Commit 3d938d71 authored by Parag Salasakar's avatar Parag Salasakar

mips msa vp8 idct optimization

average improvement ~2x-5x

Change-Id: I19e82f78772993bcd67fcf975fe180232172f86d
parent 8565a1c9
......@@ -113,4 +113,8 @@ INSTANTIATE_TEST_CASE_P(C, IDCTTest, ::testing::Values(vp8_short_idct4x4llm_c));
INSTANTIATE_TEST_CASE_P(MMX, IDCTTest,
::testing::Values(vp8_short_idct4x4llm_mmx));
#endif
#if HAVE_MSA
INSTANTIATE_TEST_CASE_P(MSA, IDCTTest,
::testing::Values(vp8_short_idct4x4llm_msa));
#endif
}
This diff is collapsed.
This diff is collapsed.
......@@ -29,21 +29,21 @@ $vp8_clear_system_state_mmx=vpx_reset_mmx_state;
# Dequant
#
add_proto qw/void vp8_dequantize_b/, "struct blockd*, short *dqc";
specialize qw/vp8_dequantize_b mmx media neon/;
specialize qw/vp8_dequantize_b mmx media neon msa/;
$vp8_dequantize_b_media=vp8_dequantize_b_v6;
add_proto qw/void vp8_dequant_idct_add/, "short *input, short *dq, unsigned char *output, int stride";
specialize qw/vp8_dequant_idct_add mmx media neon dspr2/;
specialize qw/vp8_dequant_idct_add mmx media neon dspr2 msa/;
$vp8_dequant_idct_add_media=vp8_dequant_idct_add_v6;
$vp8_dequant_idct_add_dspr2=vp8_dequant_idct_add_dspr2;
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 mmx sse2 media neon dspr2/;
specialize qw/vp8_dequant_idct_add_y_block mmx sse2 media neon dspr2 msa/;
$vp8_dequant_idct_add_y_block_media=vp8_dequant_idct_add_y_block_v6;
$vp8_dequant_idct_add_y_block_dspr2=vp8_dequant_idct_add_y_block_dspr2;
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 mmx sse2 media neon dspr2/;
specialize qw/vp8_dequant_idct_add_uv_block mmx sse2 media neon dspr2 msa/;
$vp8_dequant_idct_add_uv_block_media=vp8_dequant_idct_add_uv_block_v6;
$vp8_dequant_idct_add_y_block_dspr2=vp8_dequant_idct_add_y_block_dspr2;
......@@ -108,7 +108,7 @@ $vp8_loop_filter_simple_bh_neon=vp8_loop_filter_bhs_neon;
#
#idct16
add_proto qw/void vp8_short_idct4x4llm/, "short *input, unsigned char *pred, int pitch, unsigned char *dst, int dst_stride";
specialize qw/vp8_short_idct4x4llm mmx media neon dspr2/;
specialize qw/vp8_short_idct4x4llm mmx media neon dspr2 msa/;
$vp8_short_idct4x4llm_media=vp8_short_idct4x4llm_v6_dual;
$vp8_short_idct4x4llm_dspr2=vp8_short_idct4x4llm_dspr2;
......@@ -120,13 +120,13 @@ $vp8_short_inv_walsh4x4_1_dspr2=vp8_short_inv_walsh4x4_1_dspr2;
#iwalsh16
add_proto qw/void vp8_short_inv_walsh4x4/, "short *input, short *output";
specialize qw/vp8_short_inv_walsh4x4 mmx sse2 media neon dspr2/;
specialize qw/vp8_short_inv_walsh4x4 mmx sse2 media neon dspr2 msa/;
$vp8_short_inv_walsh4x4_media=vp8_short_inv_walsh4x4_v6;
$vp8_short_inv_walsh4x4_dspr2=vp8_short_inv_walsh4x4_dspr2;
#idct1_scalar_add
add_proto qw/void vp8_dc_only_idct_add/, "short input, unsigned char *pred, int pred_stride, unsigned char *dst, int dst_stride";
specialize qw/vp8_dc_only_idct_add mmx media neon dspr2/;
specialize qw/vp8_dc_only_idct_add mmx media neon dspr2 msa/;
$vp8_dc_only_idct_add_media=vp8_dc_only_idct_add_v6;
$vp8_dc_only_idct_add_dspr2=vp8_dc_only_idct_add_dspr2;
......
......@@ -121,6 +121,10 @@ VP8_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/reconinter_dspr2.c
VP8_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/idct_blk_dspr2.c
VP8_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/dequantize_dspr2.c
# common (c)
VP8_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/idct_msa.c
VP8_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/vp8_macros_msa.h
# common (c)
VP8_COMMON_SRCS-$(ARCH_ARM) += common/arm/filter_arm.c
VP8_COMMON_SRCS-$(ARCH_ARM) += common/arm/loopfilter_arm.c
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment