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:
Henrik Gramner 2013-09-11 17:49:22 +02:00 committed by Derek Buitenhuis
parent ad76e6e7e1
commit 63f0d62310

View File

@ -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