[x265] [PATCH] asm: avx2 code for chroma_vpp/vps/vss/vsp[8x2][i420] 16bpp

[aasaipriya at multicorewareinc.com]

# HG changeset patch
# User Aasaipriya Chandran <aasaipriya at multicorewareinc.com>
# Date 1433915765 -19800
#      Wed Jun 10 11:26:05 2015 +0530
# Node ID 436e5792cd7e5ff348f25878a8d927144a45c5e0
# Parent  b252468dde7ffca57da27575388d95ce538945d2
asm: avx2 code for chroma_vpp/vps/vss/vsp[8x2][i420] 16bpp

chroma_vpp[8x2]       6.06x    215.01          1302.73
chroma_vps[8x2]       4.89x    205.02          1003.49
chroma_vsp[8x2]       6.07x    214.95          1303.73
chroma_vss[8x2]       4.66x    209.95          978.59

diff -r b252468dde7f -r 436e5792cd7e source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Tue Jun 09 10:16:44 2015 +0530
+++ b/source/common/x86/asm-primitives.cpp	Wed Jun 10 11:26:05 2015 +0530
@@ -1699,6 +1699,10 @@
         p.chroma[X265_CSP_I444].pu[LUMA_64x64].filter_hpp = x265_interp_4tap_horiz_pp_64x64_avx2;
         p.chroma[X265_CSP_I444].pu[LUMA_48x64].filter_hpp = x265_interp_4tap_horiz_pp_48x64_avx2;
 
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x2].filter_vpp = x265_interp_4tap_vert_pp_8x2_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x2].filter_vps = x265_interp_4tap_vert_ps_8x2_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x2].filter_vsp = x265_interp_4tap_vert_sp_8x2_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x2].filter_vss = x265_interp_4tap_vert_ss_8x2_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_8x16].filter_vpp = x265_interp_4tap_vert_pp_8x16_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_8x16].filter_vps = x265_interp_4tap_vert_ps_8x16_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_8x16].filter_vsp = x265_interp_4tap_vert_sp_8x16_avx2;
diff -r b252468dde7f -r 436e5792cd7e source/common/x86/ipfilter16.asm
--- a/source/common/x86/ipfilter16.asm	Tue Jun 09 10:16:44 2015 +0530
+++ b/source/common/x86/ipfilter16.asm	Wed Jun 10 11:26:05 2015 +0530
@@ -10985,3 +10985,85 @@
 FILTER_VER_CHROMA_AVX2_8xN sp, 64
 FILTER_VER_CHROMA_AVX2_8xN ss, 64
 
+%macro PROCESS_CHROMA_AVX2_8x2 3
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m2, m2, [r5 + 1 * mmsize]
+    paddd           m0, m2
+
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m3, m3, [r5 + 1 * mmsize]
+    paddd           m1, m3
+
+%ifnidn %1,ss
+    paddd           m0, m7
+    paddd           m1, m7
+%endif
+    psrad           m0, %3
+    psrad           m1, %3
+
+    packssdw        m0, m1
+    vpermq          m0, m0, q3120
+    pxor            m4, m4
+
+%ifidn %2,doClip
+    CLIPW           m0, m4, [pw_pixel_max]
+%endif
+    vextracti128    xm1, m0, 1
+%endmacro
+
+
+%macro FILTER_VER_CHROMA_AVX2_8x2 3
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x2, 4, 6, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    add             r3d, r3d
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    vbroadcasti128  m7, [pd_32]
+%elifidn %1, sp
+    mova            m7, [pd_524800]
+%else
+    vbroadcasti128  m7, [pd_n32768]
+%endif
+
+    PROCESS_CHROMA_AVX2_8x2 %1, %2, %3
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    RET
+%endmacro
+
+FILTER_VER_CHROMA_AVX2_8x2 pp, doClip, 6
+FILTER_VER_CHROMA_AVX2_8x2 ps, 0, 2
+FILTER_VER_CHROMA_AVX2_8x2 sp, doClip, 10
+FILTER_VER_CHROMA_AVX2_8x2 ss, 0, 6