what are __ieee754_exp2l/__ieee754_log2l for?

what are __ieee754_exp2l/__ieee754_log2l for?

[David Mosberger]

On a completely unrelated note, can somebody tell me what the routines
mentioned in $SUBJECT are for?  What are they're supposed to do and
what are they needed for?

I'm asking since on ia64 "make check" keeps failing with:

__ieee754_exp2l not implemented
__ieee754_log2l not implemented

	--david

Re: what are __ieee754_exp2l/__ieee754_log2l for?

[Andreas Schwab]

David Mosberger <davidm@napali.hpl.hp.com> writes:

> On a completely unrelated note, can somebody tell me what the routines
> mentioned in $SUBJECT are for?  What are they're supposed to do and
> what are they needed for?

They are the core of the C99 exp2l and log2l functions, resp.  Nobody has
yet written generic implementations of them for 80-bit extended float.
Currently only x86, m68k and x86-64 specific versions exist, and a 128-bit
extended float version of log2l.  Search for e_exp2l.* and e_log2l.* in
the sysdeps hierarchy for the existing implementations.

> I'm asking since on ia64 "make check" keeps failing with:
>
> __ieee754_exp2l not implemented
> __ieee754_log2l not implemented

It's not failing because of this.  When a math funtion is not implemented
it is just skipped by the tests.  Actually, all math tests are working
fine here.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: what are __ieee754_exp2l/__ieee754_log2l for?

[David Mosberger]

>>>>> On Thu, 11 Nov 2004 01:14:12 +0100, Andreas Schwab <schwab@suse.de> said:

  Andreas> They are the core of the C99 exp2l and log2l functions,
  Andreas> resp.  Nobody has yet written generic implementations of
  Andreas> them for 80-bit extended float.

Ah, looks like sysdeps/ia64/fpu is in need of a refresh.  v2.1 of the
Intel libm for Itanium does have exp2l (and log2l was introduced in
verison 1.4).

Can somebody clue me in on the meaning of the various math-library
file-name prefixes?  "w_" stands for "wrapper", I figured that one.
But what's the idea behind the "s_" vs. "e_" prefixes?

	--david

Re: what are __ieee754_exp2l/__ieee754_log2l for?

[Andreas Schwab]

David Mosberger <davidm@napali.hpl.hp.com> writes:

> Can somebody clue me in on the meaning of the various math-library
> file-name prefixes?  "w_" stands for "wrapper", I figured that one.
> But what's the idea behind the "s_" vs. "e_" prefixes?

I have no idea what they stand for, but s_ is used for the functions
without a wrapper, and e_ for the wrapped functions.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: what are __ieee754_exp2l/__ieee754_log2l for?

[David Mosberger]

>>>>> On Thu, 11 Nov 2004 10:58:07 +0100, Andreas Schwab <schwab@suse.de> said:

  Andreas> David Mosberger <davidm@napali.hpl.hp.com> writes:
  >> Can somebody clue me in on the meaning of the various
  >> math-library file-name prefixes?  "w_" stands for "wrapper", I
  >> figured that one.  But what's the idea behind the "s_" vs. "e_"
  >> prefixes?

  Andreas> I have no idea what they stand for, but s_ is used for the
  Andreas> functions without a wrapper, and e_ for the wrapped
  Andreas> functions.

Ah, that helps, thanks!

	--david

Re: what are __ieee754_exp2l/__ieee754_log2l for?

[David Mosberger]

BTW: you were right that the test-ldouble failures were not caused by
the lack of exp2l/log2l.  The problem is/was due to these failures:

 Failure: yn (0, inf) == 0: Exception "Invalid operation" set
 Failure: yn (1, inf) == 0: Exception "Invalid operation" set
 Failure: yn (3, inf) == 0: Exception "Invalid operation" set
 Failure: yn (10, inf) == 0: Exception "Invalid operation" set

These appear to be due to a compiler-bug, where a division is executed
needlessly and that division ends up modifying the floating-point
status register flags.  I'm seeing these problems with:

  gcc version 3.3.5 (Debian 1:3.3.5-2)

but not with:

  gcc version 3.4.2 (Debian 3.4.2-3)

Aren't you also using a 3.3.x-based compiler?  I take it that one
doesn't show the problem?

	--david

Re: what are __ieee754_exp2l/__ieee754_log2l for?

[Andreas Schwab]

David Mosberger <davidm@napali.hpl.hp.com> writes:

> I'm seeing these problems with:
>
>   gcc version 3.3.5 (Debian 1:3.3.5-2)
>
> but not with:
>
>   gcc version 3.4.2 (Debian 3.4.2-3)
>
> Aren't you also using a 3.3.x-based compiler?

We are using the hammer-3_3-branch branch.

>  I take it that one doesn't show the problem?

Yes, it doesn't seem so.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."