[PATCH] [AArch64] Remove NEON from "generic" CPU target

[Jiangning Liu]

Hi Ana,

Thanks for sharing with us your LNT test status.

For long running tests this flow might be ok, but for SingleSource and
> MultiSource tests it caused the model to hang several times.
>

Somehow, I never see any hang for testing SingleSource and MultiSource,
and I use foundation model as well.


>
>
> We changed the flow to compile all tests first and only once ssh to the
> model to run all tests. We were more successful in getting results this way.
>

Yeah. This should be helpful to speed up the test. I ever did this for
our internal AAPCS64 random test this way. For LNT test, we are using two
foundation models in parallel, so the test can finish within 11 hours for
small mode.


>
>
> Here are some findings with SingleSource and MultiSource:
>

I got fewer failures than the list below,



> SingleSource:
>
> Runtime crashes:
>
> 2003-08-11-VaListArg.simple
>
> 2007-03-02-VaCopy.simple
>
> Timeouts: (need the timeout default value to be increased)
>

I changed timelimit to 1000, so we only see smallpt fails on timeout.


> Polybench/stencils/fdtd-2d
>
> Polybench/stencils/fdtd-apml
>
> Polybench/stencils/jacobi-1d-imper
>
> Polybench/stencils/jacobi-2d-imper
>
> Polybench/stencils/seidel-2d
>
> linpack-pc.simple
>
> smallpt.simple
>
>
>

I got 4 failures for single source,



FAIL: SingleSource/Benchmarks/SmallPT/smallpt.execution_time (304 of 606)
> This is a timeout.

FAIL: SingleSource/UnitTests/2003-08-11-VaListArg.execution_time (305 of 606)
FAIL: SingleSource/UnitTests/2007-03-02-VaCopy.execution_time (306 of 606)

> They are about va_xxx lowering, and I already filed a bug at
http://llvm.org/bugs/show_bug.cgi?id=18789

FAIL: SingleSource/UnitTests/AtomicOps.execution_time (307 of 606)
> This is an encoding bug around instruction stlxr.


> MultiSource:
>
> Build failure:
>
> Applications/ClamAV/libclamav_cvd.c - missing libz (we use Linaro GCC
> which does not include zlib).
>

You can simply copy the zlib in foundation model to cross-toolchain sysroot
in your host. It can pass.


> Runtime crashes:
>
> alacconvert-encode.simple
>

I don't see this failure yet. A random failure? I will pay more attention
to this one.


> Timeouts: (need the timeout default value to be increased)
> 
> 
>
> ControlFlow-flt.simple
>
> Symbolics-flt.simple
>
> StatementReordering-flt.simple
>
> LoopRestructuring-flt.simple
>
> NodeSplitting-flt.simple
>
> NodeSplitting-flt.simple
>
> Expansion-flt.simple
>

For multi source, I only got 1 failure,

FAIL: MultiSource/Benchmarks/TSVC/InductionVariable-dbl/InductionVariable-dbl.execution_time
(183 of 362)

> This is a time out. it took extremely long time in foundation model as
I mentioned last time.

Therefore, all failures for LNT AArch64 have nothing to do with NEON.


Thanks,
-Jiangning
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140211/cdab5e28/attachment.html>