mirror of
https://github.com/FFmpeg/FFmpeg.git
synced 2024-12-23 12:43:46 +02:00
x86inc: Use SSE instead of SSE2 for copying data
Reduces code size because movaps/movups is one byte shorter than movdqa/movdqu. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
This commit is contained in:
parent
ad76e6e7e1
commit
63f0d62310
@ -436,7 +436,7 @@ DECLARE_REG 14, R15, 120
|
||||
%assign %%i xmm_regs_used
|
||||
%rep (xmm_regs_used-6)
|
||||
%assign %%i %%i-1
|
||||
movdqa [rsp + (%%i-6)*16 + stack_size + (~stack_offset&8)], xmm %+ %%i
|
||||
movaps [rsp + (%%i-6)*16 + stack_size + (~stack_offset&8)], xmm %+ %%i
|
||||
%endrep
|
||||
%endmacro
|
||||
|
||||
@ -454,7 +454,7 @@ DECLARE_REG 14, R15, 120
|
||||
%assign %%i xmm_regs_used
|
||||
%rep (xmm_regs_used-6)
|
||||
%assign %%i %%i-1
|
||||
movdqa xmm %+ %%i, [%1 + (%%i-6)*16+stack_size+(~stack_offset&8)]
|
||||
movaps xmm %+ %%i, [%1 + (%%i-6)*16+stack_size+(~stack_offset&8)]
|
||||
%endrep
|
||||
%if stack_size_padded == 0
|
||||
add %1, (xmm_regs_used-6)*16+16
|
||||
|
Loading…
Reference in New Issue
Block a user