[parisc-linux] Unknown FPU exception, help needed, please decode.

[parisc-linux] Unknown FPU exception, help needed, please decode.

[Carlos O'Donell]

pa,

Running some more compliant fpu code in glibc causes this:

arch/parisc/math-emu/decode_exc.c(351) Unknown FPU exception 0x23
arch/parisc/math-emu/decode_exc.c(351) Unknown FPU exception 0x23
...

Could someone please decode the FPU exception and tell me if the kernel
is doing the right thing?

c.

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Unknown FPU exception, help needed, please decode.

[Randolph Chung]

> arch/parisc/math-emu/decode_exc.c(351) Unknown FPU exception 0x23
> arch/parisc/math-emu/decode_exc.c(351) Unknown FPU exception 0x23
> ...
> 
> Could someone please decode the FPU exception and tell me if the kernel
> is doing the right thing?

 From page 10-3 in the PA2.0 arch book, this seems to be the 
"unimplemented" exception.

Pages 10-8 and 10-9 goes on to say:
A conversion to a floating poiint format always causes an unimplemented 
exception when the result overflows, the result lies too far outside the 
range for the exponent to be bias-adjusted, and the overflow trap is 
enabled. [...]

Similarly, an unimplemented exception is always caused by a conversion 
to a floating-point format that underflows, lies too far outside the 
range for the exponent to be bias-adjusted, and the underflow trap is 
enabled.  [...]

Finally, the unimplemented exception is always signaled when the operand 
of a conversion to an integer format is a NaN. Low-level trap handlers 
may choose to silently deliver a result or convert it to an invalid 
exception.

Do any of these apply to your test case?

randolph
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Unknown FPU exception, help needed, please decode.

[Carlos O'Donell]

On Fri, Dec 16, 2005 at 07:54:59AM +0800, Randolph Chung wrote:
> >arch/parisc/math-emu/decode_exc.c(351) Unknown FPU exception 0x23
> >arch/parisc/math-emu/decode_exc.c(351) Unknown FPU exception 0x23
> >...
> >
> >Could someone please decode the FPU exception and tell me if the kernel
> >is doing the right thing?
> 
> From page 10-3 in the PA2.0 arch book, this seems to be the 
> "unimplemented" exception.

Cool, I assume it should print "Unimplemented" exception instead of
"Unknown."
 
> Pages 10-8 and 10-9 goes on to say:
> A conversion to a floating poiint format always causes an unimplemented 
> exception when the result overflows, the result lies too far outside the 
> range for the exponent to be bias-adjusted, and the overflow trap is 
> enabled. [...]

Might have some of this.

> Similarly, an unimplemented exception is always caused by a conversion 
> to a floating-point format that underflows, lies too far outside the 
> range for the exponent to be bias-adjusted, and the underflow trap is 
> enabled.  [...]

Might have some of this.

> Finally, the unimplemented exception is always signaled when the operand 
> of a conversion to an integer format is a NaN. Low-level trap handlers 
> may choose to silently deliver a result or convert it to an invalid 
> exception.

Might have some of this.

> Do any of these apply to your test case?

It's the glibc fp testsuite. We test for a lot of things, I need to
narrow down the insn sequence causing the exception.

I've never seen this print before. Atleast not with the older
sysdeps/hppa/fpu/* bits in place.

c.

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Unknown FPU exception, help needed, please decode.

[Grant Grundler]

On Thu, Dec 15, 2005 at 08:13:18PM -0500, Carlos O'Donell wrote:
> It's the glibc fp testsuite. We test for a lot of things, I need to
> narrow down the insn sequence causing the exception.
> 
> I've never seen this print before. Atleast not with the older
> sysdeps/hppa/fpu/* bits in place.

Maybe the glibc test has changed too?

grant
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux