[x265] [PATCH Review Only] asm code for interp_4tap_vert_pp_24x32

praveen at multicorewareinc.com praveen at multicorewareinc.com
Tue Oct 29 11:37:07 CET 2013


# HG changeset patch
# User Praveen Tiwari
# Date 1383043010 -19800
# Node ID 35bbb8426fe828174d2866154953ef20cea16286
# Parent  3306252fb160dcabe2015a502572e9ad635e48a8
asm code for interp_4tap_vert_pp_24x32

diff -r 3306252fb160 -r 35bbb8426fe8 source/common/x86/ipfilter8.asm
--- a/source/common/x86/ipfilter8.asm	Tue Oct 29 14:52:06 2013 +0530
+++ b/source/common/x86/ipfilter8.asm	Tue Oct 29 16:06:50 2013 +0530
@@ -1003,3 +1003,84 @@
 %endmacro
 
 FILTER_V4_W6_H4 6, 8
+
+;-----------------------------------------------------------------------------
+;void interp_4tap_vert_pp_24x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W24 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_24x32, 4, 6, 8
+
+mov         r4d,       r4m
+sub         r0,        r1
+
+%ifdef PIC
+lea         r5,        [tab_ChromaCoeff]
+movd        m0,        [r5 + r4 * 4]
+%else
+movd        m0,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+pshufb      m1,        m0,       [tab_Vm]
+pshufb      m0,        [tab_Vm + 16]
+
+mova        m7,        [tab_c_512]
+
+mov         r4d,       %2
+
+.loop
+movu        m2,        [r0]
+movu        m3,        [r0 + r1]
+
+punpcklbw   m4,        m2,        m3,
+punpckhbw   m2,        m3,
+
+pmaddubsw   m4,        m1
+pmaddubsw   m2,        m1
+
+movu        m5,        [r0 + 2 * r1]
+lea         r5,        [r0 + 2 * r1]
+movu        m3,        [r5 + r1]
+
+punpcklbw   m6,        m5,        m3,
+punpckhbw   m5,        m3
+
+pmaddubsw   m6,        m0
+pmaddubsw   m5,        m0
+
+paddw       m4,        m6
+paddw       m2,        m5
+
+pmulhrsw    m4,        m7
+pmulhrsw    m2,        m7
+
+packuswb    m4,        m2
+
+movu        [r2],      m4
+
+movq        m2,        [r0 + 16]
+movq        m3,        [r0 + r1 + 16]
+movq        m4,        [r0 + 2 * r1 + 16]
+movq        m5,        [r5 + r1 + 16]
+
+punpcklbw   m2,        m3
+punpcklbw   m4,        m5
+
+pmaddubsw   m2,        m1
+pmaddubsw   m4,        m0
+
+paddw       m2,        m4
+
+pmulhrsw    m2,        m7
+packuswb    m2,        m2
+movh        [r2 + 16], m2
+
+lea         r0,        [r0 + r1]
+lea         r2,        [r2 + r3]
+
+dec         r4
+jnz        .loop
+RET
+%endmacro
+
+FILTER_V4_W24 24, 32


More information about the x265-devel mailing list