Revert "Avoid use of deprecated high-bitdepth functions"
This reverts commit f9d77bd5a2541387dc435e052704b2e734307c20.
Reason for revert: 8x8 transform failures
BUG=https://bugs.chromium.org/p/aomedia/issues/detail?id=502
Change-Id: I2f6c10bc576a966bd5a878b7ee8389074bf45014
diff --git a/aom_dsp/inv_txfm.c b/aom_dsp/inv_txfm.c
index bb99585..6e7d8c9 100644
--- a/aom_dsp/inv_txfm.c
+++ b/aom_dsp/inv_txfm.c
@@ -1442,4 +1442,868 @@
}
}
+void aom_highbd_idct8_c(const tran_low_t *input, tran_low_t *output, int bd) {
+ tran_low_t step1[8], step2[8];
+ tran_high_t temp1, temp2;
+ // stage 1
+ step1[0] = input[0];
+ step1[2] = input[4];
+ step1[1] = input[2];
+ step1[3] = input[6];
+ temp1 = input[1] * cospi_28_64 - input[7] * cospi_4_64;
+ temp2 = input[1] * cospi_4_64 + input[7] * cospi_28_64;
+ step1[4] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[7] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = input[5] * cospi_12_64 - input[3] * cospi_20_64;
+ temp2 = input[5] * cospi_20_64 + input[3] * cospi_12_64;
+ step1[5] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ // stage 2 & stage 3 - even half
+ aom_highbd_idct4_c(step1, step1, bd);
+
+ // stage 2 - odd half
+ step2[4] = HIGHBD_WRAPLOW(step1[4] + step1[5], bd);
+ step2[5] = HIGHBD_WRAPLOW(step1[4] - step1[5], bd);
+ step2[6] = HIGHBD_WRAPLOW(-step1[6] + step1[7], bd);
+ step2[7] = HIGHBD_WRAPLOW(step1[6] + step1[7], bd);
+
+ // stage 3 - odd half
+ step1[4] = step2[4];
+ temp1 = (step2[6] - step2[5]) * cospi_16_64;
+ temp2 = (step2[5] + step2[6]) * cospi_16_64;
+ step1[5] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step1[7] = step2[7];
+
+ // stage 4
+ output[0] = HIGHBD_WRAPLOW(step1[0] + step1[7], bd);
+ output[1] = HIGHBD_WRAPLOW(step1[1] + step1[6], bd);
+ output[2] = HIGHBD_WRAPLOW(step1[2] + step1[5], bd);
+ output[3] = HIGHBD_WRAPLOW(step1[3] + step1[4], bd);
+ output[4] = HIGHBD_WRAPLOW(step1[3] - step1[4], bd);
+ output[5] = HIGHBD_WRAPLOW(step1[2] - step1[5], bd);
+ output[6] = HIGHBD_WRAPLOW(step1[1] - step1[6], bd);
+ output[7] = HIGHBD_WRAPLOW(step1[0] - step1[7], bd);
+}
+
+void aom_highbd_iadst4_c(const tran_low_t *input, tran_low_t *output, int bd) {
+ tran_high_t s0, s1, s2, s3, s4, s5, s6, s7;
+
+ tran_low_t x0 = input[0];
+ tran_low_t x1 = input[1];
+ tran_low_t x2 = input[2];
+ tran_low_t x3 = input[3];
+ (void)bd;
+
+ if (!(x0 | x1 | x2 | x3)) {
+ memset(output, 0, 4 * sizeof(*output));
+ return;
+ }
+
+ s0 = sinpi_1_9 * x0;
+ s1 = sinpi_2_9 * x0;
+ s2 = sinpi_3_9 * x1;
+ s3 = sinpi_4_9 * x2;
+ s4 = sinpi_1_9 * x2;
+ s5 = sinpi_2_9 * x3;
+ s6 = sinpi_4_9 * x3;
+ s7 = (tran_high_t)HIGHBD_WRAPLOW(x0 - x2 + x3, bd);
+
+ s0 = s0 + s3 + s5;
+ s1 = s1 - s4 - s6;
+ s3 = s2;
+ s2 = sinpi_3_9 * s7;
+
+ // 1-D transform scaling factor is sqrt(2).
+ // The overall dynamic range is 14b (input) + 14b (multiplication scaling)
+ // + 1b (addition) = 29b.
+ // Hence the output bit depth is 15b.
+ output[0] = HIGHBD_WRAPLOW(dct_const_round_shift(s0 + s3), bd);
+ output[1] = HIGHBD_WRAPLOW(dct_const_round_shift(s1 + s3), bd);
+ output[2] = HIGHBD_WRAPLOW(dct_const_round_shift(s2), bd);
+ output[3] = HIGHBD_WRAPLOW(dct_const_round_shift(s0 + s1 - s3), bd);
+}
+
+void aom_highbd_iadst8_c(const tran_low_t *input, tran_low_t *output, int bd) {
+ tran_high_t s0, s1, s2, s3, s4, s5, s6, s7;
+
+ tran_low_t x0 = input[7];
+ tran_low_t x1 = input[0];
+ tran_low_t x2 = input[5];
+ tran_low_t x3 = input[2];
+ tran_low_t x4 = input[3];
+ tran_low_t x5 = input[4];
+ tran_low_t x6 = input[1];
+ tran_low_t x7 = input[6];
+ (void)bd;
+
+ if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7)) {
+ memset(output, 0, 8 * sizeof(*output));
+ return;
+ }
+
+ // stage 1
+ s0 = cospi_2_64 * x0 + cospi_30_64 * x1;
+ s1 = cospi_30_64 * x0 - cospi_2_64 * x1;
+ s2 = cospi_10_64 * x2 + cospi_22_64 * x3;
+ s3 = cospi_22_64 * x2 - cospi_10_64 * x3;
+ s4 = cospi_18_64 * x4 + cospi_14_64 * x5;
+ s5 = cospi_14_64 * x4 - cospi_18_64 * x5;
+ s6 = cospi_26_64 * x6 + cospi_6_64 * x7;
+ s7 = cospi_6_64 * x6 - cospi_26_64 * x7;
+
+ x0 = HIGHBD_WRAPLOW(dct_const_round_shift(s0 + s4), bd);
+ x1 = HIGHBD_WRAPLOW(dct_const_round_shift(s1 + s5), bd);
+ x2 = HIGHBD_WRAPLOW(dct_const_round_shift(s2 + s6), bd);
+ x3 = HIGHBD_WRAPLOW(dct_const_round_shift(s3 + s7), bd);
+ x4 = HIGHBD_WRAPLOW(dct_const_round_shift(s0 - s4), bd);
+ x5 = HIGHBD_WRAPLOW(dct_const_round_shift(s1 - s5), bd);
+ x6 = HIGHBD_WRAPLOW(dct_const_round_shift(s2 - s6), bd);
+ x7 = HIGHBD_WRAPLOW(dct_const_round_shift(s3 - s7), bd);
+
+ // stage 2
+ s0 = x0;
+ s1 = x1;
+ s2 = x2;
+ s3 = x3;
+ s4 = cospi_8_64 * x4 + cospi_24_64 * x5;
+ s5 = cospi_24_64 * x4 - cospi_8_64 * x5;
+ s6 = -cospi_24_64 * x6 + cospi_8_64 * x7;
+ s7 = cospi_8_64 * x6 + cospi_24_64 * x7;
+
+ x0 = HIGHBD_WRAPLOW(s0 + s2, bd);
+ x1 = HIGHBD_WRAPLOW(s1 + s3, bd);
+ x2 = HIGHBD_WRAPLOW(s0 - s2, bd);
+ x3 = HIGHBD_WRAPLOW(s1 - s3, bd);
+ x4 = HIGHBD_WRAPLOW(dct_const_round_shift(s4 + s6), bd);
+ x5 = HIGHBD_WRAPLOW(dct_const_round_shift(s5 + s7), bd);
+ x6 = HIGHBD_WRAPLOW(dct_const_round_shift(s4 - s6), bd);
+ x7 = HIGHBD_WRAPLOW(dct_const_round_shift(s5 - s7), bd);
+
+ // stage 3
+ s2 = cospi_16_64 * (x2 + x3);
+ s3 = cospi_16_64 * (x2 - x3);
+ s6 = cospi_16_64 * (x6 + x7);
+ s7 = cospi_16_64 * (x6 - x7);
+
+ x2 = HIGHBD_WRAPLOW(dct_const_round_shift(s2), bd);
+ x3 = HIGHBD_WRAPLOW(dct_const_round_shift(s3), bd);
+ x6 = HIGHBD_WRAPLOW(dct_const_round_shift(s6), bd);
+ x7 = HIGHBD_WRAPLOW(dct_const_round_shift(s7), bd);
+
+ output[0] = HIGHBD_WRAPLOW(x0, bd);
+ output[1] = HIGHBD_WRAPLOW(-x4, bd);
+ output[2] = HIGHBD_WRAPLOW(x6, bd);
+ output[3] = HIGHBD_WRAPLOW(-x2, bd);
+ output[4] = HIGHBD_WRAPLOW(x3, bd);
+ output[5] = HIGHBD_WRAPLOW(-x7, bd);
+ output[6] = HIGHBD_WRAPLOW(x5, bd);
+ output[7] = HIGHBD_WRAPLOW(-x1, bd);
+}
+
+void aom_highbd_idct16_c(const tran_low_t *input, tran_low_t *output, int bd) {
+ tran_low_t step1[16], step2[16];
+ tran_high_t temp1, temp2;
+ (void)bd;
+
+ // stage 1
+ step1[0] = input[0 / 2];
+ step1[1] = input[16 / 2];
+ step1[2] = input[8 / 2];
+ step1[3] = input[24 / 2];
+ step1[4] = input[4 / 2];
+ step1[5] = input[20 / 2];
+ step1[6] = input[12 / 2];
+ step1[7] = input[28 / 2];
+ step1[8] = input[2 / 2];
+ step1[9] = input[18 / 2];
+ step1[10] = input[10 / 2];
+ step1[11] = input[26 / 2];
+ step1[12] = input[6 / 2];
+ step1[13] = input[22 / 2];
+ step1[14] = input[14 / 2];
+ step1[15] = input[30 / 2];
+
+ // stage 2
+ step2[0] = step1[0];
+ step2[1] = step1[1];
+ step2[2] = step1[2];
+ step2[3] = step1[3];
+ step2[4] = step1[4];
+ step2[5] = step1[5];
+ step2[6] = step1[6];
+ step2[7] = step1[7];
+
+ temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
+ temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
+ step2[8] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[15] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
+ temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
+ step2[9] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[14] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
+ temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
+ step2[10] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
+ temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
+ step2[11] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[12] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ // stage 3
+ step1[0] = step2[0];
+ step1[1] = step2[1];
+ step1[2] = step2[2];
+ step1[3] = step2[3];
+
+ temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
+ temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
+ step1[4] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[7] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
+ temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
+ step1[5] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ step1[8] = HIGHBD_WRAPLOW(step2[8] + step2[9], bd);
+ step1[9] = HIGHBD_WRAPLOW(step2[8] - step2[9], bd);
+ step1[10] = HIGHBD_WRAPLOW(-step2[10] + step2[11], bd);
+ step1[11] = HIGHBD_WRAPLOW(step2[10] + step2[11], bd);
+ step1[12] = HIGHBD_WRAPLOW(step2[12] + step2[13], bd);
+ step1[13] = HIGHBD_WRAPLOW(step2[12] - step2[13], bd);
+ step1[14] = HIGHBD_WRAPLOW(-step2[14] + step2[15], bd);
+ step1[15] = HIGHBD_WRAPLOW(step2[14] + step2[15], bd);
+
+ // stage 4
+ temp1 = (step1[0] + step1[1]) * cospi_16_64;
+ temp2 = (step1[0] - step1[1]) * cospi_16_64;
+ step2[0] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[1] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
+ temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
+ step2[2] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[3] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step2[4] = HIGHBD_WRAPLOW(step1[4] + step1[5], bd);
+ step2[5] = HIGHBD_WRAPLOW(step1[4] - step1[5], bd);
+ step2[6] = HIGHBD_WRAPLOW(-step1[6] + step1[7], bd);
+ step2[7] = HIGHBD_WRAPLOW(step1[6] + step1[7], bd);
+
+ step2[8] = step1[8];
+ step2[15] = step1[15];
+ temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
+ temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
+ step2[9] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[14] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
+ temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
+ step2[10] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step2[11] = step1[11];
+ step2[12] = step1[12];
+
+ // stage 5
+ step1[0] = HIGHBD_WRAPLOW(step2[0] + step2[3], bd);
+ step1[1] = HIGHBD_WRAPLOW(step2[1] + step2[2], bd);
+ step1[2] = HIGHBD_WRAPLOW(step2[1] - step2[2], bd);
+ step1[3] = HIGHBD_WRAPLOW(step2[0] - step2[3], bd);
+ step1[4] = step2[4];
+ temp1 = (step2[6] - step2[5]) * cospi_16_64;
+ temp2 = (step2[5] + step2[6]) * cospi_16_64;
+ step1[5] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step1[7] = step2[7];
+
+ step1[8] = HIGHBD_WRAPLOW(step2[8] + step2[11], bd);
+ step1[9] = HIGHBD_WRAPLOW(step2[9] + step2[10], bd);
+ step1[10] = HIGHBD_WRAPLOW(step2[9] - step2[10], bd);
+ step1[11] = HIGHBD_WRAPLOW(step2[8] - step2[11], bd);
+ step1[12] = HIGHBD_WRAPLOW(-step2[12] + step2[15], bd);
+ step1[13] = HIGHBD_WRAPLOW(-step2[13] + step2[14], bd);
+ step1[14] = HIGHBD_WRAPLOW(step2[13] + step2[14], bd);
+ step1[15] = HIGHBD_WRAPLOW(step2[12] + step2[15], bd);
+
+ // stage 6
+ step2[0] = HIGHBD_WRAPLOW(step1[0] + step1[7], bd);
+ step2[1] = HIGHBD_WRAPLOW(step1[1] + step1[6], bd);
+ step2[2] = HIGHBD_WRAPLOW(step1[2] + step1[5], bd);
+ step2[3] = HIGHBD_WRAPLOW(step1[3] + step1[4], bd);
+ step2[4] = HIGHBD_WRAPLOW(step1[3] - step1[4], bd);
+ step2[5] = HIGHBD_WRAPLOW(step1[2] - step1[5], bd);
+ step2[6] = HIGHBD_WRAPLOW(step1[1] - step1[6], bd);
+ step2[7] = HIGHBD_WRAPLOW(step1[0] - step1[7], bd);
+ step2[8] = step1[8];
+ step2[9] = step1[9];
+ temp1 = (-step1[10] + step1[13]) * cospi_16_64;
+ temp2 = (step1[10] + step1[13]) * cospi_16_64;
+ step2[10] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = (-step1[11] + step1[12]) * cospi_16_64;
+ temp2 = (step1[11] + step1[12]) * cospi_16_64;
+ step2[11] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[12] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step2[14] = step1[14];
+ step2[15] = step1[15];
+
+ // stage 7
+ output[0] = HIGHBD_WRAPLOW(step2[0] + step2[15], bd);
+ output[1] = HIGHBD_WRAPLOW(step2[1] + step2[14], bd);
+ output[2] = HIGHBD_WRAPLOW(step2[2] + step2[13], bd);
+ output[3] = HIGHBD_WRAPLOW(step2[3] + step2[12], bd);
+ output[4] = HIGHBD_WRAPLOW(step2[4] + step2[11], bd);
+ output[5] = HIGHBD_WRAPLOW(step2[5] + step2[10], bd);
+ output[6] = HIGHBD_WRAPLOW(step2[6] + step2[9], bd);
+ output[7] = HIGHBD_WRAPLOW(step2[7] + step2[8], bd);
+ output[8] = HIGHBD_WRAPLOW(step2[7] - step2[8], bd);
+ output[9] = HIGHBD_WRAPLOW(step2[6] - step2[9], bd);
+ output[10] = HIGHBD_WRAPLOW(step2[5] - step2[10], bd);
+ output[11] = HIGHBD_WRAPLOW(step2[4] - step2[11], bd);
+ output[12] = HIGHBD_WRAPLOW(step2[3] - step2[12], bd);
+ output[13] = HIGHBD_WRAPLOW(step2[2] - step2[13], bd);
+ output[14] = HIGHBD_WRAPLOW(step2[1] - step2[14], bd);
+ output[15] = HIGHBD_WRAPLOW(step2[0] - step2[15], bd);
+}
+
+void aom_highbd_iadst16_c(const tran_low_t *input, tran_low_t *output, int bd) {
+ tran_high_t s0, s1, s2, s3, s4, s5, s6, s7, s8;
+ tran_high_t s9, s10, s11, s12, s13, s14, s15;
+
+ tran_low_t x0 = input[15];
+ tran_low_t x1 = input[0];
+ tran_low_t x2 = input[13];
+ tran_low_t x3 = input[2];
+ tran_low_t x4 = input[11];
+ tran_low_t x5 = input[4];
+ tran_low_t x6 = input[9];
+ tran_low_t x7 = input[6];
+ tran_low_t x8 = input[7];
+ tran_low_t x9 = input[8];
+ tran_low_t x10 = input[5];
+ tran_low_t x11 = input[10];
+ tran_low_t x12 = input[3];
+ tran_low_t x13 = input[12];
+ tran_low_t x14 = input[1];
+ tran_low_t x15 = input[14];
+ (void)bd;
+
+ if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7 | x8 | x9 | x10 | x11 | x12 |
+ x13 | x14 | x15)) {
+ memset(output, 0, 16 * sizeof(*output));
+ return;
+ }
+
+ // stage 1
+ s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
+ s1 = x0 * cospi_31_64 - x1 * cospi_1_64;
+ s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
+ s3 = x2 * cospi_27_64 - x3 * cospi_5_64;
+ s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
+ s5 = x4 * cospi_23_64 - x5 * cospi_9_64;
+ s6 = x6 * cospi_13_64 + x7 * cospi_19_64;
+ s7 = x6 * cospi_19_64 - x7 * cospi_13_64;
+ s8 = x8 * cospi_17_64 + x9 * cospi_15_64;
+ s9 = x8 * cospi_15_64 - x9 * cospi_17_64;
+ s10 = x10 * cospi_21_64 + x11 * cospi_11_64;
+ s11 = x10 * cospi_11_64 - x11 * cospi_21_64;
+ s12 = x12 * cospi_25_64 + x13 * cospi_7_64;
+ s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
+ s14 = x14 * cospi_29_64 + x15 * cospi_3_64;
+ s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
+
+ x0 = HIGHBD_WRAPLOW(dct_const_round_shift(s0 + s8), bd);
+ x1 = HIGHBD_WRAPLOW(dct_const_round_shift(s1 + s9), bd);
+ x2 = HIGHBD_WRAPLOW(dct_const_round_shift(s2 + s10), bd);
+ x3 = HIGHBD_WRAPLOW(dct_const_round_shift(s3 + s11), bd);
+ x4 = HIGHBD_WRAPLOW(dct_const_round_shift(s4 + s12), bd);
+ x5 = HIGHBD_WRAPLOW(dct_const_round_shift(s5 + s13), bd);
+ x6 = HIGHBD_WRAPLOW(dct_const_round_shift(s6 + s14), bd);
+ x7 = HIGHBD_WRAPLOW(dct_const_round_shift(s7 + s15), bd);
+ x8 = HIGHBD_WRAPLOW(dct_const_round_shift(s0 - s8), bd);
+ x9 = HIGHBD_WRAPLOW(dct_const_round_shift(s1 - s9), bd);
+ x10 = HIGHBD_WRAPLOW(dct_const_round_shift(s2 - s10), bd);
+ x11 = HIGHBD_WRAPLOW(dct_const_round_shift(s3 - s11), bd);
+ x12 = HIGHBD_WRAPLOW(dct_const_round_shift(s4 - s12), bd);
+ x13 = HIGHBD_WRAPLOW(dct_const_round_shift(s5 - s13), bd);
+ x14 = HIGHBD_WRAPLOW(dct_const_round_shift(s6 - s14), bd);
+ x15 = HIGHBD_WRAPLOW(dct_const_round_shift(s7 - s15), bd);
+
+ // stage 2
+ s0 = x0;
+ s1 = x1;
+ s2 = x2;
+ s3 = x3;
+ s4 = x4;
+ s5 = x5;
+ s6 = x6;
+ s7 = x7;
+ s8 = x8 * cospi_4_64 + x9 * cospi_28_64;
+ s9 = x8 * cospi_28_64 - x9 * cospi_4_64;
+ s10 = x10 * cospi_20_64 + x11 * cospi_12_64;
+ s11 = x10 * cospi_12_64 - x11 * cospi_20_64;
+ s12 = -x12 * cospi_28_64 + x13 * cospi_4_64;
+ s13 = x12 * cospi_4_64 + x13 * cospi_28_64;
+ s14 = -x14 * cospi_12_64 + x15 * cospi_20_64;
+ s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
+
+ x0 = HIGHBD_WRAPLOW(s0 + s4, bd);
+ x1 = HIGHBD_WRAPLOW(s1 + s5, bd);
+ x2 = HIGHBD_WRAPLOW(s2 + s6, bd);
+ x3 = HIGHBD_WRAPLOW(s3 + s7, bd);
+ x4 = HIGHBD_WRAPLOW(s0 - s4, bd);
+ x5 = HIGHBD_WRAPLOW(s1 - s5, bd);
+ x6 = HIGHBD_WRAPLOW(s2 - s6, bd);
+ x7 = HIGHBD_WRAPLOW(s3 - s7, bd);
+ x8 = HIGHBD_WRAPLOW(dct_const_round_shift(s8 + s12), bd);
+ x9 = HIGHBD_WRAPLOW(dct_const_round_shift(s9 + s13), bd);
+ x10 = HIGHBD_WRAPLOW(dct_const_round_shift(s10 + s14), bd);
+ x11 = HIGHBD_WRAPLOW(dct_const_round_shift(s11 + s15), bd);
+ x12 = HIGHBD_WRAPLOW(dct_const_round_shift(s8 - s12), bd);
+ x13 = HIGHBD_WRAPLOW(dct_const_round_shift(s9 - s13), bd);
+ x14 = HIGHBD_WRAPLOW(dct_const_round_shift(s10 - s14), bd);
+ x15 = HIGHBD_WRAPLOW(dct_const_round_shift(s11 - s15), bd);
+
+ // stage 3
+ s0 = x0;
+ s1 = x1;
+ s2 = x2;
+ s3 = x3;
+ s4 = x4 * cospi_8_64 + x5 * cospi_24_64;
+ s5 = x4 * cospi_24_64 - x5 * cospi_8_64;
+ s6 = -x6 * cospi_24_64 + x7 * cospi_8_64;
+ s7 = x6 * cospi_8_64 + x7 * cospi_24_64;
+ s8 = x8;
+ s9 = x9;
+ s10 = x10;
+ s11 = x11;
+ s12 = x12 * cospi_8_64 + x13 * cospi_24_64;
+ s13 = x12 * cospi_24_64 - x13 * cospi_8_64;
+ s14 = -x14 * cospi_24_64 + x15 * cospi_8_64;
+ s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
+
+ x0 = HIGHBD_WRAPLOW(s0 + s2, bd);
+ x1 = HIGHBD_WRAPLOW(s1 + s3, bd);
+ x2 = HIGHBD_WRAPLOW(s0 - s2, bd);
+ x3 = HIGHBD_WRAPLOW(s1 - s3, bd);
+ x4 = HIGHBD_WRAPLOW(dct_const_round_shift(s4 + s6), bd);
+ x5 = HIGHBD_WRAPLOW(dct_const_round_shift(s5 + s7), bd);
+ x6 = HIGHBD_WRAPLOW(dct_const_round_shift(s4 - s6), bd);
+ x7 = HIGHBD_WRAPLOW(dct_const_round_shift(s5 - s7), bd);
+ x8 = HIGHBD_WRAPLOW(s8 + s10, bd);
+ x9 = HIGHBD_WRAPLOW(s9 + s11, bd);
+ x10 = HIGHBD_WRAPLOW(s8 - s10, bd);
+ x11 = HIGHBD_WRAPLOW(s9 - s11, bd);
+ x12 = HIGHBD_WRAPLOW(dct_const_round_shift(s12 + s14), bd);
+ x13 = HIGHBD_WRAPLOW(dct_const_round_shift(s13 + s15), bd);
+ x14 = HIGHBD_WRAPLOW(dct_const_round_shift(s12 - s14), bd);
+ x15 = HIGHBD_WRAPLOW(dct_const_round_shift(s13 - s15), bd);
+
+ // stage 4
+ s2 = (-cospi_16_64) * (x2 + x3);
+ s3 = cospi_16_64 * (x2 - x3);
+ s6 = cospi_16_64 * (x6 + x7);
+ s7 = cospi_16_64 * (-x6 + x7);
+ s10 = cospi_16_64 * (x10 + x11);
+ s11 = cospi_16_64 * (-x10 + x11);
+ s14 = (-cospi_16_64) * (x14 + x15);
+ s15 = cospi_16_64 * (x14 - x15);
+
+ x2 = HIGHBD_WRAPLOW(dct_const_round_shift(s2), bd);
+ x3 = HIGHBD_WRAPLOW(dct_const_round_shift(s3), bd);
+ x6 = HIGHBD_WRAPLOW(dct_const_round_shift(s6), bd);
+ x7 = HIGHBD_WRAPLOW(dct_const_round_shift(s7), bd);
+ x10 = HIGHBD_WRAPLOW(dct_const_round_shift(s10), bd);
+ x11 = HIGHBD_WRAPLOW(dct_const_round_shift(s11), bd);
+ x14 = HIGHBD_WRAPLOW(dct_const_round_shift(s14), bd);
+ x15 = HIGHBD_WRAPLOW(dct_const_round_shift(s15), bd);
+
+ output[0] = HIGHBD_WRAPLOW(x0, bd);
+ output[1] = HIGHBD_WRAPLOW(-x8, bd);
+ output[2] = HIGHBD_WRAPLOW(x12, bd);
+ output[3] = HIGHBD_WRAPLOW(-x4, bd);
+ output[4] = HIGHBD_WRAPLOW(x6, bd);
+ output[5] = HIGHBD_WRAPLOW(x14, bd);
+ output[6] = HIGHBD_WRAPLOW(x10, bd);
+ output[7] = HIGHBD_WRAPLOW(x2, bd);
+ output[8] = HIGHBD_WRAPLOW(x3, bd);
+ output[9] = HIGHBD_WRAPLOW(x11, bd);
+ output[10] = HIGHBD_WRAPLOW(x15, bd);
+ output[11] = HIGHBD_WRAPLOW(x7, bd);
+ output[12] = HIGHBD_WRAPLOW(x5, bd);
+ output[13] = HIGHBD_WRAPLOW(-x13, bd);
+ output[14] = HIGHBD_WRAPLOW(x9, bd);
+ output[15] = HIGHBD_WRAPLOW(-x1, bd);
+}
+
+void aom_highbd_idct32_c(const tran_low_t *input, tran_low_t *output, int bd) {
+ tran_low_t step1[32], step2[32];
+ tran_high_t temp1, temp2;
+ (void)bd;
+
+ // stage 1
+ step1[0] = input[0];
+ step1[1] = input[16];
+ step1[2] = input[8];
+ step1[3] = input[24];
+ step1[4] = input[4];
+ step1[5] = input[20];
+ step1[6] = input[12];
+ step1[7] = input[28];
+ step1[8] = input[2];
+ step1[9] = input[18];
+ step1[10] = input[10];
+ step1[11] = input[26];
+ step1[12] = input[6];
+ step1[13] = input[22];
+ step1[14] = input[14];
+ step1[15] = input[30];
+
+ temp1 = input[1] * cospi_31_64 - input[31] * cospi_1_64;
+ temp2 = input[1] * cospi_1_64 + input[31] * cospi_31_64;
+ step1[16] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[31] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = input[17] * cospi_15_64 - input[15] * cospi_17_64;
+ temp2 = input[17] * cospi_17_64 + input[15] * cospi_15_64;
+ step1[17] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[30] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = input[9] * cospi_23_64 - input[23] * cospi_9_64;
+ temp2 = input[9] * cospi_9_64 + input[23] * cospi_23_64;
+ step1[18] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[29] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = input[25] * cospi_7_64 - input[7] * cospi_25_64;
+ temp2 = input[25] * cospi_25_64 + input[7] * cospi_7_64;
+ step1[19] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[28] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = input[5] * cospi_27_64 - input[27] * cospi_5_64;
+ temp2 = input[5] * cospi_5_64 + input[27] * cospi_27_64;
+ step1[20] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[27] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = input[21] * cospi_11_64 - input[11] * cospi_21_64;
+ temp2 = input[21] * cospi_21_64 + input[11] * cospi_11_64;
+ step1[21] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[26] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = input[13] * cospi_19_64 - input[19] * cospi_13_64;
+ temp2 = input[13] * cospi_13_64 + input[19] * cospi_19_64;
+ step1[22] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[25] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = input[29] * cospi_3_64 - input[3] * cospi_29_64;
+ temp2 = input[29] * cospi_29_64 + input[3] * cospi_3_64;
+ step1[23] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[24] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ // stage 2
+ step2[0] = step1[0];
+ step2[1] = step1[1];
+ step2[2] = step1[2];
+ step2[3] = step1[3];
+ step2[4] = step1[4];
+ step2[5] = step1[5];
+ step2[6] = step1[6];
+ step2[7] = step1[7];
+
+ temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
+ temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
+ step2[8] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[15] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
+ temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
+ step2[9] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[14] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
+ temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
+ step2[10] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
+ temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
+ step2[11] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[12] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ step2[16] = HIGHBD_WRAPLOW(step1[16] + step1[17], bd);
+ step2[17] = HIGHBD_WRAPLOW(step1[16] - step1[17], bd);
+ step2[18] = HIGHBD_WRAPLOW(-step1[18] + step1[19], bd);
+ step2[19] = HIGHBD_WRAPLOW(step1[18] + step1[19], bd);
+ step2[20] = HIGHBD_WRAPLOW(step1[20] + step1[21], bd);
+ step2[21] = HIGHBD_WRAPLOW(step1[20] - step1[21], bd);
+ step2[22] = HIGHBD_WRAPLOW(-step1[22] + step1[23], bd);
+ step2[23] = HIGHBD_WRAPLOW(step1[22] + step1[23], bd);
+ step2[24] = HIGHBD_WRAPLOW(step1[24] + step1[25], bd);
+ step2[25] = HIGHBD_WRAPLOW(step1[24] - step1[25], bd);
+ step2[26] = HIGHBD_WRAPLOW(-step1[26] + step1[27], bd);
+ step2[27] = HIGHBD_WRAPLOW(step1[26] + step1[27], bd);
+ step2[28] = HIGHBD_WRAPLOW(step1[28] + step1[29], bd);
+ step2[29] = HIGHBD_WRAPLOW(step1[28] - step1[29], bd);
+ step2[30] = HIGHBD_WRAPLOW(-step1[30] + step1[31], bd);
+ step2[31] = HIGHBD_WRAPLOW(step1[30] + step1[31], bd);
+
+ // stage 3
+ step1[0] = step2[0];
+ step1[1] = step2[1];
+ step1[2] = step2[2];
+ step1[3] = step2[3];
+
+ temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
+ temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
+ step1[4] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[7] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
+ temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
+ step1[5] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+
+ step1[8] = HIGHBD_WRAPLOW(step2[8] + step2[9], bd);
+ step1[9] = HIGHBD_WRAPLOW(step2[8] - step2[9], bd);
+ step1[10] = HIGHBD_WRAPLOW(-step2[10] + step2[11], bd);
+ step1[11] = HIGHBD_WRAPLOW(step2[10] + step2[11], bd);
+ step1[12] = HIGHBD_WRAPLOW(step2[12] + step2[13], bd);
+ step1[13] = HIGHBD_WRAPLOW(step2[12] - step2[13], bd);
+ step1[14] = HIGHBD_WRAPLOW(-step2[14] + step2[15], bd);
+ step1[15] = HIGHBD_WRAPLOW(step2[14] + step2[15], bd);
+
+ step1[16] = step2[16];
+ step1[31] = step2[31];
+ temp1 = -step2[17] * cospi_4_64 + step2[30] * cospi_28_64;
+ temp2 = step2[17] * cospi_28_64 + step2[30] * cospi_4_64;
+ step1[17] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[30] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = -step2[18] * cospi_28_64 - step2[29] * cospi_4_64;
+ temp2 = -step2[18] * cospi_4_64 + step2[29] * cospi_28_64;
+ step1[18] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[29] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step1[19] = step2[19];
+ step1[20] = step2[20];
+ temp1 = -step2[21] * cospi_20_64 + step2[26] * cospi_12_64;
+ temp2 = step2[21] * cospi_12_64 + step2[26] * cospi_20_64;
+ step1[21] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[26] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = -step2[22] * cospi_12_64 - step2[25] * cospi_20_64;
+ temp2 = -step2[22] * cospi_20_64 + step2[25] * cospi_12_64;
+ step1[22] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[25] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step1[23] = step2[23];
+ step1[24] = step2[24];
+ step1[27] = step2[27];
+ step1[28] = step2[28];
+
+ // stage 4
+ temp1 = (step1[0] + step1[1]) * cospi_16_64;
+ temp2 = (step1[0] - step1[1]) * cospi_16_64;
+ step2[0] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[1] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
+ temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
+ step2[2] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[3] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step2[4] = HIGHBD_WRAPLOW(step1[4] + step1[5], bd);
+ step2[5] = HIGHBD_WRAPLOW(step1[4] - step1[5], bd);
+ step2[6] = HIGHBD_WRAPLOW(-step1[6] + step1[7], bd);
+ step2[7] = HIGHBD_WRAPLOW(step1[6] + step1[7], bd);
+
+ step2[8] = step1[8];
+ step2[15] = step1[15];
+ temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
+ temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
+ step2[9] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[14] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
+ temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
+ step2[10] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step2[11] = step1[11];
+ step2[12] = step1[12];
+
+ step2[16] = HIGHBD_WRAPLOW(step1[16] + step1[19], bd);
+ step2[17] = HIGHBD_WRAPLOW(step1[17] + step1[18], bd);
+ step2[18] = HIGHBD_WRAPLOW(step1[17] - step1[18], bd);
+ step2[19] = HIGHBD_WRAPLOW(step1[16] - step1[19], bd);
+ step2[20] = HIGHBD_WRAPLOW(-step1[20] + step1[23], bd);
+ step2[21] = HIGHBD_WRAPLOW(-step1[21] + step1[22], bd);
+ step2[22] = HIGHBD_WRAPLOW(step1[21] + step1[22], bd);
+ step2[23] = HIGHBD_WRAPLOW(step1[20] + step1[23], bd);
+
+ step2[24] = HIGHBD_WRAPLOW(step1[24] + step1[27], bd);
+ step2[25] = HIGHBD_WRAPLOW(step1[25] + step1[26], bd);
+ step2[26] = HIGHBD_WRAPLOW(step1[25] - step1[26], bd);
+ step2[27] = HIGHBD_WRAPLOW(step1[24] - step1[27], bd);
+ step2[28] = HIGHBD_WRAPLOW(-step1[28] + step1[31], bd);
+ step2[29] = HIGHBD_WRAPLOW(-step1[29] + step1[30], bd);
+ step2[30] = HIGHBD_WRAPLOW(step1[29] + step1[30], bd);
+ step2[31] = HIGHBD_WRAPLOW(step1[28] + step1[31], bd);
+
+ // stage 5
+ step1[0] = HIGHBD_WRAPLOW(step2[0] + step2[3], bd);
+ step1[1] = HIGHBD_WRAPLOW(step2[1] + step2[2], bd);
+ step1[2] = HIGHBD_WRAPLOW(step2[1] - step2[2], bd);
+ step1[3] = HIGHBD_WRAPLOW(step2[0] - step2[3], bd);
+ step1[4] = step2[4];
+ temp1 = (step2[6] - step2[5]) * cospi_16_64;
+ temp2 = (step2[5] + step2[6]) * cospi_16_64;
+ step1[5] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[6] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step1[7] = step2[7];
+
+ step1[8] = HIGHBD_WRAPLOW(step2[8] + step2[11], bd);
+ step1[9] = HIGHBD_WRAPLOW(step2[9] + step2[10], bd);
+ step1[10] = HIGHBD_WRAPLOW(step2[9] - step2[10], bd);
+ step1[11] = HIGHBD_WRAPLOW(step2[8] - step2[11], bd);
+ step1[12] = HIGHBD_WRAPLOW(-step2[12] + step2[15], bd);
+ step1[13] = HIGHBD_WRAPLOW(-step2[13] + step2[14], bd);
+ step1[14] = HIGHBD_WRAPLOW(step2[13] + step2[14], bd);
+ step1[15] = HIGHBD_WRAPLOW(step2[12] + step2[15], bd);
+
+ step1[16] = step2[16];
+ step1[17] = step2[17];
+ temp1 = -step2[18] * cospi_8_64 + step2[29] * cospi_24_64;
+ temp2 = step2[18] * cospi_24_64 + step2[29] * cospi_8_64;
+ step1[18] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[29] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = -step2[19] * cospi_8_64 + step2[28] * cospi_24_64;
+ temp2 = step2[19] * cospi_24_64 + step2[28] * cospi_8_64;
+ step1[19] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[28] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = -step2[20] * cospi_24_64 - step2[27] * cospi_8_64;
+ temp2 = -step2[20] * cospi_8_64 + step2[27] * cospi_24_64;
+ step1[20] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[27] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = -step2[21] * cospi_24_64 - step2[26] * cospi_8_64;
+ temp2 = -step2[21] * cospi_8_64 + step2[26] * cospi_24_64;
+ step1[21] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[26] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step1[22] = step2[22];
+ step1[23] = step2[23];
+ step1[24] = step2[24];
+ step1[25] = step2[25];
+ step1[30] = step2[30];
+ step1[31] = step2[31];
+
+ // stage 6
+ step2[0] = HIGHBD_WRAPLOW(step1[0] + step1[7], bd);
+ step2[1] = HIGHBD_WRAPLOW(step1[1] + step1[6], bd);
+ step2[2] = HIGHBD_WRAPLOW(step1[2] + step1[5], bd);
+ step2[3] = HIGHBD_WRAPLOW(step1[3] + step1[4], bd);
+ step2[4] = HIGHBD_WRAPLOW(step1[3] - step1[4], bd);
+ step2[5] = HIGHBD_WRAPLOW(step1[2] - step1[5], bd);
+ step2[6] = HIGHBD_WRAPLOW(step1[1] - step1[6], bd);
+ step2[7] = HIGHBD_WRAPLOW(step1[0] - step1[7], bd);
+ step2[8] = step1[8];
+ step2[9] = step1[9];
+ temp1 = (-step1[10] + step1[13]) * cospi_16_64;
+ temp2 = (step1[10] + step1[13]) * cospi_16_64;
+ step2[10] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[13] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = (-step1[11] + step1[12]) * cospi_16_64;
+ temp2 = (step1[11] + step1[12]) * cospi_16_64;
+ step2[11] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step2[12] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step2[14] = step1[14];
+ step2[15] = step1[15];
+
+ step2[16] = HIGHBD_WRAPLOW(step1[16] + step1[23], bd);
+ step2[17] = HIGHBD_WRAPLOW(step1[17] + step1[22], bd);
+ step2[18] = HIGHBD_WRAPLOW(step1[18] + step1[21], bd);
+ step2[19] = HIGHBD_WRAPLOW(step1[19] + step1[20], bd);
+ step2[20] = HIGHBD_WRAPLOW(step1[19] - step1[20], bd);
+ step2[21] = HIGHBD_WRAPLOW(step1[18] - step1[21], bd);
+ step2[22] = HIGHBD_WRAPLOW(step1[17] - step1[22], bd);
+ step2[23] = HIGHBD_WRAPLOW(step1[16] - step1[23], bd);
+
+ step2[24] = HIGHBD_WRAPLOW(-step1[24] + step1[31], bd);
+ step2[25] = HIGHBD_WRAPLOW(-step1[25] + step1[30], bd);
+ step2[26] = HIGHBD_WRAPLOW(-step1[26] + step1[29], bd);
+ step2[27] = HIGHBD_WRAPLOW(-step1[27] + step1[28], bd);
+ step2[28] = HIGHBD_WRAPLOW(step1[27] + step1[28], bd);
+ step2[29] = HIGHBD_WRAPLOW(step1[26] + step1[29], bd);
+ step2[30] = HIGHBD_WRAPLOW(step1[25] + step1[30], bd);
+ step2[31] = HIGHBD_WRAPLOW(step1[24] + step1[31], bd);
+
+ // stage 7
+ step1[0] = HIGHBD_WRAPLOW(step2[0] + step2[15], bd);
+ step1[1] = HIGHBD_WRAPLOW(step2[1] + step2[14], bd);
+ step1[2] = HIGHBD_WRAPLOW(step2[2] + step2[13], bd);
+ step1[3] = HIGHBD_WRAPLOW(step2[3] + step2[12], bd);
+ step1[4] = HIGHBD_WRAPLOW(step2[4] + step2[11], bd);
+ step1[5] = HIGHBD_WRAPLOW(step2[5] + step2[10], bd);
+ step1[6] = HIGHBD_WRAPLOW(step2[6] + step2[9], bd);
+ step1[7] = HIGHBD_WRAPLOW(step2[7] + step2[8], bd);
+ step1[8] = HIGHBD_WRAPLOW(step2[7] - step2[8], bd);
+ step1[9] = HIGHBD_WRAPLOW(step2[6] - step2[9], bd);
+ step1[10] = HIGHBD_WRAPLOW(step2[5] - step2[10], bd);
+ step1[11] = HIGHBD_WRAPLOW(step2[4] - step2[11], bd);
+ step1[12] = HIGHBD_WRAPLOW(step2[3] - step2[12], bd);
+ step1[13] = HIGHBD_WRAPLOW(step2[2] - step2[13], bd);
+ step1[14] = HIGHBD_WRAPLOW(step2[1] - step2[14], bd);
+ step1[15] = HIGHBD_WRAPLOW(step2[0] - step2[15], bd);
+
+ step1[16] = step2[16];
+ step1[17] = step2[17];
+ step1[18] = step2[18];
+ step1[19] = step2[19];
+ temp1 = (-step2[20] + step2[27]) * cospi_16_64;
+ temp2 = (step2[20] + step2[27]) * cospi_16_64;
+ step1[20] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[27] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = (-step2[21] + step2[26]) * cospi_16_64;
+ temp2 = (step2[21] + step2[26]) * cospi_16_64;
+ step1[21] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[26] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = (-step2[22] + step2[25]) * cospi_16_64;
+ temp2 = (step2[22] + step2[25]) * cospi_16_64;
+ step1[22] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[25] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ temp1 = (-step2[23] + step2[24]) * cospi_16_64;
+ temp2 = (step2[23] + step2[24]) * cospi_16_64;
+ step1[23] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
+ step1[24] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
+ step1[28] = step2[28];
+ step1[29] = step2[29];
+ step1[30] = step2[30];
+ step1[31] = step2[31];
+
+ // final stage
+ output[0] = HIGHBD_WRAPLOW(step1[0] + step1[31], bd);
+ output[1] = HIGHBD_WRAPLOW(step1[1] + step1[30], bd);
+ output[2] = HIGHBD_WRAPLOW(step1[2] + step1[29], bd);
+ output[3] = HIGHBD_WRAPLOW(step1[3] + step1[28], bd);
+ output[4] = HIGHBD_WRAPLOW(step1[4] + step1[27], bd);
+ output[5] = HIGHBD_WRAPLOW(step1[5] + step1[26], bd);
+ output[6] = HIGHBD_WRAPLOW(step1[6] + step1[25], bd);
+ output[7] = HIGHBD_WRAPLOW(step1[7] + step1[24], bd);
+ output[8] = HIGHBD_WRAPLOW(step1[8] + step1[23], bd);
+ output[9] = HIGHBD_WRAPLOW(step1[9] + step1[22], bd);
+ output[10] = HIGHBD_WRAPLOW(step1[10] + step1[21], bd);
+ output[11] = HIGHBD_WRAPLOW(step1[11] + step1[20], bd);
+ output[12] = HIGHBD_WRAPLOW(step1[12] + step1[19], bd);
+ output[13] = HIGHBD_WRAPLOW(step1[13] + step1[18], bd);
+ output[14] = HIGHBD_WRAPLOW(step1[14] + step1[17], bd);
+ output[15] = HIGHBD_WRAPLOW(step1[15] + step1[16], bd);
+ output[16] = HIGHBD_WRAPLOW(step1[15] - step1[16], bd);
+ output[17] = HIGHBD_WRAPLOW(step1[14] - step1[17], bd);
+ output[18] = HIGHBD_WRAPLOW(step1[13] - step1[18], bd);
+ output[19] = HIGHBD_WRAPLOW(step1[12] - step1[19], bd);
+ output[20] = HIGHBD_WRAPLOW(step1[11] - step1[20], bd);
+ output[21] = HIGHBD_WRAPLOW(step1[10] - step1[21], bd);
+ output[22] = HIGHBD_WRAPLOW(step1[9] - step1[22], bd);
+ output[23] = HIGHBD_WRAPLOW(step1[8] - step1[23], bd);
+ output[24] = HIGHBD_WRAPLOW(step1[7] - step1[24], bd);
+ output[25] = HIGHBD_WRAPLOW(step1[6] - step1[25], bd);
+ output[26] = HIGHBD_WRAPLOW(step1[5] - step1[26], bd);
+ output[27] = HIGHBD_WRAPLOW(step1[4] - step1[27], bd);
+ output[28] = HIGHBD_WRAPLOW(step1[3] - step1[28], bd);
+ output[29] = HIGHBD_WRAPLOW(step1[2] - step1[29], bd);
+ output[30] = HIGHBD_WRAPLOW(step1[1] - step1[30], bd);
+ output[31] = HIGHBD_WRAPLOW(step1[0] - step1[31], bd);
+}
+
#endif // CONFIG_HIGHBITDEPTH
diff --git a/aom_dsp/x86/inv_txfm_sse2.c b/aom_dsp/x86/inv_txfm_sse2.c
index 5795a18..be200df 100644
--- a/aom_dsp/x86/inv_txfm_sse2.c
+++ b/aom_dsp/x86/inv_txfm_sse2.c
@@ -3628,4 +3628,107 @@
}
}
+void aom_highbd_idct8x8_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
+ int stride, int bd) {
+ tran_low_t out[8 * 8] = { 0 };
+ tran_low_t *outptr = out;
+ int i, j, test;
+ __m128i inptr[8];
+ __m128i min_input, max_input, temp1, temp2, sign_bits;
+ uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
+ const __m128i zero = _mm_set1_epi16(0);
+ const __m128i sixteen = _mm_set1_epi16(16);
+ const __m128i max = _mm_set1_epi16(6201);
+ const __m128i min = _mm_set1_epi16(-6201);
+ int optimised_cols = 0;
+
+ // Load input into __m128i & pack to 16 bits
+ for (i = 0; i < 8; i++) {
+ temp1 = _mm_loadu_si128((const __m128i *)(input + 8 * i));
+ temp2 = _mm_loadu_si128((const __m128i *)(input + 8 * i + 4));
+ inptr[i] = _mm_packs_epi32(temp1, temp2);
+ }
+
+ // Find the min & max for the row transform
+ // only first 4 row has non-zero coefs
+ max_input = _mm_max_epi16(inptr[0], inptr[1]);
+ min_input = _mm_min_epi16(inptr[0], inptr[1]);
+ for (i = 2; i < 4; i++) {
+ max_input = _mm_max_epi16(max_input, inptr[i]);
+ min_input = _mm_min_epi16(min_input, inptr[i]);
+ }
+ max_input = _mm_cmpgt_epi16(max_input, max);
+ min_input = _mm_cmplt_epi16(min_input, min);
+ temp1 = _mm_or_si128(max_input, min_input);
+ test = _mm_movemask_epi8(temp1);
+
+ if (!test) {
+ // Do the row transform
+ aom_idct8_sse2(inptr);
+
+ // Find the min & max for the column transform
+ // N.B. Only first 4 cols contain non-zero coeffs
+ max_input = _mm_max_epi16(inptr[0], inptr[1]);
+ min_input = _mm_min_epi16(inptr[0], inptr[1]);
+ for (i = 2; i < 8; i++) {
+ max_input = _mm_max_epi16(max_input, inptr[i]);
+ min_input = _mm_min_epi16(min_input, inptr[i]);
+ }
+ max_input = _mm_cmpgt_epi16(max_input, max);
+ min_input = _mm_cmplt_epi16(min_input, min);
+ temp1 = _mm_or_si128(max_input, min_input);
+ test = _mm_movemask_epi8(temp1);
+
+ if (test) {
+ // Use fact only first 4 rows contain non-zero coeffs
+ array_transpose_4X8(inptr, inptr);
+ for (i = 0; i < 4; i++) {
+ sign_bits = _mm_cmplt_epi16(inptr[i], zero);
+ temp1 = _mm_unpackhi_epi16(inptr[i], sign_bits);
+ temp2 = _mm_unpacklo_epi16(inptr[i], sign_bits);
+ _mm_storeu_si128((__m128i *)(outptr + 4 * (2 * i + 1)), temp1);
+ _mm_storeu_si128((__m128i *)(outptr + 4 * (2 * i)), temp2);
+ }
+ } else {
+ // Set to use the optimised transform for the column
+ optimised_cols = 1;
+ }
+ } else {
+ // Run the un-optimised row transform
+ for (i = 0; i < 4; ++i) {
+ aom_highbd_idct8_c(input, outptr, bd);
+ input += 8;
+ outptr += 8;
+ }
+ }
+
+ if (optimised_cols) {
+ aom_idct8_sse2(inptr);
+
+ // Final round & shift and Reconstruction and Store
+ {
+ __m128i d[8];
+ for (i = 0; i < 8; i++) {
+ inptr[i] = _mm_add_epi16(inptr[i], sixteen);
+ d[i] = _mm_loadu_si128((const __m128i *)(dest + stride * i));
+ inptr[i] = _mm_srai_epi16(inptr[i], 5);
+ d[i] = clamp_high_sse2(_mm_adds_epi16(d[i], inptr[i]), bd);
+ // Store
+ _mm_storeu_si128((__m128i *)(dest + stride * i), d[i]);
+ }
+ }
+ } else {
+ // Run the un-optimised column transform
+ tran_low_t temp_in[8], temp_out[8];
+ for (i = 0; i < 8; ++i) {
+ for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
+ aom_highbd_idct8_c(temp_in, temp_out, bd);
+ for (j = 0; j < 8; ++j) {
+ dest[j * stride + i] = highbd_clip_pixel_add(
+ dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 5), bd);
+ }
+ }
+ }
+}
+
#endif // CONFIG_HIGHBITDEPTH
diff --git a/av1/common/av1_inv_txfm2d.c b/av1/common/av1_inv_txfm2d.c
index 688cfb4..d847ab1 100644
--- a/av1/common/av1_inv_txfm2d.c
+++ b/av1/common/av1_inv_txfm2d.c
@@ -30,115 +30,64 @@
}
#if CONFIG_EXT_TX
-const TXFM_2D_CFG *inv_txfm_cfg_ls[TX_TYPES][TX_SIZES] = {
- // DCT_DCT
+static const TXFM_2D_CFG *inv_txfm_cfg_ls[FLIPADST_ADST + 1][TX_SIZES] = {
{
#if CONFIG_CB4X4
NULL,
#endif
&inv_txfm_2d_cfg_dct_dct_4, &inv_txfm_2d_cfg_dct_dct_8,
&inv_txfm_2d_cfg_dct_dct_16, &inv_txfm_2d_cfg_dct_dct_32 },
- // ADST_DCT
{
#if CONFIG_CB4X4
NULL,
#endif
&inv_txfm_2d_cfg_adst_dct_4, &inv_txfm_2d_cfg_adst_dct_8,
&inv_txfm_2d_cfg_adst_dct_16, &inv_txfm_2d_cfg_adst_dct_32 },
- // DCT_ADST
{
#if CONFIG_CB4X4
NULL,
#endif
&inv_txfm_2d_cfg_dct_adst_4, &inv_txfm_2d_cfg_dct_adst_8,
&inv_txfm_2d_cfg_dct_adst_16, &inv_txfm_2d_cfg_dct_adst_32 },
- // ADST_ADST
{
#if CONFIG_CB4X4
NULL,
#endif
&inv_txfm_2d_cfg_adst_adst_4, &inv_txfm_2d_cfg_adst_adst_8,
&inv_txfm_2d_cfg_adst_adst_16, &inv_txfm_2d_cfg_adst_adst_32 },
- // FLIPADST_DCT
{
#if CONFIG_CB4X4
NULL,
#endif
&inv_txfm_2d_cfg_adst_dct_4, &inv_txfm_2d_cfg_adst_dct_8,
&inv_txfm_2d_cfg_adst_dct_16, &inv_txfm_2d_cfg_adst_dct_32 },
- // DCT_FLIPADST
{
#if CONFIG_CB4X4
NULL,
#endif
&inv_txfm_2d_cfg_dct_adst_4, &inv_txfm_2d_cfg_dct_adst_8,
&inv_txfm_2d_cfg_dct_adst_16, &inv_txfm_2d_cfg_dct_adst_32 },
- // FLIPADST_FLIPADST
{
#if CONFIG_CB4X4
NULL,
#endif
&inv_txfm_2d_cfg_adst_adst_4, &inv_txfm_2d_cfg_adst_adst_8,
&inv_txfm_2d_cfg_adst_adst_16, &inv_txfm_2d_cfg_adst_adst_32 },
- // ADST_FLIPADST
{
#if CONFIG_CB4X4
NULL,
#endif
&inv_txfm_2d_cfg_adst_adst_4, &inv_txfm_2d_cfg_adst_adst_8,
&inv_txfm_2d_cfg_adst_adst_16, &inv_txfm_2d_cfg_adst_adst_32 },
- // FLIPADST_ADST
{
#if CONFIG_CB4X4
NULL,
#endif
&inv_txfm_2d_cfg_adst_adst_4, &inv_txfm_2d_cfg_adst_adst_8,
&inv_txfm_2d_cfg_adst_adst_16, &inv_txfm_2d_cfg_adst_adst_32 },
- { // IDTX
-#if CONFIG_CB4X4
- NULL,
-#endif
- &inv_txfm_2d_cfg_adst_adst_4, &inv_txfm_2d_cfg_adst_adst_8,
- &inv_txfm_2d_cfg_adst_adst_16, &inv_txfm_2d_cfg_adst_adst_32 },
- { // V_DCT
-#if CONFIG_CB4X4
- NULL,
-#endif
- &inv_txfm_2d_cfg_dct_adst_4, &inv_txfm_2d_cfg_dct_adst_8,
- &inv_txfm_2d_cfg_dct_adst_16, &inv_txfm_2d_cfg_dct_adst_32 },
- { // H_DCT
-#if CONFIG_CB4X4
- NULL,
-#endif
- &inv_txfm_2d_cfg_adst_dct_4, &inv_txfm_2d_cfg_adst_dct_8,
- &inv_txfm_2d_cfg_adst_dct_16, &inv_txfm_2d_cfg_adst_dct_32 },
- { // V_ADST
-#if CONFIG_CB4X4
- NULL,
-#endif
- &inv_txfm_2d_cfg_adst_adst_4, &inv_txfm_2d_cfg_adst_adst_8,
- &inv_txfm_2d_cfg_adst_adst_16, &inv_txfm_2d_cfg_adst_adst_32 },
- { // H_ADST
-#if CONFIG_CB4X4
- NULL,
-#endif
- &inv_txfm_2d_cfg_adst_adst_4, &inv_txfm_2d_cfg_adst_adst_8,
- &inv_txfm_2d_cfg_adst_adst_16, &inv_txfm_2d_cfg_adst_adst_32 },
- { // V_FLIP_ADST
-#if CONFIG_CB4X4
- NULL,
-#endif
- &inv_txfm_2d_cfg_adst_adst_4, &inv_txfm_2d_cfg_adst_adst_8,
- &inv_txfm_2d_cfg_adst_adst_16, &inv_txfm_2d_cfg_adst_adst_32 },
- { // H_FLIP_ADST
-#if CONFIG_CB4X4
- NULL,
-#endif
- &inv_txfm_2d_cfg_adst_adst_4, &inv_txfm_2d_cfg_adst_adst_8,
- &inv_txfm_2d_cfg_adst_adst_16, &inv_txfm_2d_cfg_adst_adst_32 },
};
#else
-const TXFM_2D_CFG *inv_txfm_cfg_ls[TX_TYPES][TX_SIZES] = {
+static const TXFM_2D_CFG *inv_txfm_cfg_ls[TX_TYPES][TX_SIZES] = {
{
#if CONFIG_CB4X4
NULL,
diff --git a/av1/common/av1_inv_txfm2d_cfg.h b/av1/common/av1_inv_txfm2d_cfg.h
index 9eabc2e..badafd5 100644
--- a/av1/common/av1_inv_txfm2d_cfg.h
+++ b/av1/common/av1_inv_txfm2d_cfg.h
@@ -442,6 +442,4 @@
TXFM_TYPE_DCT32
}; // .txfm_type_row
-extern const TXFM_2D_CFG *inv_txfm_cfg_ls[TX_TYPES][TX_SIZES];
-
#endif // AV1_INV_TXFM2D_CFG_H_
diff --git a/av1/common/idct.c b/av1/common/idct.c
index 0ea58bf..2c960cd 100644
--- a/av1/common/idct.c
+++ b/av1/common/idct.c
@@ -116,100 +116,38 @@
#endif // CONFIG_TX64X64
#if CONFIG_HIGHBITDEPTH
-static void highbd_idct4(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
- int bd) {
- (void)bd;
- av1_idct4_new(input, output, cos_bit, stage_range);
-}
-
-static void highbd_idct8(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
- int bd) {
- (void)bd;
- av1_idct8_new(input, output, cos_bit, stage_range);
-}
-
-static void highbd_idct16(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
- int bd) {
- (void)bd;
- av1_idct16_new(input, output, cos_bit, stage_range);
-}
-
-static void highbd_idct32(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
- int bd) {
- (void)bd;
- av1_idct32_new(input, output, cos_bit, stage_range);
-}
-
-static void highbd_iadst4(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
- int bd) {
- (void)bd;
- av1_iadst4_new(input, output, cos_bit, stage_range);
-}
-
-static void highbd_iadst8(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
- int bd) {
- (void)bd;
- av1_iadst8_new(input, output, cos_bit, stage_range);
-}
-
-static void highbd_iadst16(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
- int bd) {
- (void)bd;
- av1_iadst16_new(input, output, cos_bit, stage_range);
-}
-
#if CONFIG_EXT_TX
static void highbd_iidtx4_c(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
int bd) {
int i;
- (void)cos_bit;
- (void)stage_range;
for (i = 0; i < 4; ++i)
output[i] = HIGHBD_WRAPLOW(dct_const_round_shift(input[i] * Sqrt2), bd);
}
static void highbd_iidtx8_c(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
int bd) {
int i;
(void)bd;
- (void)cos_bit;
- (void)stage_range;
for (i = 0; i < 8; ++i) output[i] = input[i] * 2;
}
static void highbd_iidtx16_c(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
int bd) {
int i;
- (void)cos_bit;
- (void)stage_range;
for (i = 0; i < 16; ++i)
output[i] = HIGHBD_WRAPLOW(dct_const_round_shift(input[i] * 2 * Sqrt2), bd);
}
static void highbd_iidtx32_c(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
int bd) {
int i;
(void)bd;
- (void)cos_bit;
- (void)stage_range;
for (i = 0; i < 32; ++i) output[i] = input[i] * 4;
}
#endif // CONFIG_EXT_TX
static void highbd_ihalfright32_c(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit,
- const int8_t *stage_range, int bd) {
+ int bd) {
int i;
tran_low_t inputhalf[16];
// Multiply input by sqrt(2)
@@ -219,17 +157,14 @@
for (i = 0; i < 16; ++i) {
output[i] = input[16 + i] * 4;
}
- highbd_idct16(inputhalf, output + 16, cos_bit, stage_range, bd);
+ aom_highbd_idct16_c(inputhalf, output + 16, bd);
// Note overall scaling factor is 4 times orthogonal
}
#if CONFIG_EXT_TX
#if CONFIG_TX64X64
static void highbd_iidtx64_c(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit, const int8_t *stage_range,
int bd) {
- (void)cos_bit;
- (void)stage_range;
int i;
for (i = 0; i < 64; ++i)
output[i] = HIGHBD_WRAPLOW(dct_const_round_shift(input[i] * 4 * Sqrt2), bd);
@@ -240,8 +175,7 @@
#if CONFIG_TX64X64
// For use in lieu of ADST
static void highbd_ihalfright64_c(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit,
- const int8_t *stage_range, int bd) {
+ int bd) {
int i;
tran_low_t inputhalf[32];
// Multiply input by sqrt(2)
@@ -252,17 +186,14 @@
output[i] =
HIGHBD_WRAPLOW(dct_const_round_shift(input[32 + i] * 4 * Sqrt2), bd);
}
- highbd_idct32(inputhalf, output + 32, cos_bit, stage_range, bd);
+ aom_highbd_idct32_c(inputhalf, output + 32, bd);
// Note overall scaling factor is 4 * sqrt(2) times orthogonal
}
static void highbd_idct64_col_c(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit,
- const int8_t *stage_range, int bd) {
+ int bd) {
int32_t in[64], out[64];
int i;
- (void)cos_bit;
- (void)stage_range;
(void)bd;
for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
av1_idct64_new(in, out, inv_cos_bit_col_dct_dct_64,
@@ -271,12 +202,9 @@
}
static void highbd_idct64_row_c(const tran_low_t *input, tran_low_t *output,
- const int8_t *cos_bit,
- const int8_t *stage_range, int bd) {
+ int bd) {
int32_t in[64], out[64];
int i;
- (void)cos_bit;
- (void)stage_range;
(void)bd;
for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
av1_idct64_new(in, out, inv_cos_bit_row_dct_dct_64,
@@ -1588,55 +1516,27 @@
#endif // CONFIG_TX64X64
#if CONFIG_HIGHBITDEPTH
-
-const TXFM_2D_CFG *inv_txfm_cfg_ls[TX_TYPES][TX_SIZES];
-
-typedef struct {
- const int8_t *cos_bit;
- const int8_t *stage_range;
-} tx_1d_cfg;
-
-typedef struct {
- tx_1d_cfg row;
- tx_1d_cfg col;
-} tx_2d_cfg;
-
-tx_2d_cfg inv_tx_cfg(int tx_type, int tx_size_row, int tx_size_col) {
- const TXFM_2D_CFG *cfg_row = inv_txfm_cfg_ls[tx_type][tx_size_row];
- const int8_t *stage_range_row = cfg_row->stage_range_row;
- const int8_t *cos_bit_row = cfg_row->cos_bit_row;
-
- const TXFM_2D_CFG *cfg_col = inv_txfm_cfg_ls[tx_type][tx_size_col];
- const int8_t *stage_range_col = cfg_col->stage_range_col;
- const int8_t *cos_bit_col = cfg_col->cos_bit_col;
-
- tx_2d_cfg cfg = {
- { cos_bit_row, stage_range_row }, { cos_bit_col, stage_range_col },
- };
- return cfg;
-}
-
void av1_highbd_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_4[] = {
- { highbd_idct4, highbd_idct4 }, // DCT_DCT
- { highbd_iadst4, highbd_idct4 }, // ADST_DCT
- { highbd_idct4, highbd_iadst4 }, // DCT_ADST
- { highbd_iadst4, highbd_iadst4 }, // ADST_ADST
+ { aom_highbd_idct4_c, aom_highbd_idct4_c }, // DCT_DCT
+ { aom_highbd_iadst4_c, aom_highbd_idct4_c }, // ADST_DCT
+ { aom_highbd_idct4_c, aom_highbd_iadst4_c }, // DCT_ADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst4_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_iadst4, highbd_idct4 }, // FLIPADST_DCT
- { highbd_idct4, highbd_iadst4 }, // DCT_FLIPADST
- { highbd_iadst4, highbd_iadst4 }, // FLIPADST_FLIPADST
- { highbd_iadst4, highbd_iadst4 }, // ADST_FLIPADST
- { highbd_iadst4, highbd_iadst4 }, // FLIPADST_ADST
- { highbd_iidtx4_c, highbd_iidtx4_c }, // IDTX
- { highbd_idct4, highbd_iidtx4_c }, // V_DCT
- { highbd_iidtx4_c, highbd_idct4 }, // H_DCT
- { highbd_iadst4, highbd_iidtx4_c }, // V_ADST
- { highbd_iidtx4_c, highbd_iadst4 }, // H_ADST
- { highbd_iadst4, highbd_iidtx4_c }, // V_FLIPADST
- { highbd_iidtx4_c, highbd_iadst4 }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { aom_highbd_iadst4_c, aom_highbd_idct4_c }, // FLIPADST_DCT
+ { aom_highbd_idct4_c, aom_highbd_iadst4_c }, // DCT_FLIPADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst4_c }, // FLIPADST_FLIPADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst4_c }, // ADST_FLIPADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst4_c }, // FLIPADST_ADST
+ { highbd_iidtx4_c, highbd_iidtx4_c }, // IDTX
+ { aom_highbd_idct4_c, highbd_iidtx4_c }, // V_DCT
+ { highbd_iidtx4_c, aom_highbd_idct4_c }, // H_DCT
+ { aom_highbd_iadst4_c, highbd_iidtx4_c }, // V_ADST
+ { highbd_iidtx4_c, aom_highbd_iadst4_c }, // H_ADST
+ { aom_highbd_iadst4_c, highbd_iidtx4_c }, // V_FLIPADST
+ { highbd_iidtx4_c, aom_highbd_iadst4_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
@@ -1647,12 +1547,9 @@
tran_low_t *outp = &out[0][0];
int outstride = 4;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_4X4, TX_4X4);
-
// inverse transform row vectors
for (i = 0; i < 4; ++i) {
- HIGH_IHT_4[tx_type].rows(input, out[i], cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_4[tx_type].rows(input, out[i], bd);
input += 4;
}
@@ -1667,8 +1564,7 @@
// inverse transform column vectors
for (i = 0; i < 4; ++i) {
- HIGH_IHT_4[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ HIGH_IHT_4[tx_type].cols(out[i], out[i], bd);
}
#if CONFIG_EXT_TX
@@ -1689,24 +1585,24 @@
void av1_highbd_iht4x8_32_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_4x8[] = {
- { highbd_idct8, highbd_idct4 }, // DCT_DCT
- { highbd_iadst8, highbd_idct4 }, // ADST_DCT
- { highbd_idct8, highbd_iadst4 }, // DCT_ADST
- { highbd_iadst8, highbd_iadst4 }, // ADST_ADST
+ { aom_highbd_idct8_c, aom_highbd_idct4_c }, // DCT_DCT
+ { aom_highbd_iadst8_c, aom_highbd_idct4_c }, // ADST_DCT
+ { aom_highbd_idct8_c, aom_highbd_iadst4_c }, // DCT_ADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst4_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_iadst8, highbd_idct4 }, // FLIPADST_DCT
- { highbd_idct8, highbd_iadst4 }, // DCT_FLIPADST
- { highbd_iadst8, highbd_iadst4 }, // FLIPADST_FLIPADST
- { highbd_iadst8, highbd_iadst4 }, // ADST_FLIPADST
- { highbd_iadst8, highbd_iadst4 }, // FLIPADST_ADST
- { highbd_iidtx8_c, highbd_iidtx4_c }, // IDTX
- { highbd_idct8, highbd_iidtx4_c }, // V_DCT
- { highbd_iidtx8_c, highbd_idct4 }, // H_DCT
- { highbd_iadst8, highbd_iidtx4_c }, // V_ADST
- { highbd_iidtx8_c, highbd_iadst4 }, // H_ADST
- { highbd_iadst8, highbd_iidtx4_c }, // V_FLIPADST
- { highbd_iidtx8_c, highbd_iadst4 }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { aom_highbd_iadst8_c, aom_highbd_idct4_c }, // FLIPADST_DCT
+ { aom_highbd_idct8_c, aom_highbd_iadst4_c }, // DCT_FLIPADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst4_c }, // FLIPADST_FLIPADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst4_c }, // ADST_FLIPADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst4_c }, // FLIPADST_ADST
+ { highbd_iidtx8_c, highbd_iidtx4_c }, // IDTX
+ { aom_highbd_idct8_c, highbd_iidtx4_c }, // V_DCT
+ { highbd_iidtx8_c, aom_highbd_idct4_c }, // H_DCT
+ { aom_highbd_iadst8_c, highbd_iidtx4_c }, // V_ADST
+ { highbd_iidtx8_c, aom_highbd_iadst4_c }, // H_ADST
+ { aom_highbd_iadst8_c, highbd_iidtx4_c }, // V_FLIPADST
+ { highbd_iidtx8_c, aom_highbd_iadst4_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 4;
const int n2 = 8;
@@ -1718,12 +1614,9 @@
tran_low_t *outp = &out[0][0];
int outstride = n2;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_4X4, TX_8X8);
-
// inverse transform row vectors, and transpose
for (i = 0; i < n2; ++i) {
- HIGH_IHT_4x8[tx_type].rows(input, outtmp, cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_4x8[tx_type].rows(input, outtmp, bd);
for (j = 0; j < n; ++j) {
out[j][i] = HIGHBD_WRAPLOW(dct_const_round_shift(outtmp[j] * Sqrt2), bd);
}
@@ -1732,8 +1625,7 @@
// inverse transform column vectors
for (i = 0; i < n; ++i) {
- HIGH_IHT_4x8[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ HIGH_IHT_4x8[tx_type].cols(out[i], out[i], bd);
}
#if CONFIG_EXT_TX
@@ -1754,24 +1646,24 @@
void av1_highbd_iht8x4_32_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_8x4[] = {
- { highbd_idct4, highbd_idct8 }, // DCT_DCT
- { highbd_iadst4, highbd_idct8 }, // ADST_DCT
- { highbd_idct4, highbd_iadst8 }, // DCT_ADST
- { highbd_iadst4, highbd_iadst8 }, // ADST_ADST
+ { aom_highbd_idct4_c, aom_highbd_idct8_c }, // DCT_DCT
+ { aom_highbd_iadst4_c, aom_highbd_idct8_c }, // ADST_DCT
+ { aom_highbd_idct4_c, aom_highbd_iadst8_c }, // DCT_ADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst8_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_iadst4, highbd_idct8 }, // FLIPADST_DCT
- { highbd_idct4, highbd_iadst8 }, // DCT_FLIPADST
- { highbd_iadst4, highbd_iadst8 }, // FLIPADST_FLIPADST
- { highbd_iadst4, highbd_iadst8 }, // ADST_FLIPADST
- { highbd_iadst4, highbd_iadst8 }, // FLIPADST_ADST
- { highbd_iidtx4_c, highbd_iidtx8_c }, // IDTX
- { highbd_idct4, highbd_iidtx8_c }, // V_DCT
- { highbd_iidtx4_c, highbd_idct8 }, // H_DCT
- { highbd_iadst4, highbd_iidtx8_c }, // V_ADST
- { highbd_iidtx4_c, highbd_iadst8 }, // H_ADST
- { highbd_iadst4, highbd_iidtx8_c }, // V_FLIPADST
- { highbd_iidtx4_c, highbd_iadst8 }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { aom_highbd_iadst4_c, aom_highbd_idct8_c }, // FLIPADST_DCT
+ { aom_highbd_idct4_c, aom_highbd_iadst8_c }, // DCT_FLIPADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst8_c }, // FLIPADST_FLIPADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst8_c }, // ADST_FLIPADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst8_c }, // FLIPADST_ADST
+ { highbd_iidtx4_c, highbd_iidtx8_c }, // IDTX
+ { aom_highbd_idct4_c, highbd_iidtx8_c }, // V_DCT
+ { highbd_iidtx4_c, aom_highbd_idct8_c }, // H_DCT
+ { aom_highbd_iadst4_c, highbd_iidtx8_c }, // V_ADST
+ { highbd_iidtx4_c, aom_highbd_iadst8_c }, // H_ADST
+ { aom_highbd_iadst4_c, highbd_iidtx8_c }, // V_FLIPADST
+ { highbd_iidtx4_c, aom_highbd_iadst8_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 4;
const int n2 = 8;
@@ -1783,12 +1675,9 @@
tran_low_t *outp = &out[0][0];
int outstride = n;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_8X8, TX_4X4);
-
// inverse transform row vectors, and transpose
for (i = 0; i < n; ++i) {
- HIGH_IHT_8x4[tx_type].rows(input, outtmp, cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_8x4[tx_type].rows(input, outtmp, bd);
for (j = 0; j < n2; ++j) {
out[j][i] = HIGHBD_WRAPLOW(dct_const_round_shift(outtmp[j] * Sqrt2), bd);
}
@@ -1797,8 +1686,7 @@
// inverse transform column vectors
for (i = 0; i < n2; ++i) {
- HIGH_IHT_8x4[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ HIGH_IHT_8x4[tx_type].cols(out[i], out[i], bd);
}
#if CONFIG_EXT_TX
@@ -1819,24 +1707,24 @@
void av1_highbd_iht4x16_64_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_4x16[] = {
- { highbd_idct16, highbd_idct4 }, // DCT_DCT
- { highbd_iadst16, highbd_idct4 }, // ADST_DCT
- { highbd_idct16, highbd_iadst4 }, // DCT_ADST
- { highbd_iadst16, highbd_iadst4 }, // ADST_ADST
+ { aom_highbd_idct16_c, aom_highbd_idct4_c }, // DCT_DCT
+ { aom_highbd_iadst16_c, aom_highbd_idct4_c }, // ADST_DCT
+ { aom_highbd_idct16_c, aom_highbd_iadst4_c }, // DCT_ADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst4_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_iadst16, highbd_idct4 }, // FLIPADST_DCT
- { highbd_idct16, highbd_iadst4 }, // DCT_FLIPADST
- { highbd_iadst16, highbd_iadst4 }, // FLIPADST_FLIPADST
- { highbd_iadst16, highbd_iadst4 }, // ADST_FLIPADST
- { highbd_iadst16, highbd_iadst4 }, // FLIPADST_ADST
- { highbd_iidtx16_c, highbd_iidtx4_c }, // IDTX
- { highbd_idct16, highbd_iidtx4_c }, // V_DCT
- { highbd_iidtx16_c, highbd_idct4 }, // H_DCT
- { highbd_iadst16, highbd_iidtx4_c }, // V_ADST
- { highbd_iidtx16_c, highbd_iadst4 }, // H_ADST
- { highbd_iadst16, highbd_iidtx4_c }, // V_FLIPADST
- { highbd_iidtx16_c, highbd_iadst4 }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { aom_highbd_iadst16_c, aom_highbd_idct4_c }, // FLIPADST_DCT
+ { aom_highbd_idct16_c, aom_highbd_iadst4_c }, // DCT_FLIPADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst4_c }, // FLIPADST_FLIPADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst4_c }, // ADST_FLIPADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst4_c }, // FLIPADST_ADST
+ { highbd_iidtx16_c, highbd_iidtx4_c }, // IDTX
+ { aom_highbd_idct16_c, highbd_iidtx4_c }, // V_DCT
+ { highbd_iidtx16_c, aom_highbd_idct4_c }, // H_DCT
+ { aom_highbd_iadst16_c, highbd_iidtx4_c }, // V_ADST
+ { highbd_iidtx16_c, aom_highbd_iadst4_c }, // H_ADST
+ { aom_highbd_iadst16_c, highbd_iidtx4_c }, // V_FLIPADST
+ { highbd_iidtx16_c, aom_highbd_iadst4_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 4;
const int n4 = 16;
@@ -1848,20 +1736,15 @@
tran_low_t *outp = &out[0][0];
int outstride = n4;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_4X4, TX_16X16);
-
// inverse transform row vectors, and transpose
for (i = 0; i < n4; ++i) {
- HIGH_IHT_4x16[tx_type].rows(input, outtmp, cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_4x16[tx_type].rows(input, outtmp, bd);
for (j = 0; j < n; ++j) out[j][i] = outtmp[j];
input += n;
}
// inverse transform column vectors
- for (i = 0; i < n; ++i)
- HIGH_IHT_4x16[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ for (i = 0; i < n; ++i) HIGH_IHT_4x16[tx_type].cols(out[i], out[i], bd);
#if CONFIG_EXT_TX
maybe_flip_strides16(&dest, &stride, &outp, &outstride, tx_type, n4, n);
@@ -1881,24 +1764,24 @@
void av1_highbd_iht16x4_64_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_16x4[] = {
- { highbd_idct4, highbd_idct16 }, // DCT_DCT
- { highbd_iadst4, highbd_idct16 }, // ADST_DCT
- { highbd_idct4, highbd_iadst16 }, // DCT_ADST
- { highbd_iadst4, highbd_iadst16 }, // ADST_ADST
+ { aom_highbd_idct4_c, aom_highbd_idct16_c }, // DCT_DCT
+ { aom_highbd_iadst4_c, aom_highbd_idct16_c }, // ADST_DCT
+ { aom_highbd_idct4_c, aom_highbd_iadst16_c }, // DCT_ADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst16_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_iadst4, highbd_idct16 }, // FLIPADST_DCT
- { highbd_idct4, highbd_iadst16 }, // DCT_FLIPADST
- { highbd_iadst4, highbd_iadst16 }, // FLIPADST_FLIPADST
- { highbd_iadst4, highbd_iadst16 }, // ADST_FLIPADST
- { highbd_iadst4, highbd_iadst16 }, // FLIPADST_ADST
- { highbd_iidtx4_c, highbd_iidtx16_c }, // IDTX
- { highbd_idct4, highbd_iidtx16_c }, // V_DCT
- { highbd_iidtx4_c, highbd_idct16 }, // H_DCT
- { highbd_iadst4, highbd_iidtx16_c }, // V_ADST
- { highbd_iidtx4_c, highbd_iadst16 }, // H_ADST
- { highbd_iadst4, highbd_iidtx16_c }, // V_FLIPADST
- { highbd_iidtx4_c, highbd_iadst16 }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { aom_highbd_iadst4_c, aom_highbd_idct16_c }, // FLIPADST_DCT
+ { aom_highbd_idct4_c, aom_highbd_iadst16_c }, // DCT_FLIPADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst16_c }, // FLIPADST_FLIPADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst16_c }, // ADST_FLIPADST
+ { aom_highbd_iadst4_c, aom_highbd_iadst16_c }, // FLIPADST_ADST
+ { highbd_iidtx4_c, highbd_iidtx16_c }, // IDTX
+ { aom_highbd_idct4_c, highbd_iidtx16_c }, // V_DCT
+ { highbd_iidtx4_c, aom_highbd_idct16_c }, // H_DCT
+ { aom_highbd_iadst4_c, highbd_iidtx16_c }, // V_ADST
+ { highbd_iidtx4_c, aom_highbd_iadst16_c }, // H_ADST
+ { aom_highbd_iadst4_c, highbd_iidtx16_c }, // V_FLIPADST
+ { highbd_iidtx4_c, aom_highbd_iadst16_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 4;
const int n4 = 16;
@@ -1910,20 +1793,16 @@
tran_low_t *outp = &out[0][0];
int outstride = n;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_16X16, TX_4X4);
-
// inverse transform row vectors, and transpose
for (i = 0; i < n; ++i) {
- HIGH_IHT_16x4[tx_type].rows(input, outtmp, cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_16x4[tx_type].rows(input, outtmp, bd);
for (j = 0; j < n4; ++j) out[j][i] = outtmp[j];
input += n4;
}
// inverse transform column vectors
for (i = 0; i < n4; ++i) {
- HIGH_IHT_16x4[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ HIGH_IHT_16x4[tx_type].cols(out[i], out[i], bd);
}
#if CONFIG_EXT_TX
@@ -1944,24 +1823,24 @@
void av1_highbd_iht8x16_128_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_8x16[] = {
- { highbd_idct16, highbd_idct8 }, // DCT_DCT
- { highbd_iadst16, highbd_idct8 }, // ADST_DCT
- { highbd_idct16, highbd_iadst8 }, // DCT_ADST
- { highbd_iadst16, highbd_iadst8 }, // ADST_ADST
+ { aom_highbd_idct16_c, aom_highbd_idct8_c }, // DCT_DCT
+ { aom_highbd_iadst16_c, aom_highbd_idct8_c }, // ADST_DCT
+ { aom_highbd_idct16_c, aom_highbd_iadst8_c }, // DCT_ADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst8_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_iadst16, highbd_idct8 }, // FLIPADST_DCT
- { highbd_idct16, highbd_iadst8 }, // DCT_FLIPADST
- { highbd_iadst16, highbd_iadst8 }, // FLIPADST_FLIPADST
- { highbd_iadst16, highbd_iadst8 }, // ADST_FLIPADST
- { highbd_iadst16, highbd_iadst8 }, // FLIPADST_ADST
- { highbd_iidtx16_c, highbd_iidtx8_c }, // IDTX
- { highbd_idct16, highbd_iidtx8_c }, // V_DCT
- { highbd_iidtx16_c, highbd_idct8 }, // H_DCT
- { highbd_iadst16, highbd_iidtx8_c }, // V_ADST
- { highbd_iidtx16_c, highbd_iadst8 }, // H_ADST
- { highbd_iadst16, highbd_iidtx8_c }, // V_FLIPADST
- { highbd_iidtx16_c, highbd_iadst8 }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { aom_highbd_iadst16_c, aom_highbd_idct8_c }, // FLIPADST_DCT
+ { aom_highbd_idct16_c, aom_highbd_iadst8_c }, // DCT_FLIPADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst8_c }, // FLIPADST_FLIPADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst8_c }, // ADST_FLIPADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst8_c }, // FLIPADST_ADST
+ { highbd_iidtx16_c, highbd_iidtx8_c }, // IDTX
+ { aom_highbd_idct16_c, highbd_iidtx8_c }, // V_DCT
+ { highbd_iidtx16_c, aom_highbd_idct8_c }, // H_DCT
+ { aom_highbd_iadst16_c, highbd_iidtx8_c }, // V_ADST
+ { highbd_iidtx16_c, aom_highbd_iadst8_c }, // H_ADST
+ { aom_highbd_iadst16_c, highbd_iidtx8_c }, // V_FLIPADST
+ { highbd_iidtx16_c, aom_highbd_iadst8_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 8;
const int n2 = 16;
@@ -1973,12 +1852,9 @@
tran_low_t *outp = &out[0][0];
int outstride = n2;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_8X8, TX_16X16);
-
// inverse transform row vectors, and transpose
for (i = 0; i < n2; ++i) {
- HIGH_IHT_8x16[tx_type].rows(input, outtmp, cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_8x16[tx_type].rows(input, outtmp, bd);
for (j = 0; j < n; ++j)
out[j][i] = HIGHBD_WRAPLOW(dct_const_round_shift(outtmp[j] * Sqrt2), bd);
input += n;
@@ -1986,8 +1862,7 @@
// inverse transform column vectors
for (i = 0; i < n; ++i) {
- HIGH_IHT_8x16[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ HIGH_IHT_8x16[tx_type].cols(out[i], out[i], bd);
}
#if CONFIG_EXT_TX
@@ -2008,24 +1883,24 @@
void av1_highbd_iht16x8_128_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_16x8[] = {
- { highbd_idct8, highbd_idct16 }, // DCT_DCT
- { highbd_iadst8, highbd_idct16 }, // ADST_DCT
- { highbd_idct8, highbd_iadst16 }, // DCT_ADST
- { highbd_iadst8, highbd_iadst16 }, // ADST_ADST
+ { aom_highbd_idct8_c, aom_highbd_idct16_c }, // DCT_DCT
+ { aom_highbd_iadst8_c, aom_highbd_idct16_c }, // ADST_DCT
+ { aom_highbd_idct8_c, aom_highbd_iadst16_c }, // DCT_ADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst16_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_iadst8, highbd_idct16 }, // FLIPADST_DCT
- { highbd_idct8, highbd_iadst16 }, // DCT_FLIPADST
- { highbd_iadst8, highbd_iadst16 }, // FLIPADST_FLIPADST
- { highbd_iadst8, highbd_iadst16 }, // ADST_FLIPADST
- { highbd_iadst8, highbd_iadst16 }, // FLIPADST_ADST
- { highbd_iidtx8_c, highbd_iidtx16_c }, // IDTX
- { highbd_idct8, highbd_iidtx16_c }, // V_DCT
- { highbd_iidtx8_c, highbd_idct16 }, // H_DCT
- { highbd_iadst8, highbd_iidtx16_c }, // V_ADST
- { highbd_iidtx8_c, highbd_iadst16 }, // H_ADST
- { highbd_iadst8, highbd_iidtx16_c }, // V_FLIPADST
- { highbd_iidtx8_c, highbd_iadst16 }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { aom_highbd_iadst8_c, aom_highbd_idct16_c }, // FLIPADST_DCT
+ { aom_highbd_idct8_c, aom_highbd_iadst16_c }, // DCT_FLIPADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst16_c }, // FLIPADST_FLIPADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst16_c }, // ADST_FLIPADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst16_c }, // FLIPADST_ADST
+ { highbd_iidtx8_c, highbd_iidtx16_c }, // IDTX
+ { aom_highbd_idct8_c, highbd_iidtx16_c }, // V_DCT
+ { highbd_iidtx8_c, aom_highbd_idct16_c }, // H_DCT
+ { aom_highbd_iadst8_c, highbd_iidtx16_c }, // V_ADST
+ { highbd_iidtx8_c, aom_highbd_iadst16_c }, // H_ADST
+ { aom_highbd_iadst8_c, highbd_iidtx16_c }, // V_FLIPADST
+ { highbd_iidtx8_c, aom_highbd_iadst16_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 8;
const int n2 = 16;
@@ -2037,12 +1912,9 @@
tran_low_t *outp = &out[0][0];
int outstride = n;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_16X16, TX_8X8);
-
// inverse transform row vectors, and transpose
for (i = 0; i < n; ++i) {
- HIGH_IHT_16x8[tx_type].rows(input, outtmp, cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_16x8[tx_type].rows(input, outtmp, bd);
for (j = 0; j < n2; ++j)
out[j][i] = HIGHBD_WRAPLOW(dct_const_round_shift(outtmp[j] * Sqrt2), bd);
input += n2;
@@ -2050,8 +1922,7 @@
// inverse transform column vectors
for (i = 0; i < n2; ++i) {
- HIGH_IHT_16x8[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ HIGH_IHT_16x8[tx_type].cols(out[i], out[i], bd);
}
#if CONFIG_EXT_TX
@@ -2072,24 +1943,24 @@
void av1_highbd_iht8x32_256_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_8x32[] = {
- { highbd_idct32, highbd_idct8 }, // DCT_DCT
- { highbd_ihalfright32_c, highbd_idct8 }, // ADST_DCT
- { highbd_idct32, highbd_iadst8 }, // DCT_ADST
- { highbd_ihalfright32_c, highbd_iadst8 }, // ADST_ADST
+ { aom_highbd_idct32_c, aom_highbd_idct8_c }, // DCT_DCT
+ { highbd_ihalfright32_c, aom_highbd_idct8_c }, // ADST_DCT
+ { aom_highbd_idct32_c, aom_highbd_iadst8_c }, // DCT_ADST
+ { highbd_ihalfright32_c, aom_highbd_iadst8_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_ihalfright32_c, highbd_idct8 }, // FLIPADST_DCT
- { highbd_idct32, highbd_iadst8 }, // DCT_FLIPADST
- { highbd_ihalfright32_c, highbd_iadst8 }, // FLIPADST_FLIPADST
- { highbd_ihalfright32_c, highbd_iadst8 }, // ADST_FLIPADST
- { highbd_ihalfright32_c, highbd_iadst8 }, // FLIPADST_ADST
- { highbd_iidtx32_c, highbd_iidtx8_c }, // IDTX
- { highbd_idct32, highbd_iidtx8_c }, // V_DCT
- { highbd_iidtx32_c, highbd_idct8 }, // H_DCT
- { highbd_ihalfright32_c, highbd_iidtx8_c }, // V_ADST
- { highbd_iidtx32_c, highbd_iadst8 }, // H_ADST
- { highbd_ihalfright32_c, highbd_iidtx8_c }, // V_FLIPADST
- { highbd_iidtx32_c, highbd_iadst8 }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { highbd_ihalfright32_c, aom_highbd_idct8_c }, // FLIPADST_DCT
+ { aom_highbd_idct32_c, aom_highbd_iadst8_c }, // DCT_FLIPADST
+ { highbd_ihalfright32_c, aom_highbd_iadst8_c }, // FLIPADST_FLIPADST
+ { highbd_ihalfright32_c, aom_highbd_iadst8_c }, // ADST_FLIPADST
+ { highbd_ihalfright32_c, aom_highbd_iadst8_c }, // FLIPADST_ADST
+ { highbd_iidtx32_c, highbd_iidtx8_c }, // IDTX
+ { aom_highbd_idct32_c, highbd_iidtx8_c }, // V_DCT
+ { highbd_iidtx32_c, aom_highbd_idct8_c }, // H_DCT
+ { highbd_ihalfright32_c, highbd_iidtx8_c }, // V_ADST
+ { highbd_iidtx32_c, aom_highbd_iadst8_c }, // H_ADST
+ { highbd_ihalfright32_c, highbd_iidtx8_c }, // V_FLIPADST
+ { highbd_iidtx32_c, aom_highbd_iadst8_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 8;
const int n4 = 32;
@@ -2101,20 +1972,15 @@
tran_low_t *outp = &out[0][0];
int outstride = n4;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_8X8, TX_32X32);
-
// inverse transform row vectors, and transpose
for (i = 0; i < n4; ++i) {
- HIGH_IHT_8x32[tx_type].rows(input, outtmp, cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_8x32[tx_type].rows(input, outtmp, bd);
for (j = 0; j < n; ++j) out[j][i] = outtmp[j];
input += n;
}
// inverse transform column vectors
- for (i = 0; i < n; ++i)
- HIGH_IHT_8x32[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ for (i = 0; i < n; ++i) HIGH_IHT_8x32[tx_type].cols(out[i], out[i], bd);
#if CONFIG_EXT_TX
maybe_flip_strides16(&dest, &stride, &outp, &outstride, tx_type, n4, n);
@@ -2134,24 +2000,24 @@
void av1_highbd_iht32x8_256_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_32x8[] = {
- { highbd_idct8, highbd_idct32 }, // DCT_DCT
- { highbd_iadst8, highbd_idct32 }, // ADST_DCT
- { highbd_idct8, highbd_ihalfright32_c }, // DCT_ADST
- { highbd_iadst8, highbd_ihalfright32_c }, // ADST_ADST
+ { aom_highbd_idct8_c, aom_highbd_idct32_c }, // DCT_DCT
+ { aom_highbd_iadst8_c, aom_highbd_idct32_c }, // ADST_DCT
+ { aom_highbd_idct8_c, highbd_ihalfright32_c }, // DCT_ADST
+ { aom_highbd_iadst8_c, highbd_ihalfright32_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_iadst8, highbd_idct32 }, // FLIPADST_DCT
- { highbd_idct8, highbd_ihalfright32_c }, // DCT_FLIPADST
- { highbd_iadst8, highbd_ihalfright32_c }, // FLIPADST_FLIPADST
- { highbd_iadst8, highbd_ihalfright32_c }, // ADST_FLIPADST
- { highbd_iadst8, highbd_ihalfright32_c }, // FLIPADST_ADST
- { highbd_iidtx8_c, highbd_iidtx32_c }, // IDTX
- { highbd_idct8, highbd_iidtx32_c }, // V_DCT
- { highbd_iidtx8_c, highbd_idct32 }, // H_DCT
- { highbd_iadst8, highbd_iidtx32_c }, // V_ADST
- { highbd_iidtx8_c, highbd_ihalfright32_c }, // H_ADST
- { highbd_iadst8, highbd_iidtx32_c }, // V_FLIPADST
- { highbd_iidtx8_c, highbd_ihalfright32_c }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { aom_highbd_iadst8_c, aom_highbd_idct32_c }, // FLIPADST_DCT
+ { aom_highbd_idct8_c, highbd_ihalfright32_c }, // DCT_FLIPADST
+ { aom_highbd_iadst8_c, highbd_ihalfright32_c }, // FLIPADST_FLIPADST
+ { aom_highbd_iadst8_c, highbd_ihalfright32_c }, // ADST_FLIPADST
+ { aom_highbd_iadst8_c, highbd_ihalfright32_c }, // FLIPADST_ADST
+ { highbd_iidtx8_c, highbd_iidtx32_c }, // IDTX
+ { aom_highbd_idct8_c, highbd_iidtx32_c }, // V_DCT
+ { highbd_iidtx8_c, aom_highbd_idct32_c }, // H_DCT
+ { aom_highbd_iadst8_c, highbd_iidtx32_c }, // V_ADST
+ { highbd_iidtx8_c, highbd_ihalfright32_c }, // H_ADST
+ { aom_highbd_iadst8_c, highbd_iidtx32_c }, // V_FLIPADST
+ { highbd_iidtx8_c, highbd_ihalfright32_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 8;
const int n4 = 32;
@@ -2163,20 +2029,15 @@
tran_low_t *outp = &out[0][0];
int outstride = n;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_32X32, TX_8X8);
-
// inverse transform row vectors, and transpose
for (i = 0; i < n; ++i) {
- HIGH_IHT_32x8[tx_type].rows(input, outtmp, cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_32x8[tx_type].rows(input, outtmp, bd);
for (j = 0; j < n4; ++j) out[j][i] = outtmp[j];
input += n4;
}
// inverse transform column vectors
- for (i = 0; i < n4; ++i)
- HIGH_IHT_32x8[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ for (i = 0; i < n4; ++i) HIGH_IHT_32x8[tx_type].cols(out[i], out[i], bd);
#if CONFIG_EXT_TX
maybe_flip_strides16(&dest, &stride, &outp, &outstride, tx_type, n, n4);
@@ -2196,24 +2057,24 @@
void av1_highbd_iht16x32_512_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_16x32[] = {
- { highbd_idct32, highbd_idct16 }, // DCT_DCT
- { highbd_ihalfright32_c, highbd_idct16 }, // ADST_DCT
- { highbd_idct32, highbd_iadst16 }, // DCT_ADST
- { highbd_ihalfright32_c, highbd_iadst16 }, // ADST_ADST
+ { aom_highbd_idct32_c, aom_highbd_idct16_c }, // DCT_DCT
+ { highbd_ihalfright32_c, aom_highbd_idct16_c }, // ADST_DCT
+ { aom_highbd_idct32_c, aom_highbd_iadst16_c }, // DCT_ADST
+ { highbd_ihalfright32_c, aom_highbd_iadst16_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_ihalfright32_c, highbd_idct16 }, // FLIPADST_DCT
- { highbd_idct32, highbd_iadst16 }, // DCT_FLIPADST
- { highbd_ihalfright32_c, highbd_iadst16 }, // FLIPADST_FLIPADST
- { highbd_ihalfright32_c, highbd_iadst16 }, // ADST_FLIPADST
- { highbd_ihalfright32_c, highbd_iadst16 }, // FLIPADST_ADST
- { highbd_iidtx32_c, highbd_iidtx16_c }, // IDTX
- { highbd_idct32, highbd_iidtx16_c }, // V_DCT
- { highbd_iidtx32_c, highbd_idct16 }, // H_DCT
- { highbd_ihalfright32_c, highbd_iidtx16_c }, // V_ADST
- { highbd_iidtx32_c, highbd_iadst16 }, // H_ADST
- { highbd_ihalfright32_c, highbd_iidtx16_c }, // V_FLIPADST
- { highbd_iidtx32_c, highbd_iadst16 }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { highbd_ihalfright32_c, aom_highbd_idct16_c }, // FLIPADST_DCT
+ { aom_highbd_idct32_c, aom_highbd_iadst16_c }, // DCT_FLIPADST
+ { highbd_ihalfright32_c, aom_highbd_iadst16_c }, // FLIPADST_FLIPADST
+ { highbd_ihalfright32_c, aom_highbd_iadst16_c }, // ADST_FLIPADST
+ { highbd_ihalfright32_c, aom_highbd_iadst16_c }, // FLIPADST_ADST
+ { highbd_iidtx32_c, highbd_iidtx16_c }, // IDTX
+ { aom_highbd_idct32_c, highbd_iidtx16_c }, // V_DCT
+ { highbd_iidtx32_c, aom_highbd_idct16_c }, // H_DCT
+ { highbd_ihalfright32_c, highbd_iidtx16_c }, // V_ADST
+ { highbd_iidtx32_c, aom_highbd_iadst16_c }, // H_ADST
+ { highbd_ihalfright32_c, highbd_iidtx16_c }, // V_FLIPADST
+ { highbd_iidtx32_c, aom_highbd_iadst16_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 16;
const int n2 = 32;
@@ -2225,12 +2086,9 @@
tran_low_t *outp = &out[0][0];
int outstride = n2;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_16X16, TX_32X32);
-
// inverse transform row vectors, and transpose
for (i = 0; i < n2; ++i) {
- HIGH_IHT_16x32[tx_type].rows(input, outtmp, cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_16x32[tx_type].rows(input, outtmp, bd);
for (j = 0; j < n; ++j)
out[j][i] = HIGHBD_WRAPLOW(dct_const_round_shift(outtmp[j] * Sqrt2), bd);
input += n;
@@ -2238,8 +2096,7 @@
// inverse transform column vectors
for (i = 0; i < n; ++i) {
- HIGH_IHT_16x32[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ HIGH_IHT_16x32[tx_type].cols(out[i], out[i], bd);
}
#if CONFIG_EXT_TX
@@ -2260,24 +2117,24 @@
void av1_highbd_iht32x16_512_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_32x16[] = {
- { highbd_idct16, highbd_idct32 }, // DCT_DCT
- { highbd_iadst16, highbd_idct32 }, // ADST_DCT
- { highbd_idct16, highbd_ihalfright32_c }, // DCT_ADST
- { highbd_iadst16, highbd_ihalfright32_c }, // ADST_ADST
+ { aom_highbd_idct16_c, aom_highbd_idct32_c }, // DCT_DCT
+ { aom_highbd_iadst16_c, aom_highbd_idct32_c }, // ADST_DCT
+ { aom_highbd_idct16_c, highbd_ihalfright32_c }, // DCT_ADST
+ { aom_highbd_iadst16_c, highbd_ihalfright32_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_iadst16, highbd_idct32 }, // FLIPADST_DCT
- { highbd_idct16, highbd_ihalfright32_c }, // DCT_FLIPADST
- { highbd_iadst16, highbd_ihalfright32_c }, // FLIPADST_FLIPADST
- { highbd_iadst16, highbd_ihalfright32_c }, // ADST_FLIPADST
- { highbd_iadst16, highbd_ihalfright32_c }, // FLIPADST_ADST
- { highbd_iidtx16_c, highbd_iidtx32_c }, // IDTX
- { highbd_idct16, highbd_iidtx32_c }, // V_DCT
- { highbd_iidtx16_c, highbd_idct32 }, // H_DCT
- { highbd_iadst16, highbd_iidtx32_c }, // V_ADST
- { highbd_iidtx16_c, highbd_ihalfright32_c }, // H_ADST
- { highbd_iadst16, highbd_iidtx32_c }, // V_FLIPADST
- { highbd_iidtx16_c, highbd_ihalfright32_c }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { aom_highbd_iadst16_c, aom_highbd_idct32_c }, // FLIPADST_DCT
+ { aom_highbd_idct16_c, highbd_ihalfright32_c }, // DCT_FLIPADST
+ { aom_highbd_iadst16_c, highbd_ihalfright32_c }, // FLIPADST_FLIPADST
+ { aom_highbd_iadst16_c, highbd_ihalfright32_c }, // ADST_FLIPADST
+ { aom_highbd_iadst16_c, highbd_ihalfright32_c }, // FLIPADST_ADST
+ { highbd_iidtx16_c, highbd_iidtx32_c }, // IDTX
+ { aom_highbd_idct16_c, highbd_iidtx32_c }, // V_DCT
+ { highbd_iidtx16_c, aom_highbd_idct32_c }, // H_DCT
+ { aom_highbd_iadst16_c, highbd_iidtx32_c }, // V_ADST
+ { highbd_iidtx16_c, highbd_ihalfright32_c }, // H_ADST
+ { aom_highbd_iadst16_c, highbd_iidtx32_c }, // V_FLIPADST
+ { highbd_iidtx16_c, highbd_ihalfright32_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 16;
const int n2 = 32;
@@ -2289,12 +2146,9 @@
tran_low_t *outp = &out[0][0];
int outstride = n;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_32X32, TX_16X16);
-
// inverse transform row vectors, and transpose
for (i = 0; i < n; ++i) {
- HIGH_IHT_32x16[tx_type].rows(input, outtmp, cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_32x16[tx_type].rows(input, outtmp, bd);
for (j = 0; j < n2; ++j)
out[j][i] = HIGHBD_WRAPLOW(dct_const_round_shift(outtmp[j] * Sqrt2), bd);
input += n2;
@@ -2302,8 +2156,7 @@
// inverse transform column vectors
for (i = 0; i < n2; ++i) {
- HIGH_IHT_32x16[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ HIGH_IHT_32x16[tx_type].cols(out[i], out[i], bd);
}
#if CONFIG_EXT_TX
@@ -2324,24 +2177,24 @@
void av1_highbd_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_8[] = {
- { highbd_idct8, highbd_idct8 }, // DCT_DCT
- { highbd_iadst8, highbd_idct8 }, // ADST_DCT
- { highbd_idct8, highbd_iadst8 }, // DCT_ADST
- { highbd_iadst8, highbd_iadst8 }, // ADST_ADST
+ { aom_highbd_idct8_c, aom_highbd_idct8_c }, // DCT_DCT
+ { aom_highbd_iadst8_c, aom_highbd_idct8_c }, // ADST_DCT
+ { aom_highbd_idct8_c, aom_highbd_iadst8_c }, // DCT_ADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst8_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_iadst8, highbd_idct8 }, // FLIPADST_DCT
- { highbd_idct8, highbd_iadst8 }, // DCT_FLIPADST
- { highbd_iadst8, highbd_iadst8 }, // FLIPADST_FLIPADST
- { highbd_iadst8, highbd_iadst8 }, // ADST_FLIPADST
- { highbd_iadst8, highbd_iadst8 }, // FLIPADST_ADST
- { highbd_iidtx8_c, highbd_iidtx8_c }, // IDTX
- { highbd_idct8, highbd_iidtx8_c }, // V_DCT
- { highbd_iidtx8_c, highbd_idct8 }, // H_DCT
- { highbd_iadst8, highbd_iidtx8_c }, // V_ADST
- { highbd_iidtx8_c, highbd_iadst8 }, // H_ADST
- { highbd_iadst8, highbd_iidtx8_c }, // V_FLIPADST
- { highbd_iidtx8_c, highbd_iadst8 }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { aom_highbd_iadst8_c, aom_highbd_idct8_c }, // FLIPADST_DCT
+ { aom_highbd_idct8_c, aom_highbd_iadst8_c }, // DCT_FLIPADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst8_c }, // FLIPADST_FLIPADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst8_c }, // ADST_FLIPADST
+ { aom_highbd_iadst8_c, aom_highbd_iadst8_c }, // FLIPADST_ADST
+ { highbd_iidtx8_c, highbd_iidtx8_c }, // IDTX
+ { aom_highbd_idct8_c, highbd_iidtx8_c }, // V_DCT
+ { highbd_iidtx8_c, aom_highbd_idct8_c }, // H_DCT
+ { aom_highbd_iadst8_c, highbd_iidtx8_c }, // V_ADST
+ { highbd_iidtx8_c, aom_highbd_iadst8_c }, // H_ADST
+ { aom_highbd_iadst8_c, highbd_iidtx8_c }, // V_FLIPADST
+ { highbd_iidtx8_c, aom_highbd_iadst8_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
@@ -2352,12 +2205,9 @@
tran_low_t *outp = &out[0][0];
int outstride = 8;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_8X8, TX_8X8);
-
// inverse transform row vectors
for (i = 0; i < 8; ++i) {
- HIGH_IHT_8[tx_type].rows(input, out[i], cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_8[tx_type].rows(input, out[i], bd);
input += 8;
}
@@ -2372,8 +2222,7 @@
// inverse transform column vectors
for (i = 0; i < 8; ++i) {
- HIGH_IHT_8[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ HIGH_IHT_8[tx_type].cols(out[i], out[i], bd);
}
#if CONFIG_EXT_TX
@@ -2394,24 +2243,24 @@
void av1_highbd_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_16[] = {
- { highbd_idct16, highbd_idct16 }, // DCT_DCT
- { highbd_iadst16, highbd_idct16 }, // ADST_DCT
- { highbd_idct16, highbd_iadst16 }, // DCT_ADST
- { highbd_iadst16, highbd_iadst16 }, // ADST_ADST
+ { aom_highbd_idct16_c, aom_highbd_idct16_c }, // DCT_DCT
+ { aom_highbd_iadst16_c, aom_highbd_idct16_c }, // ADST_DCT
+ { aom_highbd_idct16_c, aom_highbd_iadst16_c }, // DCT_ADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst16_c }, // ADST_ADST
#if CONFIG_EXT_TX
- { highbd_iadst16, highbd_idct16 }, // FLIPADST_DCT
- { highbd_idct16, highbd_iadst16 }, // DCT_FLIPADST
- { highbd_iadst16, highbd_iadst16 }, // FLIPADST_FLIPADST
- { highbd_iadst16, highbd_iadst16 }, // ADST_FLIPADST
- { highbd_iadst16, highbd_iadst16 }, // FLIPADST_ADST
- { highbd_iidtx16_c, highbd_iidtx16_c }, // IDTX
- { highbd_idct16, highbd_iidtx16_c }, // V_DCT
- { highbd_iidtx16_c, highbd_idct16 }, // H_DCT
- { highbd_iadst16, highbd_iidtx16_c }, // V_ADST
- { highbd_iidtx16_c, highbd_iadst16 }, // H_ADST
- { highbd_iadst16, highbd_iidtx16_c }, // V_FLIPADST
- { highbd_iidtx16_c, highbd_iadst16 }, // H_FLIPADST
-#endif // CONFIG_EXT_TX
+ { aom_highbd_iadst16_c, aom_highbd_idct16_c }, // FLIPADST_DCT
+ { aom_highbd_idct16_c, aom_highbd_iadst16_c }, // DCT_FLIPADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst16_c }, // FLIPADST_FLIPADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst16_c }, // ADST_FLIPADST
+ { aom_highbd_iadst16_c, aom_highbd_iadst16_c }, // FLIPADST_ADST
+ { highbd_iidtx16_c, highbd_iidtx16_c }, // IDTX
+ { aom_highbd_idct16_c, highbd_iidtx16_c }, // V_DCT
+ { highbd_iidtx16_c, aom_highbd_idct16_c }, // H_DCT
+ { aom_highbd_iadst16_c, highbd_iidtx16_c }, // V_ADST
+ { highbd_iidtx16_c, aom_highbd_iadst16_c }, // H_ADST
+ { aom_highbd_iadst16_c, highbd_iidtx16_c }, // V_FLIPADST
+ { highbd_iidtx16_c, aom_highbd_iadst16_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
@@ -2422,12 +2271,9 @@
tran_low_t *outp = &out[0][0];
int outstride = 16;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_16X16, TX_16X16);
-
// inverse transform row vectors
for (i = 0; i < 16; ++i) {
- HIGH_IHT_16[tx_type].rows(input, out[i], cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_16[tx_type].rows(input, out[i], bd);
input += 16;
}
@@ -2442,8 +2288,7 @@
// inverse transform column vectors
for (i = 0; i < 16; ++i) {
- HIGH_IHT_16[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ HIGH_IHT_16[tx_type].cols(out[i], out[i], bd);
}
#if CONFIG_EXT_TX
@@ -2465,18 +2310,18 @@
static void highbd_iht32x32_1024_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_32[] = {
- { highbd_idct32, highbd_idct32 }, // DCT_DCT
- { highbd_ihalfright32_c, highbd_idct32 }, // ADST_DCT
- { highbd_idct32, highbd_ihalfright32_c }, // DCT_ADST
+ { aom_highbd_idct32_c, aom_highbd_idct32_c }, // DCT_DCT
+ { highbd_ihalfright32_c, aom_highbd_idct32_c }, // ADST_DCT
+ { aom_highbd_idct32_c, highbd_ihalfright32_c }, // DCT_ADST
{ highbd_ihalfright32_c, highbd_ihalfright32_c }, // ADST_ADST
- { highbd_ihalfright32_c, highbd_idct32 }, // FLIPADST_DCT
- { highbd_idct32, highbd_ihalfright32_c }, // DCT_FLIPADST
+ { highbd_ihalfright32_c, aom_highbd_idct32_c }, // FLIPADST_DCT
+ { aom_highbd_idct32_c, highbd_ihalfright32_c }, // DCT_FLIPADST
{ highbd_ihalfright32_c, highbd_ihalfright32_c }, // FLIPADST_FLIPADST
{ highbd_ihalfright32_c, highbd_ihalfright32_c }, // ADST_FLIPADST
{ highbd_ihalfright32_c, highbd_ihalfright32_c }, // FLIPADST_ADST
{ highbd_iidtx32_c, highbd_iidtx32_c }, // IDTX
- { highbd_idct32, highbd_iidtx32_c }, // V_DCT
- { highbd_iidtx32_c, highbd_idct32 }, // H_DCT
+ { aom_highbd_idct32_c, highbd_iidtx32_c }, // V_DCT
+ { highbd_iidtx32_c, aom_highbd_idct32_c }, // H_DCT
{ highbd_ihalfright32_c, highbd_iidtx32_c }, // V_ADST
{ highbd_iidtx32_c, highbd_ihalfright32_c }, // H_ADST
{ highbd_ihalfright32_c, highbd_iidtx32_c }, // V_FLIPADST
@@ -2491,12 +2336,9 @@
tran_low_t *outp = &out[0][0];
int outstride = 32;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_32X32, TX_32X32);
-
// inverse transform row vectors
for (i = 0; i < 32; ++i) {
- HIGH_IHT_32[tx_type].rows(input, out[i], cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_32[tx_type].rows(input, out[i], bd);
input += 32;
}
@@ -2511,8 +2353,7 @@
// inverse transform column vectors
for (i = 0; i < 32; ++i) {
- HIGH_IHT_32[tx_type].cols(out[i], out[i], cfg.col.cos_bit,
- cfg.col.stage_range, bd);
+ HIGH_IHT_32[tx_type].cols(out[i], out[i], bd);
}
maybe_flip_strides16(&dest, &stride, &outp, &outstride, tx_type, 32, 32);
@@ -2561,12 +2402,9 @@
tran_low_t *outp = &out[0][0];
int outstride = 64;
- tx_2d_cfg cfg = inv_tx_cfg(tx_type, TX_64X64, TX_64X64);
-
// inverse transform row vectors
for (i = 0; i < 64; ++i) {
- HIGH_IHT_64[tx_type].rows(input, out[i], cfg.row.cos_bit,
- cfg.row.stage_range, bd);
+ HIGH_IHT_64[tx_type].rows(input, out[i], bd);
for (j = 0; j < 64; ++j) out[i][j] = ROUND_POWER_OF_TWO(out[i][j], 1);
input += 64;
}
@@ -2582,8 +2420,7 @@
// inverse transform column vectors
for (i = 0; i < 64; ++i) {
- HIGH_IHT_64[tx_type].cols(out[i], out[i], cfg.col.cos_bit_col,
- cfg.col.stage_range, bd);
+ HIGH_IHT_64[tx_type].cols(out[i], out[i], bd);
}
#if CONFIG_EXT_TX
diff --git a/av1/common/idct.h b/av1/common/idct.h
index e3a1921..d9c1507 100644
--- a/av1/common/idct.h
+++ b/av1/common/idct.h
@@ -46,9 +46,7 @@
} transform_2d;
#if CONFIG_HIGHBITDEPTH
-typedef void (*highbd_transform_1d)(const tran_low_t *, tran_low_t *,
- const int8_t *cos_bit,
- const int8_t *stage_range, int bd);
+typedef void (*highbd_transform_1d)(const tran_low_t *, tran_low_t *, int bd);
typedef struct {
highbd_transform_1d cols, rows; // vertical and horizontal
