[x265] [PATCH] asm: filter_vsp, filter_vss for 12x32 in avx2
Divya Manivannan
divya at multicorewareinc.com
Tue Apr 28 12:30:09 CEST 2015
# HG changeset patch
# User Divya Manivannan <divya at multicorewareinc.com>
# Date 1430216241 -19800
# Tue Apr 28 15:47:21 2015 +0530
# Node ID 8d34f596917997d680252647380bba1aada97d37
# Parent b143e26c2846d917c0e911708907b0e3e85a368c
asm: filter_vsp, filter_vss for 12x32 in avx2
filter_vsp[12x32]: 4587c->3164c
filter_vss[12x32]: 3632c->2919c
diff -r b143e26c2846 -r 8d34f5969179 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp Tue Apr 28 14:27:11 2015 +0530
+++ b/source/common/x86/asm-primitives.cpp Tue Apr 28 15:47:21 2015 +0530
@@ -2287,6 +2287,7 @@
p.chroma[X265_CSP_I422].pu[CHROMA_422_6x16].filter_vss = x265_interp_4tap_vert_ss_6x16_avx2;
p.chroma[X265_CSP_I422].pu[CHROMA_422_2x16].filter_vss = x265_interp_4tap_vert_ss_2x16_avx2;
p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].filter_vss = x265_interp_4tap_vert_ss_16x24_avx2;
+ p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].filter_vss = x265_interp_4tap_vert_ss_12x32_avx2;
//i444 for chroma_vss
p.chroma[X265_CSP_I444].pu[LUMA_4x4].filter_vss = x265_interp_4tap_vert_ss_4x4_avx2;
@@ -2475,6 +2476,7 @@
p.chroma[X265_CSP_I422].pu[CHROMA_422_6x16].filter_vsp = x265_interp_4tap_vert_sp_6x16_avx2;
p.chroma[X265_CSP_I422].pu[CHROMA_422_2x16].filter_vsp = x265_interp_4tap_vert_sp_2x16_avx2;
p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].filter_vsp = x265_interp_4tap_vert_sp_16x24_avx2;
+ p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].filter_vsp = x265_interp_4tap_vert_sp_12x32_avx2;
//i444 for chroma_vsp
p.chroma[X265_CSP_I444].pu[LUMA_4x4].filter_vsp = x265_interp_4tap_vert_sp_4x4_avx2;
diff -r b143e26c2846 -r 8d34f5969179 source/common/x86/ipfilter8.asm
--- a/source/common/x86/ipfilter8.asm Tue Apr 28 14:27:11 2015 +0530
+++ b/source/common/x86/ipfilter8.asm Tue Apr 28 15:47:21 2015 +0530
@@ -16138,6 +16138,53 @@
FILTER_VER_CHROMA_S_AVX2_12x16 sp
FILTER_VER_CHROMA_S_AVX2_12x16 ss
+%macro FILTER_VER_CHROMA_S_AVX2_12x32 1
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_12x32, 4, 9, 10
+ mov r4d, r4m
+ shl r4d, 6
+ add r1d, r1d
+
+%ifdef PIC
+ lea r5, [pw_ChromaCoeffV]
+ add r5, r4
+%else
+ lea r5, [pw_ChromaCoeffV + r4]
+%endif
+
+ lea r4, [r1 * 3]
+ sub r0, r1
+%ifidn %1, sp
+ mova m9, [pd_526336]
+%else
+ add r3d, r3d
+%endif
+ lea r6, [r3 * 3]
+%rep 2
+ PROCESS_CHROMA_S_AVX2_W8_16R %1
+%ifidn %1, sp
+ add r2, 8
+%else
+ add r2, 16
+%endif
+ add r0, 16
+ mova m7, m9
+ PROCESS_CHROMA_AVX2_W4_16R %1
+ sub r0, 16
+%ifidn %1, sp
+ lea r2, [r2 + r3 * 4 - 8]
+%else
+ lea r2, [r2 + r3 * 4 - 16]
+%endif
+%endrep
+ RET
+%endif
+%endmacro
+
+ FILTER_VER_CHROMA_S_AVX2_12x32 sp
+ FILTER_VER_CHROMA_S_AVX2_12x32 ss
+
%macro FILTER_VER_CHROMA_S_AVX2_16x12 1
INIT_YMM avx2
%if ARCH_X86_64 == 1
More information about the x265-devel
mailing list