|
From: Jeroen N. W. <jn...@xs...> - 2005-03-27 13:06:10
Attachments:
scalar.stderr.diff
badseg.verbose
|
I seem to have run into a problem in the portability of valgrind 3.0. On my box, valgrind 2.4.0 from cvs runs 'make regtest' as in the nightly build (audi, Red Hat 9), with only memcheck/tests/scalar failing due to a mismatch in line numbers. [diff attached] Also, a number of tests fail their prereqs: insn_mmxext: (skipping, prereq failed: ../../../tests/cputest x86-mmxext) insn_sse: (skipping, prereq failed: ../../../tests/cputest x86-sse) insn_sse2: (skipping, prereq failed: ../../../tests/cputest x86-sse2) But with valgrind 3.0.0.SVN, I get this: valgrind: fatal error: unsupported CPU. Supported CPUs are: * x86 with SSE state (Pentium II or above, AMD Athlon or above) [Full verbose output of one test attached.] $ cat /proc/cpuinfo processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 3 model name : Pentium II (Klamath) stepping : 3 cpu MHz : 233.866 cache size : 512 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov mmx bogomips : 466.94 Is this a problem in valgrind 3.0, or a missing restriction in the 3.0 documentation? Should I file a bug report about this? Jeroen. |
|
From: Julian S. <js...@ac...> - 2005-03-27 13:26:56
|
> But with valgrind 3.0.0.SVN, I get this: > > valgrind: fatal error: unsupported CPU. > Supported CPUs are: > * x86 with SSE state (Pentium II or above, AMD Athlon or above) > > Is this a problem in valgrind 3.0, or a missing restriction in the 3.0 > documentation? Should I file a bug report about this? The 3.0 line should work on PIIs, but I don't have one to test it on, hence I'm not surprised there's a problem. Have a poke around in coregrind/x86/state.c, function VGA_(getArchAndSubArch). It should be that have_sse0 is set to True and have_sse1 and have_sse2 are set to False, but I guess that is not happening right. J |
|
From: Tom H. <to...@co...> - 2005-03-29 13:38:27
|
In message <200...@ac...>
Julian Seward <js...@ac...> wrote:
>> But with valgrind 3.0.0.SVN, I get this:
>>
>> valgrind: fatal error: unsupported CPU.
>> Supported CPUs are:
>> * x86 with SSE state (Pentium II or above, AMD Athlon or above)
>>
>> Is this a problem in valgrind 3.0, or a missing restriction in the 3.0
>> documentation? Should I file a bug report about this?
>
> The 3.0 line should work on PIIs, but I don't have one to test it on,
> hence I'm not surprised there's a problem.
>
> Have a poke around in coregrind/x86/state.c,
> function VGA_(getArchAndSubArch). It should be that have_sse0 is
> set to True and have_sse1 and have_sse2 are set to False, but I
> guess that is not happening right.
For have_sse0 to get set the fxsr bit has to be set in the
capabilities and according to the cpuinfo output that was posted
the processor in question didn't have that bit set.
I have a PII here that does have it set however:
model name : Pentium II (Deschutes)
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr
Note that it is a "Deschutes" processor not a "Klamath" so it is
obviously a later model that added some extra features (specifically
the apic, pat, pse36 and fxsr flags).
Requiring fxsr will also rule P5, P5-MMX and P6 processors as well
as older AMD processors like the K6.
Tom
--
Tom Hughes (to...@co...)
http://www.compton.nu/
|
|
From: Julian S. <js...@ac...> - 2005-03-29 15:58:29
|
> Requiring fxsr will also rule P5, P5-MMX and P6 processors as well > as older AMD processors like the K6. I'm not terribly bothered about not supporting P5, P5-MMX, K6 or Pentium-Pro. At some point there's a diminishing tail which it's not worth the hassle of supporting -- I'm saying that this tail includes anything that can't at least do fxsave and fxrstor. So what you're saying is that later-model Pentium IIs do set fxsr and so will be supported, yes? I had thought that all PIIs could do that, but evidently not. J |
|
From: Tom H. <to...@co...> - 2005-03-29 16:07:10
|
In message <200...@ac...>
Julian Seward <js...@ac...> wrote:
> So what you're saying is that later-model Pentium IIs do set fxsr
> and so will be supported, yes? I had thought that all PIIs could
> do that, but evidently not.
I did some googling earlier and it was introduced with Deschutes
which was a die shrink of the original PII by the looks of it, so
it is only the very earliest PII's that don't have fxsr.
Tom
--
Tom Hughes (to...@co...)
http://www.compton.nu/
|
|
From: Julian S. <js...@ac...> - 2005-03-29 16:37:15
|
On Tuesday 29 March 2005 17:07, Tom Hughes wrote: > In message <200...@ac...> > > Julian Seward <js...@ac...> wrote: > > So what you're saying is that later-model Pentium IIs do set fxsr > > and so will be supported, yes? I had thought that all PIIs could > > do that, but evidently not. > > I did some googling earlier and it was introduced with Deschutes > which was a die shrink of the original PII by the looks of it, so > it is only the very earliest PII's that don't have fxsr. Right. So then I guess I have to say to Jeroen, sorry, it ain't going to work on PII-Klamath. J |