generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

vp8: [loongson] optimize idct with mmi

Browse Source 
1. vp8_dequant_idct_add_y_block_mmi
2. vp8_dequant_idct_add_uv_block_mmi

Change-Id: I9987147be2685ac79d4b045d1d56f6709ee1223c
This commit is contained in:
Shiyou Yin 2017-09-13 16:20:21 +08:00
parent a9248457b1
commit 577d4fa792
3 changed files with 74 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

71
vp8/common/mips/mmi/idct_blk_mmi.c Normal file
View File

@ -0,0 +1,71 @@
/*
* Copyright (c) 2017 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "./vp8_rtcd.h"
#include "vpx_mem/vpx_mem.h"
void vp8_dequant_idct_add_y_block_mmi(int16_t *q, int16_t *dq, uint8_t *dst,
int stride, int8_t *eobs) {
int i, j;
for (i = 0; i < 4; i++) {
for (j = 0; j < 4; j++) {
if (*eobs++ > 1) {
vp8_dequant_idct_add_mmi(q, dq, dst, stride);
} else {
vp8_dc_only_idct_add_mmi(q[0] * dq[0], dst, stride, dst, stride);
memset(q, 0, 2 * sizeof(q[0]));
}
q += 16;
dst += 4;
}
dst += 4 * stride - 16;
}
}
void vp8_dequant_idct_add_uv_block_mmi(int16_t *q, int16_t *dq, uint8_t *dstu,
uint8_t *dstv, int stride,
int8_t *eobs) {
int i, j;
for (i = 0; i < 2; i++) {
for (j = 0; j < 2; j++) {
if (*eobs++ > 1) {
vp8_dequant_idct_add_mmi(q, dq, dstu, stride);
} else {
vp8_dc_only_idct_add_mmi(q[0] * dq[0], dstu, stride, dstu, stride);
memset(q, 0, 2 * sizeof(q[0]));
}
q += 16;
dstu += 4;
}
dstu += 4 * stride - 8;
}
for (i = 0; i < 2; i++) {
for (j = 0; j < 2; j++) {
if (*eobs++ > 1) {
vp8_dequant_idct_add_mmi(q, dq, dstv, stride);
} else {
vp8_dc_only_idct_add_mmi(q[0] * dq[0], dstv, stride, dstv, stride);
memset(q, 0, 2 * sizeof(q[0]));
}
q += 16;
dstv += 4;
}
dstv += 4 * stride - 8;
}
}

4
vp8/common/rtcd_defs.pl
View File

@ -28,10 +28,10 @@ add_proto qw/void vp8_dequant_idct_add/, "short *input, short *dq, unsigned char
specialize qw/vp8_dequant_idct_add mmx neon dspr2 msa mmi/;
add_proto qw/void vp8_dequant_idct_add_y_block/, "short *q, short *dq, unsigned char *dst, int stride, char *eobs";
specialize qw/vp8_dequant_idct_add_y_block sse2 neon dspr2 msa/;
specialize qw/vp8_dequant_idct_add_y_block sse2 neon dspr2 msa mmi/;
add_proto qw/void vp8_dequant_idct_add_uv_block/, "short *q, short *dq, unsigned char *dst_u, unsigned char *dst_v, int stride, char *eobs";
specialize qw/vp8_dequant_idct_add_uv_block sse2 neon dspr2 msa/;
specialize qw/vp8_dequant_idct_add_uv_block sse2 neon dspr2 msa mmi/;
#
# Loopfilter

1
vp8/vp8_common.mk
View File

@ -122,6 +122,7 @@ VP8_COMMON_SRCS-$(HAVE_MMI) += common/mips/mmi/loopfilter_filters_mmi.c
VP8_COMMON_SRCS-$(HAVE_MMI) += common/mips/mmi/idctllm_mmi.c
VP8_COMMON_SRCS-$(HAVE_MMI) += common/mips/mmi/dequantize_mmi.c
VP8_COMMON_SRCS-$(HAVE_MMI) += common/mips/mmi/copymem_mmi.c
VP8_COMMON_SRCS-$(HAVE_MMI) += common/mips/mmi/idct_blk_mmi.c
ifeq ($(CONFIG_POSTPROC),yes)
VP8_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/mfqe_msa.c