[FIXED] mplayer not using SSE?

Question

Thats what i have when i play anything:
Cannot test OS support for SSE, disabling to be safe.
Can i do something about?
Last edited by kokoko3k (2013-06-13 15:19:50)

Answer 1

mplayer2 works as expected:
Detected cache-line size is 64 bytes
Testing OS support for SSE... yes.
...that's weird
-edit
i toke a look to the coudetect.c file (the message and tests about sse are there) of both mplayer and mplayer2, and compiled both with the same file (discarding the changes done by mplayer2, but the result is the same.
Also, it seems that for some reason all lies here:
if defined(_POSIX_SOURCE)
struct sigaction saved_sigill;
//kokoko3k:
mp_msg(MSGT_CPUDETECT,MSGL_V, "IF YOU READ ME, POSIX SOURCE IS DEFINED " );
/* Save the original signal handlers.
sigaction( SIGILL, NULL, &saved_sigill );
signal( SIGILL, (void (*)(int))sigill_handler_sse );
/* Emulate test for OSFXSR in CR4. The OS will set this bit if it
* supports the extended FPU save and restore required for SSE. If
* we execute an SSE instruction on a PIII and get a SIGILL, the OS
* doesn't support Streaming SIMD Exceptions, even if the processor
* does.
if ( gCpuCaps.hasSSE ) {
mp_msg(MSGT_CPUDETECT,MSGL_V, "Testing OS support for SSE... " );
// __asm__ volatile ("xorps %%xmm0, %%xmm0");
__asm__ volatile ("xorps %xmm0, %xmm0");
mp_msg(MSGT_CPUDETECT,MSGL_V, gCpuCaps.hasSSE ? "yes.\n" : "no!\n" );
/* Restore the original signal handlers.
sigaction( SIGILL, &saved_sigill, NULL );
/* If we've gotten to here and the XMM CPUID bit is still set, we're
* safe to go ahead and hook out the SSE code throughout Mesa.
mp_msg(MSGT_CPUDETECT,MSGL_V, "Tests of OS support for SSE %s\n", gCpuCaps.hasSSE ? "passed." : "failed!" );
#else
/* We can't use POSIX signal handling to test the availability of
* SSE, so we disable it by default.
mp_msg(MSGT_CPUDETECT,MSGL_WARN, "Cannot test OS support for SSE, disabling to be safe.\n" );
gCpuCaps.hasSSE=0;
#endif /* _POSIX_SOURCE */
mplayer say that _POSIX_SOURCE is not defined; mplayer2 seems to get it.
-EDIT-
I went deeper into this and found that _POSIX_SOURCE is defined in /usr/include/features.h
-EDIT-
Seems not to be a packaging issue but a mainstream one; i used the previous PKGBUILD ewith the current mplayer snapshot and the bug is still there.
I guess that as soon as mplayer2 will get the current changes it will be affected too.
Last edited by kokoko3k (2013-06-07 12:38:17)