tags 379497 + help
thanks
hi,
see attached patch, *& can be omitted, though. Another posibility is to
use:
__asm__ ( "ldmxcsr (%0)\n" : : "X" (trunc_mxcsr) );
i lack the needed SSE/ASM/Intel know-how, tagging with help, asking
upstream for what would be the best solution. Both gentoo and other rpm
packages use the attached patch for fixing this problem.
bye,
- michael
--- libjpeg-mmx-0.1.5.orig/jquant_x86simd.c
+++ libjpeg-mmx-0.1.5/jquant_x86simd.c
@@ -104,7 +104,7 @@
/* Initialise zero block flags */
/* Set up SSE rounding mode */
- __asm__ ( "ldmxcsr %0\n" : : "X" (trunc_mxcsr) );
+ __asm__ ( "ldmxcsr %0\n" : : "m" (trunc_mxcsr) );
for (i=0; i < 64 ; i+=4)
{
