llvm dev - Jan 2012 - [LLVMdev] [llvm-commits] [PATCH] BasicBlock Autovectorization Pass

On Thu, 2012-01-26 at 14:34 -0600, Sebastian Pop wrote:
> On Tue, Jan 24, 2012 at 6:41 PM, Hal Finkel <hfinkel at anl.gov>
wrote:
> >> enabling vectorization gets the performance down by 80% on ARM.
> >> I will prepare a reduced testcase and try to find out the reason.
> >> As a first shot, I would say that this comes from the
vectorization of
> >> code in a loop and the overhead of transfer between scalar and
> >> vector registers.
> >
> > This is good; as has been pointed out, we'll need to develop a
> > vectorization cost model for this kind of thing to really be
successful,
> > and so we should start thinking about that.
> >
> > The pass, as implemented, has an semi-implicit cost model which says
> > that permutations followed by another vector operation are free,
scalar
> > -> vector transfers are free, and vectorizing a memory operation is
just
> > as good as vectorizing an arithmetic operation. Depending on the
system,
> > these may all be untrue (although on some systems they are true).
> >
> > If you can generate a test case that would be great, I'd like to
look at
> > it.
> 
> Here is the testcase with calls to gettimeofday to measure time spent
> in the kernel and not in the ini/fini phases.
> On ARM I saw around 5 to 6x slowdown in the vector version.
> I haven't tried this on x86 yet but that should also produce slowdowns
> as the cost between scalar and vector regs is non null there as well.
Thanks! Did you compile with any non-default flags other than -mllvm
-vectorize?

 -Hal
> 
> Sebastian
> --
> Qualcomm Innovation Center, Inc is a member of Code Aurora Forum
-- 
Hal Finkel
Postdoctoral Appointee
Leadership Computing Facility
Argonne National Laboratory

On Thu, Jan 26, 2012 at 2:49 PM, Hal Finkel <hfinkel at anl.gov>
wrote:
> Thanks! Did you compile with any non-default flags other than -mllvm
> -vectorize?
I used -O3 and -vectorize, no other non-default flags.

Sebastian
--
Qualcomm Innovation Center, Inc is a member of Code Aurora Forum

On Thu, 2012-01-26 at 15:12 -0600, Sebastian Pop wrote:
> On Thu, Jan 26, 2012 at 2:49 PM, Hal Finkel <hfinkel at anl.gov>
wrote:
> > Thanks! Did you compile with any non-default flags other than -mllvm
> > -vectorize?
> 
> I used -O3 and -vectorize, no other non-default flags.
If I run clang -O3 -mllvm -vectorize -S -emit-llvm -o test.ll test.c
then I get no vectorization at all (the output is identical to that
without the -vectorize). What target triple is your clang targeting?

If I include -mllvm -debug-only=bb-vectorize then the relevant output
is:
BBV: fusing loop #1 for entry in main...
BBV: found 0 instructions with candidate pairs
BBV: done!
BBV: fusing loop #1 for for.body in main...
BBV: found 0 instructions with candidate pairs
BBV: done!
BBV: fusing loop #1 for for.end in main...
BBV: found 0 instructions with candidate pairs
BBV: done!
BBV: fusing loop #1 for for.cond7.preheader in main...
BBV: found 0 instructions with candidate pairs
BBV: done!
BBV: fusing loop #1 for for.body10 in main...
BBV: found 16 instructions with candidate pairs
BBV: found 62 pair connections.
BBV: selected 0 pairs.
BBV: done!
BBV: fusing loop #1 for for.inc45 in main...
BBV: found 0 instructions with candidate pairs
BBV: done!
BBV: fusing loop #1 for for.end47 in main...
BBV: found 3 instructions with candidate pairs
BBV: found 0 pair connections.
BBV: done!

 -Hal
> 
> Sebastian
> --
> Qualcomm Innovation Center, Inc is a member of Code Aurora Forum
-- 
Hal Finkel
Postdoctoral Appointee
Leadership Computing Facility
Argonne National Laboratory