[x265] [PATCH] asm: assembly code for pixel_sse_ss_24x32

[chen]

At 2013-11-26 18:45:25,yuvaraj at multicorewareinc.com wrote:
># HG changeset patch
># User Yuvaraj Venkatesh <yuvaraj at multicorewareinc.com>
># Date 1385462702 -19800
>#      Tue Nov 26 16:15:02 2013 +0530
># Node ID 52738c22dce02e8d59cc4b09f1e1b23a0a8360c5
># Parent  116d91f08fcb123d4b088df5c1400e599306b6f8
>asm: assembly code for pixel_sse_ss_24x32
>
>diff -r 116d91f08fcb -r 52738c22dce0 source/common/x86/asm-primitives.cpp
>--- a/source/common/x86/asm-primitives.cpp Tue Nov 26 14:19:27 2013 +0800
>+++ b/source/common/x86/asm-primitives.cpp Tue Nov 26 16:15:02 2013 +0530
>@@ -103,6 +103,7 @@
>     p.sse_ss[LUMA_16x16]   = x265_pixel_ssd_ss_16x16_ ## cpu; \
>     p.sse_ss[LUMA_16x32]   = x265_pixel_ssd_ss_16x32_ ## cpu; \
>     p.sse_ss[LUMA_16x64]   = x265_pixel_ssd_ss_16x64_ ## cpu; \
>+    p.sse_ss[LUMA_24x32]   = x265_pixel_ssd_ss_24x32_ ## cpu; \
>     p.sse_ss[LUMA_32x8]   = x265_pixel_ssd_ss_32x8_ ## cpu; \
>     p.sse_ss[LUMA_32x16]   = x265_pixel_ssd_ss_32x16_ ## cpu; \
>     p.sse_ss[LUMA_32x24]   = x265_pixel_ssd_ss_32x24_ ## cpu; \
>diff -r 116d91f08fcb -r 52738c22dce0 source/common/x86/pixel-a.asm
>--- a/source/common/x86/pixel-a.asm Tue Nov 26 14:19:27 2013 +0800
>+++ b/source/common/x86/pixel-a.asm Tue Nov 26 16:15:02 2013 +0530
>@@ -469,17 +469,62 @@
> SSD_SS_32 64
> %endmacro
> 
>+%macro SSD_SS_24 0
>+cglobal pixel_ssd_ss_24x32, 4,7,6
>+    FIX_STRIDES r1, r3
>+    mov    r4d, 16
>+    pxor    m0, m0
>+.loop
>+    mova    m1, [r0]
>+    psubw   m1, [r2]
this is right, but it is unsafe, I am not sure the input pointer is alignment
 
>+    pmaddwd m1, m1
>+    paddd   m0, m1
>+    mova    m1, [r0 + 16]
>+    psubw   m1, [r2 + 16]
>+    pmaddwd m1, m1
>+    paddd   m0, m1
>+    mova    m1, [r0 + 32]
>+    psubw   m1, [r2 + 32]
>+    pmaddwd m1, m1
>+    paddd   m0, m1
>+    lea       r0, [r0 + 2*r1]
>+    lea       r2, [r2 + 2*r3]
>+    mova    m1, [r0]
>+    psubw   m1, [r2]
>+    pmaddwd m1, m1
>+    paddd   m0, m1
>+    mova    m1, [r0 + 16]
>+    psubw   m1, [r2 + 16]
>+    pmaddwd m1, m1
>+    paddd   m0, m1
>+    mova    m1, [r0 + 32]
>+    psubw   m1, [r2 + 32]
>+    pmaddwd m1, m1
>+    paddd   m0, m1
>+    lea       r0, [r0 + 2*r1]
>+    lea       r2, [r2 + 2*r3]
>+    dec      r4d
>+    jnz .loop
>+    phaddd    m0, m0
>+    phaddd    m0, m0
>+    movd     eax, m0
>+    RET
>+%endmacro
>+
> INIT_XMM sse2
> SSD_SS_ONE
> SSD_SS_12x16
>+SSD_SS_24
> SSD_SS_32xN
> INIT_XMM sse4
> SSD_SS_ONE
> SSD_SS_12x16
>+SSD_SS_24
> SSD_SS_32xN
> INIT_XMM avx
> SSD_SS_ONE
> SSD_SS_12x16
>+SSD_SS_24
> SSD_SS_32xN
> %endif ; !HIGH_BIT_DEPTH
> 
>@@ -7696,9 +7741,6 @@
> %endif ; !ARCH_X86_64
> %endmacro ; SA8D
> 
>-;=============================================================================
>-; INTRA SATD
>-;=============================================================================
> %define TRANS TRANS_SSE2
> %define DIFFOP DIFF_UNPACK_SSE2
> %define LOAD_SUMSUB_8x4P LOAD_DIFF_8x4P
>_______________________________________________
>x265-devel mailing list
>x265-devel at videolan.org
>https://mailman.videolan.org/listinfo/x265-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20131126/e78ccc4b/attachment.html>