This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Improving libm-test.inc structure and maintenance
- From: David Miller <davem at davemloft dot net>
- To: aj at suse dot com
- Cc: joseph at codesourcery dot com, libc-alpha at sourceware dot org
- Date: Fri, 03 May 2013 12:22:02 -0400 (EDT)
- Subject: Re: Improving libm-test.inc structure and maintenance
- References: <Pine dot LNX dot 4 dot 64 dot 1305022244550 dot 12072 at digraph dot polyomino dot org dot uk> <20130503 dot 000329 dot 2249460654982306702 dot davem at davemloft dot net> <51837FD0 dot 7090602 at suse dot com>
From: Andreas Jaeger <aj@suse.com>
Date: Fri, 03 May 2013 11:13:52 +0200
> On 05/03/2013 06:03 AM, David Miller wrote:
>> From: "Joseph S. Myers" <joseph@codesourcery.com>
>> Date: Thu, 2 May 2013 22:46:55 +0000
>>
>>> I propose two main changes to how these tests are handled:
>>
>> I think this is a fantastic set of proposals. Definitely a
>> significant improvement from the current situation.
>>
>>> It is of course possible also to split up libm-test so that the tests
>>> for each function are run separately (with or without also splitting
>>> up the auto-libm-test-* files). I don't know if that's desirable.
>>
>> I wouldn't mind seeing them split up in a way that would allow
>> running more of the tests in parallel, but I that is very low
>> priority compared to the core parts of your proposal.
>
>
> I agree. And running them on my x86-64, they execute so quickly (less
> than a second) that splitting them up is not really worth it. I think
> building takes longer, so perhaps we could split it up into <math.h>
> and <complex.h> functions? Or something like that - but I wouldn't go
> so far as to split it for each function.
You're right, building time far exceeds execution and Joseph's changes
to a table based approach will fix that.