mozilla
/
aom
зеркало из
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность

Merge "Replace vp9_ prefix with vpx_ prefix in vpx_dsp function names"

This commit is contained in:
Jingning Han 2015-08-04 22:30:36 +00:00 коммит произвёл Gerrit Code Review
Родитель df14d9f4e7 08a453b9de
Коммит 3ad75fc623
50 изменённых файлов: 550 добавлений и 545 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

30
test/dct16x16_test.cc
Просмотреть файл

@ -277,7 +277,7 @@ void fdct16x16_ref(const int16_t *in, tran_low_t *out, int stride,
void idct16x16_ref(const tran_low_t *in, uint8_t *dest, int stride,
int /*tx_type*/) {
vp9_idct16x16_256_add_c(in, dest, stride);
vpx_idct16x16_256_add_c(in, dest, stride);
}
void fht16x16_ref(const int16_t *in, tran_low_t *out, int stride,
@ -292,11 +292,11 @@ void iht16x16_ref(const tran_low_t *in, uint8_t *dest, int stride,
#if CONFIG_VP9_HIGHBITDEPTH
void idct16x16_10(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct16x16_256_add_c(in, out, stride, 10);
vpx_highbd_idct16x16_256_add_c(in, out, stride, 10);
}
void idct16x16_12(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct16x16_256_add_c(in, out, stride, 12);
vpx_highbd_idct16x16_256_add_c(in, out, stride, 12);
}
void idct16x16_10_ref(const tran_low_t *in, uint8_t *out, int stride,
@ -318,28 +318,28 @@ void iht16x16_12(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
}
void idct16x16_10_add_10_c(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct16x16_10_add_c(in, out, stride, 10);
vpx_highbd_idct16x16_10_add_c(in, out, stride, 10);
}
void idct16x16_10_add_12_c(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct16x16_10_add_c(in, out, stride, 12);
vpx_highbd_idct16x16_10_add_c(in, out, stride, 12);
}
#if HAVE_SSE2
void idct16x16_256_add_10_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct16x16_256_add_sse2(in, out, stride, 10);
vpx_highbd_idct16x16_256_add_sse2(in, out, stride, 10);
}
void idct16x16_256_add_12_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct16x16_256_add_sse2(in, out, stride, 12);
vpx_highbd_idct16x16_256_add_sse2(in, out, stride, 12);
}
void idct16x16_10_add_10_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct16x16_10_add_sse2(in, out, stride, 10);
vpx_highbd_idct16x16_10_add_sse2(in, out, stride, 10);
}
void idct16x16_10_add_12_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct16x16_10_add_sse2(in, out, stride, 12);
vpx_highbd_idct16x16_10_add_sse2(in, out, stride, 12);
}
#endif // HAVE_SSE2
#endif // CONFIG_VP9_HIGHBITDEPTH
@ -824,12 +824,12 @@ INSTANTIATE_TEST_CASE_P(
::testing::Values(
make_tuple(&vpx_highbd_fdct16x16_c, &idct16x16_10, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct16x16_c, &idct16x16_12, 0, VPX_BITS_12),
make_tuple(&vpx_fdct16x16_c, &vp9_idct16x16_256_add_c, 0, VPX_BITS_8)));
make_tuple(&vpx_fdct16x16_c, &vpx_idct16x16_256_add_c, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
C, Trans16x16DCT,
::testing::Values(
make_tuple(&vpx_fdct16x16_c, &vp9_idct16x16_256_add_c, 0, VPX_BITS_8)));
make_tuple(&vpx_fdct16x16_c, &vpx_idct16x16_256_add_c, 0, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if CONFIG_VP9_HIGHBITDEPTH
@ -863,7 +863,7 @@ INSTANTIATE_TEST_CASE_P(
NEON, Trans16x16DCT,
::testing::Values(
make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_neon, 0, VPX_BITS_8)));
&vpx_idct16x16_256_add_neon, 0, VPX_BITS_8)));
#endif
#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -871,7 +871,7 @@ INSTANTIATE_TEST_CASE_P(
SSE2, Trans16x16DCT,
::testing::Values(
make_tuple(&vpx_fdct16x16_sse2,
&vp9_idct16x16_256_add_sse2, 0, VPX_BITS_8)));
&vpx_idct16x16_256_add_sse2, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
SSE2, Trans16x16HT,
::testing::Values(
@ -898,7 +898,7 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vpx_highbd_fdct16x16_c,
&idct16x16_256_add_12_sse2, 0, VPX_BITS_12),
make_tuple(&vpx_fdct16x16_sse2,
&vp9_idct16x16_256_add_c, 0, VPX_BITS_8)));
&vpx_idct16x16_256_add_c, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
SSE2, Trans16x16HT,
::testing::Values(
@ -927,7 +927,7 @@ INSTANTIATE_TEST_CASE_P(
MSA, Trans16x16DCT,
::testing::Values(
make_tuple(&vpx_fdct16x16_msa,
&vp9_idct16x16_256_add_msa, 0, VPX_BITS_8)));
&vpx_idct16x16_256_add_msa, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
MSA, Trans16x16HT,
::testing::Values(

34
test/dct32x32_test.cc
Просмотреть файл

@ -82,15 +82,15 @@ typedef std::tr1::tuple<FwdTxfmFunc, InvTxfmFunc, int, vpx_bit_depth_t>
#if CONFIG_VP9_HIGHBITDEPTH
void idct32x32_8(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct32x32_1024_add_c(in, out, stride, 8);
vpx_highbd_idct32x32_1024_add_c(in, out, stride, 8);
}
void idct32x32_10(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct32x32_1024_add_c(in, out, stride, 10);
vpx_highbd_idct32x32_1024_add_c(in, out, stride, 10);
}
void idct32x32_12(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct32x32_1024_add_c(in, out, stride, 12);
vpx_highbd_idct32x32_1024_add_c(in, out, stride, 12);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
@ -324,17 +324,17 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vpx_highbd_fdct32x32_rd_c,
&idct32x32_12, 1, VPX_BITS_12),
make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c, 0, VPX_BITS_8),
&vpx_idct32x32_1024_add_c, 0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_c,
&vp9_idct32x32_1024_add_c, 1, VPX_BITS_8)));
&vpx_idct32x32_1024_add_c, 1, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
C, Trans32x32Test,
::testing::Values(
make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c, 0, VPX_BITS_8),
&vpx_idct32x32_1024_add_c, 0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_c,
&vp9_idct32x32_1024_add_c, 1, VPX_BITS_8)));
&vpx_idct32x32_1024_add_c, 1, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -342,9 +342,9 @@ INSTANTIATE_TEST_CASE_P(
NEON, Trans32x32Test,
::testing::Values(
make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_neon, 0, VPX_BITS_8),
&vpx_idct32x32_1024_add_neon, 0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_c,
&vp9_idct32x32_1024_add_neon, 1, VPX_BITS_8)));
&vpx_idct32x32_1024_add_neon, 1, VPX_BITS_8)));
#endif // HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -352,9 +352,9 @@ INSTANTIATE_TEST_CASE_P(
SSE2, Trans32x32Test,
::testing::Values(
make_tuple(&vpx_fdct32x32_sse2,
&vp9_idct32x32_1024_add_sse2, 0, VPX_BITS_8),
&vpx_idct32x32_1024_add_sse2, 0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_sse2,
&vp9_idct32x32_1024_add_sse2, 1, VPX_BITS_8)));
&vpx_idct32x32_1024_add_sse2, 1, VPX_BITS_8)));
#endif // HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -367,9 +367,9 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vpx_highbd_fdct32x32_sse2, &idct32x32_12, 0, VPX_BITS_12),
make_tuple(&vpx_highbd_fdct32x32_rd_sse2, &idct32x32_12, 1,
VPX_BITS_12),
make_tuple(&vpx_fdct32x32_sse2, &vp9_idct32x32_1024_add_c, 0,
make_tuple(&vpx_fdct32x32_sse2, &vpx_idct32x32_1024_add_c, 0,
VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_sse2, &vp9_idct32x32_1024_add_c, 1,
make_tuple(&vpx_fdct32x32_rd_sse2, &vpx_idct32x32_1024_add_c, 1,
VPX_BITS_8)));
#endif // HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -378,9 +378,9 @@ INSTANTIATE_TEST_CASE_P(
AVX2, Trans32x32Test,
::testing::Values(
make_tuple(&vpx_fdct32x32_avx2,
&vp9_idct32x32_1024_add_sse2, 0, VPX_BITS_8),
&vpx_idct32x32_1024_add_sse2, 0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_avx2,
&vp9_idct32x32_1024_add_sse2, 1, VPX_BITS_8)));
&vpx_idct32x32_1024_add_sse2, 1, VPX_BITS_8)));
#endif // HAVE_AVX2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -388,8 +388,8 @@ INSTANTIATE_TEST_CASE_P(
MSA, Trans32x32Test,
::testing::Values(
make_tuple(&vpx_fdct32x32_msa,
&vp9_idct32x32_1024_add_msa, 0, VPX_BITS_8),
&vpx_idct32x32_1024_add_msa, 0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_msa,
&vp9_idct32x32_1024_add_msa, 1, VPX_BITS_8)));
&vpx_idct32x32_1024_add_msa, 1, VPX_BITS_8)));
#endif // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
} // namespace

32
test/fdct4x4_test.cc
Просмотреть файл

@ -55,11 +55,11 @@ void fwht4x4_ref(const int16_t *in, tran_low_t *out, int stride,
#if CONFIG_VP9_HIGHBITDEPTH
void idct4x4_10(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct4x4_16_add_c(in, out, stride, 10);
vpx_highbd_idct4x4_16_add_c(in, out, stride, 10);
}
void idct4x4_12(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct4x4_16_add_c(in, out, stride, 12);
vpx_highbd_idct4x4_16_add_c(in, out, stride, 12);
}
void iht4x4_10(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
@ -71,20 +71,20 @@ void iht4x4_12(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
}
void iwht4x4_10(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_iwht4x4_16_add_c(in, out, stride, 10);
vpx_highbd_iwht4x4_16_add_c(in, out, stride, 10);
}
void iwht4x4_12(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_iwht4x4_16_add_c(in, out, stride, 12);
vpx_highbd_iwht4x4_16_add_c(in, out, stride, 12);
}
#if HAVE_SSE2
void idct4x4_10_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct4x4_16_add_sse2(in, out, stride, 10);
vpx_highbd_idct4x4_16_add_sse2(in, out, stride, 10);
}
void idct4x4_12_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct4x4_16_add_sse2(in, out, stride, 12);
vpx_highbd_idct4x4_16_add_sse2(in, out, stride, 12);
}
#endif // HAVE_SSE2
#endif // CONFIG_VP9_HIGHBITDEPTH
@ -421,12 +421,12 @@ INSTANTIATE_TEST_CASE_P(
::testing::Values(
make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_10, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_12, 0, VPX_BITS_12),
make_tuple(&vpx_fdct4x4_c, &vp9_idct4x4_16_add_c, 0, VPX_BITS_8)));
make_tuple(&vpx_fdct4x4_c, &vpx_idct4x4_16_add_c, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
C, Trans4x4DCT,
::testing::Values(
make_tuple(&vpx_fdct4x4_c, &vp9_idct4x4_16_add_c, 0, VPX_BITS_8)));
make_tuple(&vpx_fdct4x4_c, &vpx_idct4x4_16_add_c, 0, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if CONFIG_VP9_HIGHBITDEPTH
@ -461,12 +461,12 @@ INSTANTIATE_TEST_CASE_P(
::testing::Values(
make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_10, 0, VPX_BITS_10),
make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_12, 0, VPX_BITS_12),
make_tuple(&vp9_fwht4x4_c, &vp9_iwht4x4_16_add_c, 0, VPX_BITS_8)));
make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_c, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
C, Trans4x4WHT,
::testing::Values(
make_tuple(&vp9_fwht4x4_c, &vp9_iwht4x4_16_add_c, 0, VPX_BITS_8)));
make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_c, 0, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -474,7 +474,7 @@ INSTANTIATE_TEST_CASE_P(
NEON, Trans4x4DCT,
::testing::Values(
make_tuple(&vpx_fdct4x4_c,
&vp9_idct4x4_16_add_neon, 0, VPX_BITS_8)));
&vpx_idct4x4_16_add_neon, 0, VPX_BITS_8)));
#endif // HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -492,7 +492,7 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P(
MMX, Trans4x4WHT,
::testing::Values(
make_tuple(&vp9_fwht4x4_mmx, &vp9_iwht4x4_16_add_c, 0, VPX_BITS_8)));
make_tuple(&vp9_fwht4x4_mmx, &vpx_iwht4x4_16_add_c, 0, VPX_BITS_8)));
#endif
#if CONFIG_USE_X86INC && HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && \
@ -500,7 +500,7 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P(
SSE2, Trans4x4WHT,
::testing::Values(
make_tuple(&vp9_fwht4x4_c, &vp9_iwht4x4_16_add_sse2, 0, VPX_BITS_8)));
make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_sse2, 0, VPX_BITS_8)));
#endif
#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -508,7 +508,7 @@ INSTANTIATE_TEST_CASE_P(
SSE2, Trans4x4DCT,
::testing::Values(
make_tuple(&vpx_fdct4x4_sse2,
&vp9_idct4x4_16_add_sse2, 0, VPX_BITS_8)));
&vpx_idct4x4_16_add_sse2, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
SSE2, Trans4x4HT,
::testing::Values(
@ -526,7 +526,7 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_10_sse2, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_12_sse2, 0, VPX_BITS_12),
make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_12_sse2, 0, VPX_BITS_12),
make_tuple(&vpx_fdct4x4_sse2, &vp9_idct4x4_16_add_c, 0,
make_tuple(&vpx_fdct4x4_sse2, &vpx_idct4x4_16_add_c, 0,
VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
@ -542,7 +542,7 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P(
MSA, Trans4x4DCT,
::testing::Values(
make_tuple(&vpx_fdct4x4_msa, &vp9_idct4x4_16_add_msa, 0, VPX_BITS_8)));
make_tuple(&vpx_fdct4x4_msa, &vpx_idct4x4_16_add_msa, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
MSA, Trans4x4HT,
::testing::Values(

30
test/fdct8x8_test.cc
Просмотреть файл

@ -92,11 +92,11 @@ void fht8x8_ref(const int16_t *in, tran_low_t *out, int stride, int tx_type) {
#if CONFIG_VP9_HIGHBITDEPTH
void idct8x8_10(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct8x8_64_add_c(in, out, stride, 10);
vpx_highbd_idct8x8_64_add_c(in, out, stride, 10);
}
void idct8x8_12(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct8x8_64_add_c(in, out, stride, 12);
vpx_highbd_idct8x8_64_add_c(in, out, stride, 12);
}
void iht8x8_10(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
@ -108,28 +108,28 @@ void iht8x8_12(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
}
void idct8x8_10_add_10_c(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct8x8_10_add_c(in, out, stride, 10);
vpx_highbd_idct8x8_10_add_c(in, out, stride, 10);
}
void idct8x8_10_add_12_c(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct8x8_10_add_c(in, out, stride, 12);
vpx_highbd_idct8x8_10_add_c(in, out, stride, 12);
}
#if HAVE_SSE2
void idct8x8_10_add_10_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct8x8_10_add_sse2(in, out, stride, 10);
vpx_highbd_idct8x8_10_add_sse2(in, out, stride, 10);
}
void idct8x8_10_add_12_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct8x8_10_add_sse2(in, out, stride, 12);
vpx_highbd_idct8x8_10_add_sse2(in, out, stride, 12);
}
void idct8x8_64_add_10_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct8x8_64_add_sse2(in, out, stride, 10);
vpx_highbd_idct8x8_64_add_sse2(in, out, stride, 10);
}
void idct8x8_64_add_12_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vp9_highbd_idct8x8_64_add_sse2(in, out, stride, 12);
vpx_highbd_idct8x8_64_add_sse2(in, out, stride, 12);
}
#endif // HAVE_SSE2
#endif // CONFIG_VP9_HIGHBITDEPTH
@ -658,14 +658,14 @@ using std::tr1::make_tuple;
INSTANTIATE_TEST_CASE_P(
C, FwdTrans8x8DCT,
::testing::Values(
make_tuple(&vpx_fdct8x8_c, &vp9_idct8x8_64_add_c, 0, VPX_BITS_8),
make_tuple(&vpx_fdct8x8_c, &vpx_idct8x8_64_add_c, 0, VPX_BITS_8),
make_tuple(&vpx_highbd_fdct8x8_c, &idct8x8_10, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct8x8_c, &idct8x8_12, 0, VPX_BITS_12)));
#else
INSTANTIATE_TEST_CASE_P(
C, FwdTrans8x8DCT,
::testing::Values(
make_tuple(&vpx_fdct8x8_c, &vp9_idct8x8_64_add_c, 0, VPX_BITS_8)));
make_tuple(&vpx_fdct8x8_c, &vpx_idct8x8_64_add_c, 0, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if CONFIG_VP9_HIGHBITDEPTH
@ -698,7 +698,7 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P(
NEON, FwdTrans8x8DCT,
::testing::Values(
make_tuple(&vpx_fdct8x8_neon, &vp9_idct8x8_64_add_neon, 0,
make_tuple(&vpx_fdct8x8_neon, &vpx_idct8x8_64_add_neon, 0,
VPX_BITS_8)));
#endif // HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -716,7 +716,7 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P(
SSE2, FwdTrans8x8DCT,
::testing::Values(
make_tuple(&vpx_fdct8x8_sse2, &vp9_idct8x8_64_add_sse2, 0,
make_tuple(&vpx_fdct8x8_sse2, &vpx_idct8x8_64_add_sse2, 0,
VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
SSE2, FwdTrans8x8HT,
@ -731,7 +731,7 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P(
SSE2, FwdTrans8x8DCT,
::testing::Values(
make_tuple(&vpx_fdct8x8_sse2, &vp9_idct8x8_64_add_c, 0, VPX_BITS_8),
make_tuple(&vpx_fdct8x8_sse2, &vpx_idct8x8_64_add_c, 0, VPX_BITS_8),
make_tuple(&vpx_highbd_fdct8x8_c,
&idct8x8_64_add_10_sse2, 12, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct8x8_sse2,
@ -769,7 +769,7 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P(
SSSE3, FwdTrans8x8DCT,
::testing::Values(
make_tuple(&vpx_fdct8x8_ssse3, &vp9_idct8x8_64_add_ssse3, 0,
make_tuple(&vpx_fdct8x8_ssse3, &vpx_idct8x8_64_add_ssse3, 0,
VPX_BITS_8)));
#endif
@ -777,7 +777,7 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P(
MSA, FwdTrans8x8DCT,
::testing::Values(
make_tuple(&vpx_fdct8x8_msa, &vp9_idct8x8_64_add_msa, 0, VPX_BITS_8)));
make_tuple(&vpx_fdct8x8_msa, &vpx_idct8x8_64_add_msa, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
MSA, FwdTrans8x8HT,
::testing::Values(

2
test/idct8x8_test.cc
Просмотреть файл

@ -124,7 +124,7 @@ TEST(VP9Idct8x8Test, AccuracyCheck) {
reference_dct_2d(input, output_r);
for (int j = 0; j < 64; ++j)
coeff[j] = round(output_r[j]);
vp9_idct8x8_64_add_c(coeff, dst, 8);
vpx_idct8x8_64_add_c(coeff, dst, 8);
for (int j = 0; j < 64; ++j) {
const int diff = dst[j] - src[j];
const int error = diff * diff;

112
test/partial_idct_test.cc
Просмотреть файл

@ -203,32 +203,32 @@ INSTANTIATE_TEST_CASE_P(
C, PartialIDctTest,
::testing::Values(
make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_34_add_c,
&vpx_idct32x32_1024_add_c,
&vpx_idct32x32_34_add_c,
TX_32X32, 34),
make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_1_add_c,
&vpx_idct32x32_1024_add_c,
&vpx_idct32x32_1_add_c,
TX_32X32, 1),
make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_10_add_c,
&vpx_idct16x16_256_add_c,
&vpx_idct16x16_10_add_c,
TX_16X16, 10),
make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_1_add_c,
&vpx_idct16x16_256_add_c,
&vpx_idct16x16_1_add_c,
TX_16X16, 1),
make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_12_add_c,
&vpx_idct8x8_64_add_c,
&vpx_idct8x8_12_add_c,
TX_8X8, 12),
make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_1_add_c,
&vpx_idct8x8_64_add_c,
&vpx_idct8x8_1_add_c,
TX_8X8, 1),
make_tuple(&vpx_fdct4x4_c,
&vp9_idct4x4_16_add_c,
&vp9_idct4x4_1_add_c,
&vpx_idct4x4_16_add_c,
&vpx_idct4x4_1_add_c,
TX_4X4, 1)));
#if HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -236,28 +236,28 @@ INSTANTIATE_TEST_CASE_P(
NEON, PartialIDctTest,
::testing::Values(
make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_1_add_neon,
&vpx_idct32x32_1024_add_c,
&vpx_idct32x32_1_add_neon,
TX_32X32, 1),
make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_10_add_neon,
&vpx_idct16x16_256_add_c,
&vpx_idct16x16_10_add_neon,
TX_16X16, 10),
make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_1_add_neon,
&vpx_idct16x16_256_add_c,
&vpx_idct16x16_1_add_neon,
TX_16X16, 1),
make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_12_add_neon,
&vpx_idct8x8_64_add_c,
&vpx_idct8x8_12_add_neon,
TX_8X8, 12),
make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_1_add_neon,
&vpx_idct8x8_64_add_c,
&vpx_idct8x8_1_add_neon,
TX_8X8, 1),
make_tuple(&vpx_fdct4x4_c,
&vp9_idct4x4_16_add_c,
&vp9_idct4x4_1_add_neon,
&vpx_idct4x4_16_add_c,
&vpx_idct4x4_1_add_neon,
TX_4X4, 1)));
#endif // HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -266,32 +266,32 @@ INSTANTIATE_TEST_CASE_P(
SSE2, PartialIDctTest,
::testing::Values(
make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_34_add_sse2,
&vpx_idct32x32_1024_add_c,
&vpx_idct32x32_34_add_sse2,
TX_32X32, 34),
make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_1_add_sse2,
&vpx_idct32x32_1024_add_c,
&vpx_idct32x32_1_add_sse2,
TX_32X32, 1),
make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_10_add_sse2,
&vpx_idct16x16_256_add_c,
&vpx_idct16x16_10_add_sse2,
TX_16X16, 10),
make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_1_add_sse2,
&vpx_idct16x16_256_add_c,
&vpx_idct16x16_1_add_sse2,
TX_16X16, 1),
make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_12_add_sse2,
&vpx_idct8x8_64_add_c,
&vpx_idct8x8_12_add_sse2,
TX_8X8, 12),
make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_1_add_sse2,
&vpx_idct8x8_64_add_c,
&vpx_idct8x8_1_add_sse2,
TX_8X8, 1),
make_tuple(&vpx_fdct4x4_c,
&vp9_idct4x4_16_add_c,
&vp9_idct4x4_1_add_sse2,
&vpx_idct4x4_16_add_c,
&vpx_idct4x4_1_add_sse2,
TX_4X4, 1)));
#endif
@ -301,8 +301,8 @@ INSTANTIATE_TEST_CASE_P(
SSSE3_64, PartialIDctTest,
::testing::Values(
make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_12_add_ssse3,
&vpx_idct8x8_64_add_c,
&vpx_idct8x8_12_add_ssse3,
TX_8X8, 12)));
#endif
@ -311,32 +311,32 @@ INSTANTIATE_TEST_CASE_P(
MSA, PartialIDctTest,
::testing::Values(
make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_34_add_msa,
&vpx_idct32x32_1024_add_c,
&vpx_idct32x32_34_add_msa,
TX_32X32, 34),
make_tuple(&vpx_fdct32x32_c,
&vp9_idct32x32_1024_add_c,
&vp9_idct32x32_1_add_msa,
&vpx_idct32x32_1024_add_c,
&vpx_idct32x32_1_add_msa,
TX_32X32, 1),
make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_10_add_msa,
&vpx_idct16x16_256_add_c,
&vpx_idct16x16_10_add_msa,
TX_16X16, 10),
make_tuple(&vpx_fdct16x16_c,
&vp9_idct16x16_256_add_c,
&vp9_idct16x16_1_add_msa,
&vpx_idct16x16_256_add_c,
&vpx_idct16x16_1_add_msa,
TX_16X16, 1),
make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_12_add_msa,
&vpx_idct8x8_64_add_c,
&vpx_idct8x8_12_add_msa,
TX_8X8, 10),
make_tuple(&vpx_fdct8x8_c,
&vp9_idct8x8_64_add_c,
&vp9_idct8x8_1_add_msa,
&vpx_idct8x8_64_add_c,
&vpx_idct8x8_1_add_msa,
TX_8X8, 1),
make_tuple(&vpx_fdct4x4_c,
&vp9_idct4x4_16_add_c,
&vp9_idct4x4_1_add_msa,
&vpx_idct4x4_16_add_c,
&vpx_idct4x4_1_add_msa,
TX_4X4, 1)));
#endif // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE

6
test/register_state_check.h
Просмотреть файл

@ -96,7 +96,7 @@ class RegisterStateCheck {
extern "C" {
// Save the d8-d15 registers into store.
void vp9_push_neon(int64_t *store);
void vpx_push_neon(int64_t *store);
}
namespace libvpx_test {
@ -111,7 +111,7 @@ class RegisterStateCheck {
private:
static bool StoreRegisters(int64_t store[8]) {
vp9_push_neon(store);
vpx_push_neon(store);
return true;
}
@ -119,7 +119,7 @@ class RegisterStateCheck {
bool Check() const {
if (!initialized_) return false;
int64_t post_store[8];
vp9_push_neon(post_store);
vpx_push_neon(post_store);
for (int i = 0; i < 8; ++i) {
EXPECT_EQ(pre_store_[i], post_store[i]) << "d"
<< i + 8 << " has been modified";

8
vp9/common/mips/dspr2/vp9_itrans4_dspr2.c
Просмотреть файл

@ -38,11 +38,11 @@ void vp9_iht4x4_16_add_dspr2(const int16_t *input, uint8_t *dest,
switch (tx_type) {
case DCT_DCT: // DCT in both horizontal and vertical
vp9_idct4_rows_dspr2(input, outptr);
vp9_idct4_columns_add_blk_dspr2(&out[0], dest, dest_stride);
vpx_idct4_rows_dspr2(input, outptr);
vpx_idct4_columns_add_blk_dspr2(&out[0], dest, dest_stride);
break;
case ADST_DCT: // ADST in vertical, DCT in horizontal
vp9_idct4_rows_dspr2(input, outptr);
vpx_idct4_rows_dspr2(input, outptr);
outptr = out;
@ -69,7 +69,7 @@ void vp9_iht4x4_16_add_dspr2(const int16_t *input, uint8_t *dest,
temp_in[i * 4 + j] = out[j * 4 + i];
}
}
vp9_idct4_columns_add_blk_dspr2(&temp_in[0], dest, dest_stride);
vpx_idct4_columns_add_blk_dspr2(&temp_in[0], dest, dest_stride);
break;
case ADST_ADST: // ADST in both directions
for (i = 0; i < 4; ++i) {

16
vp9/common/mips/msa/vp9_idct16x16_msa.c
Просмотреть файл

@ -24,13 +24,13 @@ void vp9_iht16x16_256_add_msa(const int16_t *input, uint8_t *dst,
/* transform rows */
for (i = 0; i < 2; ++i) {
/* process 16 * 8 block */
vp9_idct16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7)));
vpx_idct16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7)));
}
/* transform columns */
for (i = 0; i < 2; ++i) {
/* process 8 * 16 block */
vp9_idct16_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)),
vpx_idct16_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)),
dst_stride);
}
break;
@ -38,12 +38,12 @@ void vp9_iht16x16_256_add_msa(const int16_t *input, uint8_t *dst,
/* transform rows */
for (i = 0; i < 2; ++i) {
/* process 16 * 8 block */
vp9_idct16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7)));
vpx_idct16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7)));
}
/* transform columns */
for (i = 0; i < 2; ++i) {
vp9_iadst16_1d_columns_addblk_msa((out_ptr + (i << 3)),
vpx_iadst16_1d_columns_addblk_msa((out_ptr + (i << 3)),
(dst + (i << 3)), dst_stride);
}
break;
@ -51,13 +51,13 @@ void vp9_iht16x16_256_add_msa(const int16_t *input, uint8_t *dst,
/* transform rows */
for (i = 0; i < 2; ++i) {
/* process 16 * 8 block */
vp9_iadst16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7)));
vpx_iadst16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7)));
}
/* transform columns */
for (i = 0; i < 2; ++i) {
/* process 8 * 16 block */
vp9_idct16_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)),
vpx_idct16_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)),
dst_stride);
}
break;
@ -65,12 +65,12 @@ void vp9_iht16x16_256_add_msa(const int16_t *input, uint8_t *dst,
/* transform rows */
for (i = 0; i < 2; ++i) {
/* process 16 * 8 block */
vp9_iadst16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7)));
vpx_iadst16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7)));
}
/* transform columns */
for (i = 0; i < 2; ++i) {
vp9_iadst16_1d_columns_addblk_msa((out_ptr + (i << 3)),
vpx_iadst16_1d_columns_addblk_msa((out_ptr + (i << 3)),
(dst + (i << 3)), dst_stride);
}
break;

76
vp9/common/vp9_idct.c
Просмотреть файл

@ -123,18 +123,18 @@ void vp9_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest, int stride,
void vp9_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob) {
if (eob > 1)
vp9_idct4x4_16_add(input, dest, stride);
vpx_idct4x4_16_add(input, dest, stride);
else
vp9_idct4x4_1_add(input, dest, stride);
vpx_idct4x4_1_add(input, dest, stride);
}
void vp9_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob) {
if (eob > 1)
vp9_iwht4x4_16_add(input, dest, stride);
vpx_iwht4x4_16_add(input, dest, stride);
else
vp9_iwht4x4_1_add(input, dest, stride);
vpx_iwht4x4_1_add(input, dest, stride);
}
void vp9_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
@ -148,11 +148,11 @@ void vp9_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
// Combine that with code here.
if (eob == 1)
// DC only DCT coefficient
vp9_idct8x8_1_add(input, dest, stride);
vpx_idct8x8_1_add(input, dest, stride);
else if (eob <= 12)
vp9_idct8x8_12_add(input, dest, stride);
vpx_idct8x8_12_add(input, dest, stride);
else
vp9_idct8x8_64_add(input, dest, stride);
vpx_idct8x8_64_add(input, dest, stride);
}
void vp9_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride,
@ -161,22 +161,22 @@ void vp9_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride,
* coefficients. Use eobs to separate different cases. */
if (eob == 1)
/* DC only DCT coefficient. */
vp9_idct16x16_1_add(input, dest, stride);
vpx_idct16x16_1_add(input, dest, stride);
else if (eob <= 10)
vp9_idct16x16_10_add(input, dest, stride);
vpx_idct16x16_10_add(input, dest, stride);
else
vp9_idct16x16_256_add(input, dest, stride);
vpx_idct16x16_256_add(input, dest, stride);
}
void vp9_idct32x32_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob) {
if (eob == 1)
vp9_idct32x32_1_add(input, dest, stride);
vpx_idct32x32_1_add(input, dest, stride);
else if (eob <= 34)
// non-zero coeff only in upper-left 8x8
vp9_idct32x32_34_add(input, dest, stride);
vpx_idct32x32_34_add(input, dest, stride);
else
vp9_idct32x32_1024_add(input, dest, stride);
vpx_idct32x32_1024_add(input, dest, stride);
}
// iht
@ -210,10 +210,10 @@ void vp9_iht16x16_add(TX_TYPE tx_type, const tran_low_t *input, uint8_t *dest,
void vp9_highbd_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
const highbd_transform_2d IHT_4[] = {
{ vp9_highbd_idct4_c, vp9_highbd_idct4_c }, // DCT_DCT = 0
{ highbd_iadst4_c, vp9_highbd_idct4_c }, // ADST_DCT = 1
{ vp9_highbd_idct4_c, highbd_iadst4_c }, // DCT_ADST = 2
{ highbd_iadst4_c, highbd_iadst4_c } // ADST_ADST = 3
{ vpx_highbd_idct4_c, vpx_highbd_idct4_c }, // DCT_DCT = 0
{ vpx_highbd_iadst4_c, vpx_highbd_idct4_c }, // ADST_DCT = 1
{ vpx_highbd_idct4_c, vpx_highbd_iadst4_c }, // DCT_ADST = 2
{ vpx_highbd_iadst4_c, vpx_highbd_iadst4_c } // ADST_ADST = 3
};
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
@ -242,10 +242,10 @@ void vp9_highbd_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
}
static const highbd_transform_2d HIGH_IHT_8[] = {
{ vp9_highbd_idct8_c, vp9_highbd_idct8_c }, // DCT_DCT = 0
{ highbd_iadst8_c, vp9_highbd_idct8_c }, // ADST_DCT = 1
{ vp9_highbd_idct8_c, highbd_iadst8_c }, // DCT_ADST = 2
{ highbd_iadst8_c, highbd_iadst8_c } // ADST_ADST = 3
{ vpx_highbd_idct8_c, vpx_highbd_idct8_c }, // DCT_DCT = 0
{ vpx_highbd_iadst8_c, vpx_highbd_idct8_c }, // ADST_DCT = 1
{ vpx_highbd_idct8_c, vpx_highbd_iadst8_c }, // DCT_ADST = 2
{ vpx_highbd_iadst8_c, vpx_highbd_iadst8_c } // ADST_ADST = 3
};
void vp9_highbd_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
@ -277,10 +277,10 @@ void vp9_highbd_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
}
static const highbd_transform_2d HIGH_IHT_16[] = {
{ vp9_highbd_idct16_c, vp9_highbd_idct16_c }, // DCT_DCT = 0
{ highbd_iadst16_c, vp9_highbd_idct16_c }, // ADST_DCT = 1
{ vp9_highbd_idct16_c, highbd_iadst16_c }, // DCT_ADST = 2
{ highbd_iadst16_c, highbd_iadst16_c } // ADST_ADST = 3
{ vpx_highbd_idct16_c, vpx_highbd_idct16_c }, // DCT_DCT = 0
{ vpx_highbd_iadst16_c, vpx_highbd_idct16_c }, // ADST_DCT = 1
{ vpx_highbd_idct16_c, vpx_highbd_iadst16_c }, // DCT_ADST = 2
{ vpx_highbd_iadst16_c, vpx_highbd_iadst16_c } // ADST_ADST = 3
};
void vp9_highbd_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
@ -315,18 +315,18 @@ void vp9_highbd_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
void vp9_highbd_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob, int bd) {
if (eob > 1)
vp9_highbd_idct4x4_16_add(input, dest, stride, bd);
vpx_highbd_idct4x4_16_add(input, dest, stride, bd);
else
vp9_highbd_idct4x4_1_add(input, dest, stride, bd);
vpx_highbd_idct4x4_1_add(input, dest, stride, bd);
}
void vp9_highbd_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob, int bd) {
if (eob > 1)
vp9_highbd_iwht4x4_16_add(input, dest, stride, bd);
vpx_highbd_iwht4x4_16_add(input, dest, stride, bd);
else
vp9_highbd_iwht4x4_1_add(input, dest, stride, bd);
vpx_highbd_iwht4x4_1_add(input, dest, stride, bd);
}
void vp9_highbd_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
@ -340,11 +340,11 @@ void vp9_highbd_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
// Combine that with code here.
// DC only DCT coefficient
if (eob == 1) {
vp9_highbd_idct8x8_1_add(input, dest, stride, bd);
vpx_highbd_idct8x8_1_add(input, dest, stride, bd);
} else if (eob <= 10) {
vp9_highbd_idct8x8_10_add(input, dest, stride, bd);
vpx_highbd_idct8x8_10_add(input, dest, stride, bd);
} else {
vp9_highbd_idct8x8_64_add(input, dest, stride, bd);
vpx_highbd_idct8x8_64_add(input, dest, stride, bd);
}
}
@ -354,11 +354,11 @@ void vp9_highbd_idct16x16_add(const tran_low_t *input, uint8_t *dest,
// coefficients. Use eobs to separate different cases.
// DC only DCT coefficient.
if (eob == 1) {
vp9_highbd_idct16x16_1_add(input, dest, stride, bd);
vpx_highbd_idct16x16_1_add(input, dest, stride, bd);
} else if (eob <= 10) {
vp9_highbd_idct16x16_10_add(input, dest, stride, bd);
vpx_highbd_idct16x16_10_add(input, dest, stride, bd);
} else {
vp9_highbd_idct16x16_256_add(input, dest, stride, bd);
vpx_highbd_idct16x16_256_add(input, dest, stride, bd);
}
}
@ -366,11 +366,11 @@ void vp9_highbd_idct32x32_add(const tran_low_t *input, uint8_t *dest,
int stride, int eob, int bd) {
// Non-zero coeff only in upper-left 8x8
if (eob == 1) {
vp9_highbd_idct32x32_1_add(input, dest, stride, bd);
vpx_highbd_idct32x32_1_add(input, dest, stride, bd);
} else if (eob <= 34) {
vp9_highbd_idct32x32_34_add(input, dest, stride, bd);
vpx_highbd_idct32x32_34_add(input, dest, stride, bd);
} else {
vp9_highbd_idct32x32_1024_add(input, dest, stride, bd);
vpx_highbd_idct32x32_1024_add(input, dest, stride, bd);
}
}

4
vp9/common/vp9_idct.h
Просмотреть файл

@ -44,8 +44,8 @@ void vp9_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob);
void vp9_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob);
void vp9_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride, int
eob);
void vp9_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob);
void vp9_idct32x32_add(const tran_low_t *input, uint8_t *dest, int stride,
int eob);

2
vp9/vp9_common.mk
Просмотреть файл

@ -88,8 +88,6 @@ endif
VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_idct_intrin_sse2.c
VP9_COMMON_SRCS-$(HAVE_NEON_ASM) += common/arm/neon/vp9_save_reg_neon$(ASM)
ifneq ($(CONFIG_VP9_HIGHBITDEPTH),yes)
VP9_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/vp9_iht4x4_add_neon.c
VP9_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/vp9_iht8x8_add_neon.c

2
vpx_dsp/arm/fwd_txfm_neon.c
Просмотреть файл

@ -170,7 +170,7 @@ void vpx_fdct8x8_neon(const int16_t *input, int16_t *final_output, int stride) {
}
} // for
{
// from vp9_dct_sse2.c
// from vpx_dct_sse2.c
// Post-condition (division by two)
// division of two 16 bits signed numbers using shifts
// n / 2 = (n - (n >> 15)) >> 1

8
vpx_dsp/arm/idct16x16_1_add_neon.asm
Просмотреть файл

@ -8,21 +8,21 @@
;
EXPORT |vp9_idct16x16_1_add_neon|
EXPORT |vpx_idct16x16_1_add_neon|
ARM
REQUIRE8
PRESERVE8
AREA ||.text||, CODE, READONLY, ALIGN=2
;void vp9_idct16x16_1_add_neon(int16_t *input, uint8_t *dest,
;void vpx_idct16x16_1_add_neon(int16_t *input, uint8_t *dest,
; int dest_stride)
;
; r0 int16_t input
; r1 uint8_t *dest
; r2 int dest_stride)
|vp9_idct16x16_1_add_neon| PROC
|vpx_idct16x16_1_add_neon| PROC
ldrsh r0, [r0]
; generate cospi_16_64 = 11585
@ -193,6 +193,6 @@
vst1.64 {d31}, [r12], r2
bx lr
ENDP ; |vp9_idct16x16_1_add_neon|
ENDP ; |vpx_idct16x16_1_add_neon|
END

2
vpx_dsp/arm/idct16x16_1_add_neon.c
Просмотреть файл

@ -13,7 +13,7 @@
#include "vpx_dsp/inv_txfm.h"
#include "vpx_ports/mem.h"
void vp9_idct16x16_1_add_neon(
void vpx_idct16x16_1_add_neon(
int16_t *input,
uint8_t *dest,
int dest_stride) {

32
vpx_dsp/arm/idct16x16_add_neon.asm
Просмотреть файл

@ -8,10 +8,10 @@
; be found in the AUTHORS file in the root of the source tree.
;
EXPORT |vp9_idct16x16_256_add_neon_pass1|
EXPORT |vp9_idct16x16_256_add_neon_pass2|
EXPORT |vp9_idct16x16_10_add_neon_pass1|
EXPORT |vp9_idct16x16_10_add_neon_pass2|
EXPORT |vpx_idct16x16_256_add_neon_pass1|
EXPORT |vpx_idct16x16_256_add_neon_pass2|
EXPORT |vpx_idct16x16_10_add_neon_pass1|
EXPORT |vpx_idct16x16_10_add_neon_pass2|
ARM
REQUIRE8
PRESERVE8
@ -36,7 +36,7 @@
MEND
AREA Block, CODE, READONLY ; name this block of code
;void |vp9_idct16x16_256_add_neon_pass1|(int16_t *input,
;void |vpx_idct16x16_256_add_neon_pass1|(int16_t *input,
; int16_t *output, int output_stride)
;
; r0 int16_t input
@ -46,7 +46,7 @@
; idct16 stage1 - stage6 on all the elements loaded in q8-q15. The output
; will be stored back into q8-q15 registers. This function will touch q0-q7
; registers and use them as buffer during calculation.
|vp9_idct16x16_256_add_neon_pass1| PROC
|vpx_idct16x16_256_add_neon_pass1| PROC
; TODO(hkuang): Find a better way to load the elements.
; load elements of 0, 2, 4, 6, 8, 10, 12, 14 into q8 - q15
@ -273,9 +273,9 @@
vst1.64 {d31}, [r1], r2
bx lr
ENDP ; |vp9_idct16x16_256_add_neon_pass1|
ENDP ; |vpx_idct16x16_256_add_neon_pass1|
;void vp9_idct16x16_256_add_neon_pass2(int16_t *src,
;void vpx_idct16x16_256_add_neon_pass2(int16_t *src,
; int16_t *output,
; int16_t *pass1Output,
; int16_t skip_adding,
@ -292,7 +292,7 @@
; idct16 stage1 - stage7 on all the elements loaded in q8-q15. The output
; will be stored back into q8-q15 registers. This function will touch q0-q7
; registers and use them as buffer during calculation.
|vp9_idct16x16_256_add_neon_pass2| PROC
|vpx_idct16x16_256_add_neon_pass2| PROC
push {r3-r9}
; TODO(hkuang): Find a better way to load the elements.
@ -784,9 +784,9 @@ skip_adding_dest
end_idct16x16_pass2
pop {r3-r9}
bx lr
ENDP ; |vp9_idct16x16_256_add_neon_pass2|
ENDP ; |vpx_idct16x16_256_add_neon_pass2|
;void |vp9_idct16x16_10_add_neon_pass1|(int16_t *input,
;void |vpx_idct16x16_10_add_neon_pass1|(int16_t *input,
; int16_t *output, int output_stride)
;
; r0 int16_t input
@ -796,7 +796,7 @@ end_idct16x16_pass2
; idct16 stage1 - stage6 on all the elements loaded in q8-q15. The output
; will be stored back into q8-q15 registers. This function will touch q0-q7
; registers and use them as buffer during calculation.
|vp9_idct16x16_10_add_neon_pass1| PROC
|vpx_idct16x16_10_add_neon_pass1| PROC
; TODO(hkuang): Find a better way to load the elements.
; load elements of 0, 2, 4, 6, 8, 10, 12, 14 into q8 - q15
@ -905,9 +905,9 @@ end_idct16x16_pass2
vst1.64 {d31}, [r1], r2
bx lr
ENDP ; |vp9_idct16x16_10_add_neon_pass1|
ENDP ; |vpx_idct16x16_10_add_neon_pass1|
;void vp9_idct16x16_10_add_neon_pass2(int16_t *src,
;void vpx_idct16x16_10_add_neon_pass2(int16_t *src,
; int16_t *output,
; int16_t *pass1Output,
; int16_t skip_adding,
@ -924,7 +924,7 @@ end_idct16x16_pass2
; idct16 stage1 - stage7 on all the elements loaded in q8-q15. The output
; will be stored back into q8-q15 registers. This function will touch q0-q7
; registers and use them as buffer during calculation.
|vp9_idct16x16_10_add_neon_pass2| PROC
|vpx_idct16x16_10_add_neon_pass2| PROC
push {r3-r9}
; TODO(hkuang): Find a better way to load the elements.
@ -1175,5 +1175,5 @@ end_idct16x16_pass2
end_idct10_16x16_pass2
pop {r3-r9}
bx lr
ENDP ; |vp9_idct16x16_10_add_neon_pass2|
ENDP ; |vpx_idct16x16_10_add_neon_pass2|
END

8
vpx_dsp/arm/idct16x16_add_neon.c
Просмотреть файл

@ -82,7 +82,7 @@ static INLINE void TRANSPOSE8X8(
return;
}
void vp9_idct16x16_256_add_neon_pass1(
void vpx_idct16x16_256_add_neon_pass1(
int16_t *in,
int16_t *out,
int output_stride) {
@ -320,7 +320,7 @@ void vp9_idct16x16_256_add_neon_pass1(
return;
}
void vp9_idct16x16_256_add_neon_pass2(
void vpx_idct16x16_256_add_neon_pass2(
int16_t *src,
int16_t *out,
int16_t *pass1Output,
@ -879,7 +879,7 @@ void vp9_idct16x16_256_add_neon_pass2(
return;
}
void vp9_idct16x16_10_add_neon_pass1(
void vpx_idct16x16_10_add_neon_pass1(
int16_t *in,
int16_t *out,
int output_stride) {
@ -1017,7 +1017,7 @@ void vp9_idct16x16_10_add_neon_pass1(
return;
}
void vp9_idct16x16_10_add_neon_pass2(
void vpx_idct16x16_10_add_neon_pass2(
int16_t *src,
int16_t *out,
int16_t *pass1Output,

52
vpx_dsp/arm/idct16x16_neon.c
Просмотреть файл

@ -10,19 +10,19 @@
#include "vpx_dsp/vpx_dsp_common.h"
void vp9_idct16x16_256_add_neon_pass1(const int16_t *input,
void vpx_idct16x16_256_add_neon_pass1(const int16_t *input,
int16_t *output,
int output_stride);
void vp9_idct16x16_256_add_neon_pass2(const int16_t *src,
void vpx_idct16x16_256_add_neon_pass2(const int16_t *src,
int16_t *output,
int16_t *pass1Output,
int16_t skip_adding,
uint8_t *dest,
int dest_stride);
void vp9_idct16x16_10_add_neon_pass1(const int16_t *input,
void vpx_idct16x16_10_add_neon_pass1(const int16_t *input,
int16_t *output,
int output_stride);
void vp9_idct16x16_10_add_neon_pass2(const int16_t *src,
void vpx_idct16x16_10_add_neon_pass2(const int16_t *src,
int16_t *output,
int16_t *pass1Output,
int16_t skip_adding,
@ -31,11 +31,11 @@ void vp9_idct16x16_10_add_neon_pass2(const int16_t *src,
#if HAVE_NEON_ASM
/* For ARM NEON, d8-d15 are callee-saved registers, and need to be saved. */
extern void vp9_push_neon(int64_t *store);
extern void vp9_pop_neon(int64_t *store);
extern void vpx_push_neon(int64_t *store);
extern void vpx_pop_neon(int64_t *store);
#endif // HAVE_NEON_ASM
void vp9_idct16x16_256_add_neon(const int16_t *input,
void vpx_idct16x16_256_add_neon(const int16_t *input,
uint8_t *dest, int dest_stride) {
#if HAVE_NEON_ASM
int64_t store_reg[8];
@ -45,18 +45,18 @@ void vp9_idct16x16_256_add_neon(const int16_t *input,
#if HAVE_NEON_ASM
// save d8-d15 register values.
vp9_push_neon(store_reg);
vpx_push_neon(store_reg);
#endif
/* Parallel idct on the upper 8 rows */
// First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
// stage 6 result in pass1_output.
vp9_idct16x16_256_add_neon_pass1(input, pass1_output, 8);
vpx_idct16x16_256_add_neon_pass1(input, pass1_output, 8);
// Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
// with result in pass1(pass1_output) to calculate final result in stage 7
// which will be saved into row_idct_output.
vp9_idct16x16_256_add_neon_pass2(input+1,
vpx_idct16x16_256_add_neon_pass2(input+1,
row_idct_output,
pass1_output,
0,
@ -66,12 +66,12 @@ void vp9_idct16x16_256_add_neon(const int16_t *input,
/* Parallel idct on the lower 8 rows */
// First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
// stage 6 result in pass1_output.
vp9_idct16x16_256_add_neon_pass1(input+8*16, pass1_output, 8);
vpx_idct16x16_256_add_neon_pass1(input+8*16, pass1_output, 8);
// Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
// with result in pass1(pass1_output) to calculate final result in stage 7
// which will be saved into row_idct_output.
vp9_idct16x16_256_add_neon_pass2(input+8*16+1,
vpx_idct16x16_256_add_neon_pass2(input+8*16+1,
row_idct_output+8,
pass1_output,
0,
@ -81,12 +81,12 @@ void vp9_idct16x16_256_add_neon(const int16_t *input,
/* Parallel idct on the left 8 columns */
// First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
// stage 6 result in pass1_output.
vp9_idct16x16_256_add_neon_pass1(row_idct_output, pass1_output, 8);
vpx_idct16x16_256_add_neon_pass1(row_idct_output, pass1_output, 8);
// Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
// with result in pass1(pass1_output) to calculate final result in stage 7.
// Then add the result to the destination data.
vp9_idct16x16_256_add_neon_pass2(row_idct_output+1,
vpx_idct16x16_256_add_neon_pass2(row_idct_output+1,
row_idct_output,
pass1_output,
1,
@ -96,12 +96,12 @@ void vp9_idct16x16_256_add_neon(const int16_t *input,
/* Parallel idct on the right 8 columns */
// First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
// stage 6 result in pass1_output.
vp9_idct16x16_256_add_neon_pass1(row_idct_output+8*16, pass1_output, 8);
vpx_idct16x16_256_add_neon_pass1(row_idct_output+8*16, pass1_output, 8);
// Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
// with result in pass1(pass1_output) to calculate final result in stage 7.
// Then add the result to the destination data.
vp9_idct16x16_256_add_neon_pass2(row_idct_output+8*16+1,
vpx_idct16x16_256_add_neon_pass2(row_idct_output+8*16+1,
row_idct_output+8,
pass1_output,
1,
@ -110,13 +110,13 @@ void vp9_idct16x16_256_add_neon(const int16_t *input,
#if HAVE_NEON_ASM
// restore d8-d15 register values.
vp9_pop_neon(store_reg);
vpx_pop_neon(store_reg);
#endif
return;
}
void vp9_idct16x16_10_add_neon(const int16_t *input,
void vpx_idct16x16_10_add_neon(const int16_t *input,
uint8_t *dest, int dest_stride) {
#if HAVE_NEON_ASM
int64_t store_reg[8];
@ -126,18 +126,18 @@ void vp9_idct16x16_10_add_neon(const int16_t *input,
#if HAVE_NEON_ASM
// save d8-d15 register values.
vp9_push_neon(store_reg);
vpx_push_neon(store_reg);
#endif
/* Parallel idct on the upper 8 rows */
// First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
// stage 6 result in pass1_output.
vp9_idct16x16_10_add_neon_pass1(input, pass1_output, 8);
vpx_idct16x16_10_add_neon_pass1(input, pass1_output, 8);
// Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
// with result in pass1(pass1_output) to calculate final result in stage 7
// which will be saved into row_idct_output.
vp9_idct16x16_10_add_neon_pass2(input+1,
vpx_idct16x16_10_add_neon_pass2(input+1,
row_idct_output,
pass1_output,
0,
@ -149,12 +149,12 @@ void vp9_idct16x16_10_add_neon(const int16_t *input,
/* Parallel idct on the left 8 columns */
// First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
// stage 6 result in pass1_output.
vp9_idct16x16_256_add_neon_pass1(row_idct_output, pass1_output, 8);
vpx_idct16x16_256_add_neon_pass1(row_idct_output, pass1_output, 8);
// Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
// with result in pass1(pass1_output) to calculate final result in stage 7.
// Then add the result to the destination data.
vp9_idct16x16_256_add_neon_pass2(row_idct_output+1,
vpx_idct16x16_256_add_neon_pass2(row_idct_output+1,
row_idct_output,
pass1_output,
1,
@ -164,12 +164,12 @@ void vp9_idct16x16_10_add_neon(const int16_t *input,
/* Parallel idct on the right 8 columns */
// First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
// stage 6 result in pass1_output.
vp9_idct16x16_256_add_neon_pass1(row_idct_output+8*16, pass1_output, 8);
vpx_idct16x16_256_add_neon_pass1(row_idct_output+8*16, pass1_output, 8);
// Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
// with result in pass1(pass1_output) to calculate final result in stage 7.
// Then add the result to the destination data.
vp9_idct16x16_256_add_neon_pass2(row_idct_output+8*16+1,
vpx_idct16x16_256_add_neon_pass2(row_idct_output+8*16+1,
row_idct_output+8,
pass1_output,
1,
@ -178,7 +178,7 @@ void vp9_idct16x16_10_add_neon(const int16_t *input,
#if HAVE_NEON_ASM
// restore d8-d15 register values.
vp9_pop_neon(store_reg);
vpx_pop_neon(store_reg);
#endif
return;

8
vpx_dsp/arm/idct32x32_1_add_neon.asm
Просмотреть файл

@ -7,7 +7,7 @@
; file in the root of the source tree.
;
EXPORT |vp9_idct32x32_1_add_neon|
EXPORT |vpx_idct32x32_1_add_neon|
ARM
REQUIRE8
PRESERVE8
@ -64,14 +64,14 @@
vst1.8 {q15},[$dst], $stride
MEND
;void vp9_idct32x32_1_add_neon(int16_t *input, uint8_t *dest,
;void vpx_idct32x32_1_add_neon(int16_t *input, uint8_t *dest,
; int dest_stride)
;
; r0 int16_t input
; r1 uint8_t *dest
; r2 int dest_stride
|vp9_idct32x32_1_add_neon| PROC
|vpx_idct32x32_1_add_neon| PROC
push {lr}
pld [r1]
add r3, r1, #16 ; r3 dest + 16 for second loop
@ -140,5 +140,5 @@ diff_positive_32_32_loop
bne diff_positive_32_32_loop
pop {pc}
ENDP ; |vp9_idct32x32_1_add_neon|
ENDP ; |vpx_idct32x32_1_add_neon|
END

2
vpx_dsp/arm/idct32x32_1_add_neon.c
Просмотреть файл

@ -115,7 +115,7 @@ static INLINE void ST_16x8(
return;
}
void vp9_idct32x32_1_add_neon(
void vpx_idct32x32_1_add_neon(
int16_t *input,
uint8_t *dest,
int dest_stride) {

8
vpx_dsp/arm/idct32x32_add_neon.asm
Просмотреть файл

@ -43,7 +43,7 @@ cospi_30_64 EQU 1606
cospi_31_64 EQU 804
EXPORT |vp9_idct32x32_1024_add_neon|
EXPORT |vpx_idct32x32_1024_add_neon|
ARM
REQUIRE8
PRESERVE8
@ -288,7 +288,7 @@ cospi_31_64 EQU 804
MEND
; --------------------------------------------------------------------------
;void vp9_idct32x32_1024_add_neon(int16_t *input, uint8_t *dest, int dest_stride);
;void vpx_idct32x32_1024_add_neon(int16_t *input, uint8_t *dest, int dest_stride);
;
; r0 int16_t *input,
; r1 uint8_t *dest,
@ -303,7 +303,7 @@ cospi_31_64 EQU 804
; r9 dest + 15 * dest_stride, descending (14, 13, 12, ...)
; r10 dest + 16 * dest_stride, ascending (17, 18, 19, ...)
|vp9_idct32x32_1024_add_neon| PROC
|vpx_idct32x32_1024_add_neon| PROC
; This function does one pass of idct32x32 transform.
;
; This is done by transposing the input and then doing a 1d transform on
@ -1295,5 +1295,5 @@ idct32_bands_end_2nd_pass
vpop {d8-d15}
pop {r4-r11}
bx lr
ENDP ; |vp9_idct32x32_1024_add_neon|
ENDP ; |vpx_idct32x32_1024_add_neon|
END

2
vpx_dsp/arm/idct32x32_add_neon.c
Просмотреть файл

@ -454,7 +454,7 @@ static INLINE void idct32_bands_end_2nd_pass(
return;
}
void vp9_idct32x32_1024_add_neon(
void vpx_idct32x32_1024_add_neon(
int16_t *input,
uint8_t *dest,
int stride) {

8
vpx_dsp/arm/idct4x4_1_add_neon.asm
Просмотреть файл

@ -8,21 +8,21 @@
;
EXPORT |vp9_idct4x4_1_add_neon|
EXPORT |vpx_idct4x4_1_add_neon|
ARM
REQUIRE8
PRESERVE8
AREA ||.text||, CODE, READONLY, ALIGN=2
;void vp9_idct4x4_1_add_neon(int16_t *input, uint8_t *dest,
;void vpx_idct4x4_1_add_neon(int16_t *input, uint8_t *dest,
; int dest_stride)
;
; r0 int16_t input
; r1 uint8_t *dest
; r2 int dest_stride)
|vp9_idct4x4_1_add_neon| PROC
|vpx_idct4x4_1_add_neon| PROC
ldrsh r0, [r0]
; generate cospi_16_64 = 11585
@ -63,6 +63,6 @@
vst1.32 {d7[1]}, [r12]
bx lr
ENDP ; |vp9_idct4x4_1_add_neon|
ENDP ; |vpx_idct4x4_1_add_neon|
END

2
vpx_dsp/arm/idct4x4_1_add_neon.c
Просмотреть файл

@ -13,7 +13,7 @@
#include "vpx_dsp/inv_txfm.h"
#include "vpx_ports/mem.h"
void vp9_idct4x4_1_add_neon(
void vpx_idct4x4_1_add_neon(
int16_t *input,
uint8_t *dest,
int dest_stride) {

8
vpx_dsp/arm/idct4x4_add_neon.asm
Просмотреть файл

@ -8,7 +8,7 @@
; be found in the AUTHORS file in the root of the source tree.
;
EXPORT |vp9_idct4x4_16_add_neon|
EXPORT |vpx_idct4x4_16_add_neon|
ARM
REQUIRE8
PRESERVE8
@ -16,13 +16,13 @@
AREA ||.text||, CODE, READONLY, ALIGN=2
AREA Block, CODE, READONLY ; name this block of code
;void vp9_idct4x4_16_add_neon(int16_t *input, uint8_t *dest, int dest_stride)
;void vpx_idct4x4_16_add_neon(int16_t *input, uint8_t *dest, int dest_stride)
;
; r0 int16_t input
; r1 uint8_t *dest
; r2 int dest_stride)
|vp9_idct4x4_16_add_neon| PROC
|vpx_idct4x4_16_add_neon| PROC
; The 2D transform is done with two passes which are actually pretty
; similar. We first transform the rows. This is done by transposing
@ -185,6 +185,6 @@
vst1.32 {d26[1]}, [r1], r2
vst1.32 {d26[0]}, [r1] ; no post-increment
bx lr
ENDP ; |vp9_idct4x4_16_add_neon|
ENDP ; |vpx_idct4x4_16_add_neon|
END

2
vpx_dsp/arm/idct4x4_add_neon.c
Просмотреть файл

@ -10,7 +10,7 @@
#include <arm_neon.h>
void vp9_idct4x4_16_add_neon(
void vpx_idct4x4_16_add_neon(
int16_t *input,
uint8_t *dest,
int dest_stride) {

8
vpx_dsp/arm/idct8x8_1_add_neon.asm
Просмотреть файл

@ -8,21 +8,21 @@
;
EXPORT |vp9_idct8x8_1_add_neon|
EXPORT |vpx_idct8x8_1_add_neon|
ARM
REQUIRE8
PRESERVE8
AREA ||.text||, CODE, READONLY, ALIGN=2
;void vp9_idct8x8_1_add_neon(int16_t *input, uint8_t *dest,
;void vpx_idct8x8_1_add_neon(int16_t *input, uint8_t *dest,
; int dest_stride)
;
; r0 int16_t input
; r1 uint8_t *dest
; r2 int dest_stride)
|vp9_idct8x8_1_add_neon| PROC
|vpx_idct8x8_1_add_neon| PROC
ldrsh r0, [r0]
; generate cospi_16_64 = 11585
@ -83,6 +83,6 @@
vst1.64 {d31}, [r12], r2
bx lr
ENDP ; |vp9_idct8x8_1_add_neon|
ENDP ; |vpx_idct8x8_1_add_neon|
END

2
vpx_dsp/arm/idct8x8_1_add_neon.c
Просмотреть файл

@ -13,7 +13,7 @@
#include "vpx_dsp/inv_txfm.h"
#include "vpx_ports/mem.h"
void vp9_idct8x8_1_add_neon(
void vpx_idct8x8_1_add_neon(
int16_t *input,
uint8_t *dest,
int dest_stride) {

16
vpx_dsp/arm/idct8x8_add_neon.asm
Просмотреть файл

@ -8,8 +8,8 @@
; be found in the AUTHORS file in the root of the source tree.
;
EXPORT |vp9_idct8x8_64_add_neon|
EXPORT |vp9_idct8x8_12_add_neon|
EXPORT |vpx_idct8x8_64_add_neon|
EXPORT |vpx_idct8x8_12_add_neon|
ARM
REQUIRE8
PRESERVE8
@ -198,13 +198,13 @@
MEND
AREA Block, CODE, READONLY ; name this block of code
;void vp9_idct8x8_64_add_neon(int16_t *input, uint8_t *dest, int dest_stride)
;void vpx_idct8x8_64_add_neon(int16_t *input, uint8_t *dest, int dest_stride)
;
; r0 int16_t input
; r1 uint8_t *dest
; r2 int dest_stride)
|vp9_idct8x8_64_add_neon| PROC
|vpx_idct8x8_64_add_neon| PROC
push {r4-r9}
vpush {d8-d15}
vld1.s16 {q8,q9}, [r0]!
@ -308,15 +308,15 @@
vpop {d8-d15}
pop {r4-r9}
bx lr
ENDP ; |vp9_idct8x8_64_add_neon|
ENDP ; |vpx_idct8x8_64_add_neon|
;void vp9_idct8x8_12_add_neon(int16_t *input, uint8_t *dest, int dest_stride)
;void vpx_idct8x8_12_add_neon(int16_t *input, uint8_t *dest, int dest_stride)
;
; r0 int16_t input
; r1 uint8_t *dest
; r2 int dest_stride)
|vp9_idct8x8_12_add_neon| PROC
|vpx_idct8x8_12_add_neon| PROC
push {r4-r9}
vpush {d8-d15}
vld1.s16 {q8,q9}, [r0]!
@ -514,6 +514,6 @@
vpop {d8-d15}
pop {r4-r9}
bx lr
ENDP ; |vp9_idct8x8_12_add_neon|
ENDP ; |vpx_idct8x8_12_add_neon|
END

4
vpx_dsp/arm/idct8x8_add_neon.c
Просмотреть файл

@ -238,7 +238,7 @@ static INLINE void IDCT8x8_1D(
return;
}
void vp9_idct8x8_64_add_neon(
void vpx_idct8x8_64_add_neon(
int16_t *input,
uint8_t *dest,
int dest_stride) {
@ -351,7 +351,7 @@ void vp9_idct8x8_64_add_neon(
return;
}
void vp9_idct8x8_12_add_neon(
void vpx_idct8x8_12_add_neon(
int16_t *input,
uint8_t *dest,
int dest_stride) {

8
vp9/common/arm/neon/vp9_save_reg_neon.asm → vpx_dsp/arm/save_reg_neon.asm
Просмотреть файл

@ -9,8 +9,8 @@
;
EXPORT |vp9_push_neon|
EXPORT |vp9_pop_neon|
EXPORT |vpx_push_neon|
EXPORT |vpx_pop_neon|
ARM
REQUIRE8
@ -18,14 +18,14 @@
AREA ||.text||, CODE, READONLY, ALIGN=2
|vp9_push_neon| PROC
|vpx_push_neon| PROC
vst1.i64 {d8, d9, d10, d11}, [r0]!
vst1.i64 {d12, d13, d14, d15}, [r0]!
bx lr
ENDP
|vp9_pop_neon| PROC
|vpx_pop_neon| PROC
vld1.i64 {d8, d9, d10, d11}, [r0]!
vld1.i64 {d12, d13, d14, d15}, [r0]!
bx lr

2
vpx_dsp/fwd_txfm.c
Просмотреть файл

@ -754,7 +754,7 @@ void vpx_fdct32x32_rd_c(const int16_t *input, tran_low_t *out, int stride) {
for (j = 0; j < 32; ++j)
// TODO(cd): see quality impact of only doing
// output[j * 32 + i] = (temp_out[j] + 1) >> 2;
// PS: also change code in vp9/encoder/x86/vp9_dct_sse2.c
// PS: also change code in vpx_dsp/x86/vpx_dct_sse2.c
output[j * 32 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2;
}

89
vpx_dsp/inv_txfm.c
Просмотреть файл

@ -13,7 +13,7 @@
#include "vpx_dsp/inv_txfm.h"
void vp9_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
void vpx_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
/* 4-point reversible, orthonormal inverse Walsh-Hadamard in 3.5 adds,
0.5 shifts per pixel. */
int i;
@ -65,7 +65,7 @@ void vp9_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
}
}
void vp9_iwht4x4_1_add_c(const tran_low_t *in, uint8_t *dest, int dest_stride) {
void vpx_iwht4x4_1_add_c(const tran_low_t *in, uint8_t *dest, int dest_stride) {
int i;
tran_high_t a1, e1;
tran_low_t tmp[4];
@ -111,7 +111,7 @@ void idct4_c(const tran_low_t *input, tran_low_t *output) {
output[3] = WRAPLOW(step[0] - step[3], 8);
}
void vp9_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
void vpx_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
tran_low_t out[4 * 4];
tran_low_t *outptr = out;
int i, j;
@ -136,7 +136,7 @@ void vp9_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
}
}
void vp9_idct4x4_1_add_c(const tran_low_t *input, uint8_t *dest,
void vpx_idct4x4_1_add_c(const tran_low_t *input, uint8_t *dest,
int dest_stride) {
int i;
tran_high_t a1;
@ -198,7 +198,7 @@ void idct8_c(const tran_low_t *input, tran_low_t *output) {
output[7] = WRAPLOW(step1[0] - step1[7], 8);
}
void vp9_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
void vpx_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
tran_low_t out[8 * 8];
tran_low_t *outptr = out;
int i, j;
@ -223,7 +223,7 @@ void vp9_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
}
}
void vp9_idct8x8_1_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
void vpx_idct8x8_1_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
int i, j;
tran_high_t a1;
tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64), 8);
@ -350,7 +350,7 @@ void iadst8_c(const tran_low_t *input, tran_low_t *output) {
output[7] = WRAPLOW(-x1, 8);
}
void vp9_idct8x8_12_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
void vpx_idct8x8_12_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
tran_low_t out[8 * 8] = { 0 };
tran_low_t *outptr = out;
int i, j;
@ -541,7 +541,7 @@ void idct16_c(const tran_low_t *input, tran_low_t *output) {
output[15] = WRAPLOW(step2[0] - step2[15], 8);
}
void vp9_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest,
void vpx_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest,
int stride) {
tran_low_t out[16 * 16];
tran_low_t *outptr = out;
@ -739,7 +739,7 @@ void iadst16_c(const tran_low_t *input, tran_low_t *output) {
output[15] = WRAPLOW(-x1, 8);
}
void vp9_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest,
void vpx_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest,
int stride) {
tran_low_t out[16 * 16] = { 0 };
tran_low_t *outptr = out;
@ -766,7 +766,7 @@ void vp9_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest,
}
}
void vp9_idct16x16_1_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
void vpx_idct16x16_1_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
int i, j;
tran_high_t a1;
tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64), 8);
@ -1146,7 +1146,7 @@ void idct32_c(const tran_low_t *input, tran_low_t *output) {
output[31] = WRAPLOW(step1[0] - step1[31], 8);
}
void vp9_idct32x32_1024_add_c(const tran_low_t *input, uint8_t *dest,
void vpx_idct32x32_1024_add_c(const tran_low_t *input, uint8_t *dest,
int stride) {
tran_low_t out[32 * 32];
tran_low_t *outptr = out;
@ -1185,7 +1185,7 @@ void vp9_idct32x32_1024_add_c(const tran_low_t *input, uint8_t *dest,
}
}
void vp9_idct32x32_34_add_c(const tran_low_t *input, uint8_t *dest,
void vpx_idct32x32_34_add_c(const tran_low_t *input, uint8_t *dest,
int stride) {
tran_low_t out[32 * 32] = {0};
tran_low_t *outptr = out;
@ -1212,7 +1212,7 @@ void vp9_idct32x32_34_add_c(const tran_low_t *input, uint8_t *dest,
}
}
void vp9_idct32x32_1_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
void vpx_idct32x32_1_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
int i, j;
tran_high_t a1;
@ -1228,7 +1228,7 @@ void vp9_idct32x32_1_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
}
#if CONFIG_VP9_HIGHBITDEPTH
void vp9_highbd_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
/* 4-point reversible, orthonormal inverse Walsh-Hadamard in 3.5 adds,
0.5 shifts per pixel. */
@ -1282,7 +1282,7 @@ void vp9_highbd_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_iwht4x4_1_add_c(const tran_low_t *in, uint8_t *dest8,
void vpx_highbd_iwht4x4_1_add_c(const tran_low_t *in, uint8_t *dest8,
int dest_stride, int bd) {
int i;
tran_high_t a1, e1;
@ -1315,7 +1315,7 @@ void vp9_highbd_iwht4x4_1_add_c(const tran_low_t *in, uint8_t *dest8,
}
}
void vp9_highbd_idct4_c(const tran_low_t *input, tran_low_t *output, int bd) {
void vpx_highbd_idct4_c(const tran_low_t *input, tran_low_t *output, int bd) {
tran_low_t step[4];
tran_high_t temp1, temp2;
(void) bd;
@ -1336,7 +1336,7 @@ void vp9_highbd_idct4_c(const tran_low_t *input, tran_low_t *output, int bd) {
output[3] = WRAPLOW(step[0] - step[3], bd);
}
void vp9_highbd_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
tran_low_t out[4 * 4];
tran_low_t *outptr = out;
@ -1346,7 +1346,7 @@ void vp9_highbd_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
// Rows
for (i = 0; i < 4; ++i) {
vp9_highbd_idct4_c(input, outptr, bd);
vpx_highbd_idct4_c(input, outptr, bd);
input += 4;
outptr += 4;
}
@ -1355,7 +1355,7 @@ void vp9_highbd_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
for (i = 0; i < 4; ++i) {
for (j = 0; j < 4; ++j)
temp_in[j] = out[j * 4 + i];
vp9_highbd_idct4_c(temp_in, temp_out, bd);
vpx_highbd_idct4_c(temp_in, temp_out, bd);
for (j = 0; j < 4; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 4), bd);
@ -1363,7 +1363,7 @@ void vp9_highbd_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_idct4x4_1_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct4x4_1_add_c(const tran_low_t *input, uint8_t *dest8,
int dest_stride, int bd) {
int i;
tran_high_t a1;
@ -1383,7 +1383,7 @@ void vp9_highbd_idct4x4_1_add_c(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_idct8_c(const tran_low_t *input, tran_low_t *output, int bd) {
void vpx_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
@ -1401,7 +1401,7 @@ void vp9_highbd_idct8_c(const tran_low_t *input, tran_low_t *output, int bd) {
step1[6] = WRAPLOW(highbd_dct_const_round_shift(temp2, bd), bd);
// stage 2 & stage 3 - even half
vp9_highbd_idct4_c(step1, step1, bd);
vpx_highbd_idct4_c(step1, step1, bd);
// stage 2 - odd half
step2[4] = WRAPLOW(step1[4] + step1[5], bd);
@ -1428,7 +1428,7 @@ void vp9_highbd_idct8_c(const tran_low_t *input, tran_low_t *output, int bd) {
output[7] = WRAPLOW(step1[0] - step1[7], bd);
}
void vp9_highbd_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
tran_low_t out[8 * 8];
tran_low_t *outptr = out;
@ -1438,7 +1438,7 @@ void vp9_highbd_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
// First transform rows.
for (i = 0; i < 8; ++i) {
vp9_highbd_idct8_c(input, outptr, bd);
vpx_highbd_idct8_c(input, outptr, bd);
input += 8;
outptr += 8;
}
@ -1447,7 +1447,7 @@ void vp9_highbd_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
for (i = 0; i < 8; ++i) {
for (j = 0; j < 8; ++j)
temp_in[j] = out[j * 8 + i];
vp9_highbd_idct8_c(temp_in, temp_out, bd);
vpx_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);
@ -1455,7 +1455,7 @@ void vp9_highbd_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_idct8x8_1_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct8x8_1_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
int i, j;
tran_high_t a1;
@ -1471,7 +1471,7 @@ void vp9_highbd_idct8x8_1_add_c(const tran_low_t *input, uint8_t *dest8,
}
}
void highbd_iadst4_c(const tran_low_t *input, tran_low_t *output, int bd) {
void vpx_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];
@ -1509,7 +1509,7 @@ void highbd_iadst4_c(const tran_low_t *input, tran_low_t *output, int bd) {
output[3] = WRAPLOW(highbd_dct_const_round_shift(s0 + s1 - s3, bd), bd);
}
void highbd_iadst8_c(const tran_low_t *input, tran_low_t *output, int bd) {
void vpx_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];
@ -1586,7 +1586,7 @@ void highbd_iadst8_c(const tran_low_t *input, tran_low_t *output, int bd) {
output[7] = WRAPLOW(-x1, bd);
}
void vp9_highbd_idct8x8_10_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct8x8_10_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
tran_low_t out[8 * 8] = { 0 };
tran_low_t *outptr = out;
@ -1597,7 +1597,7 @@ void vp9_highbd_idct8x8_10_add_c(const tran_low_t *input, uint8_t *dest8,
// First transform rows.
// Only first 4 row has non-zero coefs.
for (i = 0; i < 4; ++i) {
vp9_highbd_idct8_c(input, outptr, bd);
vpx_highbd_idct8_c(input, outptr, bd);
input += 8;
outptr += 8;
}
@ -1605,7 +1605,7 @@ void vp9_highbd_idct8x8_10_add_c(const tran_low_t *input, uint8_t *dest8,
for (i = 0; i < 8; ++i) {
for (j = 0; j < 8; ++j)
temp_in[j] = out[j * 8 + i];
vp9_highbd_idct8_c(temp_in, temp_out, bd);
vpx_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);
@ -1613,7 +1613,7 @@ void vp9_highbd_idct8x8_10_add_c(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_idct16_c(const tran_low_t *input, tran_low_t *output, int bd) {
void vpx_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;
@ -1779,7 +1779,7 @@ void vp9_highbd_idct16_c(const tran_low_t *input, tran_low_t *output, int bd) {
output[15] = WRAPLOW(step2[0] - step2[15], bd);
}
void vp9_highbd_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
tran_low_t out[16 * 16];
tran_low_t *outptr = out;
@ -1789,7 +1789,7 @@ void vp9_highbd_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
// First transform rows.
for (i = 0; i < 16; ++i) {
vp9_highbd_idct16_c(input, outptr, bd);
vpx_highbd_idct16_c(input, outptr, bd);
input += 16;
outptr += 16;
}
@ -1798,7 +1798,7 @@ void vp9_highbd_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
for (i = 0; i < 16; ++i) {
for (j = 0; j < 16; ++j)
temp_in[j] = out[j * 16 + i];
vp9_highbd_idct16_c(temp_in, temp_out, bd);
vpx_highbd_idct16_c(temp_in, temp_out, bd);
for (j = 0; j < 16; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 6), bd);
@ -1806,7 +1806,7 @@ void vp9_highbd_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
}
}
void highbd_iadst16_c(const tran_low_t *input, tran_low_t *output, int bd) {
void vpx_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;
@ -1976,7 +1976,7 @@ void highbd_iadst16_c(const tran_low_t *input, tran_low_t *output, int bd) {
output[15] = WRAPLOW(-x1, bd);
}
void vp9_highbd_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
tran_low_t out[16 * 16] = { 0 };
tran_low_t *outptr = out;
@ -1987,7 +1987,7 @@ void vp9_highbd_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest8,
// First transform rows. Since all non-zero dct coefficients are in
// upper-left 4x4 area, we only need to calculate first 4 rows here.
for (i = 0; i < 4; ++i) {
vp9_highbd_idct16_c(input, outptr, bd);
vpx_highbd_idct16_c(input, outptr, bd);
input += 16;
outptr += 16;
}
@ -1996,7 +1996,7 @@ void vp9_highbd_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest8,
for (i = 0; i < 16; ++i) {
for (j = 0; j < 16; ++j)
temp_in[j] = out[j*16 + i];
vp9_highbd_idct16_c(temp_in, temp_out, bd);
vpx_highbd_idct16_c(temp_in, temp_out, bd);
for (j = 0; j < 16; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 6), bd);
@ -2004,7 +2004,7 @@ void vp9_highbd_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_idct16x16_1_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct16x16_1_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
int i, j;
tran_high_t a1;
@ -2021,7 +2021,8 @@ void vp9_highbd_idct16x16_1_add_c(const tran_low_t *input, uint8_t *dest8,
}
}
void highbd_idct32_c(const tran_low_t *input, tran_low_t *output, int bd) {
static void 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;
@ -2389,7 +2390,7 @@ void highbd_idct32_c(const tran_low_t *input, tran_low_t *output, int bd) {
output[31] = WRAPLOW(step1[0] - step1[31], bd);
}
void vp9_highbd_idct32x32_1024_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct32x32_1024_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
tran_low_t out[32 * 32];
tran_low_t *outptr = out;
@ -2429,7 +2430,7 @@ void vp9_highbd_idct32x32_1024_add_c(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_idct32x32_34_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct32x32_34_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
tran_low_t out[32 * 32] = {0};
tran_low_t *outptr = out;
@ -2456,7 +2457,7 @@ void vp9_highbd_idct32x32_34_add_c(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_idct32x32_1_add_c(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct32x32_1_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
int i, j;
int a1;

13
vpx_dsp/inv_txfm.h
Просмотреть файл

@ -97,14 +97,13 @@ void iadst8_c(const tran_low_t *input, tran_low_t *output);
void iadst16_c(const tran_low_t *input, tran_low_t *output);
#if CONFIG_VP9_HIGHBITDEPTH
void vp9_highbd_idct4_c(const tran_low_t *input, tran_low_t *output, int bd);
void vp9_highbd_idct8_c(const tran_low_t *input, tran_low_t *output, int bd);
void vp9_highbd_idct16_c(const tran_low_t *input, tran_low_t *output, int bd);
void highbd_idct32_c(const tran_low_t *input, tran_low_t *output, int bd);
void vpx_highbd_idct4_c(const tran_low_t *input, tran_low_t *output, int bd);
void vpx_highbd_idct8_c(const tran_low_t *input, tran_low_t *output, int bd);
void vpx_highbd_idct16_c(const tran_low_t *input, tran_low_t *output, int bd);
void highbd_iadst4_c(const tran_low_t *input, tran_low_t *output, int bd);
void highbd_iadst8_c(const tran_low_t *input, tran_low_t *output, int bd);
void highbd_iadst16_c(const tran_low_t *input, tran_low_t *output, int bd);
void vpx_highbd_iadst4_c(const tran_low_t *input, tran_low_t *output, int bd);
void vpx_highbd_iadst8_c(const tran_low_t *input, tran_low_t *output, int bd);
void vpx_highbd_iadst16_c(const tran_low_t *input, tran_low_t *output, int bd);
static INLINE uint16_t highbd_clip_pixel_add(uint16_t dest, tran_high_t trans,
int bd) {

22
vpx_dsp/mips/idct16x16_msa.c
Просмотреть файл

@ -10,7 +10,7 @@
#include "vpx_dsp/mips/inv_txfm_msa.h"
void vp9_idct16_1d_rows_msa(const int16_t *input, int16_t *output) {
void vpx_idct16_1d_rows_msa(const int16_t *input, int16_t *output) {
v8i16 loc0, loc1, loc2, loc3;
v8i16 reg0, reg2, reg4, reg6, reg8, reg10, reg12, reg14;
v8i16 reg3, reg13, reg11, reg5, reg7, reg9, reg1, reg15;
@ -103,7 +103,7 @@ void vp9_idct16_1d_rows_msa(const int16_t *input, int16_t *output) {
ST_SH8(reg3, reg13, reg11, reg5, reg7, reg9, reg1, reg15, (output + 8), 16);
}
void vp9_idct16_1d_columns_addblk_msa(int16_t *input, uint8_t *dst,
void vpx_idct16_1d_columns_addblk_msa(int16_t *input, uint8_t *dst,
int32_t dst_stride) {
v8i16 loc0, loc1, loc2, loc3;
v8i16 reg0, reg2, reg4, reg6, reg8, reg10, reg12, reg14;
@ -201,7 +201,7 @@ void vp9_idct16_1d_columns_addblk_msa(int16_t *input, uint8_t *dst,
VP9_ADDBLK_ST8x4_UB(dst, dst_stride, reg7, reg9, reg1, reg15);
}
void vp9_idct16x16_256_add_msa(const int16_t *input, uint8_t *dst,
void vpx_idct16x16_256_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
int32_t i;
DECLARE_ALIGNED(32, int16_t, out_arr[16 * 16]);
@ -210,25 +210,25 @@ void vp9_idct16x16_256_add_msa(const int16_t *input, uint8_t *dst,
/* transform rows */
for (i = 0; i < 2; ++i) {
/* process 16 * 8 block */
vp9_idct16_1d_rows_msa((input + (i << 7)), (out + (i << 7)));
vpx_idct16_1d_rows_msa((input + (i << 7)), (out + (i << 7)));
}
/* transform columns */
for (i = 0; i < 2; ++i) {
/* process 8 * 16 block */
vp9_idct16_1d_columns_addblk_msa((out + (i << 3)), (dst + (i << 3)),
vpx_idct16_1d_columns_addblk_msa((out + (i << 3)), (dst + (i << 3)),
dst_stride);
}
}
void vp9_idct16x16_10_add_msa(const int16_t *input, uint8_t *dst,
void vpx_idct16x16_10_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
uint8_t i;
DECLARE_ALIGNED(32, int16_t, out_arr[16 * 16]);
int16_t *out = out_arr;
/* process 16 * 8 block */
vp9_idct16_1d_rows_msa(input, out);
vpx_idct16_1d_rows_msa(input, out);
/* short case just considers top 4 rows as valid output */
out += 4 * 16;
@ -255,12 +255,12 @@ void vp9_idct16x16_10_add_msa(const int16_t *input, uint8_t *dst,
/* transform columns */
for (i = 0; i < 2; ++i) {
/* process 8 * 16 block */
vp9_idct16_1d_columns_addblk_msa((out + (i << 3)), (dst + (i << 3)),
vpx_idct16_1d_columns_addblk_msa((out + (i << 3)), (dst + (i << 3)),
dst_stride);
}
}
void vp9_idct16x16_1_add_msa(const int16_t *input, uint8_t *dst,
void vpx_idct16x16_1_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
uint8_t i;
int16_t out;
@ -290,7 +290,7 @@ void vp9_idct16x16_1_add_msa(const int16_t *input, uint8_t *dst,
}
}
void vp9_iadst16_1d_rows_msa(const int16_t *input, int16_t *output) {
void vpx_iadst16_1d_rows_msa(const int16_t *input, int16_t *output) {
v8i16 r0, r1, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, r13, r14, r15;
v8i16 l0, l1, l2, l3, l4, l5, l6, l7, l8, l9, l10, l11, l12, l13, l14, l15;
@ -321,7 +321,7 @@ void vp9_iadst16_1d_rows_msa(const int16_t *input, int16_t *output) {
ST_SH8(l8, l9, l10, l11, l12, l13, l14, l15, (output + 8), 16);
}
void vp9_iadst16_1d_columns_addblk_msa(int16_t *input, uint8_t *dst,
void vpx_iadst16_1d_columns_addblk_msa(int16_t *input, uint8_t *dst,
int32_t dst_stride) {
v8i16 v0, v2, v4, v6, k0, k1, k2, k3;
v8i16 r0, r1, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, r13, r14, r15;

46
vpx_dsp/mips/idct32x32_msa.c
Просмотреть файл

@ -10,7 +10,7 @@
#include "vpx_dsp/mips/inv_txfm_msa.h"
static void vp9_idct32x8_row_transpose_store(const int16_t *input,
static void idct32x8_row_transpose_store(const int16_t *input,
int16_t *tmp_buf) {
v8i16 m0, m1, m2, m3, m4, m5, m6, m7, n0, n1, n2, n3, n4, n5, n6, n7;
@ -38,7 +38,7 @@ static void vp9_idct32x8_row_transpose_store(const int16_t *input,
ST_SH4(m6, n6, m7, n7, (tmp_buf + 28 * 8), 8);
}
static void vp9_idct32x8_row_even_process_store(int16_t *tmp_buf,
static void idct32x8_row_even_process_store(int16_t *tmp_buf,
int16_t *tmp_eve_buf) {
v8i16 vec0, vec1, vec2, vec3, loc0, loc1, loc2, loc3;
v8i16 reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
@ -122,7 +122,7 @@ static void vp9_idct32x8_row_even_process_store(int16_t *tmp_buf,
ST_SH(loc3, (tmp_eve_buf + 7 * 8));
}
static void vp9_idct32x8_row_odd_process_store(int16_t *tmp_buf,
static void idct32x8_row_odd_process_store(int16_t *tmp_buf,
int16_t *tmp_odd_buf) {
v8i16 vec0, vec1, vec2, vec3, loc0, loc1, loc2, loc3;
v8i16 reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
@ -240,7 +240,7 @@ static void vp9_idct32x8_row_odd_process_store(int16_t *tmp_buf,
ST_SH4(loc0, loc1, loc2, loc3, (tmp_odd_buf + 12 * 8), 8);
}
static void vp9_idct_butterfly_transpose_store(int16_t *tmp_buf,
static void idct_butterfly_transpose_store(int16_t *tmp_buf,
int16_t *tmp_eve_buf,
int16_t *tmp_odd_buf,
int16_t *dst) {
@ -341,19 +341,19 @@ static void vp9_idct_butterfly_transpose_store(int16_t *tmp_buf,
ST_SH4(m6, n6, m7, n7, (dst + 24 + 4 * 32), 32);
}
static void vp9_idct32x8_1d_rows_msa(const int16_t *input, int16_t *output) {
static void idct32x8_1d_rows_msa(const int16_t *input, int16_t *output) {
DECLARE_ALIGNED(32, int16_t, tmp_buf[8 * 32]);
DECLARE_ALIGNED(32, int16_t, tmp_odd_buf[16 * 8]);
DECLARE_ALIGNED(32, int16_t, tmp_eve_buf[16 * 8]);
vp9_idct32x8_row_transpose_store(input, &tmp_buf[0]);
vp9_idct32x8_row_even_process_store(&tmp_buf[0], &tmp_eve_buf[0]);
vp9_idct32x8_row_odd_process_store(&tmp_buf[0], &tmp_odd_buf[0]);
vp9_idct_butterfly_transpose_store(&tmp_buf[0], &tmp_eve_buf[0],
idct32x8_row_transpose_store(input, &tmp_buf[0]);
idct32x8_row_even_process_store(&tmp_buf[0], &tmp_eve_buf[0]);
idct32x8_row_odd_process_store(&tmp_buf[0], &tmp_odd_buf[0]);
idct_butterfly_transpose_store(&tmp_buf[0], &tmp_eve_buf[0],
&tmp_odd_buf[0], output);
}
static void vp9_idct8x32_column_even_process_store(int16_t *tmp_buf,
static void idct8x32_column_even_process_store(int16_t *tmp_buf,
int16_t *tmp_eve_buf) {
v8i16 vec0, vec1, vec2, vec3, loc0, loc1, loc2, loc3;
v8i16 reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
@ -433,7 +433,7 @@ static void vp9_idct8x32_column_even_process_store(int16_t *tmp_buf,
ST_SH2(loc2, loc0, (tmp_eve_buf + 8 * 8), 8);
}
static void vp9_idct8x32_column_odd_process_store(int16_t *tmp_buf,
static void idct8x32_column_odd_process_store(int16_t *tmp_buf,
int16_t *tmp_odd_buf) {
v8i16 vec0, vec1, vec2, vec3, loc0, loc1, loc2, loc3;
v8i16 reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
@ -540,7 +540,7 @@ static void vp9_idct8x32_column_odd_process_store(int16_t *tmp_buf,
ST_SH4(loc0, loc1, loc2, loc3, (tmp_odd_buf + 12 * 8), 8);
}
static void vp9_idct8x32_column_butterfly_addblk(int16_t *tmp_eve_buf,
static void idct8x32_column_butterfly_addblk(int16_t *tmp_eve_buf,
int16_t *tmp_odd_buf,
uint8_t *dst,
int32_t dst_stride) {
@ -627,18 +627,18 @@ static void vp9_idct8x32_column_butterfly_addblk(int16_t *tmp_eve_buf,
n1, n3, n5, n7);
}
static void vp9_idct8x32_1d_columns_addblk_msa(int16_t *input, uint8_t *dst,
static void idct8x32_1d_columns_addblk_msa(int16_t *input, uint8_t *dst,
int32_t dst_stride) {
DECLARE_ALIGNED(32, int16_t, tmp_odd_buf[16 * 8]);
DECLARE_ALIGNED(32, int16_t, tmp_eve_buf[16 * 8]);
vp9_idct8x32_column_even_process_store(input, &tmp_eve_buf[0]);
vp9_idct8x32_column_odd_process_store(input, &tmp_odd_buf[0]);
vp9_idct8x32_column_butterfly_addblk(&tmp_eve_buf[0], &tmp_odd_buf[0],
idct8x32_column_even_process_store(input, &tmp_eve_buf[0]);
idct8x32_column_odd_process_store(input, &tmp_odd_buf[0]);
idct8x32_column_butterfly_addblk(&tmp_eve_buf[0], &tmp_odd_buf[0],
dst, dst_stride);
}
void vp9_idct32x32_1024_add_msa(const int16_t *input, uint8_t *dst,
void vpx_idct32x32_1024_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
int32_t i;
DECLARE_ALIGNED(32, int16_t, out_arr[32 * 32]);
@ -647,18 +647,18 @@ void vp9_idct32x32_1024_add_msa(const int16_t *input, uint8_t *dst,
/* transform rows */
for (i = 0; i < 4; ++i) {
/* process 32 * 8 block */
vp9_idct32x8_1d_rows_msa((input + (i << 8)), (out_ptr + (i << 8)));
idct32x8_1d_rows_msa((input + (i << 8)), (out_ptr + (i << 8)));
}
/* transform columns */
for (i = 0; i < 4; ++i) {
/* process 8 * 32 block */
vp9_idct8x32_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)),
idct8x32_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)),
dst_stride);
}
}
void vp9_idct32x32_34_add_msa(const int16_t *input, uint8_t *dst,
void vpx_idct32x32_34_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
int32_t i;
DECLARE_ALIGNED(32, int16_t, out_arr[32 * 32]);
@ -693,17 +693,17 @@ void vp9_idct32x32_34_add_msa(const int16_t *input, uint8_t *dst,
out_ptr = out_arr;
/* rows: only upper-left 8x8 has non-zero coeff */
vp9_idct32x8_1d_rows_msa(input, out_ptr);
idct32x8_1d_rows_msa(input, out_ptr);
/* transform columns */
for (i = 0; i < 4; ++i) {
/* process 8 * 32 block */
vp9_idct8x32_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)),
idct8x32_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)),
dst_stride);
}
}
void vp9_idct32x32_1_add_msa(const int16_t *input, uint8_t *dst,
void vpx_idct32x32_1_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
int32_t i;
int16_t out;

8
vpx_dsp/mips/idct4x4_msa.c
Просмотреть файл

@ -10,7 +10,7 @@
#include "vpx_dsp/mips/inv_txfm_msa.h"
void vp9_iwht4x4_16_add_msa(const int16_t *input, uint8_t *dst,
void vpx_iwht4x4_16_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
v8i16 in0, in1, in2, in3;
v4i32 in0_r, in1_r, in2_r, in3_r, in4_r;
@ -47,7 +47,7 @@ void vp9_iwht4x4_16_add_msa(const int16_t *input, uint8_t *dst,
ADDBLK_ST4x4_UB(in0, in3, in1, in2, dst, dst_stride);
}
void vp9_iwht4x4_1_add_msa(const int16_t *input, uint8_t *dst,
void vpx_iwht4x4_1_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
int16_t a1, e1;
v8i16 in1, in0 = { 0 };
@ -67,7 +67,7 @@ void vp9_iwht4x4_1_add_msa(const int16_t *input, uint8_t *dst,
ADDBLK_ST4x4_UB(in0, in1, in1, in1, dst, dst_stride);
}
void vp9_idct4x4_16_add_msa(const int16_t *input, uint8_t *dst,
void vpx_idct4x4_16_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
v8i16 in0, in1, in2, in3;
@ -84,7 +84,7 @@ void vp9_idct4x4_16_add_msa(const int16_t *input, uint8_t *dst,
ADDBLK_ST4x4_UB(in0, in1, in2, in3, dst, dst_stride);
}
void vp9_idct4x4_1_add_msa(const int16_t *input, uint8_t *dst,
void vpx_idct4x4_1_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
int16_t out;
v8i16 vec;

6
vpx_dsp/mips/idct8x8_msa.c
Просмотреть файл

@ -10,7 +10,7 @@
#include "vpx_dsp/mips/inv_txfm_msa.h"
void vp9_idct8x8_64_add_msa(const int16_t *input, uint8_t *dst,
void vpx_idct8x8_64_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
@ -38,7 +38,7 @@ void vp9_idct8x8_64_add_msa(const int16_t *input, uint8_t *dst,
VP9_ADDBLK_ST8x4_UB(dst, dst_stride, in4, in5, in6, in7);
}
void vp9_idct8x8_12_add_msa(const int16_t *input, uint8_t *dst,
void vpx_idct8x8_12_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v8i16 s0, s1, s2, s3, s4, s5, s6, s7, k0, k1, k2, k3, m0, m1, m2, m3;
@ -99,7 +99,7 @@ void vp9_idct8x8_12_add_msa(const int16_t *input, uint8_t *dst,
VP9_ADDBLK_ST8x4_UB(dst, dst_stride, in4, in5, in6, in7);
}
void vp9_idct8x8_1_add_msa(const int16_t *input, uint8_t *dst,
void vpx_idct8x8_1_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
int16_t out;
int32_t val;

6
vpx_dsp/mips/inv_txfm_dspr2.h
Просмотреть файл

@ -49,10 +49,10 @@ extern "C" {
); \
out; })
void vp9_idct32_cols_add_blk_dspr2(int16_t *input, uint8_t *dest,
void vpx_idct32_cols_add_blk_dspr2(int16_t *input, uint8_t *dest,
int dest_stride);
void vp9_idct4_rows_dspr2(const int16_t *input, int16_t *output);
void vp9_idct4_columns_add_blk_dspr2(int16_t *input, uint8_t *dest,
void vpx_idct4_rows_dspr2(const int16_t *input, int16_t *output);
void vpx_idct4_columns_add_blk_dspr2(int16_t *input, uint8_t *dest,
int dest_stride);
void iadst4_dspr2(const int16_t *input, int16_t *output);
void idct8_rows_dspr2(const int16_t *input, int16_t *output, uint32_t no_rows);

8
vpx_dsp/mips/inv_txfm_msa.h
Просмотреть файл

@ -401,10 +401,10 @@
MADD_SHORT(out14, out15, k1_m, k2_m, out14, out15); \
}
void vp9_idct16_1d_columns_addblk_msa(int16_t *input, uint8_t *dst,
void vpx_idct16_1d_columns_addblk_msa(int16_t *input, uint8_t *dst,
int32_t dst_stride);
void vp9_idct16_1d_rows_msa(const int16_t *input, int16_t *output);
void vp9_iadst16_1d_columns_addblk_msa(int16_t *input, uint8_t *dst,
void vpx_idct16_1d_rows_msa(const int16_t *input, int16_t *output);
void vpx_iadst16_1d_columns_addblk_msa(int16_t *input, uint8_t *dst,
int32_t dst_stride);
void vp9_iadst16_1d_rows_msa(const int16_t *input, int16_t *output);
void vpx_iadst16_1d_rows_msa(const int16_t *input, int16_t *output);
#endif // VPX_DSP_MIPS_INV_TXFM_MSA_H_

6
vpx_dsp/mips/itrans16_dspr2.c
Просмотреть файл

@ -887,7 +887,7 @@ void idct16_cols_add_blk_dspr2(int16_t *input, uint8_t *dest,
}
}
void vp9_idct16x16_256_add_dspr2(const int16_t *input, uint8_t *dest,
void vpx_idct16x16_256_add_dspr2(const int16_t *input, uint8_t *dest,
int dest_stride) {
DECLARE_ALIGNED(32, int16_t, out[16 * 16]);
uint32_t pos = 45;
@ -906,7 +906,7 @@ void vp9_idct16x16_256_add_dspr2(const int16_t *input, uint8_t *dest,
idct16_cols_add_blk_dspr2(out, dest, dest_stride);
}
void vp9_idct16x16_10_add_dspr2(const int16_t *input, uint8_t *dest,
void vpx_idct16x16_10_add_dspr2(const int16_t *input, uint8_t *dest,
int dest_stride) {
DECLARE_ALIGNED(32, int16_t, out[16 * 16]);
int16_t *outptr = out;
@ -955,7 +955,7 @@ void vp9_idct16x16_10_add_dspr2(const int16_t *input, uint8_t *dest,
idct16_cols_add_blk_dspr2(out, dest, dest_stride);
}
void vp9_idct16x16_1_add_dspr2(const int16_t *input, uint8_t *dest,
void vpx_idct16x16_1_add_dspr2(const int16_t *input, uint8_t *dest,
int dest_stride) {
uint32_t pos = 45;
int32_t out;

2
vpx_dsp/mips/itrans32_cols_dspr2.c
Просмотреть файл

@ -13,7 +13,7 @@
#include "vpx_dsp/txfm_common.h"
#if HAVE_DSPR2
void vp9_idct32_cols_add_blk_dspr2(int16_t *input, uint8_t *dest,
void vpx_idct32_cols_add_blk_dspr2(int16_t *input, uint8_t *dest,
int dest_stride) {
int16_t step1_0, step1_1, step1_2, step1_3, step1_4, step1_5, step1_6;
int16_t step1_7, step1_8, step1_9, step1_10, step1_11, step1_12, step1_13;

10
vpx_dsp/mips/itrans32_dspr2.c
Просмотреть файл

@ -865,7 +865,7 @@ static void idct32_rows_dspr2(const int16_t *input, int16_t *output,
}
}
void vp9_idct32x32_1024_add_dspr2(const int16_t *input, uint8_t *dest,
void vpx_idct32x32_1024_add_dspr2(const int16_t *input, uint8_t *dest,
int dest_stride) {
DECLARE_ALIGNED(32, int16_t, out[32 * 32]);
int16_t *outptr = out;
@ -882,10 +882,10 @@ void vp9_idct32x32_1024_add_dspr2(const int16_t *input, uint8_t *dest,
idct32_rows_dspr2(input, outptr, 32);
// Columns
vp9_idct32_cols_add_blk_dspr2(out, dest, dest_stride);
vpx_idct32_cols_add_blk_dspr2(out, dest, dest_stride);
}
void vp9_idct32x32_34_add_dspr2(const int16_t *input, uint8_t *dest,
void vpx_idct32x32_34_add_dspr2(const int16_t *input, uint8_t *dest,
int stride) {
DECLARE_ALIGNED(32, int16_t, out[32 * 32]);
int16_t *outptr = out;
@ -944,10 +944,10 @@ void vp9_idct32x32_34_add_dspr2(const int16_t *input, uint8_t *dest,
}
// Columns
vp9_idct32_cols_add_blk_dspr2(out, dest, stride);
vpx_idct32_cols_add_blk_dspr2(out, dest, stride);
}
void vp9_idct32x32_1_add_dspr2(const int16_t *input, uint8_t *dest,
void vpx_idct32x32_1_add_dspr2(const int16_t *input, uint8_t *dest,
int stride) {
int r, out;
int32_t a1, absa1;

12
vpx_dsp/mips/itrans4_dspr2.c
Просмотреть файл

@ -14,7 +14,7 @@
#include "vpx_dsp/txfm_common.h"
#if HAVE_DSPR2
void vp9_idct4_rows_dspr2(const int16_t *input, int16_t *output) {
void vpx_idct4_rows_dspr2(const int16_t *input, int16_t *output) {
int16_t step_0, step_1, step_2, step_3;
int Temp0, Temp1, Temp2, Temp3;
const int const_2_power_13 = 8192;
@ -99,7 +99,7 @@ void vp9_idct4_rows_dspr2(const int16_t *input, int16_t *output) {
}
}
void vp9_idct4_columns_add_blk_dspr2(int16_t *input, uint8_t *dest,
void vpx_idct4_columns_add_blk_dspr2(int16_t *input, uint8_t *dest,
int dest_stride) {
int16_t step_0, step_1, step_2, step_3;
int Temp0, Temp1, Temp2, Temp3;
@ -221,7 +221,7 @@ void vp9_idct4_columns_add_blk_dspr2(int16_t *input, uint8_t *dest,
}
}
void vp9_idct4x4_16_add_dspr2(const int16_t *input, uint8_t *dest,
void vpx_idct4x4_16_add_dspr2(const int16_t *input, uint8_t *dest,
int dest_stride) {
DECLARE_ALIGNED(32, int16_t, out[4 * 4]);
int16_t *outptr = out;
@ -235,13 +235,13 @@ void vp9_idct4x4_16_add_dspr2(const int16_t *input, uint8_t *dest,
);
// Rows
vp9_idct4_rows_dspr2(input, outptr);
vpx_idct4_rows_dspr2(input, outptr);
// Columns
vp9_idct4_columns_add_blk_dspr2(&out[0], dest, dest_stride);
vpx_idct4_columns_add_blk_dspr2(&out[0], dest, dest_stride);
}
void vp9_idct4x4_1_add_dspr2(const int16_t *input, uint8_t *dest,
void vpx_idct4x4_1_add_dspr2(const int16_t *input, uint8_t *dest,
int dest_stride) {
int a1, absa1;
int r;

6
vpx_dsp/mips/itrans8_dspr2.c
Просмотреть файл

@ -442,7 +442,7 @@ void idct8_columns_add_blk_dspr2(int16_t *input, uint8_t *dest,
}
}
void vp9_idct8x8_64_add_dspr2(const int16_t *input, uint8_t *dest,
void vpx_idct8x8_64_add_dspr2(const int16_t *input, uint8_t *dest,
int dest_stride) {
DECLARE_ALIGNED(32, int16_t, out[8 * 8]);
int16_t *outptr = out;
@ -462,7 +462,7 @@ void vp9_idct8x8_64_add_dspr2(const int16_t *input, uint8_t *dest,
idct8_columns_add_blk_dspr2(&out[0], dest, dest_stride);
}
void vp9_idct8x8_12_add_dspr2(const int16_t *input, uint8_t *dest,
void vpx_idct8x8_12_add_dspr2(const int16_t *input, uint8_t *dest,
int dest_stride) {
DECLARE_ALIGNED(32, int16_t, out[8 * 8]);
int16_t *outptr = out;
@ -507,7 +507,7 @@ void vp9_idct8x8_12_add_dspr2(const int16_t *input, uint8_t *dest,
idct8_columns_add_blk_dspr2(&out[0], dest, dest_stride);
}
void vp9_idct8x8_1_add_dspr2(const int16_t *input, uint8_t *dest,
void vpx_idct8x8_1_add_dspr2(const int16_t *input, uint8_t *dest,
int dest_stride) {
uint32_t pos = 45;
int32_t out;

1
vpx_dsp/vpx_dsp.mk
Просмотреть файл

@ -186,6 +186,7 @@ endif # ARCH_X86_64
endif # CONFIG_USE_X86INC
ifeq ($(HAVE_NEON_ASM),yes)
DSP_SRCS-yes += arm/save_reg_neon$(ASM)
DSP_SRCS-yes += arm/idct4x4_1_add_neon$(ASM)
DSP_SRCS-yes += arm/idct4x4_add_neon$(ASM)
DSP_SRCS-yes += arm/idct8x8_1_add_neon$(ASM)

230
vpx_dsp/vpx_dsp_rtcd_defs.pl
Просмотреть файл

@ -598,183 +598,183 @@ if (vpx_config("CONFIG_VP9") eq "yes") {
if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
# Note as optimized versions of these functions are added we need to add a check to ensure
# that when CONFIG_EMULATE_HARDWARE is on, it defaults to the C versions only.
add_proto qw/void vp9_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct4x4_1_add/;
add_proto qw/void vpx_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct4x4_1_add/;
add_proto qw/void vp9_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct4x4_16_add/;
add_proto qw/void vpx_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct4x4_16_add/;
add_proto qw/void vp9_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct8x8_1_add/;
add_proto qw/void vpx_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct8x8_1_add/;
add_proto qw/void vp9_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct8x8_64_add/;
add_proto qw/void vpx_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct8x8_64_add/;
add_proto qw/void vp9_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct8x8_12_add/;
add_proto qw/void vpx_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct8x8_12_add/;
add_proto qw/void vp9_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct16x16_1_add/;
add_proto qw/void vpx_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct16x16_1_add/;
add_proto qw/void vp9_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct16x16_256_add/;
add_proto qw/void vpx_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct16x16_256_add/;
add_proto qw/void vp9_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct16x16_10_add/;
add_proto qw/void vpx_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct16x16_10_add/;
add_proto qw/void vp9_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct32x32_1024_add/;
add_proto qw/void vpx_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct32x32_1024_add/;
add_proto qw/void vp9_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct32x32_34_add/;
add_proto qw/void vpx_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct32x32_34_add/;
add_proto qw/void vp9_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct32x32_1_add/;
add_proto qw/void vpx_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct32x32_1_add/;
add_proto qw/void vp9_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_iwht4x4_1_add/;
add_proto qw/void vpx_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_iwht4x4_1_add/;
add_proto qw/void vp9_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_iwht4x4_16_add/;
add_proto qw/void vpx_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_iwht4x4_16_add/;
add_proto qw/void vp9_highbd_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct4x4_1_add/;
add_proto qw/void vpx_highbd_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct4x4_1_add/;
add_proto qw/void vp9_highbd_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct8x8_1_add/;
add_proto qw/void vpx_highbd_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct8x8_1_add/;
add_proto qw/void vp9_highbd_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct16x16_1_add/;
add_proto qw/void vpx_highbd_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct16x16_1_add/;
add_proto qw/void vp9_highbd_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct32x32_1024_add/;
add_proto qw/void vpx_highbd_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct32x32_1024_add/;
add_proto qw/void vp9_highbd_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct32x32_34_add/;
add_proto qw/void vpx_highbd_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct32x32_34_add/;
add_proto qw/void vp9_highbd_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct32x32_1_add/;
add_proto qw/void vpx_highbd_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct32x32_1_add/;
add_proto qw/void vp9_highbd_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_iwht4x4_1_add/;
add_proto qw/void vpx_highbd_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_iwht4x4_1_add/;
add_proto qw/void vp9_highbd_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_iwht4x4_16_add/;
add_proto qw/void vpx_highbd_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_iwht4x4_16_add/;
# Force C versions if CONFIG_EMULATE_HARDWARE is 1
if (vpx_config("CONFIG_EMULATE_HARDWARE") eq "yes") {
add_proto qw/void vp9_highbd_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct4x4_16_add/;
add_proto qw/void vpx_highbd_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct4x4_16_add/;
add_proto qw/void vp9_highbd_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct8x8_64_add/;
add_proto qw/void vpx_highbd_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct8x8_64_add/;
add_proto qw/void vp9_highbd_idct8x8_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct8x8_10_add/;
add_proto qw/void vpx_highbd_idct8x8_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct8x8_10_add/;
add_proto qw/void vp9_highbd_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct16x16_256_add/;
add_proto qw/void vpx_highbd_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct16x16_256_add/;
add_proto qw/void vp9_highbd_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct16x16_10_add/;
add_proto qw/void vpx_highbd_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct16x16_10_add/;
} else {
add_proto qw/void vp9_highbd_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct4x4_16_add sse2/;
add_proto qw/void vpx_highbd_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct4x4_16_add sse2/;
add_proto qw/void vp9_highbd_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct8x8_64_add sse2/;
add_proto qw/void vpx_highbd_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct8x8_64_add sse2/;
add_proto qw/void vp9_highbd_idct8x8_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct8x8_10_add sse2/;
add_proto qw/void vpx_highbd_idct8x8_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct8x8_10_add sse2/;
add_proto qw/void vp9_highbd_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct16x16_256_add sse2/;
add_proto qw/void vpx_highbd_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct16x16_256_add sse2/;
add_proto qw/void vp9_highbd_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vp9_highbd_idct16x16_10_add sse2/;
add_proto qw/void vpx_highbd_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct16x16_10_add sse2/;
} # CONFIG_EMULATE_HARDWARE
} else {
# Force C versions if CONFIG_EMULATE_HARDWARE is 1
if (vpx_config("CONFIG_EMULATE_HARDWARE") eq "yes") {
add_proto qw/void vp9_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct4x4_1_add/;
add_proto qw/void vpx_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct4x4_1_add/;
add_proto qw/void vp9_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct4x4_16_add/;
add_proto qw/void vpx_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct4x4_16_add/;
add_proto qw/void vp9_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct8x8_1_add/;
add_proto qw/void vpx_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct8x8_1_add/;
add_proto qw/void vp9_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct8x8_64_add/;
add_proto qw/void vpx_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct8x8_64_add/;
add_proto qw/void vp9_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct8x8_12_add/;
add_proto qw/void vpx_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct8x8_12_add/;
add_proto qw/void vp9_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct16x16_1_add/;
add_proto qw/void vpx_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct16x16_1_add/;
add_proto qw/void vp9_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct16x16_256_add/;
add_proto qw/void vpx_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct16x16_256_add/;
add_proto qw/void vp9_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct16x16_10_add/;
add_proto qw/void vpx_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct16x16_10_add/;
add_proto qw/void vp9_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct32x32_1024_add/;
add_proto qw/void vpx_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct32x32_1024_add/;
add_proto qw/void vp9_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct32x32_34_add/;
add_proto qw/void vpx_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct32x32_34_add/;
add_proto qw/void vp9_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct32x32_1_add/;
add_proto qw/void vpx_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct32x32_1_add/;
add_proto qw/void vp9_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_iwht4x4_1_add/;
add_proto qw/void vpx_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_iwht4x4_1_add/;
add_proto qw/void vp9_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_iwht4x4_16_add/;
add_proto qw/void vpx_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_iwht4x4_16_add/;
} else {
add_proto qw/void vp9_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct4x4_1_add sse2 neon dspr2 msa/;
add_proto qw/void vpx_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct4x4_1_add sse2 neon dspr2 msa/;
add_proto qw/void vp9_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct4x4_16_add sse2 neon dspr2 msa/;
add_proto qw/void vpx_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct4x4_16_add sse2 neon dspr2 msa/;
add_proto qw/void vp9_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct8x8_1_add sse2 neon dspr2 msa/;
add_proto qw/void vpx_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct8x8_1_add sse2 neon dspr2 msa/;
add_proto qw/void vp9_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct8x8_64_add sse2 neon dspr2 msa/, "$ssse3_x86_64_x86inc";
add_proto qw/void vpx_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct8x8_64_add sse2 neon dspr2 msa/, "$ssse3_x86_64_x86inc";
add_proto qw/void vp9_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct8x8_12_add sse2 neon dspr2 msa/, "$ssse3_x86_64_x86inc";
add_proto qw/void vpx_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct8x8_12_add sse2 neon dspr2 msa/, "$ssse3_x86_64_x86inc";
add_proto qw/void vp9_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct16x16_1_add sse2 neon dspr2 msa/;
add_proto qw/void vpx_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct16x16_1_add sse2 neon dspr2 msa/;
add_proto qw/void vp9_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct16x16_256_add sse2 neon dspr2 msa/;
add_proto qw/void vpx_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct16x16_256_add sse2 neon dspr2 msa/;
add_proto qw/void vp9_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct16x16_10_add sse2 neon dspr2 msa/;
add_proto qw/void vpx_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct16x16_10_add sse2 neon dspr2 msa/;
add_proto qw/void vp9_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct32x32_1024_add sse2 neon dspr2 msa/;
add_proto qw/void vpx_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct32x32_1024_add sse2 neon dspr2 msa/;
add_proto qw/void vp9_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct32x32_34_add sse2 neon_asm dspr2 msa/;
add_proto qw/void vpx_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct32x32_34_add sse2 neon_asm dspr2 msa/;
# Need to add 34 eob idct32x32 neon implementation.
$vp9_idct32x32_34_add_neon_asm=vp9_idct32x32_1024_add_neon;
$vpx_idct32x32_34_add_neon_asm=vpx_idct32x32_1024_add_neon;
add_proto qw/void vp9_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_idct32x32_1_add sse2 neon dspr2 msa/;
add_proto qw/void vpx_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct32x32_1_add sse2 neon dspr2 msa/;
add_proto qw/void vp9_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_iwht4x4_1_add msa/;
add_proto qw/void vpx_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_iwht4x4_1_add msa/;
add_proto qw/void vp9_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vp9_iwht4x4_16_add msa/, "$sse2_x86inc";
add_proto qw/void vpx_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_iwht4x4_16_add msa/, "$sse2_x86inc";
} # CONFIG_EMULATE_HARDWARE
} # CONFIG_VP9_HIGHBITDEPTH
} # CONFIG_VP9

3
vpx_dsp/x86/inv_txfm_sse2.asm
Просмотреть файл

@ -7,6 +7,9 @@
; in the file PATENTS. All contributing project authors may
; be found in the AUTHORS file in the root of the source tree.
;
%define program_name vpx
%include "third_party/x86inc/x86inc.asm"
SECTION .text

56
vpx_dsp/x86/inv_txfm_sse2.c
Просмотреть файл

@ -21,7 +21,7 @@
*(int *)(dest) = _mm_cvtsi128_si32(d0); \
}
void vp9_idct4x4_16_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
void vpx_idct4x4_16_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
const __m128i zero = _mm_setzero_si128();
const __m128i eight = _mm_set1_epi16(8);
const __m128i cst = _mm_setr_epi16(
@ -151,7 +151,7 @@ void vp9_idct4x4_16_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
}
}
void vp9_idct4x4_1_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
void vpx_idct4x4_1_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
__m128i dc_value;
const __m128i zero = _mm_setzero_si128();
int a;
@ -449,7 +449,7 @@ void iadst4_sse2(__m128i *in) {
out7 = _mm_subs_epi16(stp1_0, stp2_7); \
}
void vp9_idct8x8_64_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
void vpx_idct8x8_64_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
const __m128i zero = _mm_setzero_si128();
const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
const __m128i final_rounding = _mm_set1_epi16(1 << 4);
@ -480,7 +480,7 @@ void vp9_idct8x8_64_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
// 2-D
for (i = 0; i < 2; i++) {
// 8x8 Transpose is copied from vp9_fdct8x8_sse2()
// 8x8 Transpose is copied from vpx_fdct8x8_sse2()
TRANSPOSE_8X8(in0, in1, in2, in3, in4, in5, in6, in7,
in0, in1, in2, in3, in4, in5, in6, in7);
@ -518,7 +518,7 @@ void vp9_idct8x8_64_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
RECON_AND_STORE(dest + 7 * stride, in7);
}
void vp9_idct8x8_1_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
void vpx_idct8x8_1_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
__m128i dc_value;
const __m128i zero = _mm_setzero_si128();
int a;
@ -555,7 +555,7 @@ void idct8_sse2(__m128i *in) {
__m128i stp2_0, stp2_1, stp2_2, stp2_3, stp2_4, stp2_5, stp2_6, stp2_7;
__m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
// 8x8 Transpose is copied from vp9_fdct8x8_sse2()
// 8x8 Transpose is copied from vpx_fdct8x8_sse2()
TRANSPOSE_8X8(in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7],
in0, in1, in2, in3, in4, in5, in6, in7);
@ -792,7 +792,7 @@ void iadst8_sse2(__m128i *in) {
in[7] = _mm_sub_epi16(k__const_0, s1);
}
void vp9_idct8x8_12_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
void vpx_idct8x8_12_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
const __m128i zero = _mm_setzero_si128();
const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
const __m128i final_rounding = _mm_set1_epi16(1 << 4);
@ -1169,7 +1169,7 @@ void vp9_idct8x8_12_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
stp2_10, stp2_13, stp2_11, stp2_12) \
}
void vp9_idct16x16_256_add_sse2(const int16_t *input, uint8_t *dest,
void vpx_idct16x16_256_add_sse2(const int16_t *input, uint8_t *dest,
int stride) {
const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
const __m128i final_rounding = _mm_set1_epi16(1 << 5);
@ -1294,7 +1294,7 @@ void vp9_idct16x16_256_add_sse2(const int16_t *input, uint8_t *dest,
}
}
void vp9_idct16x16_1_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
void vpx_idct16x16_1_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
__m128i dc_value;
const __m128i zero = _mm_setzero_si128();
int a, i;
@ -2152,7 +2152,7 @@ void iadst16_sse2(__m128i *in0, __m128i *in1) {
iadst16_8col(in1);
}
void vp9_idct16x16_10_add_sse2(const int16_t *input, uint8_t *dest,
void vpx_idct16x16_10_add_sse2(const int16_t *input, uint8_t *dest,
int stride) {
const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
const __m128i final_rounding = _mm_set1_epi16(1 << 5);
@ -3029,7 +3029,7 @@ void vp9_idct16x16_10_add_sse2(const int16_t *input, uint8_t *dest,
}
// Only upper-left 8x8 has non-zero coeff
void vp9_idct32x32_34_add_sse2(const int16_t *input, uint8_t *dest,
void vpx_idct32x32_34_add_sse2(const int16_t *input, uint8_t *dest,
int stride) {
const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
const __m128i final_rounding = _mm_set1_epi16(1<<5);
@ -3188,7 +3188,7 @@ void vp9_idct32x32_34_add_sse2(const int16_t *input, uint8_t *dest,
}
}
void vp9_idct32x32_1024_add_sse2(const int16_t *input, uint8_t *dest,
void vpx_idct32x32_1024_add_sse2(const int16_t *input, uint8_t *dest,
int stride) {
const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
const __m128i final_rounding = _mm_set1_epi16(1 << 5);
@ -3464,7 +3464,7 @@ void vp9_idct32x32_1024_add_sse2(const int16_t *input, uint8_t *dest,
}
}
void vp9_idct32x32_1_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
void vpx_idct32x32_1_add_sse2(const int16_t *input, uint8_t *dest, int stride) {
__m128i dc_value;
const __m128i zero = _mm_setzero_si128();
int a, i;
@ -3498,7 +3498,7 @@ static INLINE __m128i clamp_high_sse2(__m128i value, int bd) {
return retval;
}
void vp9_highbd_idct4x4_16_add_sse2(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct4x4_16_add_sse2(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
tran_low_t out[4 * 4];
tran_low_t *outptr = out;
@ -3561,7 +3561,7 @@ void vp9_highbd_idct4x4_16_add_sse2(const tran_low_t *input, uint8_t *dest8,
} else {
// Run the un-optimised row transform
for (i = 0; i < 4; ++i) {
vp9_highbd_idct4_c(input, outptr, bd);
vpx_highbd_idct4_c(input, outptr, bd);
input += 4;
outptr += 4;
}
@ -3605,7 +3605,7 @@ void vp9_highbd_idct4x4_16_add_sse2(const tran_low_t *input, uint8_t *dest8,
for (i = 0; i < 4; ++i) {
for (j = 0; j < 4; ++j)
temp_in[j] = out[j * 4 + i];
vp9_highbd_idct4_c(temp_in, temp_out, bd);
vpx_highbd_idct4_c(temp_in, temp_out, bd);
for (j = 0; j < 4; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 4), bd);
@ -3614,7 +3614,7 @@ void vp9_highbd_idct4x4_16_add_sse2(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_idct8x8_64_add_sse2(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct8x8_64_add_sse2(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
tran_low_t out[8 * 8];
tran_low_t *outptr = out;
@ -3679,7 +3679,7 @@ void vp9_highbd_idct8x8_64_add_sse2(const tran_low_t *input, uint8_t *dest8,
} else {
// Run the un-optimised row transform
for (i = 0; i < 8; ++i) {
vp9_highbd_idct8_c(input, outptr, bd);
vpx_highbd_idct8_c(input, outptr, bd);
input += 8;
outptr += 8;
}
@ -3706,7 +3706,7 @@ void vp9_highbd_idct8x8_64_add_sse2(const tran_low_t *input, uint8_t *dest8,
for (i = 0; i < 8; ++i) {
for (j = 0; j < 8; ++j)
temp_in[j] = out[j * 8 + i];
vp9_highbd_idct8_c(temp_in, temp_out, bd);
vpx_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);
@ -3715,7 +3715,7 @@ void vp9_highbd_idct8x8_64_add_sse2(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_idct8x8_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
void vpx_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;
@ -3783,7 +3783,7 @@ void vp9_highbd_idct8x8_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
} else {
// Run the un-optimised row transform
for (i = 0; i < 4; ++i) {
vp9_highbd_idct8_c(input, outptr, bd);
vpx_highbd_idct8_c(input, outptr, bd);
input += 8;
outptr += 8;
}
@ -3810,7 +3810,7 @@ void vp9_highbd_idct8x8_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
for (i = 0; i < 8; ++i) {
for (j = 0; j < 8; ++j)
temp_in[j] = out[j * 8 + i];
vp9_highbd_idct8_c(temp_in, temp_out, bd);
vpx_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);
@ -3819,7 +3819,7 @@ void vp9_highbd_idct8x8_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_idct16x16_256_add_sse2(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct16x16_256_add_sse2(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
tran_low_t out[16 * 16];
tran_low_t *outptr = out;
@ -3892,7 +3892,7 @@ void vp9_highbd_idct16x16_256_add_sse2(const tran_low_t *input, uint8_t *dest8,
} else {
// Run the un-optimised row transform
for (i = 0; i < 16; ++i) {
vp9_highbd_idct16_c(input, outptr, bd);
vpx_highbd_idct16_c(input, outptr, bd);
input += 16;
outptr += 16;
}
@ -3924,7 +3924,7 @@ void vp9_highbd_idct16x16_256_add_sse2(const tran_low_t *input, uint8_t *dest8,
for (i = 0; i < 16; ++i) {
for (j = 0; j < 16; ++j)
temp_in[j] = out[j * 16 + i];
vp9_highbd_idct16_c(temp_in, temp_out, bd);
vpx_highbd_idct16_c(temp_in, temp_out, bd);
for (j = 0; j < 16; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 6), bd);
@ -3933,7 +3933,7 @@ void vp9_highbd_idct16x16_256_add_sse2(const tran_low_t *input, uint8_t *dest8,
}
}
void vp9_highbd_idct16x16_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
void vpx_highbd_idct16x16_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
int stride, int bd) {
tran_low_t out[16 * 16] = { 0 };
tran_low_t *outptr = out;
@ -4011,7 +4011,7 @@ void vp9_highbd_idct16x16_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
} else {
// Run the un-optimised row transform
for (i = 0; i < 4; ++i) {
vp9_highbd_idct16_c(input, outptr, bd);
vpx_highbd_idct16_c(input, outptr, bd);
input += 16;
outptr += 16;
}
@ -4043,7 +4043,7 @@ void vp9_highbd_idct16x16_10_add_sse2(const tran_low_t *input, uint8_t *dest8,
for (i = 0; i < 16; ++i) {
for (j = 0; j < 16; ++j)
temp_in[j] = out[j * 16 + i];
vp9_highbd_idct16_c(temp_in, temp_out, bd);
vpx_highbd_idct16_c(temp_in, temp_out, bd);
for (j = 0; j < 16; ++j) {
dest[j * stride + i] = highbd_clip_pixel_add(
dest[j * stride + i], ROUND_POWER_OF_TWO(temp_out[j], 6), bd);

3
vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm
Просмотреть файл

@ -7,6 +7,9 @@
; in the file PATENTS. All contributing project authors may
; be found in the AUTHORS file in the root of the source tree.
;
%define program_name vpx
%include "third_party/x86inc/x86inc.asm"
; This file provides SSSE3 version of the inverse transformation. Part