@@ -692,7 +692,7 @@
intra_predict_128dc_4x4_msa(dst, y_stride);
}
-void vp9_dc_128_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_128_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
(void)left;
@@ -700,7 +700,7 @@
intra_predict_128dc_8x8_msa(dst, y_stride);
}
-void vp9_dc_128_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_128_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
(void)left;
@@ -708,7 +708,7 @@
intra_predict_128dc_16x16_msa(dst, y_stride);
}
-void vp9_dc_128_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_dc_128_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
(void)above;
(void)left;
@@ -716,22 +716,22 @@
intra_predict_128dc_32x32_msa(dst, y_stride);
}
-void vp9_tm_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_tm_predictor_4x4_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
intra_predict_tm_4x4_msa(above, left, dst, y_stride);
}
-void vp9_tm_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_tm_predictor_8x8_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
intra_predict_tm_8x8_msa(above, left, dst, y_stride);
}
-void vp9_tm_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_tm_predictor_16x16_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
intra_predict_tm_16x16_msa(above, left, dst, y_stride);
}
-void vp9_tm_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
+void vpx_tm_predictor_32x32_msa(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left) {
intra_predict_tm_32x32_msa(above, left, dst, y_stride);
}
diff --git a/vpx_dsp/vpx_dsp.mk b/vpx_dsp/vpx_dsp.mk
index e3bf5a2..2919e14 100644
--- a/vpx_dsp/vpx_dsp.mk
+++ b/vpx_dsp/vpx_dsp.mk
@@ -31,6 +31,27 @@
DSP_SRCS-yes += bitreader_buffer.h
endif
+# intra predictions
+DSP_SRCS-yes += intrapred.c
+
+ifeq ($(CONFIG_USE_X86INC),yes)
+DSP_SRCS-$(HAVE_SSE2) += x86/intrapred_sse2.asm
+DSP_SRCS-$(HAVE_SSSE3) += x86/intrapred_ssse3.asm
+endif
+
+ifeq ($(CONFIG_VP9_HIGHBITDEPTH),yes)
+ifeq ($(CONFIG_USE_X86INC),yes)
+DSP_SRCS-$(HAVE_SSE2) += x86/highbd_intrapred_sse2.asm
+endif
+endif
+
+DSP_SRCS-$(HAVE_NEON_ASM) += arm/intrapred_neon_asm$(ASM)
+DSP_SRCS-$(HAVE_NEON) += arm/intrapred_neon.c
+DSP_SRCS-$(HAVE_MSA) += mips/intrapred_msa.c
+DSP_SRCS-$(HAVE_DSPR2) += mips/intrapred4_dspr2.c
+DSP_SRCS-$(HAVE_DSPR2) += mips/intrapred8_dspr2.c
+DSP_SRCS-$(HAVE_DSPR2) += mips/intrapred16_dspr2.c
+
# loop filters
DSP_SRCS-yes += loopfilter.c
@@ -71,14 +92,18 @@
DSP_SRCS-$(HAVE_SSE2) += x86/fwd_txfm_sse2.h
DSP_SRCS-$(HAVE_SSE2) += x86/fwd_txfm_sse2.c
DSP_SRCS-$(HAVE_SSE2) += x86/fwd_txfm_impl_sse2.h
+DSP_SRCS-$(HAVE_SSE2) += x86/fwd_dct32x32_impl_sse2.h
ifeq ($(ARCH_X86_64),yes)
ifeq ($(CONFIG_USE_X86INC),yes)
DSP_SRCS-$(HAVE_SSSE3) += x86/fwd_txfm_ssse3_x86_64.asm
endif
endif
+DSP_SRCS-$(HAVE_AVX2) += x86/fwd_txfm_avx2.c
+DSP_SRCS-$(HAVE_AVX2) += x86/fwd_dct32x32_impl_avx2.h
DSP_SRCS-$(HAVE_NEON) += arm/fwd_txfm_neon.c
DSP_SRCS-$(HAVE_MSA) += mips/fwd_txfm_msa.h
DSP_SRCS-$(HAVE_MSA) += mips/fwd_txfm_msa.c
+DSP_SRCS-$(HAVE_MSA) += mips/fwd_dct32x32_msa.c
endif # CONFIG_VP9_ENCODER
# quantization
diff --git a/vpx_dsp/vpx_dsp_common.h b/vpx_dsp/vpx_dsp_common.h
index 6793036..1fd7c15 100644
--- a/vpx_dsp/vpx_dsp_common.h
+++ b/vpx_dsp/vpx_dsp_common.h
@@ -50,6 +50,20 @@
return value < low ? low : (value > high ? high : value);
}
+#if CONFIG_VP9_HIGHBITDEPTH
+static INLINE uint16_t clip_pixel_highbd(int val, int bd) {
+ switch (bd) {
+ case 8:
+ default:
+ return (uint16_t)clamp(val, 0, 255);
+ case 10:
+ return (uint16_t)clamp(val, 0, 1023);
+ case 12:
+ return (uint16_t)clamp(val, 0, 4095);
+ }
+}
+#endif // CONFIG_VP9_HIGHBITDEPTH
+
#ifdef __cplusplus
} // extern "C"
#endif
diff --git a/vpx_dsp/vpx_dsp_rtcd_defs.pl b/vpx_dsp/vpx_dsp_rtcd_defs.pl
index 6dc4423..c656934 100644
--- a/vpx_dsp/vpx_dsp_rtcd_defs.pl
+++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl
@@ -45,6 +45,324 @@
}
#
+# Intra prediction
+#
+add_proto qw/void vpx_d207_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d207_predictor_4x4/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d45_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d45_predictor_4x4 neon/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d63_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d63_predictor_4x4/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_h_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_h_predictor_4x4 neon dspr2 msa/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d117_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d117_predictor_4x4/;
+
+add_proto qw/void vpx_d135_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d135_predictor_4x4 neon/;
+
+add_proto qw/void vpx_d153_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d153_predictor_4x4/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_v_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_v_predictor_4x4 neon msa/, "$sse_x86inc";
+
+add_proto qw/void vpx_tm_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_tm_predictor_4x4 neon dspr2 msa/, "$sse_x86inc";
+
+add_proto qw/void vpx_dc_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_predictor_4x4 dspr2 msa neon/, "$sse_x86inc";
+
+add_proto qw/void vpx_dc_top_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_top_predictor_4x4 msa neon/, "$sse_x86inc";
+
+add_proto qw/void vpx_dc_left_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_left_predictor_4x4 msa neon/, "$sse_x86inc";
+
+add_proto qw/void vpx_dc_128_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_128_predictor_4x4 msa neon/, "$sse_x86inc";
+
+add_proto qw/void vpx_d207_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d207_predictor_8x8/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d45_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d45_predictor_8x8 neon/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d63_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d63_predictor_8x8/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_h_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_h_predictor_8x8 neon dspr2 msa/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d117_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d117_predictor_8x8/;
+
+add_proto qw/void vpx_d135_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d135_predictor_8x8/;
+
+add_proto qw/void vpx_d153_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d153_predictor_8x8/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_v_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_v_predictor_8x8 neon msa/, "$sse_x86inc";
+
+add_proto qw/void vpx_tm_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_tm_predictor_8x8 neon dspr2 msa/, "$sse2_x86inc";
+
+add_proto qw/void vpx_dc_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_predictor_8x8 dspr2 neon msa/, "$sse_x86inc";
+
+add_proto qw/void vpx_dc_top_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_top_predictor_8x8 neon msa/, "$sse_x86inc";
+
+add_proto qw/void vpx_dc_left_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_left_predictor_8x8 neon msa/, "$sse_x86inc";
+
+add_proto qw/void vpx_dc_128_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_128_predictor_8x8 neon msa/, "$sse_x86inc";
+
+add_proto qw/void vpx_d207_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d207_predictor_16x16/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d45_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d45_predictor_16x16 neon/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d63_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d63_predictor_16x16/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_h_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_h_predictor_16x16 neon dspr2 msa/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d117_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d117_predictor_16x16/;
+
+add_proto qw/void vpx_d135_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d135_predictor_16x16/;
+
+add_proto qw/void vpx_d153_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d153_predictor_16x16/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_v_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_v_predictor_16x16 neon msa/, "$sse2_x86inc";
+
+add_proto qw/void vpx_tm_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_tm_predictor_16x16 neon msa/, "$sse2_x86inc";
+
+add_proto qw/void vpx_dc_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_predictor_16x16 dspr2 neon msa/, "$sse2_x86inc";
+
+add_proto qw/void vpx_dc_top_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_top_predictor_16x16 neon msa/, "$sse2_x86inc";
+
+add_proto qw/void vpx_dc_left_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_left_predictor_16x16 neon msa/, "$sse2_x86inc";
+
+add_proto qw/void vpx_dc_128_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_128_predictor_16x16 neon msa/, "$sse2_x86inc";
+
+add_proto qw/void vpx_d207_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d207_predictor_32x32/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d45_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d45_predictor_32x32/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d63_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d63_predictor_32x32/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_h_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_h_predictor_32x32 neon msa/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_d117_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d117_predictor_32x32/;
+
+add_proto qw/void vpx_d135_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d135_predictor_32x32/;
+
+add_proto qw/void vpx_d153_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d153_predictor_32x32/, "$ssse3_x86inc";
+
+add_proto qw/void vpx_v_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_v_predictor_32x32 neon msa/, "$sse2_x86inc";
+
+add_proto qw/void vpx_tm_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_tm_predictor_32x32 neon msa/, "$sse2_x86_64_x86inc";
+
+add_proto qw/void vpx_dc_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_predictor_32x32 msa neon/, "$sse2_x86inc";
+
+add_proto qw/void vpx_dc_top_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_top_predictor_32x32 msa neon/, "$sse2_x86inc";
+
+add_proto qw/void vpx_dc_left_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_left_predictor_32x32 msa neon/, "$sse2_x86inc";
+
+add_proto qw/void vpx_dc_128_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_128_predictor_32x32 msa neon/, "$sse2_x86inc";
+
+# High bitdepth functions
+if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
+ add_proto qw/void vpx_highbd_d207_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d207_predictor_4x4/;
+
+ add_proto qw/void vpx_highbd_d45_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d45_predictor_4x4/;
+
+ add_proto qw/void vpx_highbd_d63_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d63_predictor_4x4/;
+
+ add_proto qw/void vpx_highbd_h_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_h_predictor_4x4/;
+
+ add_proto qw/void vpx_highbd_d117_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d117_predictor_4x4/;
+
+ add_proto qw/void vpx_highbd_d135_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d135_predictor_4x4/;
+
+ add_proto qw/void vpx_highbd_d153_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d153_predictor_4x4/;
+
+ add_proto qw/void vpx_highbd_v_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_v_predictor_4x4/, "$sse_x86inc";
+
+ add_proto qw/void vpx_highbd_tm_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_tm_predictor_4x4/, "$sse_x86inc";
+
+ add_proto qw/void vpx_highbd_dc_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_predictor_4x4/, "$sse_x86inc";
+
+ add_proto qw/void vpx_highbd_dc_top_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_top_predictor_4x4/;
+
+ add_proto qw/void vpx_highbd_dc_left_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_left_predictor_4x4/;
+
+ add_proto qw/void vpx_highbd_dc_128_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_128_predictor_4x4/;
+
+ add_proto qw/void vpx_highbd_d207_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d207_predictor_8x8/;
+
+ add_proto qw/void vpx_highbd_d45_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d45_predictor_8x8/;
+
+ add_proto qw/void vpx_highbd_d63_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d63_predictor_8x8/;
+
+ add_proto qw/void vpx_highbd_h_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_h_predictor_8x8/;
+
+ add_proto qw/void vpx_highbd_d117_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d117_predictor_8x8/;
+
+ add_proto qw/void vpx_highbd_d135_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d135_predictor_8x8/;
+
+ add_proto qw/void vpx_highbd_d153_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d153_predictor_8x8/;
+
+ add_proto qw/void vpx_highbd_v_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_v_predictor_8x8/, "$sse2_x86inc";
+
+ add_proto qw/void vpx_highbd_tm_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_tm_predictor_8x8/, "$sse2_x86inc";
+
+ add_proto qw/void vpx_highbd_dc_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_predictor_8x8/, "$sse2_x86inc";;
+
+ add_proto qw/void vpx_highbd_dc_top_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_top_predictor_8x8/;
+
+ add_proto qw/void vpx_highbd_dc_left_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_left_predictor_8x8/;
+
+ add_proto qw/void vpx_highbd_dc_128_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_128_predictor_8x8/;
+
+ add_proto qw/void vpx_highbd_d207_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d207_predictor_16x16/;
+
+ add_proto qw/void vpx_highbd_d45_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d45_predictor_16x16/;
+
+ add_proto qw/void vpx_highbd_d63_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d63_predictor_16x16/;
+
+ add_proto qw/void vpx_highbd_h_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_h_predictor_16x16/;
+
+ add_proto qw/void vpx_highbd_d117_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d117_predictor_16x16/;
+
+ add_proto qw/void vpx_highbd_d135_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d135_predictor_16x16/;
+
+ add_proto qw/void vpx_highbd_d153_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d153_predictor_16x16/;
+
+ add_proto qw/void vpx_highbd_v_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_v_predictor_16x16/, "$sse2_x86inc";
+
+ add_proto qw/void vpx_highbd_tm_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_tm_predictor_16x16/, "$sse2_x86_64_x86inc";
+
+ add_proto qw/void vpx_highbd_dc_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_predictor_16x16/, "$sse2_x86inc";
+
+ add_proto qw/void vpx_highbd_dc_top_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_top_predictor_16x16/;
+
+ add_proto qw/void vpx_highbd_dc_left_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_left_predictor_16x16/;
+
+ add_proto qw/void vpx_highbd_dc_128_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_128_predictor_16x16/;
+
+ add_proto qw/void vpx_highbd_d207_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d207_predictor_32x32/;
+
+ add_proto qw/void vpx_highbd_d45_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d45_predictor_32x32/;
+
+ add_proto qw/void vpx_highbd_d63_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d63_predictor_32x32/;
+
+ add_proto qw/void vpx_highbd_h_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_h_predictor_32x32/;
+
+ add_proto qw/void vpx_highbd_d117_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d117_predictor_32x32/;
+
+ add_proto qw/void vpx_highbd_d135_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d135_predictor_32x32/;
+
+ add_proto qw/void vpx_highbd_d153_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_d153_predictor_32x32/;
+
+ add_proto qw/void vpx_highbd_v_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_v_predictor_32x32/, "$sse2_x86inc";
+
+ add_proto qw/void vpx_highbd_tm_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_tm_predictor_32x32/, "$sse2_x86_64_x86inc";
+
+ add_proto qw/void vpx_highbd_dc_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_predictor_32x32/, "$sse2_x86_64_x86inc";
+
+ add_proto qw/void vpx_highbd_dc_top_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_top_predictor_32x32/;
+
+ add_proto qw/void vpx_highbd_dc_left_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_left_predictor_32x32/;
+
+ add_proto qw/void vpx_highbd_dc_128_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+ specialize qw/vpx_highbd_dc_128_predictor_32x32/;
+}
+
+#
# Loopfilter
#
add_proto qw/void vpx_lpf_vertical_16/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
@@ -129,32 +447,83 @@
#
if (vpx_config("CONFIG_VP9_ENCODER") eq "yes") {
if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
- add_proto qw/void vp9_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct4x4 sse2/;
+ add_proto qw/void vpx_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct4x4 sse2/;
- add_proto qw/void vp9_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct8x8 sse2/;
+ add_proto qw/void vpx_fdct4x4_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct4x4_1 sse2/;
- add_proto qw/void vp9_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct16x16 sse2/;
+ add_proto qw/void vpx_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct8x8 sse2/;
- add_proto qw/void vp9_highbd_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_highbd_fdct4x4 sse2/;
+ add_proto qw/void vpx_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct8x8_1 sse2/;
- add_proto qw/void vp9_highbd_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_highbd_fdct8x8 sse2/;
+ add_proto qw/void vpx_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct16x16 sse2/;
- add_proto qw/void vp9_highbd_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_highbd_fdct16x16 sse2/;
+ add_proto qw/void vpx_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct16x16_1 sse2/;
+
+ add_proto qw/void vpx_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct32x32 sse2/;
+
+ add_proto qw/void vpx_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct32x32_rd sse2/;
+
+ add_proto qw/void vpx_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct32x32_1 sse2/;
+
+ add_proto qw/void vpx_highbd_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_highbd_fdct4x4 sse2/;
+
+ add_proto qw/void vpx_highbd_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_highbd_fdct8x8 sse2/;
+
+ add_proto qw/void vpx_highbd_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_highbd_fdct8x8_1/;
+
+ add_proto qw/void vpx_highbd_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_highbd_fdct16x16 sse2/;
+
+ add_proto qw/void vpx_highbd_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_highbd_fdct16x16_1/;
+
+ add_proto qw/void vpx_highbd_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_highbd_fdct32x32 sse2/;
+
+ add_proto qw/void vpx_highbd_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_highbd_fdct32x32_rd sse2/;
+
+ add_proto qw/void vpx_highbd_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_highbd_fdct32x32_1/;
} else {
- add_proto qw/void vp9_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct4x4 sse2 msa/;
+ add_proto qw/void vpx_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct4x4 sse2 msa/;
- add_proto qw/void vp9_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct8x8 sse2 neon msa/, "$ssse3_x86_64_x86inc";
+ add_proto qw/void vpx_fdct4x4_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct4x4_1 sse2/;
- add_proto qw/void vp9_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
- specialize qw/vp9_fdct16x16 sse2 msa/;
+ add_proto qw/void vpx_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct8x8 sse2 neon msa/, "$ssse3_x86_64_x86inc";
+
+ add_proto qw/void vpx_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct8x8_1 sse2 neon msa/;
+
+ add_proto qw/void vpx_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct16x16 sse2 msa/;
+
+ add_proto qw/void vpx_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct16x16_1 sse2 msa/;
+
+ add_proto qw/void vpx_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct32x32 sse2 avx2 msa/;
+
+ add_proto qw/void vpx_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct32x32_rd sse2 avx2 msa/;
+
+ add_proto qw/void vpx_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
+ specialize qw/vpx_fdct32x32_1 sse2 msa/;
} # CONFIG_VP9_HIGHBITDEPTH
} # CONFIG_VP9_ENCODER
diff --git a/vp9/encoder/x86/vp9_dct32x32_avx2_impl.h b/vpx_dsp/x86/fwd_dct32x32_impl_avx2.h
similarity index 99%
rename from vp9/encoder/x86/vp9_dct32x32_avx2_impl.h
rename to vpx_dsp/x86/fwd_dct32x32_impl_avx2.h
index ba5e059..4df39df 100644
--- a/vp9/encoder/x86/vp9_dct32x32_avx2_impl.h
+++ b/vpx_dsp/x86/fwd_dct32x32_impl_avx2.h
@@ -10,9 +10,7 @@
#include <immintrin.h> // AVX2
-#include "./vp9_rtcd.h"
#include "vpx_dsp/txfm_common.h"
-#include "vpx_ports/mem.h"
#define pair256_set_epi16(a, b) \
_mm256_set_epi16((int16_t)(b), (int16_t)(a), (int16_t)(b), (int16_t)(a), \
diff --git a/vp9/encoder/x86/vp9_dct32x32_sse2_impl.h b/vpx_dsp/x86/fwd_dct32x32_impl_sse2.h
similarity index 98%
rename from vp9/encoder/x86/vp9_dct32x32_sse2_impl.h
rename to vpx_dsp/x86/fwd_dct32x32_impl_sse2.h
index f0707ea..b85ae10 100644
--- a/vp9/encoder/x86/vp9_dct32x32_sse2_impl.h
+++ b/vpx_dsp/x86/fwd_dct32x32_impl_sse2.h
@@ -10,43 +10,45 @@
#include <emmintrin.h> // SSE2
-#include "vp9/encoder/vp9_dct.h"
+#include "vpx_dsp/fwd_txfm.h"
#include "vpx_dsp/txfm_common.h"
#include "vpx_dsp/x86/txfm_common_sse2.h"
-#include "vpx_ports/mem.h"
+// TODO(jingning) The high bit-depth version needs re-work for performance.
+// The current SSE2 implementation also causes cross reference to the static
+// functions in the C implementation file.
#if DCT_HIGH_BIT_DEPTH
#define ADD_EPI16 _mm_adds_epi16
#define SUB_EPI16 _mm_subs_epi16
#if FDCT32x32_HIGH_PRECISION
-void vp9_fdct32x32_rows_c(const int16_t *intermediate, tran_low_t *out) {
+void vpx_fdct32x32_rows_c(const int16_t *intermediate, tran_low_t *out) {
int i, j;
for (i = 0; i < 32; ++i) {
tran_high_t temp_in[32], temp_out[32];
for (j = 0; j < 32; ++j)
temp_in[j] = intermediate[j * 32 + i];
- vp9_fdct32(temp_in, temp_out, 0);
+ vpx_fdct32(temp_in, temp_out, 0);
for (j = 0; j < 32; ++j)
out[j + i * 32] =
(tran_low_t)((temp_out[j] + 1 + (temp_out[j] < 0)) >> 2);
}
}
- #define HIGH_FDCT32x32_2D_C vp9_highbd_fdct32x32_c
- #define HIGH_FDCT32x32_2D_ROWS_C vp9_fdct32x32_rows_c
+ #define HIGH_FDCT32x32_2D_C vpx_highbd_fdct32x32_c
+ #define HIGH_FDCT32x32_2D_ROWS_C vpx_fdct32x32_rows_c
#else
-void vp9_fdct32x32_rd_rows_c(const int16_t *intermediate, tran_low_t *out) {
+void vpx_fdct32x32_rd_rows_c(const int16_t *intermediate, tran_low_t *out) {
int i, j;
for (i = 0; i < 32; ++i) {
tran_high_t temp_in[32], temp_out[32];
for (j = 0; j < 32; ++j)
temp_in[j] = intermediate[j * 32 + i];
- vp9_fdct32(temp_in, temp_out, 1);
+ vpx_fdct32(temp_in, temp_out, 1);
for (j = 0; j < 32; ++j)
out[j + i * 32] = (tran_low_t)temp_out[j];
}
}
- #define HIGH_FDCT32x32_2D_C vp9_highbd_fdct32x32_rd_c
- #define HIGH_FDCT32x32_2D_ROWS_C vp9_fdct32x32_rd_rows_c
+ #define HIGH_FDCT32x32_2D_C vpx_highbd_fdct32x32_rd_c
+ #define HIGH_FDCT32x32_2D_ROWS_C vpx_fdct32x32_rd_rows_c
#endif // FDCT32x32_HIGH_PRECISION
#else
#define ADD_EPI16 _mm_add_epi16
diff --git a/vp9/encoder/x86/vp9_dct_avx2.c b/vpx_dsp/x86/fwd_txfm_avx2.c
similarity index 64%
rename from vp9/encoder/x86/vp9_dct_avx2.c
rename to vpx_dsp/x86/fwd_txfm_avx2.c
index 8f3b61a..6d9da6a 100644
--- a/vp9/encoder/x86/vp9_dct_avx2.c
+++ b/vpx_dsp/x86/fwd_txfm_avx2.c
@@ -8,19 +8,16 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-#include <immintrin.h> // AVX2
-#include "vp9/common/vp9_idct.h" // for cospi constants
-#include "vpx_ports/mem.h"
+#include "./vpx_config.h"
-
-#define FDCT32x32_2D_AVX2 vp9_fdct32x32_rd_avx2
+#define FDCT32x32_2D_AVX2 vpx_fdct32x32_rd_avx2
#define FDCT32x32_HIGH_PRECISION 0
-#include "vp9/encoder/x86/vp9_dct32x32_avx2_impl.h"
+#include "vpx_dsp/x86/fwd_dct32x32_impl_avx2.h"
#undef FDCT32x32_2D_AVX2
#undef FDCT32x32_HIGH_PRECISION
-#define FDCT32x32_2D_AVX2 vp9_fdct32x32_avx2
+#define FDCT32x32_2D_AVX2 vpx_fdct32x32_avx2
#define FDCT32x32_HIGH_PRECISION 1
-#include "vp9/encoder/x86/vp9_dct32x32_avx2_impl.h" // NOLINT
+#include "vpx_dsp/x86/fwd_dct32x32_impl_avx2.h" // NOLINT
#undef FDCT32x32_2D_AVX2
#undef FDCT32x32_HIGH_PRECISION
diff --git a/vpx_dsp/x86/fwd_txfm_impl_sse2.h b/vpx_dsp/x86/fwd_txfm_impl_sse2.h
index 1f6b302..69889e2 100644
--- a/vpx_dsp/x86/fwd_txfm_impl_sse2.h
+++ b/vpx_dsp/x86/fwd_txfm_impl_sse2.h
@@ -106,7 +106,7 @@
_mm_cmplt_epi16(in1, _mm_set1_epi16(0xfc00)));
test = _mm_movemask_epi8(_mm_or_si128(cmp0, cmp1));
if (test) {
- vp9_highbd_fdct4x4_c(input, output, stride);
+ vpx_highbd_fdct4x4_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -177,7 +177,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x2(&x0, &x1);
if (overflow) {
- vp9_highbd_fdct4x4_c(input, output, stride);
+ vpx_highbd_fdct4x4_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -199,7 +199,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x2(&t0, &t1);
if (overflow) {
- vp9_highbd_fdct4x4_c(input, output, stride);
+ vpx_highbd_fdct4x4_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -238,7 +238,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x2(&x0, &x1);
if (overflow) {
- vp9_highbd_fdct4x4_c(input, output, stride);
+ vpx_highbd_fdct4x4_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -322,7 +322,7 @@
overflow = check_epi16_overflow_x8(&q0, &q1, &q2, &q3,
&q4, &q5, &q6, &q7);
if (overflow) {
- vp9_highbd_fdct8x8_c(input, output, stride);
+ vpx_highbd_fdct8x8_c(input, output, stride);
return;
}
}
@@ -337,7 +337,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x4(&r0, &r1, &r2, &r3);
if (overflow) {
- vp9_highbd_fdct8x8_c(input, output, stride);
+ vpx_highbd_fdct8x8_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -380,7 +380,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x4(&res0, &res4, &res2, &res6);
if (overflow) {
- vp9_highbd_fdct8x8_c(input, output, stride);
+ vpx_highbd_fdct8x8_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -410,7 +410,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x2(&r0, &r1);
if (overflow) {
- vp9_highbd_fdct8x8_c(input, output, stride);
+ vpx_highbd_fdct8x8_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -423,7 +423,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x4(&x0, &x1, &x2, &x3);
if (overflow) {
- vp9_highbd_fdct8x8_c(input, output, stride);
+ vpx_highbd_fdct8x8_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -466,7 +466,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x4(&res1, &res7, &res5, &res3);
if (overflow) {
- vp9_highbd_fdct8x8_c(input, output, stride);
+ vpx_highbd_fdct8x8_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -729,7 +729,7 @@
overflow = check_epi16_overflow_x8(&input0, &input1, &input2, &input3,
&input4, &input5, &input6, &input7);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -750,7 +750,7 @@
&step1_4, &step1_5,
&step1_6, &step1_7);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -770,7 +770,7 @@
overflow = check_epi16_overflow_x8(&q0, &q1, &q2, &q3,
&q4, &q5, &q6, &q7);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -784,7 +784,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x4(&r0, &r1, &r2, &r3);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -806,7 +806,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x4(&res00, &res08, &res04, &res12);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -827,7 +827,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x2(&r0, &r1);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -840,7 +840,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x4(&x0, &x1, &x2, &x3);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -863,7 +863,7 @@
overflow = check_epi16_overflow_x4(&res02, &res14,
&res10, &res06);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -891,7 +891,7 @@
overflow = check_epi16_overflow_x4(&step2_2, &step2_3, &step2_5,
&step2_4);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -912,7 +912,7 @@
&step3_4, &step3_5,
&step3_6, &step3_7);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -935,7 +935,7 @@
overflow = check_epi16_overflow_x4(&step2_1, &step2_2, &step2_6,
&step2_5);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -956,7 +956,7 @@
&step1_4, &step1_5,
&step1_6, &step1_7);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -978,7 +978,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x4(&res01, &res09, &res15, &res07);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
@@ -999,7 +999,7 @@
#if DCT_HIGH_BIT_DEPTH
overflow = check_epi16_overflow_x4(&res05, &res13, &res11, &res03);
if (overflow) {
- vp9_highbd_fdct16x16_c(input, output, stride);
+ vpx_highbd_fdct16x16_c(input, output, stride);
return;
}
#endif // DCT_HIGH_BIT_DEPTH
diff --git a/vpx_dsp/x86/fwd_txfm_sse2.c b/vpx_dsp/x86/fwd_txfm_sse2.c
index a868d22..bca72e8 100644
--- a/vpx_dsp/x86/fwd_txfm_sse2.c
+++ b/vpx_dsp/x86/fwd_txfm_sse2.c
@@ -8,27 +8,264 @@
* be found in the AUTHORS file in the root of the source tree.
*/
+#include <emmintrin.h> // SSE2
+
#include "./vpx_config.h"
+#include "vpx_dsp/vpx_dsp_common.h"
+#include "vpx_dsp/x86/fwd_txfm_sse2.h"
+
+void vpx_fdct4x4_1_sse2(const int16_t *input, tran_low_t *output, int stride) {
+ __m128i in0, in1;
+ __m128i tmp;
+ const __m128i zero = _mm_setzero_si128();
+ in0 = _mm_loadl_epi64((const __m128i *)(input + 0 * stride));
+ in1 = _mm_loadl_epi64((const __m128i *)(input + 1 * stride));
+ in1 = _mm_unpacklo_epi64(in1, _mm_loadl_epi64((const __m128i *)
+ (input + 2 * stride)));
+ in0 = _mm_unpacklo_epi64(in0, _mm_loadl_epi64((const __m128i *)
+ (input + 3 * stride)));
+
+ tmp = _mm_add_epi16(in0, in1);
+ in0 = _mm_unpacklo_epi16(zero, tmp);
+ in1 = _mm_unpackhi_epi16(zero, tmp);
+ in0 = _mm_srai_epi32(in0, 16);
+ in1 = _mm_srai_epi32(in1, 16);
+
+ tmp = _mm_add_epi32(in0, in1);
+ in0 = _mm_unpacklo_epi32(tmp, zero);
+ in1 = _mm_unpackhi_epi32(tmp, zero);
+
+ tmp = _mm_add_epi32(in0, in1);
+ in0 = _mm_srli_si128(tmp, 8);
+
+ in1 = _mm_add_epi32(tmp, in0);
+ in0 = _mm_slli_epi32(in1, 1);
+ store_output(&in0, output);
+}
+
+void vpx_fdct8x8_1_sse2(const int16_t *input, tran_low_t *output, int stride) {
+ __m128i in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
+ __m128i in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
+ __m128i in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
+ __m128i in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
+ __m128i u0, u1, sum;
+
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
+
+ in0 = _mm_load_si128((const __m128i *)(input + 4 * stride));
+ in1 = _mm_load_si128((const __m128i *)(input + 5 * stride));
+ in2 = _mm_load_si128((const __m128i *)(input + 6 * stride));
+ in3 = _mm_load_si128((const __m128i *)(input + 7 * stride));
+
+ sum = _mm_add_epi16(u0, u1);
+
+ in0 = _mm_add_epi16(in0, in1);
+ in2 = _mm_add_epi16(in2, in3);
+ sum = _mm_add_epi16(sum, in0);
+
+ u0 = _mm_setzero_si128();
+ sum = _mm_add_epi16(sum, in2);
+
+ in0 = _mm_unpacklo_epi16(u0, sum);
+ in1 = _mm_unpackhi_epi16(u0, sum);
+ in0 = _mm_srai_epi32(in0, 16);
+ in1 = _mm_srai_epi32(in1, 16);
+
+ sum = _mm_add_epi32(in0, in1);
+ in0 = _mm_unpacklo_epi32(sum, u0);
+ in1 = _mm_unpackhi_epi32(sum, u0);
+
+ sum = _mm_add_epi32(in0, in1);
+ in0 = _mm_srli_si128(sum, 8);
+
+ in1 = _mm_add_epi32(sum, in0);
+ store_output(&in1, output);
+}
+
+void vpx_fdct16x16_1_sse2(const int16_t *input, tran_low_t *output,
+ int stride) {
+ __m128i in0, in1, in2, in3;
+ __m128i u0, u1;
+ __m128i sum = _mm_setzero_si128();
+ int i;
+
+ for (i = 0; i < 2; ++i) {
+ input += 8 * i;
+ in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
+ in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
+ in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
+ in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
+
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
+ sum = _mm_add_epi16(sum, u0);
+
+ in0 = _mm_load_si128((const __m128i *)(input + 4 * stride));
+ in1 = _mm_load_si128((const __m128i *)(input + 5 * stride));
+ in2 = _mm_load_si128((const __m128i *)(input + 6 * stride));
+ in3 = _mm_load_si128((const __m128i *)(input + 7 * stride));
+
+ sum = _mm_add_epi16(sum, u1);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
+ sum = _mm_add_epi16(sum, u0);
+
+ in0 = _mm_load_si128((const __m128i *)(input + 8 * stride));
+ in1 = _mm_load_si128((const __m128i *)(input + 9 * stride));
+ in2 = _mm_load_si128((const __m128i *)(input + 10 * stride));
+ in3 = _mm_load_si128((const __m128i *)(input + 11 * stride));
+
+ sum = _mm_add_epi16(sum, u1);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
+ sum = _mm_add_epi16(sum, u0);
+
+ in0 = _mm_load_si128((const __m128i *)(input + 12 * stride));
+ in1 = _mm_load_si128((const __m128i *)(input + 13 * stride));
+ in2 = _mm_load_si128((const __m128i *)(input + 14 * stride));
+ in3 = _mm_load_si128((const __m128i *)(input + 15 * stride));
+
+ sum = _mm_add_epi16(sum, u1);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
+ sum = _mm_add_epi16(sum, u0);
+
+ sum = _mm_add_epi16(sum, u1);
+ }
+
+ u0 = _mm_setzero_si128();
+ in0 = _mm_unpacklo_epi16(u0, sum);
+ in1 = _mm_unpackhi_epi16(u0, sum);
+ in0 = _mm_srai_epi32(in0, 16);
+ in1 = _mm_srai_epi32(in1, 16);
+
+ sum = _mm_add_epi32(in0, in1);
+ in0 = _mm_unpacklo_epi32(sum, u0);
+ in1 = _mm_unpackhi_epi32(sum, u0);
+
+ sum = _mm_add_epi32(in0, in1);
+ in0 = _mm_srli_si128(sum, 8);
+
+ in1 = _mm_add_epi32(sum, in0);
+ in1 = _mm_srai_epi32(in1, 1);
+ store_output(&in1, output);
+}
+
+void vpx_fdct32x32_1_sse2(const int16_t *input, tran_low_t *output,
+ int stride) {
+ __m128i in0, in1, in2, in3;
+ __m128i u0, u1;
+ __m128i sum = _mm_setzero_si128();
+ int i;
+
+ for (i = 0; i < 8; ++i) {
+ in0 = _mm_load_si128((const __m128i *)(input + 0));
+ in1 = _mm_load_si128((const __m128i *)(input + 8));
+ in2 = _mm_load_si128((const __m128i *)(input + 16));
+ in3 = _mm_load_si128((const __m128i *)(input + 24));
+
+ input += stride;
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
+ sum = _mm_add_epi16(sum, u0);
+
+ in0 = _mm_load_si128((const __m128i *)(input + 0));
+ in1 = _mm_load_si128((const __m128i *)(input + 8));
+ in2 = _mm_load_si128((const __m128i *)(input + 16));
+ in3 = _mm_load_si128((const __m128i *)(input + 24));
+
+ input += stride;
+ sum = _mm_add_epi16(sum, u1);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
+ sum = _mm_add_epi16(sum, u0);
+
+ in0 = _mm_load_si128((const __m128i *)(input + 0));
+ in1 = _mm_load_si128((const __m128i *)(input + 8));
+ in2 = _mm_load_si128((const __m128i *)(input + 16));
+ in3 = _mm_load_si128((const __m128i *)(input + 24));
+
+ input += stride;
+ sum = _mm_add_epi16(sum, u1);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
+ sum = _mm_add_epi16(sum, u0);
+
+ in0 = _mm_load_si128((const __m128i *)(input + 0));
+ in1 = _mm_load_si128((const __m128i *)(input + 8));
+ in2 = _mm_load_si128((const __m128i *)(input + 16));
+ in3 = _mm_load_si128((const __m128i *)(input + 24));
+
+ input += stride;
+ sum = _mm_add_epi16(sum, u1);
+ u0 = _mm_add_epi16(in0, in1);
+ u1 = _mm_add_epi16(in2, in3);
+ sum = _mm_add_epi16(sum, u0);
+
+ sum = _mm_add_epi16(sum, u1);
+ }
+
+ u0 = _mm_setzero_si128();
+ in0 = _mm_unpacklo_epi16(u0, sum);
+ in1 = _mm_unpackhi_epi16(u0, sum);
+ in0 = _mm_srai_epi32(in0, 16);
+ in1 = _mm_srai_epi32(in1, 16);
+
+ sum = _mm_add_epi32(in0, in1);
+ in0 = _mm_unpacklo_epi32(sum, u0);
+ in1 = _mm_unpackhi_epi32(sum, u0);
+
+ sum = _mm_add_epi32(in0, in1);
+ in0 = _mm_srli_si128(sum, 8);
+
+ in1 = _mm_add_epi32(sum, in0);
+ in1 = _mm_srai_epi32(in1, 3);
+ store_output(&in1, output);
+}
#define DCT_HIGH_BIT_DEPTH 0
-
-#define FDCT4x4_2D vp9_fdct4x4_sse2
-#define FDCT8x8_2D vp9_fdct8x8_sse2
-#define FDCT16x16_2D vp9_fdct16x16_sse2
+#define FDCT4x4_2D vpx_fdct4x4_sse2
+#define FDCT8x8_2D vpx_fdct8x8_sse2
+#define FDCT16x16_2D vpx_fdct16x16_sse2
#include "vpx_dsp/x86/fwd_txfm_impl_sse2.h"
#undef FDCT4x4_2D
#undef FDCT8x8_2D
#undef FDCT16x16_2D
+
+#define FDCT32x32_2D vpx_fdct32x32_rd_sse2
+#define FDCT32x32_HIGH_PRECISION 0
+#include "vpx_dsp/x86/fwd_dct32x32_impl_sse2.h"
+#undef FDCT32x32_2D
+#undef FDCT32x32_HIGH_PRECISION
+
+#define FDCT32x32_2D vpx_fdct32x32_sse2
+#define FDCT32x32_HIGH_PRECISION 1
+#include "vpx_dsp/x86/fwd_dct32x32_impl_sse2.h" // NOLINT
+#undef FDCT32x32_2D
+#undef FDCT32x32_HIGH_PRECISION
#undef DCT_HIGH_BIT_DEPTH
#if CONFIG_VP9_HIGHBITDEPTH
#define DCT_HIGH_BIT_DEPTH 1
-#define FDCT4x4_2D vp9_highbd_fdct4x4_sse2
-#define FDCT8x8_2D vp9_highbd_fdct8x8_sse2
-#define FDCT16x16_2D vp9_highbd_fdct16x16_sse2
+#define FDCT4x4_2D vpx_highbd_fdct4x4_sse2
+#define FDCT8x8_2D vpx_highbd_fdct8x8_sse2
+#define FDCT16x16_2D vpx_highbd_fdct16x16_sse2
#include "vpx_dsp/x86/fwd_txfm_impl_sse2.h" // NOLINT
#undef FDCT4x4_2D
#undef FDCT8x8_2D
#undef FDCT16x16_2D
+
+#define FDCT32x32_2D vpx_highbd_fdct32x32_rd_sse2
+#define FDCT32x32_HIGH_PRECISION 0
+#include "vpx_dsp/x86/fwd_dct32x32_impl_sse2.h" // NOLINT
+#undef FDCT32x32_2D
+#undef FDCT32x32_HIGH_PRECISION
+
+#define FDCT32x32_2D vpx_highbd_fdct32x32_sse2
+#define FDCT32x32_HIGH_PRECISION 1
+#include "vpx_dsp/x86/fwd_dct32x32_impl_sse2.h" // NOLINT
+#undef FDCT32x32_2D
+#undef FDCT32x32_HIGH_PRECISION
#undef DCT_HIGH_BIT_DEPTH
#endif // CONFIG_VP9_HIGHBITDEPTH
diff --git a/vpx_dsp/x86/fwd_txfm_ssse3_x86_64.asm b/vpx_dsp/x86/fwd_txfm_ssse3_x86_64.asm
index 5f63546..0fa8ea1 100644
--- a/vpx_dsp/x86/fwd_txfm_ssse3_x86_64.asm
+++ b/vpx_dsp/x86/fwd_txfm_ssse3_x86_64.asm
@@ -7,6 +7,9 @@
; in the file PATENTS. All contributing project authors may
; be found in the AUTHORS file in the root of the source tree.
;
+
+%define program_name vpx
+
%include "third_party/x86inc/x86inc.asm"
; This file provides SSSE3 version of the forward transformation. Part
diff --git a/vp9/common/x86/vp9_high_intrapred_sse2.asm b/vpx_dsp/x86/highbd_intrapred_sse2.asm
similarity index 99%
rename from vp9/common/x86/vp9_high_intrapred_sse2.asm
rename to vpx_dsp/x86/highbd_intrapred_sse2.asm
index b12d29c..5ad1768 100644
--- a/vp9/common/x86/vp9_high_intrapred_sse2.asm
+++ b/vpx_dsp/x86/highbd_intrapred_sse2.asm
@@ -8,6 +8,8 @@
; be found in the AUTHORS file in the root of the source tree.
;
+%define program_name vpx
+
%include "third_party/x86inc/x86inc.asm"
SECTION_RODATA
diff --git a/vpx_dsp/x86/highbd_subpel_variance_impl_sse2.asm b/vpx_dsp/x86/highbd_subpel_variance_impl_sse2.asm
index 2be6446..f558770 100644
--- a/vpx_dsp/x86/highbd_subpel_variance_impl_sse2.asm
+++ b/vpx_dsp/x86/highbd_subpel_variance_impl_sse2.asm
@@ -111,7 +111,7 @@
cglobal highbd_sub_pixel_variance%1xh, 7, 8, 13, src, src_stride, x_offset, \
y_offset, dst, dst_stride, height, sse
%endif
- %define h heightd
+ %define block_height heightd
%define bilin_filter sseq
%else
%if ARCH_X86=1 && CONFIG_PIC=1
@@ -121,7 +121,7 @@
dst, dst_stride, \
sec, sec_stride, \
height, sse, g_bilin_filter, g_pw_8
- %define h dword heightm
+ %define block_height dword heightm
%define sec_str sec_stridemp
; Store bilin_filter and pw_8 location in stack
@@ -139,7 +139,7 @@
cglobal highbd_sub_pixel_variance%1xh, 7, 7, 13, src, src_stride, \
x_offset, y_offset, dst, dst_stride, height, \
sse, g_bilin_filter, g_pw_8
- %define h heightd
+ %define block_height heightd
; Store bilin_filter and pw_8 location in stack
GET_GOT eax
@@ -162,16 +162,16 @@
sec, sec_stride, \
height, sse
%if ARCH_X86_64
- %define h heightd
+ %define block_height heightd
%define sec_str sec_strideq
%else
- %define h dword heightm
+ %define block_height dword heightm
%define sec_str sec_stridemp
%endif
%else
cglobal highbd_sub_pixel_variance%1xh, 7, 7, 13, src, src_stride, \
x_offset, y_offset, dst, dst_stride, height, sse
- %define h heightd
+ %define block_height heightd
%endif
%define bilin_filter bilin_filter_m
@@ -183,7 +183,7 @@
pxor m7, m7 ; sse
%if %1 < 16
- sar h, 1
+ sar block_height, 1
%endif
%if %2 == 1 ; avg
shl sec_str, 1
@@ -232,7 +232,7 @@
add secq, sec_str
%endif
%endif
- dec h
+ dec block_height
jg .x_zero_y_zero_loop
STORE_AND_RET
@@ -283,7 +283,7 @@
add secq, sec_str
%endif
%endif
- dec h
+ dec block_height
jg .x_zero_y_half_loop
STORE_AND_RET
@@ -381,7 +381,7 @@
add secq, sec_str
%endif
%endif
- dec h
+ dec block_height
jg .x_zero_y_other_loop
%undef filter_y_a
%undef filter_y_b
@@ -439,7 +439,7 @@
add secq, sec_str
%endif
%endif
- dec h
+ dec block_height
jg .x_half_y_zero_loop
STORE_AND_RET
@@ -510,7 +510,7 @@
add secq, sec_str
%endif
%endif
- dec h
+ dec block_height
jg .x_half_y_half_loop
STORE_AND_RET
@@ -626,7 +626,7 @@
add secq, sec_str
%endif
%endif
- dec h
+ dec block_height
jg .x_half_y_other_loop
%undef filter_y_a
%undef filter_y_b
@@ -725,7 +725,7 @@
add secq, sec_str
%endif
%endif
- dec h
+ dec block_height
jg .x_other_y_zero_loop
%undef filter_x_a
%undef filter_x_b
@@ -855,7 +855,7 @@
add secq, sec_str
%endif
%endif
- dec h
+ dec block_height
jg .x_other_y_half_loop
%undef filter_x_a
%undef filter_x_b
@@ -1022,7 +1022,7 @@
add secq, sec_str
%endif
%endif
- dec h
+ dec block_height
jg .x_other_y_other_loop
%undef filter_x_a
%undef filter_x_b
diff --git a/vp9/common/x86/vp9_intrapred_sse2.asm b/vpx_dsp/x86/intrapred_sse2.asm
similarity index 99%
rename from vp9/common/x86/vp9_intrapred_sse2.asm
rename to vpx_dsp/x86/intrapred_sse2.asm
index 22b5731..b15f6fb 100644
--- a/vp9/common/x86/vp9_intrapred_sse2.asm
+++ b/vpx_dsp/x86/intrapred_sse2.asm
@@ -8,6 +8,8 @@
; be found in the AUTHORS file in the root of the source tree.
;
+%define program_name vpx
+
%include "third_party/x86inc/x86inc.asm"
SECTION_RODATA
diff --git a/vp9/common/x86/vp9_intrapred_ssse3.asm b/vpx_dsp/x86/intrapred_ssse3.asm
similarity index 99%
rename from vp9/common/x86/vp9_intrapred_ssse3.asm
rename to vpx_dsp/x86/intrapred_ssse3.asm
index 88df9b2..fa53126 100644
--- a/vp9/common/x86/vp9_intrapred_ssse3.asm
+++ b/vpx_dsp/x86/intrapred_ssse3.asm
@@ -8,6 +8,8 @@
; be found in the AUTHORS file in the root of the source tree.
;
+%define program_name vpx
+
%include "third_party/x86inc/x86inc.asm"
SECTION_RODATA
diff --git a/vpx_dsp/x86/subpel_variance_sse2.asm b/vpx_dsp/x86/subpel_variance_sse2.asm
index 294f54f..0046ce8 100644
--- a/vpx_dsp/x86/subpel_variance_sse2.asm
+++ b/vpx_dsp/x86/subpel_variance_sse2.asm
@@ -127,7 +127,7 @@
cglobal sub_pixel_variance%1xh, 7, 8, 13, src, src_stride, x_offset, \
y_offset, dst, dst_stride, height, sse
%endif
- %define h heightd
+ %define block_height heightd
%define bilin_filter sseq
%else
%if ARCH_X86=1 && CONFIG_PIC=1
@@ -137,7 +137,7 @@
dst, dst_stride, \
sec, sec_stride, \
height, sse, g_bilin_filter, g_pw_8
- %define h dword heightm
+ %define block_height dword heightm
%define sec_str sec_stridemp
;Store bilin_filter and pw_8 location in stack
@@ -155,7 +155,7 @@
cglobal sub_pixel_variance%1xh, 7, 7, 13, src, src_stride, x_offset, \
y_offset, dst, dst_stride, height, sse, \
g_bilin_filter, g_pw_8
- %define h heightd
+ %define block_height heightd
;Store bilin_filter and pw_8 location in stack
GET_GOT eax
@@ -178,16 +178,16 @@
sec, sec_stride, \
height, sse
%if ARCH_X86_64
- %define h heightd
+ %define block_height heightd
%define sec_str sec_strideq
%else
- %define h dword heightm
+ %define block_height dword heightm
%define sec_str sec_stridemp
%endif
%else
cglobal sub_pixel_variance%1xh, 7, 7, 13, src, src_stride, x_offset, \
y_offset, dst, dst_stride, height, sse
- %define h heightd
+ %define block_height heightd
%endif
%define bilin_filter bilin_filter_m
@@ -201,7 +201,7 @@
; could perhaps use it for something more productive then
pxor m5, m5 ; dedicated zero register
%if %1 < 16
- sar h, 1
+ sar block_height, 1
%if %2 == 1 ; avg
shl sec_str, 1
%endif
@@ -267,7 +267,7 @@
%if %2 == 1 ; avg
add secq, sec_str
%endif
- dec h
+ dec block_height
jg .x_zero_y_zero_loop
STORE_AND_RET
@@ -339,7 +339,7 @@
%if %2 == 1 ; avg
add secq, sec_str
%endif
- dec h
+ dec block_height
jg .x_zero_y_half_loop
STORE_AND_RET
@@ -468,7 +468,7 @@
%if %2 == 1 ; avg
add secq, sec_str
%endif
- dec h
+ dec block_height
jg .x_zero_y_other_loop
%undef filter_y_a
%undef filter_y_b
@@ -539,7 +539,7 @@
%if %2 == 1 ; avg
add secq, sec_str
%endif
- dec h
+ dec block_height
jg .x_half_y_zero_loop
STORE_AND_RET
@@ -637,7 +637,7 @@
%if %2 == 1 ; avg
add secq, sec_str
%endif
- dec h
+ dec block_height
jg .x_half_y_half_loop
STORE_AND_RET
@@ -779,7 +779,7 @@
%if %2 == 1 ; avg
add secq, sec_str
%endif
- dec h
+ dec block_height
jg .x_half_y_other_loop
%undef filter_y_a
%undef filter_y_b
@@ -911,7 +911,7 @@
%if %2 == 1 ; avg
add secq, sec_str
%endif
- dec h
+ dec block_height
jg .x_other_y_zero_loop
%undef filter_x_a
%undef filter_x_b
@@ -1103,7 +1103,7 @@
%if %2 == 1 ; avg
add secq, sec_str
%endif
- dec h
+ dec block_height
jg .x_other_y_half_loop
%undef filter_x_a
%undef filter_x_b
@@ -1357,7 +1357,7 @@
%if %2 == 1 ; avg
add secq, sec_str
%endif
- dec h
+ dec block_height
jg .x_other_y_other_loop
%undef filter_x_a
%undef filter_x_b
