similar to: [LLVMdev] [PATCH] Split LoopUnroll pass into mechanism and policy

Hi, please find an updated patch attached that incorporates the (trivial) changes introduced by r50696 and which makes this patch apply cleanly again. Gr. Matthijs -------------- next part -------------- A non-text attachment was scrubbed... Name: split-unroll.diff Type: text/x-diff Size: 14660 bytes Desc: not available URL:

Hello Matthijs, Separating mechanism from policy is a good thing for the LoopUnroll pass. Instead of moving the policy to a subclass though, I think it'd be better to move the mechanism, the unrollLoop function, out to be a standalone utility function, with the LoopInfo object passed in explicitly. FoldBlockIntoPredecessor would also be good to make into a standalone utility function, since

Hi All, the attached patch performs the splitting in the proposed manner. before applying the patch, please execute svn cp lib/Transforms/Scalar/LoopUnroll.cpp lib/Transforms/Utils/UnrollLoop.cpp to make the patch apply and preserve proper history. Transforms/Utils/UnrollLoop.cpp contains the unrollLoop function, which is now used by the LoopUnroll pass. I've also moved the

Hello Matthijs, On May 9, 2008, at 3:47 AM, Matthijs Kooijman wrote: > Hi All, > > the attached patch performs the splitting in the proposed manner. > before applying the patch, please execute > svn cp lib/Transforms/Scalar/LoopUnroll.cpp lib/Transforms/Utils/ > UnrollLoop.cpp > to make the patch apply and preserve proper history. > > Transforms/Utils/UnrollLoop.cpp

Hi Dan, please find an updated patch attached (don't forget to run the svn cp command I posted earlier, before applying this patch). I've incorporated all the changes we've discussed. Gr. Matthijs -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL:

On May 6, 2008, at 8:30 PM, Devang Patel wrote: > Dan, > > On May 6, 2008, at 5:56 PM, Dan Gohman wrote: > >> This should still make it easy to write new unrolling passes with >> custom heuristics, but it would also be more flexible for passes to >> do unrolling in combination with other transformations. > > Do you mean unrolling during other transformation or

Hi Dan, > in the AMD 5.5 What's that? Got a link? Google only finds some kind of golf cart powered by the "Advanced Motors & Drivers 5.5" engine :-) Gr. Matthijs -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL:

On May 13, 2008, at 3:17 AM, Matthijs Kooijman wrote: > Hi Dan, > > please find an updated patch attached (don't forget to run the svn > cp command > I posted earlier, before applying this patch). > > I've incorporated all the changes we've discussed. Looks good. I just committed it as r51083. Thanks, Dan

On May 7, 2008, at 1:08 AM, Matthijs Kooijman wrote: > Hi Dan, > >> This should still make it easy to write new unrolling passes with >> custom heuristics, but it would also be more flexible for passes to >> do unrolling in combination with other transformations. > Agreed. > > To shed some light on my specific requirements, I am working with an > architecture

Hi Dan, > This should still make it easy to write new unrolling passes with > custom heuristics, but it would also be more flexible for passes to > do unrolling in combination with other transformations. Agreed. To shed some light on my specific requirements, I am working with an architecture that has only limited support for loops and control flow. This means that loop unrolling is not

Hi Dan, > I'm not sure that's very pretty either :-}. My sense is that it's > better to just duplicate those few lines. Then I'll just leave it at that. > An assert at the end that checks that if the loop > wasn't completely unrolled that it's still isLCSSAForm would be a > good addition in any case. Good idea. > I don't think it's desirable

Dan, On May 6, 2008, at 5:56 PM, Dan Gohman wrote: > This should still make it easy to write new unrolling passes with > custom heuristics, but it would also be more flexible for passes to > do unrolling in combination with other transformations. Do you mean unrolling during other transformation or before/after other transformation ? Do you have specific example in mind ? Thanks, -

On Aug 6, 2012, at 6:04 AM, Jorge Navas <navas at comp.nus.edu.sg> wrote: > > Hello, > > I wrote my own pass which needs to do some loop unrolling. > > I can perform loop unrolling via opt: > > opt -mem2reg -loops -loop-simplify -loop-rotate -lcssa -loop-unroll > -unroll-count=50 mytest.bc -o mytest.bc > > This command works perfectly. > >

Hello, I wrote my own pass which needs to do some loop unrolling. I can perform loop unrolling via opt: opt -mem2reg -loops -loop-simplify -loop-rotate -lcssa -loop-unroll -unroll-count=50 mytest.bc -o mytest.bc This command works perfectly. However, what I really want is to produce the **same behavior** but from my own pass (i.e., I don't want to use opt). I wrote a Module pass which

> However, the proper include is #include <climits>, since we're in c++, not C. > > Could someone add this include to lib/Transforms/Scalar/LoopUnroll.cpp ? Done (r51145). Maybe this should go into the release too. Ciao, Duncan.

Hi, > > Could someone add this include to lib/Transforms/Scalar/LoopUnroll.cpp ? > Done (r51145). Maybe this should go into the release too. I don't think the unroll changes ended up in the release, so this fix shouldn't need to either. Gr. Matthijs -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature

Hi all, I was looking around in the lib/Transforms/Utils dir and ran into the BasicInliner class. Seeing a large similarity between it and the SimpleInliner Pass, I grep'd around for BasicInliner, but it doesn't seem to be used. The difference between them seems that SimpleInliner is a pass that looks at all functions, while BasicInliner is not a pass and looks only at explicitely given

Hi, I am trying to use llvm to translate ftp.c(a file of wget-1.11.2.tar.bz2 ) to msil. But it shows errors like: [dayin at localhost src]$ llvm-gcc -emit-llvm ftp.c -c -o ftp.bc [dayin at localhost src]$ llc ftp.bc -march=msil -o ftp.s Intrinsic ID = 32 llc: /home/dayin/llvm/llvm-2.2/lib/Target/MSIL/MSILWriter.cpp:802: void <unnamed>::MSILWriter::printIntrinsicCall(const

I tried to compile LLVM on Fedora 9 with GCC 4.3. The compilation fails because UINT_MAX is not defined in file lib/Transforms/Scalar/LoopUnroll.cpp. I included <limits.h> and everything works. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20080515/4ba7997d/attachment.html>

Hi, I recently cleaned up the includes from the loop unrolling code, and this also removed this include. It seems it's included implicitely somehow here, because I have not problems compiling (running gcc 4.1). However, the proper include is #include <climits>, since we're in c++, not C. Could someone add this include to lib/Transforms/Scalar/LoopUnroll.cpp ? Gr. Matthijs