ref: 2ae93a776b4b5ba864363a7de36f3e4b6aa99eab
parent: 272adbbec48eb6e16eaa8ee5addb34733419a397
parent: c6ad70d5f1a79231d177df6323f35b658e45edd0
author: Dmitry Kovalev <dkovalev@google.com>
date: Mon Oct 7 10:19:50 EDT 2013
Merge "Giving consistent names to IDCT 8x8 functions."
--- a/test/fdct8x8_test.cc
+++ b/test/fdct8x8_test.cc
@@ -21,7 +21,7 @@
extern "C" {#include "vp9/common/vp9_entropy.h"
#include "./vp9_rtcd.h"
-void vp9_short_idct8x8_add_c(int16_t *input, uint8_t *output, int pitch);
+void vp9_idct8x8_64_add_c(int16_t *input, uint8_t *output, int pitch);
}
#include "vpx/vpx_integer.h"
@@ -296,7 +296,7 @@
INSTANTIATE_TEST_CASE_P(
C, FwdTrans8x8DCT,
::testing::Values(
- make_tuple(&vp9_short_fdct8x8_c, &vp9_short_idct8x8_add_c, 0)));
+ make_tuple(&vp9_short_fdct8x8_c, &vp9_idct8x8_64_add_c, 0)));
INSTANTIATE_TEST_CASE_P(
C, FwdTrans8x8HT,
::testing::Values(
@@ -309,7 +309,7 @@
INSTANTIATE_TEST_CASE_P(
SSE2, FwdTrans8x8DCT,
::testing::Values(
- make_tuple(&vp9_short_fdct8x8_sse2, &vp9_short_idct8x8_add_sse2, 0)));
+ make_tuple(&vp9_short_fdct8x8_sse2, &vp9_idct8x8_64_add_sse2, 0)));
INSTANTIATE_TEST_CASE_P(
SSE2, FwdTrans8x8HT,
::testing::Values(
--- a/test/idct8x8_test.cc
+++ b/test/idct8x8_test.cc
@@ -126,7 +126,7 @@
reference_dct_2d(input, output_r);
for (int j = 0; j < 64; ++j)
coeff[j] = round(output_r[j]);
- vp9_short_idct8x8_add_c(coeff, dst, 8);
+ vp9_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;
--- a/vp9/common/arm/neon/vp9_short_idct8x8_1_add_neon.asm
+++ b/vp9/common/arm/neon/vp9_short_idct8x8_1_add_neon.asm
@@ -8,7 +8,7 @@
;
- EXPORT |vp9_short_idct8x8_1_add_neon|
+ EXPORT |vp9_idct8x8_1_add_neon|
ARM
REQUIRE8
PRESERVE8
@@ -15,7 +15,7 @@
AREA ||.text||, CODE, READONLY, ALIGN=2
-;void vp9_short_idct8x8_1_add_neon(int16_t *input, uint8_t *dest,
+;void vp9_idct8x8_1_add_neon(int16_t *input, uint8_t *dest,
; int dest_stride)
;
; r0 int16_t input
@@ -22,7 +22,7 @@
; r1 uint8_t *dest
; r2 int dest_stride)
-|vp9_short_idct8x8_1_add_neon| PROC
+|vp9_idct8x8_1_add_neon| PROC
ldrsh r0, [r0]
; generate cospi_16_64 = 11585
@@ -83,6 +83,6 @@
vst1.64 {d31}, [r12], r2bx lr
- ENDP ; |vp9_short_idct8x8_1_add_neon|
+ ENDP ; |vp9_idct8x8_1_add_neon|
END
--- a/vp9/common/arm/neon/vp9_short_idct8x8_add_neon.asm
+++ b/vp9/common/arm/neon/vp9_short_idct8x8_add_neon.asm
@@ -8,8 +8,8 @@
; be found in the AUTHORS file in the root of the source tree.
;
- EXPORT |vp9_short_idct8x8_add_neon|
- EXPORT |vp9_short_idct8x8_10_add_neon|
+ EXPORT |vp9_idct8x8_64_add_neon|
+ EXPORT |vp9_idct8x8_10_add_neon|
ARM
REQUIRE8
PRESERVE8
@@ -198,13 +198,13 @@
MEND
AREA Block, CODE, READONLY ; name this block of code
-;void vp9_short_idct8x8_add_neon(int16_t *input, uint8_t *dest, int dest_stride)
+;void vp9_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_short_idct8x8_add_neon| PROC
+|vp9_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_short_idct8x8_add_neon|
+ ENDP ; |vp9_idct8x8_64_add_neon|
-;void vp9_short_idct8x8_10_add_neon(int16_t *input, uint8_t *dest, int dest_stride)
+;void vp9_idct8x8_10_add_neon(int16_t *input, uint8_t *dest, int dest_stride)
;
; r0 int16_t input
; r1 uint8_t *dest
; r2 int dest_stride)
-|vp9_short_idct8x8_10_add_neon| PROC
+|vp9_idct8x8_10_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_short_idct8x8_10_add_neon|
+ ENDP ; |vp9_idct8x8_10_add_neon|
END
--- a/vp9/common/vp9_idct.c
+++ b/vp9/common/vp9_idct.c
@@ -201,7 +201,7 @@
output[7] = step1[0] - step1[7];
}
-void vp9_short_idct8x8_add_c(int16_t *input, uint8_t *dest, int dest_stride) {+void vp9_idct8x8_64_add_c(int16_t *input, uint8_t *dest, int dest_stride) {int16_t out[8 * 8];
int16_t *outptr = out;
int i, j;
@@ -225,7 +225,7 @@
}
}
-void vp9_short_idct8x8_1_add_c(int16_t *input, uint8_t *dest, int dest_stride) {+void vp9_idct8x8_1_add_c(int16_t *input, uint8_t *dest, int dest_stride) {int i, j;
int a1;
int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
@@ -420,7 +420,7 @@
+ dest[j * dest_stride + i]); }
}
-void vp9_short_idct8x8_10_add_c(int16_t *input, uint8_t *dest,
+void vp9_idct8x8_10_add_c(int16_t *input, uint8_t *dest,
int dest_stride) { int16_t out[8 * 8] = { 0 };int16_t *outptr = out;
@@ -1301,7 +1301,7 @@
vp9_iwht4x4_1_add(input, dest, stride);
}
-void vp9_idct_add_8x8(int16_t *input, uint8_t *dest, int stride, int eob) {+void vp9_idct8x8_add(int16_t *input, uint8_t *dest, int stride, int eob) {// If dc is 1, then input[0] is the reconstructed value, do not need
// dequantization. Also, when dc is 1, dc is counted in eobs, namely eobs >=1.
@@ -1312,11 +1312,11 @@
if (eob) {if (eob == 1)
// DC only DCT coefficient
- vp9_short_idct8x8_1_add(input, dest, stride);
+ vp9_idct8x8_1_add(input, dest, stride);
else if (eob <= 10)
- vp9_short_idct8x8_10_add(input, dest, stride);
+ vp9_idct8x8_10_add(input, dest, stride);
else
- vp9_short_idct8x8_add(input, dest, stride);
+ vp9_idct8x8_64_add(input, dest, stride);
}
}
@@ -1355,7 +1355,7 @@
void vp9_iht_add_8x8(TX_TYPE tx_type, int16_t *input, uint8_t *dest,
int stride, int eob) { if (tx_type == DCT_DCT) {- vp9_idct_add_8x8(input, dest, stride, eob);
+ vp9_idct8x8_add(input, dest, stride, eob);
} else { if (eob > 0) {vp9_short_iht8x8_add(input, dest, stride, tx_type);
--- a/vp9/common/vp9_idct.h
+++ b/vp9/common/vp9_idct.h
@@ -90,7 +90,7 @@
void vp9_idct4x4_add(int16_t *input, uint8_t *dest, int stride, int eob);
void vp9_iwht4x4_add(int16_t *input, uint8_t *dest, int stride, int eob);
-void vp9_idct_add_8x8(int16_t *input, uint8_t *dest, int stride, int eob);
+void vp9_idct8x8_add(int16_t *input, uint8_t *dest, int stride, int eob);
void vp9_idct_add_16x16(int16_t *input, uint8_t *dest, int stride, int eob);
void vp9_idct_add_32x32(int16_t *input, uint8_t *dest, int stride, int eob);
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -273,14 +273,14 @@
prototype void vp9_idct4x4_16_add "int16_t *input, uint8_t *dest, int dest_stride"
specialize vp9_idct4x4_16_add sse2 neon
-prototype void vp9_short_idct8x8_1_add "int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_short_idct8x8_1_add sse2 neon
+prototype void vp9_idct8x8_1_add "int16_t *input, uint8_t *dest, int dest_stride"
+specialize vp9_idct8x8_1_add sse2 neon
-prototype void vp9_short_idct8x8_add "int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_short_idct8x8_add sse2 neon
+prototype void vp9_idct8x8_64_add "int16_t *input, uint8_t *dest, int dest_stride"
+specialize vp9_idct8x8_64_add sse2 neon
-prototype void vp9_short_idct8x8_10_add "int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_short_idct8x8_10_add sse2 neon
+prototype void vp9_idct8x8_10_add "int16_t *input, uint8_t *dest, int dest_stride"
+specialize vp9_idct8x8_10_add sse2 neon
prototype void vp9_short_idct16x16_1_add "int16_t *input, uint8_t *dest, int dest_stride"
specialize vp9_short_idct16x16_1_add sse2 neon
--- a/vp9/common/x86/vp9_idct_intrin_sse2.c
+++ b/vp9/common/x86/vp9_idct_intrin_sse2.c
@@ -529,7 +529,7 @@
dest += stride; \
}
-void vp9_short_idct8x8_add_sse2(int16_t *input, uint8_t *dest, int stride) {+void vp9_idct8x8_64_add_sse2(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);
@@ -597,7 +597,7 @@
RECON_AND_STORE(dest, in7);
}
-void vp9_short_idct8x8_1_add_sse2(int16_t *input, uint8_t *dest, int stride) {+void vp9_idct8x8_1_add_sse2(int16_t *input, uint8_t *dest, int stride) {__m128i dc_value;
const __m128i zero = _mm_setzero_si128();
int a;
@@ -985,7 +985,7 @@
RECON_AND_STORE(dest, in[7]);
}
-void vp9_short_idct8x8_10_add_sse2(int16_t *input, uint8_t *dest, int stride) {+void vp9_idct8x8_10_add_sse2(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);
--- a/vp9/encoder/vp9_encodemb.c
+++ b/vp9/encoder/vp9_encodemb.c
@@ -460,7 +460,7 @@
vp9_idct_add_16x16(dqcoeff, dst, pd->dst.stride, pd->eobs[block]);
break;
case TX_8X8:
- vp9_idct_add_8x8(dqcoeff, dst, pd->dst.stride, pd->eobs[block]);
+ vp9_idct8x8_add(dqcoeff, dst, pd->dst.stride, pd->eobs[block]);
break;
case TX_4X4:
// this is like vp9_short_idct4x4 but has a special case around eob<=1