[FFmpeg-cvslog] x86/aacpsdsp: optimize add_squares loop

[James Almer]

ffmpeg | branch: master | James Almer <jamrial at gmail.com> | Tue Jun 14 12:41:23 2016 -0300| [ede4ec1f8f5fd94dccd880199419a1f1b8137ab6] | committer: James Almer

x86/aacpsdsp: optimize add_squares loop

Signed-off-by: James Almer <jamrial at gmail.com>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ede4ec1f8f5fd94dccd880199419a1f1b8137ab6
---

 libavcodec/x86/aacpsdsp.asm |   14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/libavcodec/x86/aacpsdsp.asm b/libavcodec/x86/aacpsdsp.asm
index d7d7a9a..e92cbbc 100644
--- a/libavcodec/x86/aacpsdsp.asm
+++ b/libavcodec/x86/aacpsdsp.asm
@@ -33,18 +33,22 @@ SECTION .text
 ;*************************************************************************
 %macro PS_ADD_SQUARES 1
 cglobal ps_add_squares, 3, 3, %1, dst, src, n
+    shl    nd, 3
+    add  srcq, nq
+    neg    nq
+
+align 16
 .loop:
-    movaps m0, [srcq]
-    movaps m1, [srcq+mmsize]
+    movaps m0, [srcq+nq]
+    movaps m1, [srcq+nq+mmsize]
     mulps  m0, m0
     mulps  m1, m1
     HADDPS m0, m1, m2
     addps  m0, [dstq]
     movaps [dstq], m0
     add  dstq, mmsize
-    add  srcq, mmsize*2
-    sub    nd, mmsize/4
-    jg .loop
+    add    nq, mmsize*2
+    jl .loop
     REP_RET
 %endmacro