[llvm-commits] [PATCH] Adding the Linpack test to the test suite

[David Blaikie]

On Sun, Nov 25, 2012 at 1:25 PM, Renato Golin <rengolin at systemcall.org> wrote:
> On 25 November 2012 21:08, David Blaikie <dblaikie at gmail.com> wrote:
>> Thanks - not exactly a rush job (this being a weekend & all & it's
>> been broken for quite a while so a day or two isn't going to change
>> that) but I wanted to make sure it was brought up.
>
> I can't see what's wrong with Livermore from the output. Some llc
> invocations succeed, others "fail" with this message:
>
> TEST-RESULT-nat-time: user       0.3480
>
> TEST-FAIL: llc /home/baldrick/osuosl/slave/clang-x86_64-debian-fnt/test-suite.obj/SingleSource/Benchmarks/LivermoreLoops/kernel23
>
> TEST-PASS: compile
> /home/baldrick/osuosl/slave/clang-x86_64-debian-fnt/test-suite.obj/SingleSource/Benchmarks/LivermoreLoops/kernel23
> TEST-RESULT-compile:   Total Execution Time: 0.0280 seconds (0.0270 wall clock)
>
> TEST-RESULT-compile: 4896 Output/kernel23.llvm.bc
>
> It fails and pass, and the output is generated, with no error message?
> I'll try to reproduce it here, but my machine is dead, it may take a
> while... :(

If you look up a bit further in the logs you'll find things like:

******************** TEST (llc) 'kernel01' FAILED! ********************
Execution Context Diff:
/home/baldrick/osuosl/slave/clang-x86_64-debian-fnt/test-suite.obj/tools/fpcmp:
Compared: 1.143052e+02 and 1.000876e+02
abs. diff = 1.421754e+01 rel.diff = 1.420509e-01
Out of tolerance: rel/abs: 1.000000e-04/0.000000e+00

which seem relevant. Perhaps some info before that too.