[x265] [PATCH] asm: pixelavg_pp[8xN] sse2 code for 10bpp

[rajesh at multicorewareinc.com]

# HG changeset patch
# User Rajesh Paulraj<rajesh at multicorewareinc.com>
# Date 1435735579 -19800
#      Wed Jul 01 12:56:19 2015 +0530
# Node ID ab8cfdf88aad580381f2fcc11533c4cd1d2f1250
# Parent  2f345c1c0d8e2351e5aaae5f3e0e017b5810f32e
asm: pixelavg_pp[8xN] sse2 code for 10bpp

     avg_pp[  8x4]  5.12x    125.34          641.61
     avg_pp[  8x8]  5.95x    202.53          1205.34
     avg_pp[ 8x16]  6.94x    334.54          2322.57
     avg_pp[ 8x32]  8.15x    589.39          4806.23

diff -r 2f345c1c0d8e -r ab8cfdf88aad source/common/x86/mc-a.asm
--- a/source/common/x86/mc-a.asm	Tue Jun 30 13:08:15 2015 -0500
+++ b/source/common/x86/mc-a.asm	Wed Jul 01 12:56:19 2015 +0530
@@ -4009,6 +4009,87 @@
     AVG_END
 %endmacro
 
+%macro  pixel_avg_W8 0
+    movu    m0, [r2]
+    movu    m1, [r4]
+    pavgw   m0, m1
+    movu    [r0], m0
+    movu    m2, [r2 + r3]
+    movu    m3, [r4 + r5]
+    pavgw   m2, m3
+    movu    [r0 + r1], m2
+
+    movu    m0, [r2 + r3 * 2]
+    movu    m1, [r4 + r5 * 2]
+    pavgw   m0, m1
+    movu    [r0 + r1 * 2], m0
+    movu    m2, [r2 + r6]
+    movu    m3, [r4 + r7]
+    pavgw   m2, m3
+    movu    [r0 + r8], m2
+
+    lea     r0, [r0 + 4 * r1]
+    lea     r2, [r2 + 4 * r3]
+    lea     r4, [r4 + 4 * r5]
+%endmacro
+
+;-------------------------------------------------------------------------------------------------------------------------------
+;void pixelavg_pp(pixel dst, intptr_t dstride, const pixel src0, intptr_t sstride0, const pixel* src1, intptr_t sstride1, int)
+;-------------------------------------------------------------------------------------------------------------------------------
+%if HIGH_BIT_DEPTH
+%if ARCH_X86_64
+INIT_XMM sse2
+cglobal pixel_avg_8x4, 6,9,4
+    add     r1d, r1d
+    add     r3d, r3d
+    add     r5d, r5d
+    lea     r6, [r3 * 3]
+    lea     r7, [r5 * 3]
+    lea     r8, [r1 * 3]
+    pixel_avg_W8
+    RET
+
+cglobal pixel_avg_8x8, 6,9,4
+    add     r1d, r1d
+    add     r3d, r3d
+    add     r5d, r5d
+    lea     r6, [r3 * 3]
+    lea     r7, [r5 * 3]
+    lea     r8, [r1 * 3]
+    pixel_avg_W8
+    pixel_avg_W8
+    RET
+
+cglobal pixel_avg_8x16, 6,10,4
+    add     r1d, r1d
+    add     r3d, r3d
+    add     r5d, r5d
+    lea     r6, [r3 * 3]
+    lea     r7, [r5 * 3]
+    lea     r8, [r1 * 3]
+    mov     r9d, 4
+.loop
+    pixel_avg_W8
+    dec     r9d
+    jnz     .loop
+    RET
+
+cglobal pixel_avg_8x32, 6,10,4
+    add     r1d, r1d
+    add     r3d, r3d
+    add     r5d, r5d
+    lea     r6, [r3 * 3]
+    lea     r7, [r5 * 3]
+    lea     r8, [r1 * 3]
+    mov     r9d, 8
+.loop
+    pixel_avg_W8
+    dec     r9d
+    jnz     .loop
+    RET
+%endif
+%endif
+
 %if HIGH_BIT_DEPTH
 
 INIT_MMX mmx2
@@ -4060,11 +4141,6 @@
 AVGH  4, 4
 AVGH  4, 2
 
-AVG_FUNC 8, movdqu, movdqa
-AVGH  8, 32
-AVGH  8, 16
-AVGH  8,  8
-AVGH  8,  4
 
 AVG_FUNC 16, movdqu, movdqa
 AVGH  16, 64