swr: int16->int32: use the old index negate trick to avoid 2 adds

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Michael Niedermayer 2012-04-28 17:03:20 +02:00
parent 113738d6c2
commit 95057b1972
1 changed files with 13 additions and 11 deletions

View File

@ -35,18 +35,20 @@ cglobal int16_to_int32_%1, 3, 3, 3, dst, src, len
%else
int16_to_int32_u_int %+ SUFFIX
%endif
add dstq, lenq
shr lenq, 1
add srcq, lenq
neg lenq
.next
mov%1 m4, [srcq]
mov%1 m2, [srcq+lenq]
pxor m0, m0
pxor m1, m1
punpcklwd m0, m4
punpckhwd m1, m4
mov%1 [ dstq], m0
mov%1 [mmsize + dstq], m1
add srcq, mmsize
add dstq, 2*mmsize
sub lenq, 2*mmsize
jg .next
punpcklwd m0, m2
punpckhwd m1, m2
mov%1 [ dstq+2*lenq], m0
mov%1 [mmsize + dstq+2*lenq], m1
add lenq, mmsize
jl .next
%if mmsize == 8
emms
%endif