# HG changeset patch
# User Sumalatha Polureddy
# Date 1431581856 -19800
# Thu May 14 11:07:36 2015 +0530
# Node ID 8f66730a49c9c679f81cf54dd52a94e40e51b86c
# Parent 6502653ac351cea28a741753942d32d0720545f5
asm: avx2 code for sad[64x48] (10791 -> 4053) for 10 bpp
sse2
sad[64x48] 2.34x 10791.39 25291.58
avx2
sad[64x48] 6.45x 4053.05 26139.05
diff -r 6502653ac351 -r 8f66730a49c9 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp Thu May 14 10:49:41 2015 +0530
+++ b/source/common/x86/asm-primitives.cpp Thu May 14 11:07:36 2015 +0530
@@ -1286,6 +1286,7 @@
p.pu[LUMA_32x64].sad = x265_pixel_sad_32x64_avx2;
p.pu[LUMA_64x16].sad = x265_pixel_sad_64x16_avx2;
p.pu[LUMA_64x32].sad = x265_pixel_sad_64x32_avx2;
+ p.pu[LUMA_64x48].sad = x265_pixel_sad_64x48_avx2;
p.pu[LUMA_16x4].convert_p2s = x265_filterPixelToShort_16x4_avx2;
p.pu[LUMA_16x8].convert_p2s = x265_filterPixelToShort_16x8_avx2;
diff -r 6502653ac351 -r 8f66730a49c9 source/common/x86/sad16-a.asm
--- a/source/common/x86/sad16-a.asm Thu May 14 10:49:41 2015 +0530
+++ b/source/common/x86/sad16-a.asm Thu May 14 11:07:36 2015 +0530
@@ -868,6 +868,67 @@
movd eax, xm0
RET
+INIT_YMM avx2
+cglobal pixel_sad_64x48, 4, 5, 8
+ pxor m0, m0
+ pxor m5, m5
+ pxor m6, m6
+ pxor m7, m7
+ mov r4d, 48 / 2
+ add r3d, r3d
+ add r1d, r1d
+.loop:
+ movu m1, [r2 + 0]
+ movu m2, [r2 + 32]
+ movu m3, [r2 + 64]
+ movu m4, [r2 + 96]
+ psubw m1, [r0 + 0]
+ psubw m2, [r0 + 32]
+ psubw m3, [r0 + 64]
+ psubw m4, [r0 + 96]
+ pabsw m1, m1
+ pabsw m2, m2
+ pabsw m3, m3
+ pabsw m4, m4
+ paddw m0, m1
+ paddw m5, m2
+ paddw m6, m3
+ paddw m7, m4
+
+ movu m1, [r2 + r3]
+ movu m2, [r2 + r3 + 32]
+ movu m3, [r2 + r3 + 64]
+ movu m4, [r2 + r3 + 96]
+ psubw m1, [r0 + r1]
+ psubw m2, [r0 + r1 + 32]
+ psubw m3, [r0 + r1 + 64]
+ psubw m4, [r0 + r1 + 96]
+ pabsw m1, m1
+ pabsw m2, m2
+ pabsw m3, m3
+ pabsw m4, m4
+ paddw m0, m1
+ paddw m5, m2
+ paddw m6, m3
+ paddw m7, m4
+
+ lea r0, [r0 + 2 * r1]
+ lea r2, [r2 + 2 * r3]
+
+ dec r4d
+ jg .loop
+
+ HADDUWD m0, m1
+ HADDUWD m5, m1
+ HADDUWD m6, m1
+ HADDUWD m7, m1
+ paddd m0, m5
+ paddd m0, m6
+ paddd m0, m7
+ HADDD m0, m1
+ movd eax, xm0
+ RET
+
;------------------------------------------------------------------
; int pixel_sad_32xN( uint16_t *, intptr_t, uint16_t *, intptr_t )
;------------------------------------------------------------------
_______________________________________________
x265-devel mailing list
x265-devel@videolan.org
https://mailman.videolan.org/listinfo/x265-devel
