similar to: [LLVMdev] dragonegg vs xplor-nih

Try -mllvm -disable-unroll-scev if you're curious. There can be some luck involved. If you have the bitcode for the important function, I may be able to convert it into a test case to avoid regressing. I usually grab the unoptimized bitcode as follows: -emit-llvm -mllvm -disable-llvm-optzns -o module.bc -Andy On Sep 6, 2011, at 12:03 PM, Owen Anderson wrote: > Seems very likely to be

I'm not certain yet which commit in the last couple of days caused this, but the current llvm/dragonegg svn shows a major improvement in the runtime of the xplor-nih testsuite when xplor-nih is built with FSF gcc 4.6.1 and the dragonegg plugin at -O3 -ffast-math -funroll-loops. Previously the xplor-nih testsuite always executed in ~40 sec but now it is coming it at 34.5 sec which is about the

Seems very likely to be related to Andy's SCEV-unroll-loops changes. --Owen On Sep 6, 2011, at 11:56 AM, Jack Howarth wrote: > I'm not certain yet which commit in the last couple of days caused this, > but the current llvm/dragonegg svn shows a major improvement in the runtime > of the xplor-nih testsuite when xplor-nih is built with FSF gcc 4.6.1 and the > dragonegg plugin

I am happy to report that current llvm/llvm-gcc-4.2 svn builds all of xplor-nih (a complex mix of c, c++ and fortran) with -O3 -fPIC -msse4 -ffast-math. A single fortran file exposes PR3376 which is triggered by -O3 -ffinite-math-only. The resulting build of xplor-nih completely passes its testsuite and compares very well to the same build against gcc trunk for gcc 4.4 in terms of execution time.

Now that dragoneegg is robust in its default usage and the dragonegg svn is moderately stable with -fplugin-arg-dragonegg-enable-gcc-optzns, it is possible to gauge the impact of that feature. Comparing clang 2.9, FSF gcc 4.5.3svn, FSF gcc 4.6.0 and dragonegg svn with FSF gcc 4.5.3svn using the himenoBMTxpa benchmark, the enhancement to code performance from

Hi Jack, > Now that dragoneegg is robust in its default usage and the dragonegg svn > is moderately stable with -fplugin-arg-dragonegg-enable-gcc-optzns, it is > possible to gauge the impact of that feature. Comparing clang 2.9, FSF gcc 4.5.3svn, > FSF gcc 4.6.0 and dragonegg svn with FSF gcc 4.5.3svn using the himenoBMTxpa benchmark, > the enhancement to code performance from

On Fri, Apr 15, 2011 at 08:53:19AM +0200, Duncan Sands wrote: > Hi Jack, > > > Now that dragoneegg is robust in its default usage and the dragonegg svn > > is moderately stable with -fplugin-arg-dragonegg-enable-gcc-optzns, it is > > possible to gauge the impact of that feature. Comparing clang 2.9, FSF gcc 4.5.3svn, > > FSF gcc 4.6.0 and dragonegg svn with FSF

On Thu, Jun 09, 2011 at 08:47:26PM -0400, Jack Howarth wrote: > Duncan, > Here are the complete benchmarks rerun against gcc 4.5.4 built with... > > Using built-in specs. > COLLECT_GCC=gfortran-fsf-4.5 > COLLECT_LTO_WRAPPER=/sw/lib/gcc4.5/libexec/gcc/x86_64-apple-darwin11.0.0/4.5.4/lto-wrapper > Target: x86_64-apple-darwin11.0.0 > Configured with:

Hi Jack, thanks for these numbers. Can you also please measure compile times? I'm thinking of enabling gcc optimizations by default, but I don't want to increase compile times, which means choosing a value for the -fplugin-arg-dragonegg-llvm-ir-optimize option that is low enough to get good compile times, yet high enough to get fast code. It would be great if you could play around with

Current dragonegg svn has all of the -fplugin-arg-dragonegg-enable-gcc-optzns bugs for usage with -ffast-math -O3 addressed except for those related to PR2314. Using the -fno-tree-vectorize option, we can evaluate the current state of -fplugin-arg-dragonegg-enable-gcc-optzns with the Polyhedron 2005 benchmarks compared to stock dragonegg and stock gcc 4.5.4. The runtime benchmarks below show that

Duncan, Here are the complete benchmarks rerun against gcc 4.5.4 built with... Using built-in specs. COLLECT_GCC=gfortran-fsf-4.5 COLLECT_LTO_WRAPPER=/sw/lib/gcc4.5/libexec/gcc/x86_64-apple-darwin11.0.0/4.5.4/lto-wrapper Target: x86_64-apple-darwin11.0.0 Configured with: ../gcc-4.5.4/configure --prefix=/sw --prefix=/sw/lib/gcc4.5 --mandir=/sw/share/man --infodir=/sw/lib/gcc4.5/info

On Thu, Jun 09, 2011 at 03:44:40PM +0200, Duncan Sands wrote: > Hi Jack, thanks for doing this. > >> Below are the tabulated compile times and executable sizes. >> >> A) gcc 4.5.4svn using -msse3 -ffast-math -O3 -fno-tree-vectorize >> B) gcc 4.5.4svn/dragonegg using -msse3 -ffast-math -O3 -fno-tree-vectorize -fplugin-arg-dragonegg-enable-gcc-optzns >> C)

Hi Jack, thanks for doing this. > Below are the tabulated compile times and executable sizes. > > A) gcc 4.5.4svn using -msse3 -ffast-math -O3 -fno-tree-vectorize > B) gcc 4.5.4svn/dragonegg using -msse3 -ffast-math -O3 -fno-tree-vectorize -fplugin-arg-dragonegg-enable-gcc-optzns > C) gcc 4.5.4svn/dragonegg using -msse3 -ffast-math -O3 -fno-tree-vectorize These numbers really

Duncan, Below are the tabulated compile times and executable sizes. A) gcc 4.5.4svn using -msse3 -ffast-math -O3 -fno-tree-vectorize B) gcc 4.5.4svn/dragonegg using -msse3 -ffast-math -O3 -fno-tree-vectorize -fplugin-arg-dragonegg-enable-gcc-optzns C) gcc 4.5.4svn/dragonegg using -msse3 -ffast-math -O3 -fno-tree-vectorize Compile time (seconds) Benchmark A) stock B) gcc 4.5.4/ C)

On Sat, Jun 01, 2013 at 06:45:48AM +0200, Duncan Sands wrote: > > These results are very disappointing, I was hoping to see a big improvement > somewhere instead of no real improvement anywhere (except for gas_dyn) or a > regression (eg: mdbx). I think LLVM now has a reasonable array of fast-math > optimizations. I will try to find time to poke at gas_dyn and induct: since >

Hi Jack, On 29/05/13 22:04, Jack Howarth wrote: > On Wed, May 29, 2013 at 03:25:30PM +0200, Duncan Sands wrote: >> Hi Jack, I pulled the loop vectorizer and fast math changes into the 3.3 branch, >> so hopefully they will be part of 3.3 rc3 (and 3.3 final!). It would be great >> if you could redo the benchmarks rc3. >> > > Duncan, > As requested, appended

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. > > thanks for the numbers. How does this compare to LLVM 3.0 - were there any > regressions? The results from just before

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 Wed, May 29, 2013 at 03:25:30PM +0200, Duncan Sands wrote: > Hi Jack, I pulled the loop vectorizer and fast math changes into the 3.3 branch, > so hopefully they will be part of 3.3 rc3 (and 3.3 final!). It would be great > if you could redo the benchmarks rc3. > Duncan, As requested, appended are the updated Polyhedron 2005 benchmark results with both RC1 and RC3 llvm 3.3

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. thanks for the numbers. How does this compare to LLVM 3.0 - were there any regressions? Ciao, Duncan. The benchmarks > for -msse3 and -msse4 appear identical (at least for degg+optnz). This is fortunate > since