f1f42d12d4
require you to use movd (instead of movq) when transferring data between reg32/64 and an mmx register. No PKGREVISION bump since it failed to compile on amd64 meaning there was no binary package.
13 lines
526 B
C
13 lines
526 B
C
$NetBSD: patch-src_gsm0610__rpe.c,v 1.1 2011/09/01 09:22:30 jnemeth Exp $
|
|
|
|
--- src/gsm0610_rpe.c.orig 2010-07-24 16:28:34.000000000 +0000
|
|
+++ src/gsm0610_rpe.c
|
|
@@ -82,7 +82,7 @@ static void weighting_filter(int16_t x[4
|
|
" movq 8(%%rax),%%mm2;\n"
|
|
" movq 16(%%rax),%%mm3;\n"
|
|
" movq $0x1000,%%rax;\n"
|
|
- " movq %%rax,%%mm5;\n" /* For rounding */
|
|
+ " movd %%rax,%%mm5;\n" /* For rounding */
|
|
" xorq %%rsi,%%rsi;\n"
|
|
" .p2align 2;\n"
|
|
"1:\n"
|