x87 instructions are largely unimplemented in gem5. Unless you'd like to implement them for us, you should compile your binary to use SSE rather than x87 for floating point (which should happen automatically if you compile in 64-bit mode).
Steve On Sun, Aug 19, 2012 at 5:00 AM, Yi Liu <cnliu...@gmail.com> wrote: > Hi, > recently when I run fft program of splash2 benchmark on gem5 x86 full > system, I found fft cannot work right. According to the trace file, because > of some instruction unimplemented, fft stuck in infinite loop test loop, > part of trace as follows: > * > * > *fprem1 (unimplemented)* > *fnstsw (unimplemented)* > *test eax, 0x400* > *jnz 0xfffffffffffffff5* > *fprem1 (unimplemented)* > *fnstsw (unimplemented)* > *test eax, 0x400* > *jnz 0xfffffffffffffff5* > *fprem1 (unimplemented)* > *fnstsw (unimplemented)* > ... > > and gem5 also shows: > > **** REAL SIMULATION **** > info: Entering event queue @ 0. Starting simulation... > warn: Don't know what interrupt to clear for console. > warn: instruction 'fxsave' unimplemented > warn: Tried to clear PCI interrupt 14 > warn: Unknown mouse command 0xe1. > warn: instruction 'wbinvd' unimplemented > hack: be nice to actually delete the event here > Writing checkpoint > info: Entering event queue @ 5113183552500. Starting simulation... > warn: instruction 'fnstcw_Mw' unimplemented > warn: instruction 'fldcw_Mw' unimplemented > warn: instruction 'fsincos' unimplemented > warn: instruction 'fnstsw' unimplemented > warn: instruction 'fldpi' unimplemented > warn: instruction 'fadd' unimplemented > warn: instruction 'fxch' unimplemented > warn: instruction 'fprem1' unimplemented > > Is it possible to solve this problem? > > > > _______________________________________________ > gem5-users mailing list > gem5-users@gem5.org > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >
_______________________________________________ gem5-users mailing list gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
