On Mon, May 9, 2011 at 8:54 PM, Joseph S. Myers <jos...@codesourcery.com> wrote:
> On Mon, 9 May 2011, H.J. Lu wrote:
>
>> > Index: lib/target-supports.exp
>> > ===================================================================
>> > --- lib/target-supports.exp (revision 173569)
>> > +++ lib/target-supports.exp (working copy)
>> > @@ -3845,6 +3845,8 @@
>> > set dg-do-what-default run
>> > } elseif { [istarget "i?86-*-*"] || [istarget "x86_64-*-*"] } {
>> > lappend DEFAULT_VECTCFLAGS "-msse2"
>> > + # FIXME: Vectorizer testsuite assumes 128bit vector widths.
>> > + lappend DEFAULT_VECTCFLAGS "-mprefer-avx128"
>> > if { [check_effective_target_sse2_runtime] } {
>> > set dg-do-what-default run
>> > } else {
>> >
>>
>> That means 256bit vectorizer won't be tested. I think we
>> should investigate each testcase and update it if needed.
>
> I think we should work out how to get the various vectorizer testsuites to
> run multiple times, with each vector ISA variant that's available on the
> target architecture (so you'd test SSE; 128-bit AVX; 256-bit AVX; and
> maybe other variants - each variant tested with execution testing if
> there's hardware support, compile testing otherwise), like the torture
> testsuites run each test multiple times with different options. Though
> that certainly complicates all the effective target tests for
> vectorization support, since the results may depend on the options as well
> as the target.
Indeed - the above change makes testing with -mavx useless, no? ISTR
there are not many fails with -mavx in the vectorizer testsuite.
Richard.
> --
> Joseph S. Myers
> jos...@codesourcery.com
