Skip to content

F
ffmpeg
Commit ede4ec1f authored by James Almer's avatar James Almer
Browse files
Options

x86/aacpsdsp: optimize add_squares loop 

Signed-off-by: 's avatarJames Almer <jamrial@gmail.com>
parent d5ded429
Hide whitespace changes
Inline Side-by-side
Showing with 9 additions and 5 deletions
libavcodec/x86/aacpsdsp.asm
View file @ ede4ec1f
......@@ -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
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment