[x265] [PATCH] asm: filter_vsp[16x32, 24x32, 32x32], filter_vss[16x32, 24x32, 32x32] in avx2

[Divya Manivannan]

# HG changeset patch
# User Divya Manivannan <divya at multicorewareinc.com>
# Date 1426161705 -19800
#      Thu Mar 12 17:31:45 2015 +0530
# Node ID 17d37a2b315c4ead360461e7ce3dfd86d4b21287
# Parent  45d498a1a59263fd758bf71c910e7d4289e771d3
asm: filter_vsp[16x32, 24x32, 32x32], filter_vss[16x32, 24x32, 32x32] in avx2

filter_vsp[16x32, 24x32, 32x32]: 6015c->3693c, 8710c->5692c, 11284c->7731c
filter_vss[16x32, 24x32, 32x32]: 4702c->4024c, 7013c->6132c, 9046c->7926c

diff -r 45d498a1a592 -r 17d37a2b315c source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Thu Mar 12 16:55:03 2015 +0530
+++ b/source/common/x86/asm-primitives.cpp	Thu Mar 12 17:31:45 2015 +0530
@@ -1625,11 +1625,17 @@
         p.chroma[X265_CSP_I420].pu[CHROMA_420_4x4].filter_vsp = x265_interp_4tap_vert_sp_4x4_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_8x8].filter_vsp = x265_interp_4tap_vert_sp_8x8_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_16x16].filter_vsp = x265_interp_4tap_vert_sp_16x16_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x32].filter_vsp = x265_interp_4tap_vert_sp_32x32_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x32].filter_vsp = x265_interp_4tap_vert_sp_16x32_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_24x32].filter_vsp = x265_interp_4tap_vert_sp_24x32_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x16].filter_vsp = x265_interp_4tap_vert_sp_32x16_avx2;
 
         p.chroma[X265_CSP_I420].pu[CHROMA_420_4x4].filter_vss = x265_interp_4tap_vert_ss_4x4_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_8x8].filter_vss = x265_interp_4tap_vert_ss_8x8_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_16x16].filter_vss = x265_interp_4tap_vert_ss_16x16_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x32].filter_vss = x265_interp_4tap_vert_ss_32x32_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x32].filter_vss = x265_interp_4tap_vert_ss_16x32_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_24x32].filter_vss = x265_interp_4tap_vert_ss_24x32_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x16].filter_vss = x265_interp_4tap_vert_ss_32x16_avx2;
     }
 #endif
diff -r 45d498a1a592 -r 17d37a2b315c source/common/x86/ipfilter8.asm
--- a/source/common/x86/ipfilter8.asm	Thu Mar 12 16:55:03 2015 +0530
+++ b/source/common/x86/ipfilter8.asm	Thu Mar 12 17:31:45 2015 +0530
@@ -12220,6 +12220,60 @@
 FILTER_VER_CHROMA_S_AVX2_Nx16 ss, 16
 FILTER_VER_CHROMA_S_AVX2_Nx16 ss, 32
 
+%macro FILTER_VER_CHROMA_S_AVX2_NxN 3
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%3_%1x%2, 4, 11, 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 %3,sp
+    mova            m9, [pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+    mov             r9d, %2 / 16
+.loopH:
+    mov             r10d, %1 / 8
+.loopW:
+    PROCESS_CHROMA_S_AVX2_W8_16R %3
+%ifidn %3,sp
+    add             r2, 8
+%else
+    add             r2, 16
+%endif
+    add             r0, 16
+    dec             r10d
+    jnz             .loopW
+    lea             r0, [r7 - 2 * %1 + 16]
+%ifidn %3,sp
+    lea             r2, [r8 + r3 * 4 - %1 + 8]
+%else
+    lea             r2, [r8 + r3 * 4 - 2 * %1 + 16]
+%endif
+    dec             r9d
+    jnz             .loopH
+    RET
+%endif
+%endmacro
+
+FILTER_VER_CHROMA_S_AVX2_NxN 16, 32, sp
+FILTER_VER_CHROMA_S_AVX2_NxN 24, 32, sp
+FILTER_VER_CHROMA_S_AVX2_NxN 32, 32, sp
+FILTER_VER_CHROMA_S_AVX2_NxN 16, 32, ss
+FILTER_VER_CHROMA_S_AVX2_NxN 24, 32, ss
+FILTER_VER_CHROMA_S_AVX2_NxN 32, 32, ss
+
 ;---------------------------------------------------------------------------------------------------------------------
 ; void interp_4tap_vertical_ss_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
 ;---------------------------------------------------------------------------------------------------------------------