similar to: [LLVMdev] Running the test suite with special opt flags

Hi Hal, I think the testsuite has support for "opt-beta" which means running opt with special flags side by side with the usual opt. TEST.nightly.Makefile turns this on with ENABLE_OPTBETA. I took a peek at Makefile.programs to see how it is implemented, and as far as I can see it can't possibly actually work :( So probably the simplest thing is to run the testsuite once to get a

Hi Hal, > As some of you may know, I committed my basic-block autovectorization > pass a few days ago. I encourage anyone interested to try it out (pass > -vectorize to opt or -mllvm -vectorize to clang) and provide feedback. > Especially in combination with -unroll-allow-partial, I have observed > some significant benchmark speedups, but, I have also observed some > significant

As some of you may know, I committed my basic-block autovectorization pass a few days ago. I encourage anyone interested to try it out (pass -vectorize to opt or -mllvm -vectorize to clang) and provide feedback. Especially in combination with -unroll-allow-partial, I have observed some significant benchmark speedups, but, I have also observed some significant slowdowns. I would like to share my

On Fri, Feb 03, 2012 at 09:49:30AM +0100, Duncan Sands wrote: > Hi Hal, > > > As some of you may know, I committed my basic-block autovectorization > > pass a few days ago. I encourage anyone interested to try it out (pass > > -vectorize to opt or -mllvm -vectorize to clang) and provide feedback. > > Especially in combination with -unroll-allow-partial, I have

Ralf, et al., Attached is the latest version of my autovectorization patch. llvmdev has been CC'd (as had been suggested to me); this e-mail contains additional benchmark results. First, these are preliminary results because I did not do the things necessary to make them real (explicitly quiet the machine, bind the processes to one cpu, etc.). But they should be good enough for discussion.

Carl-Philip, The reason that this does not vectorize is that it cannot vectorize the stores; this leaves only the mul-add chains (and some chains with loads), and they only have a depth of 2 (the threshold is 6). If you give clang -mllvm -bb-vectorize-req-chain-depth=2 then it will vectorize. The reason the heuristic has such a large default value is to prevent cases where it costs more to

On Sat, 2011-10-29 at 12:30 -0500, Hal Finkel wrote: > Ralf, et al., > > Attached is the latest version of my autovectorization patch. llvmdev > has been CC'd (as had been suggested to me); this e-mail contains > additional benchmark results. > > First, these are preliminary results because I did not do the things > necessary to make them real (explicitly quiet the

I have a super-simple test case 4x4 matrix * 4-vector which gets correctly unrolled, but is not vectorized by -bb-vectorize. (I used llvm 3.1svn) I attached the test case so you can see what is going wrong there. 2012/2/3 Hal Finkel <hfinkel at anl.gov> > As some of you may know, I committed my basic-block autovectorization > pass a few days ago. I encourage anyone interested to try

Duncan, I also noticed cases where vector IR is scalariezd by the codegen. From what I have seen (which is based on a different vectorizer with a different code model, etc) there are two main areas for improvements: 1. Complex instructions - Instructions such as shuffles are very sensitive to the ability of the codegen to lower them. If a vectorizer generates shuffle instructions which are not

I will test your suggestion, but I designed the test case to load the memory directly into <4 x float> registers. So there is absolutely no permutation and other swizzle or move operations. Maybe the heuristic should not only count the depth but also the surrounding load/store operations. Are the load/store operations vectorized, too? (I designed the test case to completely fit the SSE

On Tue, 3 Apr 2012 08:57:51 -0400 Jack Howarth <howarth at bromo.med.uc.edu> wrote: > On Tue, Apr 03, 2012 at 09:26:38AM +0200, Duncan Sands wrote: > > Hi Jack, > > > >> Attached are the Polyhedron 2005 benchmark results for current > >> llvm/dragonegg svn on x86_64-apple-darwin11 built against Xcode > >> 4.3.2 and FSF gcc 4.6.3. > > >

On Sat, 2011-10-29 at 14:02 -0500, Hal Finkel wrote: > On Sat, 2011-10-29 at 12:30 -0500, Hal Finkel wrote: > > Ralf, et al., > > > > Attached is the latest version of my autovectorization patch. llvmdev > > has been CC'd (as had been suggested to me); this e-mail contains > > additional benchmark results. > > > > First, these are preliminary

That works. Thank you. Will -vectorize become default later? 2012/2/14 Hal Finkel <hfinkel at anl.gov> > If you run with -vectorize instead of -bb-vectorize it will schedule the > cleanup passes for you. > > -Hal > > *Sent from my Verizon Wireless Droid* > > > -----Original message----- > > *From: *"Carl-Philip Hänsch" <cphaensch at

How can I install the patch? Any step-by-step guide? Thanks, Pablo On 24/11/2011, at 15:33, Hal Finkel wrote: > On Thu, 2011-11-24 at 15:09 +0100, Pablo Barrio López-Cortijo wrote: >> Hi everybody, >> >> I'm trying to run some examples with different backend options (turning >> on/off SSE, 3dnow, or MMX instructions). However, I don't see any >>

On Mon, 2012-02-13 at 11:11 +0100, Carl-Philip Hänsch wrote: > I will test your suggestion, but I designed the test case to load the > memory directly into <4 x float> registers. So there is absolutely no > permutation and other swizzle or move operations. Maybe the heuristic > should not only count the depth but also the surrounding load/store > operations. I've attached

On 28/11/2011, at 18:52, Hal Finkel wrote: > On Mon, 2011-11-28 at 17:49 +0100, Pablo Barrio wrote: >> How can I install the patch? Any step-by-step guide? > > First, grab the trunk versions of llvm and clang. This is detailed on > http://clang.llvm.org/get_started.html > > Then you apply the patch. On a Unix-like system, this is something like: > cd llvm > patch

I've attached the latest version of my autovectorization patch. Working through the test suite has proved to be a productive experience ;) -- And almost all of the bugs that it revealed have now been fixed. There are still two programs that don't compile with vectorization turned on, and I'm working on those now, but in case anyone feels like playing with vectorization, this patch

I applied the patch to the trunk version successfully, although I get an error in between: 1 out of 1 hunk FAILED -- saving rejects to file lib/Transforms/IPO/CMakeLists.txt.rej Can I ignore the error? The patch exits normally except for that error. Also, I tried to apply the patch to the LLVM 3.0 but does not work at all. Is the trunk version the only one where the patch can be applied?

On Thu, 2011-11-24 at 15:09 +0100, Pablo Barrio López-Cortijo wrote: > Hi everybody, > > I'm trying to run some examples with different backend options (turning > on/off SSE, 3dnow, or MMX instructions). However, I don't see any > difference in terms of execution time. Does anybody know which language > constructs (in C/C++ preferably) should I try to make these

On Feb 2, 2012, at 7:56 PM, Hal Finkel wrote: > As some of you may know, I committed my basic-block autovectorization > pass a few days ago. I encourage anyone interested to try it out (pass > -vectorize to opt or -mllvm -vectorize to clang) and provide feedback. > Especially in combination with -unroll-allow-partial, I have observed > some significant benchmark speedups, but, I