Exclude hbd functions in av1_rtcd_defs.pl from HBD flag.
~1% binary size reduction.
BUG=aomedia:2397
Change-Id: I751a901f902faa907224a962fee3e794ef69113f
diff --git a/av1/av1.cmake b/av1/av1.cmake
index 647d0b0..56c169bf 100644
--- a/av1/av1.cmake
+++ b/av1/av1.cmake
@@ -272,7 +272,8 @@
if(NOT CONFIG_AV1_HIGHBITDEPTH)
list(REMOVE_ITEM AOM_AV1_COMMON_INTRIN_SSE4_1
- "${AOM_ROOT}/av1/common/x86/highbd_convolve_2d_sse4.c")
+ "${AOM_ROOT}/av1/common/x86/highbd_convolve_2d_sse4.c"
+ "${AOM_ROOT}/av1/common/x86/highbd_warp_plane_sse4.c")
endif()
list(APPEND AOM_AV1_COMMON_INTRIN_AVX2
@@ -308,6 +309,12 @@
"${AOM_ROOT}/av1/encoder/x86/highbd_block_error_intrin_sse2.c"
"${AOM_ROOT}/av1/encoder/x86/wedge_utils_sse2.c")
+if(NOT CONFIG_AV1_HIGHBITDEPTH)
+ list(
+ REMOVE_ITEM AOM_AV1_ENCODER_INTRIN_SSE2
+ "${AOM_ROOT}/av1/encoder/x86/highbd_block_error_intrin_sse2.c")
+endif()
+
list(APPEND AOM_AV1_ENCODER_INTRIN_SSE3 "${AOM_ROOT}/av1/encoder/x86/ml_sse3.c")
list(APPEND AOM_AV1_ENCODER_ASM_SSSE3_X86_64
@@ -340,6 +347,12 @@
"${AOM_ROOT}/av1/encoder/x86/rdopt_avx2.c"
"${AOM_ROOT}/av1/encoder/x86/pickrst_avx2.c")
+if(NOT CONFIG_AV1_HIGHBITDEPTH)
+ list(
+ REMOVE_ITEM AOM_AV1_ENCODER_INTRIN_AVX2
+ "${AOM_ROOT}/av1/encoder/x86/highbd_block_error_intrin_avx2.c")
+endif()
+
list(APPEND AOM_AV1_ENCODER_INTRIN_NEON
"${AOM_ROOT}/av1/encoder/arm/neon/quantize_neon.c"
"${AOM_ROOT}/av1/encoder/arm/neon/av1_error_neon.c")
diff --git a/av1/common/arm/cfl_neon.c b/av1/common/arm/cfl_neon.c
index 39025b5..371be5f 100644
--- a/av1/common/arm/cfl_neon.c
+++ b/av1/common/arm/cfl_neon.c
@@ -131,6 +131,7 @@
} while ((pred_buf_q3 += CFL_BUF_LINE) < end);
}
+#if CONFIG_AV1_HIGHBITDEPTH
#ifndef __aarch64__
uint16x8_t vpaddq_u16(uint16x8_t a, uint16x8_t b) {
return vcombine_u16(vpadd_u16(vget_low_u16(a), vget_high_u16(a)),
@@ -247,6 +248,7 @@
input += input_stride;
} while ((pred_buf_q3 += CFL_BUF_LINE) < end);
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
CFL_GET_SUBSAMPLE_FUNCTION(neon)
@@ -511,6 +513,7 @@
CFL_PREDICT_FN(neon, lbd)
+#if CONFIG_AV1_HIGHBITDEPTH
static INLINE uint16x4_t clamp_s16(int16x4_t a, int16x4_t max) {
return vreinterpret_u16_s16(vmax_s16(vmin_s16(a, max), vdup_n_s16(0)));
}
@@ -582,3 +585,4 @@
}
CFL_PREDICT_FN(neon, hbd)
+#endif // CONFIG_AV1_HIGHBITDEPTH
diff --git a/av1/common/av1_rtcd_defs.pl b/av1/common/av1_rtcd_defs.pl
index 15bb5cc..792f171 100644
--- a/av1/common/av1_rtcd_defs.pl
+++ b/av1/common/av1_rtcd_defs.pl
@@ -60,16 +60,19 @@
typedef void (*cfl_subsample_lbd_fn)(const uint8_t *input, int input_stride,
uint16_t *output_q3);
+#if CONFIG_AV1_HIGHBITDEPTH
typedef void (*cfl_subsample_hbd_fn)(const uint16_t *input, int input_stride,
uint16_t *output_q3);
+typedef void (*cfl_predict_hbd_fn)(const int16_t *src, uint16_t *dst,
+ int dst_stride, int alpha_q3, int bd);
+#endif
+
typedef void (*cfl_subtract_average_fn)(const uint16_t *src, int16_t *dst);
typedef void (*cfl_predict_lbd_fn)(const int16_t *src, uint8_t *dst,
int dst_stride, int alpha_q3);
-typedef void (*cfl_predict_hbd_fn)(const int16_t *src, uint16_t *dst,
- int dst_stride, int alpha_q3, int bd);
EOF
}
forward_decls qw/av1_common_forward_decls/;
@@ -283,9 +286,10 @@
add_proto qw/void av1_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan, const qm_val_t * qm_ptr, const qm_val_t * iqm_ptr, int log_scale";
# ENCODEMB INVOKE
-
- add_proto qw/int64_t av1_highbd_block_error/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz, int bd";
- specialize qw/av1_highbd_block_error sse2 avx2/;
+ if (aom_config("CONFIG_AV1_HIGHBITDEPTH") eq "yes") {
+ add_proto qw/int64_t av1_highbd_block_error/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz, int bd";
+ specialize qw/av1_highbd_block_error sse2 avx2/;
+ }
if (aom_config("CONFIG_REALTIME_ONLY") ne "yes") {
add_proto qw/void av1_highbd_apply_temporal_filter/, "const uint8_t *yf, int y_stride, const uint8_t *yp, int y_buf_stride, const uint8_t *uf, const uint8_t *vf, int uv_stride, const uint8_t *up, const uint8_t *vp, int uv_buf_stride, unsigned int block_width, unsigned int block_height, int ss_x, int ss_y, int strength, const int *blk_fw, int use_32x32, uint32_t *y_accumulator, uint16_t *y_count, uint32_t *u_accumulator, uint16_t *u_count, uint32_t *v_accumulator, uint16_t *v_count";
@@ -321,15 +325,17 @@
add_proto qw/void av1_compute_stats/, "int wiener_win, const uint8_t *dgd8, const uint8_t *src8, int h_start, int h_end, int v_start, int v_end, int dgd_stride, int src_stride, int64_t *M, int64_t *H";
specialize qw/av1_compute_stats sse4_1 avx2/;
- add_proto qw/void av1_compute_stats_highbd/, "int wiener_win, const uint8_t *dgd8, const uint8_t *src8, int h_start, int h_end, int v_start, int v_end, int dgd_stride, int src_stride, int64_t *M, int64_t *H, aom_bit_depth_t bit_depth";
- specialize qw/av1_compute_stats_highbd sse4_1 avx2/;
-
+ if (aom_config("CONFIG_AV1_HIGHBITDEPTH") eq "yes") {
+ add_proto qw/void av1_compute_stats_highbd/, "int wiener_win, const uint8_t *dgd8, const uint8_t *src8, int h_start, int h_end, int v_start, int v_end, int dgd_stride, int src_stride, int64_t *M, int64_t *H, aom_bit_depth_t bit_depth";
+ specialize qw/av1_compute_stats_highbd sse4_1 avx2/;
+ }
add_proto qw/int64_t av1_lowbd_pixel_proj_error/, " const uint8_t *src8, int width, int height, int src_stride, const uint8_t *dat8, int dat_stride, int32_t *flt0, int flt0_stride, int32_t *flt1, int flt1_stride, int xq[2], const sgr_params_type *params";
specialize qw/av1_lowbd_pixel_proj_error sse4_1 avx2/;
- add_proto qw/int64_t av1_highbd_pixel_proj_error/, " const uint8_t *src8, int width, int height, int src_stride, const uint8_t *dat8, int dat_stride, int32_t *flt0, int flt0_stride, int32_t *flt1, int flt1_stride, int xq[2], const sgr_params_type *params";
- specialize qw/av1_highbd_pixel_proj_error sse4_1 avx2/;
-
+ if (aom_config("CONFIG_AV1_HIGHBITDEPTH") eq "yes") {
+ add_proto qw/int64_t av1_highbd_pixel_proj_error/, " const uint8_t *src8, int width, int height, int src_stride, const uint8_t *dat8, int dat_stride, int32_t *flt0, int flt0_stride, int32_t *flt1, int flt1_stride, int xq[2], const sgr_params_type *params";
+ specialize qw/av1_highbd_pixel_proj_error sse4_1 avx2/;
+ }
add_proto qw/void av1_get_horver_correlation_full/, " const int16_t *diff, int stride, int w, int h, float *hcorr, float *vcorr";
specialize qw/av1_get_horver_correlation_full sse4_1 avx2/;
@@ -370,8 +376,10 @@
add_proto qw/void av1_warp_affine/, "const int32_t *mat, const uint8_t *ref, int width, int height, int stride, uint8_t *pred, int p_col, int p_row, int p_width, int p_height, int p_stride, int subsampling_x, int subsampling_y, ConvolveParams *conv_params, int16_t alpha, int16_t beta, int16_t gamma, int16_t delta";
specialize qw/av1_warp_affine sse4_1 avx2 neon/;
-add_proto qw/void av1_highbd_warp_affine/, "const int32_t *mat, const uint16_t *ref, int width, int height, int stride, uint16_t *pred, int p_col, int p_row, int p_width, int p_height, int p_stride, int subsampling_x, int subsampling_y, int bd, ConvolveParams *conv_params, int16_t alpha, int16_t beta, int16_t gamma, int16_t delta";
-specialize qw/av1_highbd_warp_affine sse4_1/;
+if (aom_config("CONFIG_AV1_HIGHBITDEPTH") eq "yes") {
+ add_proto qw/void av1_highbd_warp_affine/, "const int32_t *mat, const uint16_t *ref, int width, int height, int stride, uint16_t *pred, int p_col, int p_row, int p_width, int p_height, int p_stride, int subsampling_x, int subsampling_y, int bd, ConvolveParams *conv_params, int16_t alpha, int16_t beta, int16_t gamma, int16_t delta";
+ specialize qw/av1_highbd_warp_affine sse4_1/;
+}
add_proto qw/int64_t av1_calc_frame_error/, "const uint8_t *const ref, int stride, const uint8_t *const dst, int p_width, int p_height, int p_stride";
specialize qw/av1_calc_frame_error sse2 avx2/;
@@ -460,19 +468,21 @@
add_proto qw/cfl_subsample_lbd_fn cfl_get_luma_subsampling_444_lbd/, "TX_SIZE tx_size";
specialize qw/cfl_get_luma_subsampling_444_lbd ssse3 avx2 neon/;
-add_proto qw/cfl_subsample_hbd_fn cfl_get_luma_subsampling_420_hbd/, "TX_SIZE tx_size";
-specialize qw/cfl_get_luma_subsampling_420_hbd ssse3 avx2 neon/;
+if (aom_config("CONFIG_AV1_HIGHBITDEPTH") eq "yes") {
+ add_proto qw/cfl_subsample_hbd_fn cfl_get_luma_subsampling_420_hbd/, "TX_SIZE tx_size";
+ specialize qw/cfl_get_luma_subsampling_420_hbd ssse3 avx2 neon/;
-add_proto qw/cfl_subsample_hbd_fn cfl_get_luma_subsampling_422_hbd/, "TX_SIZE tx_size";
-specialize qw/cfl_get_luma_subsampling_422_hbd ssse3 avx2 neon/;
+ add_proto qw/cfl_subsample_hbd_fn cfl_get_luma_subsampling_422_hbd/, "TX_SIZE tx_size";
+ specialize qw/cfl_get_luma_subsampling_422_hbd ssse3 avx2 neon/;
-add_proto qw/cfl_subsample_hbd_fn cfl_get_luma_subsampling_444_hbd/, "TX_SIZE tx_size";
-specialize qw/cfl_get_luma_subsampling_444_hbd ssse3 avx2 neon/;
+ add_proto qw/cfl_subsample_hbd_fn cfl_get_luma_subsampling_444_hbd/, "TX_SIZE tx_size";
+ specialize qw/cfl_get_luma_subsampling_444_hbd ssse3 avx2 neon/;
+
+ add_proto qw/cfl_predict_hbd_fn cfl_get_predict_hbd_fn/, "TX_SIZE tx_size";
+ specialize qw/cfl_get_predict_hbd_fn ssse3 avx2 neon/;
+}
add_proto qw/cfl_predict_lbd_fn cfl_get_predict_lbd_fn/, "TX_SIZE tx_size";
specialize qw/cfl_get_predict_lbd_fn ssse3 avx2 neon/;
-add_proto qw/cfl_predict_hbd_fn cfl_get_predict_hbd_fn/, "TX_SIZE tx_size";
-specialize qw/cfl_get_predict_hbd_fn ssse3 avx2 neon/;
-
1;
diff --git a/av1/common/cfl.c b/av1/common/cfl.c
index 5d449a3..db84ebb 100644
--- a/av1/common/cfl.c
+++ b/av1/common/cfl.c
@@ -160,6 +160,7 @@
CFL_PREDICT_FN(c, lbd)
+#if CONFIG_AV1_HIGHBITDEPTH
void cfl_predict_hbd_c(const int16_t *ac_buf_q3, uint16_t *dst, int dst_stride,
int alpha_q3, int bit_depth, int width, int height) {
for (int j = 0; j < height; j++) {
@@ -173,6 +174,7 @@
}
CFL_PREDICT_FN(c, hbd)
+#endif
static void cfl_compute_parameters(MACROBLOCKD *const xd, TX_SIZE tx_size) {
CFL_CTX *const cfl = &xd->cfl;
@@ -196,12 +198,14 @@
cfl_idx_to_alpha(mbmi->cfl_alpha_idx, mbmi->cfl_alpha_signs, plane - 1);
assert((tx_size_high[tx_size] - 1) * CFL_BUF_LINE + tx_size_wide[tx_size] <=
CFL_BUF_SQUARE);
+#if CONFIG_AV1_HIGHBITDEPTH
if (is_cur_buf_hbd(xd)) {
uint16_t *dst_16 = CONVERT_TO_SHORTPTR(dst);
cfl_get_predict_hbd_fn(tx_size)(cfl->ac_buf_q3, dst_16, dst_stride,
alpha_q3, xd->bd);
return;
}
+#endif
cfl_get_predict_lbd_fn(tx_size)(cfl->ac_buf_q3, dst, dst_stride, alpha_q3);
}
@@ -248,6 +252,7 @@
}
}
+#if CONFIG_AV1_HIGHBITDEPTH
static void cfl_luma_subsampling_420_hbd_c(const uint16_t *input,
int input_stride,
uint16_t *output_q3, int width,
@@ -290,9 +295,11 @@
output_q3 += CFL_BUF_LINE;
}
}
+#endif
CFL_GET_SUBSAMPLE_FUNCTION(c)
+#if CONFIG_AV1_HIGHBITDEPTH
static INLINE cfl_subsample_hbd_fn cfl_subsampling_hbd(TX_SIZE tx_size,
int sub_x, int sub_y) {
if (sub_x == 1) {
@@ -303,6 +310,7 @@
}
return cfl_get_luma_subsampling_444_hbd(tx_size);
}
+#endif
static INLINE cfl_subsample_lbd_fn cfl_subsampling_lbd(TX_SIZE tx_size,
int sub_x, int sub_y) {
@@ -348,7 +356,7 @@
// Store the input into the CfL pixel buffer
uint16_t *recon_buf_q3 =
cfl->recon_buf_q3 + (store_row * CFL_BUF_LINE + store_col);
-
+#if CONFIG_AV1_HIGHBITDEPTH
if (use_hbd) {
cfl_subsampling_hbd(tx_size, sub_x, sub_y)(CONVERT_TO_SHORTPTR(input),
input_stride, recon_buf_q3);
@@ -356,6 +364,10 @@
cfl_subsampling_lbd(tx_size, sub_x, sub_y)(input, input_stride,
recon_buf_q3);
}
+#else
+ (void)use_hbd;
+ cfl_subsampling_lbd(tx_size, sub_x, sub_y)(input, input_stride, recon_buf_q3);
+#endif
}
// Adjust the row and column of blocks smaller than 8X8, as chroma-referenced
diff --git a/av1/common/cfl.h b/av1/common/cfl.h
index f6965ee..2acad26 100644
--- a/av1/common/cfl.h
+++ b/av1/common/cfl.h
@@ -144,6 +144,7 @@
// The RTCD script does not support passing in an array, so we wrap it in this
// function.
+#if CONFIG_AV1_HIGHBITDEPTH
#define CFL_GET_SUBSAMPLE_FUNCTION(arch) \
CFL_SUBSAMPLE_FUNCTIONS(arch, 420, lbd) \
CFL_SUBSAMPLE_FUNCTIONS(arch, 422, lbd) \
@@ -151,6 +152,12 @@
CFL_SUBSAMPLE_FUNCTIONS(arch, 420, hbd) \
CFL_SUBSAMPLE_FUNCTIONS(arch, 422, hbd) \
CFL_SUBSAMPLE_FUNCTIONS(arch, 444, hbd)
+#else
+#define CFL_GET_SUBSAMPLE_FUNCTION(arch) \
+ CFL_SUBSAMPLE_FUNCTIONS(arch, 420, lbd) \
+ CFL_SUBSAMPLE_FUNCTIONS(arch, 422, lbd) \
+ CFL_SUBSAMPLE_FUNCTIONS(arch, 444, lbd)
+#endif
// Declare a size-specific wrapper for the size-generic function. The compiler
// will inline the size generic function in here, the advantage is that the size
@@ -221,6 +228,7 @@
height); \
}
+#if CONFIG_AV1_HIGHBITDEPTH
#define CFL_PREDICT_hbd(arch, width, height) \
void cfl_predict_hbd_##width##x##height##_##arch( \
const int16_t *pred_buf_q3, uint16_t *dst, int dst_stride, int alpha_q3, \
@@ -228,6 +236,7 @@
cfl_predict_hbd_##arch(pred_buf_q3, dst, dst_stride, alpha_q3, bd, width, \
height); \
}
+#endif
// This wrapper exists because clang format does not like calling macros with
// lowercase letters.
diff --git a/av1/common/warped_motion.c b/av1/common/warped_motion.c
index d6dcde1..8997d81 100644
--- a/av1/common/warped_motion.c
+++ b/av1/common/warped_motion.c
@@ -304,6 +304,7 @@
return 1;
}
+#if CONFIG_AV1_HIGHBITDEPTH
static INLINE int highbd_error_measure(int err, int bd) {
const int b = bd - 8;
const int bmask = (1 << b) - 1;
@@ -503,6 +504,7 @@
}
return sum_error;
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
/* The warp filter for ROTZOOM and AFFINE models works as follows:
* Split the input into 8x8 blocks
@@ -779,11 +781,15 @@
int64_t av1_frame_error(int use_hbd, int bd, const uint8_t *ref, int stride,
uint8_t *dst, int p_width, int p_height, int p_stride) {
+#if CONFIG_AV1_HIGHBITDEPTH
if (use_hbd) {
return av1_calc_highbd_frame_error(CONVERT_TO_SHORTPTR(ref), stride,
CONVERT_TO_SHORTPTR(dst), p_width,
p_height, p_stride, bd);
}
+#endif
+ (void)use_hbd;
+ (void)bd;
return av1_calc_frame_error(ref, stride, dst, p_width, p_height, p_stride);
}
@@ -792,11 +798,15 @@
int p_height, int p_stride,
uint8_t *segment_map,
int segment_map_stride) {
+#if CONFIG_AV1_HIGHBITDEPTH
if (use_hbd) {
return highbd_segmented_frame_error(
CONVERT_TO_SHORTPTR(ref), stride, CONVERT_TO_SHORTPTR(dst), p_width,
p_height, p_stride, bd, segment_map, segment_map_stride);
}
+#endif
+ (void)use_hbd;
+ (void)bd;
return segmented_frame_error(ref, stride, dst, p_width, p_height, p_stride,
segment_map, segment_map_stride);
}
@@ -806,6 +816,7 @@
uint8_t *pred, int p_col, int p_row, int p_width,
int p_height, int p_stride, int subsampling_x,
int subsampling_y, ConvolveParams *conv_params) {
+#if CONFIG_AV1_HIGHBITDEPTH
if (use_hbd)
highbd_warp_plane(wm, CONVERT_TO_SHORTPTR(ref), width, height, stride,
CONVERT_TO_SHORTPTR(pred), p_col, p_row, p_width,
@@ -814,6 +825,12 @@
else
warp_plane(wm, ref, width, height, stride, pred, p_col, p_row, p_width,
p_height, p_stride, subsampling_x, subsampling_y, conv_params);
+#else
+ (void)use_hbd;
+ (void)bd;
+ warp_plane(wm, ref, width, height, stride, pred, p_col, p_row, p_width,
+ p_height, p_stride, subsampling_x, subsampling_y, conv_params);
+#endif
}
#define LS_MV_MAX 256 // max mv in 1/8-pel
diff --git a/av1/common/x86/cfl_avx2.c b/av1/common/x86/cfl_avx2.c
index 3356984..d9c6f99 100644
--- a/av1/common/x86/cfl_avx2.c
+++ b/av1/common/x86/cfl_avx2.c
@@ -147,6 +147,7 @@
CFL_GET_SUBSAMPLE_FUNCTION_AVX2(444, lbd)
+#if CONFIG_AV1_HIGHBITDEPTH
/**
* Adds 4 pixels (in a 2x2 grid) and multiplies them by 2. Resulting in a more
* precise version of a box filter 4:2:0 pixel subsampling in Q3.
@@ -238,6 +239,7 @@
}
CFL_GET_SUBSAMPLE_FUNCTION_AVX2(444, hbd)
+#endif // CONFIG_AV1_HIGHBITDEPTH
static INLINE __m256i predict_unclipped(const __m256i *input, __m256i alpha_q12,
__m256i alpha_sign, __m256i dc_q0) {
@@ -300,6 +302,7 @@
return pred[tx_size % TX_SIZES_ALL];
}
+#if CONFIG_AV1_HIGHBITDEPTH
static __m256i highbd_max_epi16(int bd) {
const __m256i neg_one = _mm256_set1_epi16(-1);
// (1 << bd) - 1 => -(-1 << bd) -1 => -1 - (-1 << bd) => -1 ^ (-1 << bd)
@@ -372,6 +375,7 @@
// function pointer array out of bounds.
return pred[tx_size % TX_SIZES_ALL];
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
// Returns a vector where all the (32-bits) elements are the sum of all the
// lanes in a.
diff --git a/av1/common/x86/cfl_simd.h b/av1/common/x86/cfl_simd.h
index f2d4ac7..03ae02a 100644
--- a/av1/common/x86/cfl_simd.h
+++ b/av1/common/x86/cfl_simd.h
@@ -98,6 +98,7 @@
void cfl_subsample_lbd_444_16x32_ssse3(const uint8_t *cfl_type,
int input_stride, uint16_t *output_q3);
+#if CONFIG_AV1_HIGHBITDEPTH
void cfl_subsample_hbd_420_4x4_ssse3(const uint16_t *cfl_type, int input_stride,
uint16_t *output_q3);
void cfl_subsample_hbd_420_4x8_ssse3(const uint16_t *cfl_type, int input_stride,
@@ -178,6 +179,7 @@
int input_stride, uint16_t *output_q3);
void cfl_subsample_hbd_444_16x32_ssse3(const uint16_t *cfl_type,
int input_stride, uint16_t *output_q3);
+#endif // CONFIG_AV1_HIGHBITDEPTH
// SSE2 version is optimal for with == 4, we reuse them in AVX2
void cfl_subtract_average_4x4_sse2(const uint16_t *src, int16_t *dst);
@@ -215,6 +217,7 @@
void cfl_predict_lbd_16x32_ssse3(const int16_t *pred_buf_q3, uint8_t *dst,
int dst_stride, int alpha_q3);
+#if CONFIG_AV1_HIGHBITDEPTH
void cfl_predict_hbd_4x4_ssse3(const int16_t *pred_buf_q3, uint16_t *dst,
int dst_stride, int alpha_q3, int bd);
void cfl_predict_hbd_4x8_ssse3(const int16_t *pred_buf_q3, uint16_t *dst,
@@ -239,5 +242,5 @@
int dst_stride, int alpha_q3, int bd);
void cfl_predict_hbd_16x32_ssse3(const int16_t *pred_buf_q3, uint16_t *dst,
int dst_stride, int alpha_q3, int bd);
-
+#endif // CONFIG_AV1_HIGHBITDEPTH
#endif // AOM_AV1_COMMON_X86_CFL_SIMD_H_
diff --git a/av1/common/x86/cfl_ssse3.c b/av1/common/x86/cfl_ssse3.c
index bbf0072..476b660 100644
--- a/av1/common/x86/cfl_ssse3.c
+++ b/av1/common/x86/cfl_ssse3.c
@@ -168,6 +168,7 @@
} while (pred_buf_m128i < end);
}
+#if CONFIG_AV1_HIGHBITDEPTH
/**
* Adds 4 pixels (in a 2x2 grid) and multiplies them by 2. Resulting in a more
* precise version of a box filter 4:2:0 pixel subsampling in Q3.
@@ -296,6 +297,7 @@
pred_buf_q3 += CFL_BUF_LINE;
} while (pred_buf_q3 < end);
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
CFL_GET_SUBSAMPLE_FUNCTION(ssse3)
@@ -341,6 +343,7 @@
CFL_PREDICT_FN(ssse3, lbd)
+#if CONFIG_AV1_HIGHBITDEPTH
static INLINE __m128i highbd_max_epi16(int bd) {
const __m128i neg_one = _mm_set1_epi16(-1);
// (1 << bd) - 1 => -(-1 << bd) -1 => -1 - (-1 << bd) => -1 ^ (-1 << bd)
@@ -391,3 +394,4 @@
}
CFL_PREDICT_FN(ssse3, hbd)
+#endif // CONFIG_AV1_HIGHBITDEPTH
diff --git a/av1/encoder/global_motion.c b/av1/encoder/global_motion.c
index 9354bc5..9623ec3 100644
--- a/av1/encoder/global_motion.c
+++ b/av1/encoder/global_motion.c
@@ -161,6 +161,7 @@
wm->wmtype = wmtype;
}
+#if CONFIG_AV1_HIGHBITDEPTH
static int64_t highbd_warp_error(
WarpedMotionParams *wm, const uint16_t *const ref, int width, int height,
int stride, const uint16_t *const dst, int p_col, int p_row, int p_width,
@@ -195,6 +196,7 @@
}
return gm_sumerr;
}
+#endif
static int64_t warp_error(WarpedMotionParams *wm, const uint8_t *const ref,
int width, int height, int stride,
@@ -242,12 +244,16 @@
uint8_t *segment_map, int segment_map_stride) {
if (wm->wmtype <= AFFINE)
if (!av1_get_shear_params(wm)) return INT64_MAX;
+#if CONFIG_AV1_HIGHBITDEPTH
if (use_hbd)
return highbd_warp_error(wm, CONVERT_TO_SHORTPTR(ref), width, height,
stride, CONVERT_TO_SHORTPTR(dst), p_col, p_row,
p_width, p_height, p_stride, subsampling_x,
subsampling_y, bd, best_error, segment_map,
segment_map_stride);
+#endif
+ (void)use_hbd;
+ (void)bd;
return warp_error(wm, ref, width, height, stride, dst, p_col, p_row, p_width,
p_height, p_stride, subsampling_x, subsampling_y,
best_error, segment_map, segment_map_stride);
diff --git a/av1/encoder/pickrst.c b/av1/encoder/pickrst.c
index 6908a72..26aa211 100644
--- a/av1/encoder/pickrst.c
+++ b/av1/encoder/pickrst.c
@@ -267,6 +267,7 @@
return err;
}
+#if CONFIG_AV1_HIGHBITDEPTH
int64_t av1_highbd_pixel_proj_error_c(const uint8_t *src8, int width,
int height, int src_stride,
const uint8_t *dat8, int dat_stride,
@@ -340,6 +341,7 @@
}
return err;
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
static int64_t get_pixel_proj_error(const uint8_t *src8, int width, int height,
int src_stride, const uint8_t *dat8,
@@ -349,15 +351,24 @@
const sgr_params_type *params) {
int xq[2];
av1_decode_xq(xqd, xq, params);
- if (!use_highbitdepth) {
- return av1_lowbd_pixel_proj_error(src8, width, height, src_stride, dat8,
- dat_stride, flt0, flt0_stride, flt1,
- flt1_stride, xq, params);
- } else {
+
+#if CONFIG_AV1_HIGHBITDEPTH
+ if (use_highbitdepth) {
return av1_highbd_pixel_proj_error(src8, width, height, src_stride, dat8,
dat_stride, flt0, flt0_stride, flt1,
flt1_stride, xq, params);
+
+ } else {
+ return av1_lowbd_pixel_proj_error(src8, width, height, src_stride, dat8,
+ dat_stride, flt0, flt0_stride, flt1,
+ flt1_stride, xq, params);
}
+#else
+ (void)use_highbitdepth;
+ return av1_lowbd_pixel_proj_error(src8, width, height, src_stride, dat8,
+ dat_stride, flt0, flt0_stride, flt1,
+ flt1_stride, xq, params);
+#endif
}
#define USE_SGRPROJ_REFINEMENT_SEARCH 1
@@ -777,6 +788,7 @@
}
}
+#if CONFIG_AV1_HIGHBITDEPTH
void av1_compute_stats_highbd_c(int wiener_win, const uint8_t *dgd8,
const uint8_t *src8, int h_start, int h_end,
int v_start, int v_end, int dgd_stride,
@@ -830,6 +842,7 @@
}
}
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
static INLINE int wrap_index(int i, int wiener_win) {
const int wiener_halfwin1 = (wiener_win >> 1) + 1;
@@ -1272,6 +1285,7 @@
int64_t H[WIENER_WIN2 * WIENER_WIN2];
int32_t vfilter[WIENER_WIN], hfilter[WIENER_WIN];
+#if CONFIG_AV1_HIGHBITDEPTH
const AV1_COMMON *const cm = rsc->cm;
if (cm->seq_params.use_highbitdepth) {
av1_compute_stats_highbd(reduced_wiener_win, rsc->dgd_buffer,
@@ -1283,7 +1297,11 @@
limits->h_start, limits->h_end, limits->v_start,
limits->v_end, rsc->dgd_stride, rsc->src_stride, M, H);
}
-
+#else
+ av1_compute_stats(reduced_wiener_win, rsc->dgd_buffer, rsc->src_buffer,
+ limits->h_start, limits->h_end, limits->v_start,
+ limits->v_end, rsc->dgd_stride, rsc->src_stride, M, H);
+#endif
const MACROBLOCK *const x = rsc->x;
const int64_t bits_none = x->wiener_restore_cost[0];
diff --git a/av1/encoder/pickrst.h b/av1/encoder/pickrst.h
index f34359c..eee3055 100644
--- a/av1/encoder/pickrst.h
+++ b/av1/encoder/pickrst.h
@@ -42,6 +42,7 @@
return (uint8_t)avg;
}
+#if CONFIG_AV1_HIGHBITDEPTH
static INLINE uint16_t find_average_highbd(const uint16_t *src, int h_start,
int h_end, int v_start, int v_end,
int stride) {
@@ -54,6 +55,7 @@
uint64_t avg = sum / ((v_end - v_start) * (h_end - h_start));
return (uint16_t)avg;
}
+#endif
void av1_pick_filter_restoration(const YV12_BUFFER_CONFIG *sd, AV1_COMP *cpi);
diff --git a/av1/encoder/rdopt.c b/av1/encoder/rdopt.c
index 7b2ae0e..ab3e631 100644
--- a/av1/encoder/rdopt.c
+++ b/av1/encoder/rdopt.c
@@ -1881,6 +1881,7 @@
return error;
}
+#if CONFIG_AV1_HIGHBITDEPTH
int64_t av1_highbd_block_error_c(const tran_low_t *coeff,
const tran_low_t *dqcoeff, intptr_t block_size,
int64_t *ssz, int bd) {
@@ -1901,6 +1902,7 @@
*ssz = sqcoeff;
return error;
}
+#endif
// Get transform block visible dimensions cropped to the MI units.
static AOM_INLINE void get_txb_dimensions(const MACROBLOCKD *xd, int plane,
@@ -2110,13 +2112,15 @@
int shift = (MAX_TX_SCALE - av1_get_tx_scale(tx_size)) * 2;
tran_low_t *const coeff = BLOCK_OFFSET(p->coeff, block);
tran_low_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
-
+#if CONFIG_AV1_HIGHBITDEPTH
if (is_cur_buf_hbd(xd))
*out_dist = av1_highbd_block_error(coeff, dqcoeff, buffer_length, &this_sse,
xd->bd);
else
*out_dist = av1_block_error(coeff, dqcoeff, buffer_length, &this_sse);
-
+#else
+ *out_dist = av1_block_error(coeff, dqcoeff, buffer_length, &this_sse);
+#endif
*out_dist = RIGHT_SIGNED_SHIFT(*out_dist, shift);
*out_sse = RIGHT_SIGNED_SHIFT(this_sse, shift);
}
diff --git a/av1/encoder/x86/pickrst_avx2.c b/av1/encoder/x86/pickrst_avx2.c
index 5571165..e011c26 100644
--- a/av1/encoder/x86/pickrst_avx2.c
+++ b/av1/encoder/x86/pickrst_avx2.c
@@ -122,6 +122,7 @@
}
}
+#if CONFIG_AV1_HIGHBITDEPTH
static INLINE void acc_stat_highbd_avx2(int64_t *dst, const uint16_t *dgd,
const __m256i *shuffle,
const __m256i *dgd_ijkl) {
@@ -380,6 +381,7 @@
v_end, dgd_stride, src_stride, M, H, bit_depth);
}
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
static INLINE void acc_stat_win5_one_line_avx2(
const uint8_t *dgd, const uint8_t *src, int h_start, int h_end,
@@ -628,6 +630,7 @@
return err;
}
+#if CONFIG_AV1_HIGHBITDEPTH
int64_t av1_highbd_pixel_proj_error_avx2(
const uint8_t *src8, int width, int height, int src_stride,
const uint8_t *dat8, int dat_stride, int32_t *flt0, int flt0_stride,
@@ -847,3 +850,4 @@
err += sum[0] + sum[1] + sum[2] + sum[3];
return err;
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
diff --git a/av1/encoder/x86/pickrst_sse4.c b/av1/encoder/x86/pickrst_sse4.c
index a94e169..f41694b 100644
--- a/av1/encoder/x86/pickrst_sse4.c
+++ b/av1/encoder/x86/pickrst_sse4.c
@@ -125,6 +125,7 @@
}
}
+#if CONFIG_AV1_HIGHBITDEPTH
static INLINE void acc_stat_highbd_sse41(int64_t *dst, const uint16_t *dgd,
const __m128i *shuffle,
const __m128i *dgd_ijkl) {
@@ -386,6 +387,7 @@
v_end, dgd_stride, src_stride, M, H, bit_depth);
}
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
static INLINE void acc_stat_win5_one_line_sse4_1(
const uint8_t *dgd, const uint8_t *src, int h_start, int h_end,
@@ -622,6 +624,7 @@
return err;
}
+#if CONFIG_AV1_HIGHBITDEPTH
int64_t av1_highbd_pixel_proj_error_sse4_1(
const uint8_t *src8, int width, int height, int src_stride,
const uint8_t *dat8, int dat_stride, int32_t *flt0, int flt0_stride,
@@ -827,3 +830,4 @@
err += sum[0] + sum[1];
return err;
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
diff --git a/test/cfl_test.cc b/test/cfl_test.cc
index fccc9f2..b2d6137 100644
--- a/test/cfl_test.cc
+++ b/test/cfl_test.cc
@@ -310,6 +310,7 @@
subsampleSpeedTest(fun_444, fun_444_ref, &ACMRandom::Rand8);
}
+#if CONFIG_AV1_HIGHBITDEPTH
typedef cfl_subsample_hbd_fn (*get_subsample_hbd_fn)(TX_SIZE tx_size);
typedef ::testing::tuple<TX_SIZE, get_subsample_hbd_fn, get_subsample_hbd_fn,
get_subsample_hbd_fn>
@@ -351,6 +352,7 @@
TEST_P(CFLSubsampleHBDTest, DISABLED_SubsampleHBD444SpeedTest) {
subsampleSpeedTest(fun_444, fun_444_ref, &ACMRandom::Rand12);
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
typedef cfl_predict_lbd_fn (*get_predict_fn)(TX_SIZE tx_size);
typedef ::testing::tuple<TX_SIZE, get_predict_fn> predict_param;
@@ -398,6 +400,7 @@
assertFaster(ref_elapsed_time, elapsed_time);
}
+#if CONFIG_AV1_HIGHBITDEPTH
typedef cfl_predict_hbd_fn (*get_predict_fn_hbd)(TX_SIZE tx_size);
typedef ::testing::tuple<TX_SIZE, get_predict_fn_hbd> predict_param_hbd;
class CFLPredictHBDTest : public ::testing::TestWithParam<predict_param_hbd>,
@@ -445,6 +448,7 @@
printSpeed(ref_elapsed_time, elapsed_time, width, height);
assertFaster(ref_elapsed_time, elapsed_time);
}
+#endif // CONFIG_AV1_HIGHBITDEPTH
#if HAVE_SSE2
const sub_avg_param sub_avg_sizes_sse2[] = { ALL_CFL_TX_SIZES(
@@ -462,30 +466,32 @@
cfl_get_luma_subsampling_444_lbd_ssse3)
};
+const predict_param predict_sizes_ssse3[] = { ALL_CFL_TX_SIZES(
+ cfl_get_predict_lbd_fn_ssse3) };
+
+INSTANTIATE_TEST_CASE_P(SSSE3, CFLSubsampleLBDTest,
+ ::testing::ValuesIn(subsample_lbd_sizes_ssse3));
+
+INSTANTIATE_TEST_CASE_P(SSSE3, CFLPredictTest,
+ ::testing::ValuesIn(predict_sizes_ssse3));
+
+#if CONFIG_AV1_HIGHBITDEPTH
const subsample_hbd_param subsample_hbd_sizes_ssse3[] = {
ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_hbd_ssse3,
cfl_get_luma_subsampling_422_hbd_ssse3,
cfl_get_luma_subsampling_444_hbd_ssse3)
};
-const predict_param predict_sizes_ssse3[] = { ALL_CFL_TX_SIZES(
- cfl_get_predict_lbd_fn_ssse3) };
-
const predict_param_hbd predict_sizes_hbd_ssse3[] = { ALL_CFL_TX_SIZES(
cfl_get_predict_hbd_fn_ssse3) };
-INSTANTIATE_TEST_CASE_P(SSSE3, CFLSubsampleLBDTest,
- ::testing::ValuesIn(subsample_lbd_sizes_ssse3));
-
INSTANTIATE_TEST_CASE_P(SSSE3, CFLSubsampleHBDTest,
::testing::ValuesIn(subsample_hbd_sizes_ssse3));
-INSTANTIATE_TEST_CASE_P(SSSE3, CFLPredictTest,
- ::testing::ValuesIn(predict_sizes_ssse3));
-
INSTANTIATE_TEST_CASE_P(SSSE3, CFLPredictHBDTest,
::testing::ValuesIn(predict_sizes_hbd_ssse3));
-#endif
+#endif // CONFIG_AV1_HIGHBITDEPTH
+#endif // HAVE_SSSE3
#if HAVE_AVX2
const sub_avg_param sub_avg_sizes_avx2[] = { ALL_CFL_TX_SIZES(
@@ -497,72 +503,75 @@
cfl_get_luma_subsampling_444_lbd_avx2)
};
-const subsample_hbd_param subsample_hbd_sizes_avx2[] = {
- ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_hbd_avx2,
- cfl_get_luma_subsampling_422_hbd_avx2,
- cfl_get_luma_subsampling_444_hbd_avx2)
-};
-
const predict_param predict_sizes_avx2[] = { ALL_CFL_TX_SIZES(
cfl_get_predict_lbd_fn_avx2) };
-const predict_param_hbd predict_sizes_hbd_avx2[] = { ALL_CFL_TX_SIZES(
- cfl_get_predict_hbd_fn_avx2) };
-
INSTANTIATE_TEST_CASE_P(AVX2, CFLSubAvgTest,
::testing::ValuesIn(sub_avg_sizes_avx2));
INSTANTIATE_TEST_CASE_P(AVX2, CFLSubsampleLBDTest,
::testing::ValuesIn(subsample_lbd_sizes_avx2));
-INSTANTIATE_TEST_CASE_P(AVX2, CFLSubsampleHBDTest,
- ::testing::ValuesIn(subsample_hbd_sizes_avx2));
-
INSTANTIATE_TEST_CASE_P(AVX2, CFLPredictTest,
::testing::ValuesIn(predict_sizes_avx2));
+#if CONFIG_AV1_HIGHBITDEPTH
+const subsample_hbd_param subsample_hbd_sizes_avx2[] = {
+ ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_hbd_avx2,
+ cfl_get_luma_subsampling_422_hbd_avx2,
+ cfl_get_luma_subsampling_444_hbd_avx2)
+};
+
+const predict_param_hbd predict_sizes_hbd_avx2[] = { ALL_CFL_TX_SIZES(
+ cfl_get_predict_hbd_fn_avx2) };
+
+INSTANTIATE_TEST_CASE_P(AVX2, CFLSubsampleHBDTest,
+ ::testing::ValuesIn(subsample_hbd_sizes_avx2));
+
INSTANTIATE_TEST_CASE_P(AVX2, CFLPredictHBDTest,
::testing::ValuesIn(predict_sizes_hbd_avx2));
-#endif
+#endif // CONFIG_AV1_HIGHBITDEPTH
+#endif // HAVE_AVX2
#if HAVE_NEON
-
const sub_avg_param sub_avg_sizes_neon[] = { ALL_CFL_TX_SIZES(
cfl_get_subtract_average_fn_neon) };
+const predict_param predict_sizes_neon[] = { ALL_CFL_TX_SIZES(
+ cfl_get_predict_lbd_fn_neon) };
+
const subsample_lbd_param subsample_lbd_sizes_neon[] = {
ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_lbd_neon,
cfl_get_luma_subsampling_422_lbd_neon,
cfl_get_luma_subsampling_444_lbd_neon)
};
-const subsample_hbd_param subsample_hbd_sizes_neon[] = {
- ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_hbd_neon,
- cfl_get_luma_subsampling_422_hbd_neon,
- cfl_get_luma_subsampling_444_hbd_neon)
-};
-
-const predict_param predict_sizes_neon[] = { ALL_CFL_TX_SIZES(
- cfl_get_predict_lbd_fn_neon) };
-
-const predict_param_hbd predict_sizes_hbd_neon[] = { ALL_CFL_TX_SIZES(
- cfl_get_predict_hbd_fn_neon) };
-
INSTANTIATE_TEST_CASE_P(NEON, CFLSubAvgTest,
::testing::ValuesIn(sub_avg_sizes_neon));
INSTANTIATE_TEST_CASE_P(NEON, CFLSubsampleLBDTest,
::testing::ValuesIn(subsample_lbd_sizes_neon));
-INSTANTIATE_TEST_CASE_P(NEON, CFLSubsampleHBDTest,
- ::testing::ValuesIn(subsample_hbd_sizes_neon));
-
INSTANTIATE_TEST_CASE_P(NEON, CFLPredictTest,
::testing::ValuesIn(predict_sizes_neon));
+#if CONFIG_AV1_HIGHBITDEPTH
+const subsample_hbd_param subsample_hbd_sizes_neon[] = {
+ ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_hbd_neon,
+ cfl_get_luma_subsampling_422_hbd_neon,
+ cfl_get_luma_subsampling_444_hbd_neon)
+};
+
+const predict_param_hbd predict_sizes_hbd_neon[] = { ALL_CFL_TX_SIZES(
+ cfl_get_predict_hbd_fn_neon) };
+
+INSTANTIATE_TEST_CASE_P(NEON, CFLSubsampleHBDTest,
+ ::testing::ValuesIn(subsample_hbd_sizes_neon));
+
INSTANTIATE_TEST_CASE_P(NEON, CFLPredictHBDTest,
::testing::ValuesIn(predict_sizes_hbd_neon));
-#endif
+#endif // CONFIG_AV1_HIGHBITDEPTH
+#endif // HAVE_NEON
#if HAVE_VSX
const sub_avg_param sub_avg_sizes_vsx[] = { ALL_CFL_TX_SIZES(
diff --git a/test/error_block_test.cc b/test/error_block_test.cc
index da2ceba..85ec076 100644
--- a/test/error_block_test.cc
+++ b/test/error_block_test.cc
@@ -237,29 +237,33 @@
#if (HAVE_SSE2)
INSTANTIATE_TEST_CASE_P(
SSE2, ErrorBlockTest,
- ::testing::Values(make_tuple(&av1_highbd_block_error_sse2,
- &av1_highbd_block_error_c, AOM_BITS_10),
- make_tuple(&av1_highbd_block_error_sse2,
- &av1_highbd_block_error_c, AOM_BITS_12),
- make_tuple(&av1_highbd_block_error_sse2,
- &av1_highbd_block_error_c, AOM_BITS_8),
- make_tuple(&BlockError8BitWrapper<av1_block_error_sse2>,
- &BlockError8BitWrapper<av1_block_error_c>,
- AOM_BITS_8)));
+ ::testing::Values(
+#if CONFIG_AV1_HIGHBITDEPTH
+ make_tuple(&av1_highbd_block_error_sse2, &av1_highbd_block_error_c,
+ AOM_BITS_10),
+ make_tuple(&av1_highbd_block_error_sse2, &av1_highbd_block_error_c,
+ AOM_BITS_12),
+ make_tuple(&av1_highbd_block_error_sse2, &av1_highbd_block_error_c,
+ AOM_BITS_8),
+#endif
+ make_tuple(&BlockError8BitWrapper<av1_block_error_sse2>,
+ &BlockError8BitWrapper<av1_block_error_c>, AOM_BITS_8)));
#endif // HAVE_SSE2
#if (HAVE_AVX2)
INSTANTIATE_TEST_CASE_P(
AVX2, ErrorBlockTest,
- ::testing::Values(make_tuple(&av1_highbd_block_error_avx2,
- &av1_highbd_block_error_c, AOM_BITS_10),
- make_tuple(&av1_highbd_block_error_avx2,
- &av1_highbd_block_error_c, AOM_BITS_12),
- make_tuple(&av1_highbd_block_error_avx2,
- &av1_highbd_block_error_c, AOM_BITS_8),
- make_tuple(&BlockError8BitWrapper<av1_block_error_avx2>,
- &BlockError8BitWrapper<av1_block_error_c>,
- AOM_BITS_8)));
+ ::testing::Values(
+#if CONFIG_AV1_HIGHBITDEPTH
+ make_tuple(&av1_highbd_block_error_avx2, &av1_highbd_block_error_c,
+ AOM_BITS_10),
+ make_tuple(&av1_highbd_block_error_avx2, &av1_highbd_block_error_c,
+ AOM_BITS_12),
+ make_tuple(&av1_highbd_block_error_avx2, &av1_highbd_block_error_c,
+ AOM_BITS_8),
+#endif
+ make_tuple(&BlockError8BitWrapper<av1_block_error_avx2>,
+ &BlockError8BitWrapper<av1_block_error_c>, AOM_BITS_8)));
#endif // HAVE_AVX2
#if (HAVE_MSA)
diff --git a/test/pickrst_test.cc b/test/pickrst_test.cc
index 0aa49b6..a4da28c 100644
--- a/test/pickrst_test.cc
+++ b/test/pickrst_test.cc
@@ -189,6 +189,7 @@
} // namespace pickrst_test_lowbd
+#if CONFIG_AV1_HIGHBITDEPTH
namespace pickrst_test_highbd {
static const int kIterations = 100;
@@ -355,3 +356,4 @@
#endif // HAVE_AVX2
} // namespace pickrst_test_highbd
+#endif // CONFIG_AV1_HIGHBITDEPTH
diff --git a/test/warp_filter_test.cc b/test/warp_filter_test.cc
index cafb344..cb7ba04 100644
--- a/test/warp_filter_test.cc
+++ b/test/warp_filter_test.cc
@@ -11,7 +11,9 @@
#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
#include "test/warp_filter_test_util.h"
using libaom_test::ACMRandom;
+#if CONFIG_AV1_HIGHBITDEPTH
using libaom_test::AV1HighbdWarpFilter::AV1HighbdWarpFilterTest;
+#endif
using libaom_test::AV1WarpFilter::AV1WarpFilterTest;
using ::testing::make_tuple;
using ::testing::tuple;
@@ -34,6 +36,7 @@
SSE4_1, AV1WarpFilterTest,
libaom_test::AV1WarpFilter::BuildParams(av1_warp_affine_sse4_1));
+#if CONFIG_AV1_HIGHBITDEPTH
TEST_P(AV1HighbdWarpFilterTest, CheckOutput) {
RunCheckOutput(::testing::get<4>(GET_PARAM(0)));
}
@@ -44,7 +47,7 @@
INSTANTIATE_TEST_CASE_P(SSE4_1, AV1HighbdWarpFilterTest,
libaom_test::AV1HighbdWarpFilter::BuildParams(
av1_highbd_warp_affine_sse4_1));
-
+#endif // CONFIG_AV1_HIGHBITDEPTH
#endif // HAVE_SSE4_1
#if HAVE_AVX2
diff --git a/test/warp_filter_test_util.cc b/test/warp_filter_test_util.cc
index 9208af8..91d040a 100644
--- a/test/warp_filter_test_util.cc
+++ b/test/warp_filter_test_util.cc
@@ -276,6 +276,7 @@
}
} // namespace AV1WarpFilter
+#if CONFIG_AV1_HIGHBITDEPTH
namespace AV1HighbdWarpFilter {
::testing::internal::ParamGenerator<HighbdWarpTestParams> BuildParams(
highbd_warp_affine_func filter) {
@@ -477,4 +478,5 @@
delete[] dstb;
}
} // namespace AV1HighbdWarpFilter
+#endif // CONFIG_AV1_HIGHBITDEPTH
} // namespace libaom_test
diff --git a/test/warp_filter_test_util.h b/test/warp_filter_test_util.h
index b8998e5..23749fb 100644
--- a/test/warp_filter_test_util.h
+++ b/test/warp_filter_test_util.h
@@ -63,6 +63,7 @@
} // namespace AV1WarpFilter
+#if CONFIG_AV1_HIGHBITDEPTH
namespace AV1HighbdWarpFilter {
typedef void (*highbd_warp_affine_func)(const int32_t *mat, const uint16_t *ref,
int width, int height, int stride,
@@ -97,6 +98,7 @@
};
} // namespace AV1HighbdWarpFilter
+#endif // CONFIG_AV1_HIGHBITDEPTH
} // namespace libaom_test
diff --git a/test/wiener_test.cc b/test/wiener_test.cc
index 8f49af62..68bf3ad 100644
--- a/test/wiener_test.cc
+++ b/test/wiener_test.cc
@@ -284,6 +284,7 @@
} // namespace wiener_lowbd
+#if CONFIG_AV1_HIGHBITDEPTH
// High bit-depth tests:
namespace wiener_highbd {
@@ -582,3 +583,4 @@
#endif // HAVE_AVX2
} // namespace wiener_highbd
+#endif // CONFIG_AV1_HIGHBITDEPTH
