Re: [Mingw-users] An error regarding AMD-XOP and emmintrin.h?

[Andrew S. <ver...@ya...>]

Hi Gisle,

"I believe 'vmovd xmm0..' is an Intel AVX instruction. Running this on a CPU w/o AVX
would off-course trigger an 'illegal instruction'." 

I see. I'm using an AMD-8150 FX processor. Do you know if there is a flag I could set
to generate code for AMD specific XOP instructions? I'm a little confused why the AVX 
instruction even shows up though (I thought I was using XOP instructions not AVX), plus
I thought AVX was supported by AMD too. I checked CPUID through CPU-Z, the 8150 
has:

MMX(+), SSE(1,2,3,3S,4.1,4.2,4A), x86-64, AMD-V, AES,AVX,XOP

Also should I wrap my code around with ( is this even possible for different manufacturers btw)?

#ifdef SOME INTEL TAG FOR INTEL CPUS
//insert intel cpu only code here.
#endif

Any other suggestions would be helpful.

thanks,

Andrew   




________________________________
 From: Gisle Vanem <gv...@br...>
To: MinGW Users List <min...@li...> 
Sent: Tuesday, July 9, 2013 3:56 AM
Subject: Re: [Mingw-users] An error regarding AMD-XOP and emmintrin.h?
 

"Andrew Somorjai" <ver...@ya...> wrote:

------------

I'm using GCC-TDM 4.7.1-2 and I tried to compile the following code using GCC with the compiler flags -msse3 -msse4 -mxop but there
seems to be a problem only
if the header x86intrin.h is included.


If I run the code through GDB I get a seg fault in emmintrin.h on line 603 where it says:

return __extension__ (__m128i)(__v16qi){

If I comment out x86intrin.h I can run the program with the line


__m128i HADDQinput = _mm_set_epi8(4, 3, 2, 1, 8, 7, 5, 3, 7, 6, 5, 4, 9, 0, 3, 3); //line 14


no trouble. Its only when x86intrin.h is compiled with the source below a crash occurs.



#include <emmintrin.h>
#include <x86intrin.h>//code compiles and runs fine without this header but otherwise causes a seg fault in emmintrin.h line 603
#include <stdio.h>
#include <stdint.h>

void print_2_64_bit_ints(const char * label, __m128i m64_r)
{
int *val = (int *) &m64_r;
printf("%s: %d %d\n",
label, val[0], val[1]);
}
int main()
{
__m128i HADDQinput = _mm_set_epi8(4, 3, 2, 1, 8, 7, 5, 3, 7, 6, 5, 4, 9, 0, 3, 3);//line 14
//__m128i RESULT = _mm_haddq_epi8(HADDQinput);
//_mm_empty();

//print_2_64_bit_ints("Result ", RESULT);

return 0;
}

thanks,

------------

I tried your program here. WinDbg says:

  (14f8.1524): Illegal instruction - code c000001d
  ...
  mmx_gcc+0x1454:
  00401454 c5f96e44243c    vmovd   xmm0,dword ptr [esp+3Ch] ss:0023:0022fefc=00000003


I believe 'vmovd xmm0..' is an Intel AVX instruction. Running this on a CPU w/o AVX
would off-course trigger an 'illegal instruction'.

Ref. http://en.wikipedia.org/wiki/Advanced_Vector_Extensions

--gv 


------------------------------------------------------------------------------
See everything from the browser to the database with AppDynamics
Get end-to-end visibility with application monitoring from AppDynamics
Isolate bottlenecks and diagnose root cause in seconds.
Start your free trial of AppDynamics Pro today!
http://pubads.g.doubleclick.net/gampad/clk?id=48808831&iu=/4140/ostg.clktrk
_______________________________________________
MinGW-users mailing list
Min...@li...

This list observes the Etiquette found at 
http://www.mingw.org/Mailing_Lists.
We ask that you be polite and do the same.  Disregard for the list etiquette may cause your account to be moderated.

_______________________________________________
You may change your MinGW Account Options or unsubscribe at:
https://lists.sourceforge.net/lists/listinfo/mingw-users
Also: mailto:min...@li...?subject=unsubscribe