[x265] [PATCH Review only] asm: pixel_add_pp routine for block sizes 32x8, 32x16, 32x24, 32x32 and 32x64
murugan at multicorewareinc.com
murugan at multicorewareinc.com
Wed Nov 6 08:47:50 CET 2013
# HG changeset patch
# User Murugan Vairavel <murugan at multicorewareinc.com>
# Date 1383724042 -19800
# Wed Nov 06 13:17:22 2013 +0530
# Node ID d14d18379efbeb4b77e84e66e809c8d02715530e
# Parent e025d0e905a28d8517bd13303d157a6139ded651
asm: pixel_add_pp routine for block sizes 32x8, 32x16, 32x24, 32x32 and 32x64
diff -r e025d0e905a2 -r d14d18379efb source/common/x86/pixel-add8.asm
--- a/source/common/x86/pixel-add8.asm Wed Nov 06 12:59:57 2013 +0530
+++ b/source/common/x86/pixel-add8.asm Wed Nov 06 13:17:22 2013 +0530
@@ -484,3 +484,54 @@
%endmacro
PIXELADD_PP_W24_H2 24, 32
+
+;-----------------------------------------------------------------------------
+; void pixel_add_pp_c_%1x%2(pixel *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);
+;-----------------------------------------------------------------------------
+%macro PIXELADD_PP_W32_H2 2
+INIT_XMM sse2
+cglobal pixel_add_pp_%1x%2, 4, 7, 8, dest, deststride, src0, src1
+
+mov r4d, r4m
+mov r5d, r5m
+mov r6d, %2
+
+.loop
+
+ movu m0, [r2]
+ movu m1, [r2 + 16]
+ movu m2, [r3]
+ movu m3, [r3 + 16]
+
+ movu m4, [r2 + r4]
+ movu m5, [r2 + r4 + 16]
+ movu m6, [r3 + r5]
+ movu m7, [r3 + r5 + 16]
+
+ paddusb m0, m2
+ paddusb m1, m3
+ paddusb m4, m6
+ paddusb m5, m7
+
+ movu [r0], m0
+ movu [r0 + 16], m1
+ movu [r0 + r1], m4
+ movu [r0 + r1 + 16], m5
+
+ lea r2, [r2 + 2 * r4]
+ lea r3, [r3 + 2 * r5]
+ lea r0, [r0 + 2 * r1]
+
+ sub r6d, 2
+
+jnz .loop
+
+RET
+%endmacro
+
+PIXELADD_PP_W32_H2 32, 8
+PIXELADD_PP_W32_H2 32, 16
+PIXELADD_PP_W32_H2 32, 24
+PIXELADD_PP_W32_H2 32, 32
+PIXELADD_PP_W32_H2 32, 64
+
More information about the x265-devel
mailing list