[Beignet] Two Failure in Utest

[Zhigang Gong]

On Thu, Jun 27, 2013 at 07:17:03AM +0000, Sun, Yi wrote:
> Hi all,
> 
> After talking a bit with Homer, we found it may be a GT1 only issue.
> My test machine is GT1 desktop with pciID 0x0152.

I don't have a GT1 machine in hand. But the reason why the "culprit" commit
is the test case for double float should be that the double test case will
force the OCL_SIMD_WIDTH to 8. So all the following test case will run with
SIMD8, and maybe the two cases should be fail with SIMD8 all the time on
your machine. Could you set OCL_SIMD_WIDTH to 8 and then run the test cases
separately?

To avoid the side effect by the current double implementation which will force
OCL_SIMD_WIDTH to 8, one easiest way is to move the double related test cases
to the last of test list. Or is there any better idea?

> 
> Thanks
>   --Sun, Yi
> 
> On Thu, 2013-06-27 at 05:24 +0000, Sun, Yi wrote:
> > After bisecting, I got the culprit commit as following:
> > 
> > db199abf207daad52b34fac600f3388472a9966c is the first bad commit
> > commit db199abf207daad52b34fac600f3388472a9966c
> > Author: Homer Hsing <homer.xing at intel.com>
> > Date:   Fri Jun 21 12:26:32 2013 +0800
> > 
> >     test cases for 64-bit float
> > 
> >     Signed-off-by: Homer Hsing <homer.xing at intel.com>
> >     Reviewed-by: Zhigang Gong <zhigang.gong at linux.intel.com>
> >     Tested-by: Yang Rong <rong.r.yang at intel.com>
> > 
> > Thanks
> >   --Sun, Yi
> > 
> > On Thu, 2013-06-27 at 02:08 +0000, Sun, Yi wrote:
> > > Zhigang,
> > > 
> > > Yes, I double checked it. I'm using the top commit now.
> > > And a new finding: if I run the test cases separately both two would be
> > > passed by command:
> > > ./utest_run compiler_global_memory_barrier
> > > ./utest_run compiler_local_memory_barrier_wg64
> > > 
> > > But it's a bit strange that if I run all the test cases by
> > > command ./utest_run without sub-test name, these two failures would turn
> > > out.
> > > 
> > > Thanks
> > >   --Sun, Yi
> > > 
> > > 
> > > On Thu, 2013-06-27 at 09:46 +0800, Zhigang Gong wrote:
> > > > Yi，
> > > > 
> > > > Were you using the latest master branch to do the testing? There should be
> > > > no failure.
> > > > 
> > > > > -----Original Message-----
> > > > > From: beignet-bounces+zhigang.gong=linux.intel.com at lists.freedesktop.org
> > > > >
> > > > [mailto:beignet-bounces+zhigang.gong=linux.intel.com at lists.freedesktop.org]
> > > > > On Behalf Of Sun, Yi
> > > > > Sent: Thursday, June 27, 2013 9:08 AM
> > > > > To: beignet at lists.freedesktop.org
> > > > > Subject: [Beignet] Two Failure in Utest
> > > > > 
> > > > > 
> > > > > Hi all
> > > > > 
> > > > > I found two failed cases in utest as following. Are they known ones?
> > > > > 
> > > > > compiler_global_memory_barrier:
> > > > >   compiler_global_memory_barrier()    [FAILED]
> > > > >     Error: dst[i+j] == locals[0] - 1 -j
> > > > >   at
> > > > > file /home/works/opencl/beignet/utests/compiler_global_memory_barrier.cpp,
> > > > > function compiler_global_memory_barrier, line 24
> > > > > 
> > > > > 
> > > > > compiler_local_memory_barrier_wg64:
> > > > >   compiler_local_memory_barrier_wg64()    [FAILED]
> > > > >     Error: dst[i+j] == 63-j
> > > > >   at
> > > > > file
> > > > > /home/works/opencl/beignet/utests/compiler_local_memory_barrier_wg64.cp
> > > > > p, function compiler_local_memory_barrier_wg64, line 42
> > > > > 
> > > > > 
> > > > > Thanks
> > > > >   --Sun, Yi
> > > > > _______________________________________________
> > > > > Beignet mailing list
> > > > > Beignet at lists.freedesktop.org
> > > > > http://lists.freedesktop.org/mailman/listinfo/beignet
> > > > 
> > > 
> > > _______________________________________________
> > > Beignet mailing list
> > > Beignet at lists.freedesktop.org
> > > http://lists.freedesktop.org/mailman/listinfo/beignet
> > 
> > _______________________________________________
> > Beignet mailing list
> > Beignet at lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/beignet
> 
> _______________________________________________
> Beignet mailing list
> Beignet at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/beignet