similar to: [LLVMdev] Lowering switch statements with PGO

> On Mon, Dec 15, 2014 at 9:57 AM, Chad Rosier <mcrosier at codeaurora.org> > wrote: >> All, >> About two months ago I posted a patch that hoisted the hottest case >> statement from a switch statement during ISelLowering. >> >> See: http://reviews.llvm.org/D5786 >> >> Sean was rather adamant about using a Huffman tree (and I agree this is

> On Mon, Dec 15, 2014 at 2:26 PM, Chad Rosier <mcrosier at codeaurora.org> > wrote: >> >> > On Mon, Dec 15, 2014 at 9:57 AM, Chad Rosier <mcrosier at codeaurora.org> >> > wrote: >> >> All, >> >> About two months ago I posted a patch that hoisted the hottest case >> >> statement from a switch statement during

After messing around with this a bit recently, I've come to the following conclusions: 1. One issue is at what granularity to consider the PGO weightings. For example, should a set of 50 consecutive cases with 65% of the total weight distributed approximately evenly and which can be lowered to a table lookup be handled before the 3 remaining cases that 5%, 10%, and 20% probability,

+LLVMDev (prev email hit the wrong list..) All, Do we have any logic that hoists hot case statements out of switches based on profile information? Chad

Thanks alot Chad for quick response. Does this means that, we can not use LLVM optimizations except O1, O2, O3, O4 and unroll-loops with clang? One more thing I would like to know that If I want to process multiple modules with opt at the same time like opt -adce *.bc then how is it possible with opt in one go, if I process all the bytecode files within Makefile. Thanks. Shahzad On Thu, Jun

Hi Shahzad, > Is it possible that we can use LLVM optimization beside O1, O2, O3 > along with dragonegg plugin? sure, try this: gcc -fplugin=path/dragonegg.so ...other_options_here... -S -o - -fplugin-arg-dragonegg-emit-ir -fplugin-arg-dragonegg-llvm-ir-optimize=0 | opt -pass1 -pass2 ... Here -fplugin-arg-dragonegg-emit-ir tells it to output LLVM IR rather than target assembler.

Hi Shahzad, > I tried your method and it works fine. What would be the next step to > produce the final executable? I have tried the following but it is > producing an error > > $ gcc -fplugin=/path/to/dragonegg.so -S *.c > -fplugin-arg-dragonegg-emit-ir | opt -adce this won't work because you aren't passing the IR to opt (you need -o - for that if using a pipe) and you

Hi, > I tried it with -o - but its producing an error > > gcc: fatal error: cannot specify -o with -c, -S or -E with multiple files > > What you suggest? what I wrote: >> for F in *.c ; do B=`basename $F .c` ; gcc -fplugin=/path/to/dragonegg.so >> -S -o - $F -fplugin-arg-dragonegg-emit-ir | opt -adce -o $B.ll ; done >> clang *.ll Thanks to the for loop and

Hello Duncan Is it possible that we can use LLVM optimization beside O1, O2, O3 along with dragonegg plugin? Regards Shahzad On Thu, Jun 7, 2012 at 10:59 PM, Abdul Wahid Memon <engrwahidmemon at gmail.com> wrote: > Thanks alot Chad for these quick and fine responses. > > Regards > > Abdul > > On Thu, Jun 7, 2012 at 10:57 PM, Chad Rosier <mcrosier at apple.com>

Hi, > If I compile the program using the following command line i.e. > > $ clang -O3 -lm *.c this may be doing link time optimization. > > then > > $ time ./a.out > > real 0m2.606s > user 0m2.584s > sys 0m0.012s > > BUT, if I use all the optimizations enabled with -O3 but specify them > explicity i.e. you can just use "opt -O3"

Hello Duncan I tried your method and it works fine. What would be the next step to produce the final executable? I have tried the following but it is producing an error $ gcc -fplugin=/path/to/dragonegg.so -S *.c -fplugin-arg-dragonegg-emit-ir | opt -adce $ clang *.s Regards Shahzad On Fri, Jun 8, 2012 at 9:10 AM, Duncan Sands <baldrick at free.fr> wrote: > Hi Shahzad, > >

Hello Duncan On Fri, Jun 8, 2012 at 2:58 PM, Duncan Sands <baldrick at free.fr> wrote: > Hi Shahzad, > > >> I tried your method and it works fine. What would be the next step to >> produce the final executable? I have tried the following but it is >> producing an error >> >> $ gcc -fplugin=/path/to/dragonegg.so -S *.c >>

Hello I need some help here please. If we compile source files directly in to native code: $ clang -O3 -lm *.c then the runtime is like following real 0m2.807s user 0m2.784s sys 0m0.012s and If we emit LLVM bytcode and apply optimizations $ clang -O3 -c -emit-llvm *.c $ llvm-link *.o -o comb.ll $ time lli ./comb.ll then the runtime is real 0m2.671s user 0m2.640s sys 0m0.020s But, if I

Hi Yes, they both are exactly the same. Regards Shahzad On Tue, Jun 12, 2012 at 9:38 AM, Duncan Sands <baldrick at free.fr> wrote: > Hi, is the comb.ll used here: > > >> $ time lli ./comb.ll >> >> then the runtime is >> >> real    0m2.671s >> user    0m2.640s >> sys     0m0.020s >> >> But, if I convert this same file comb,ll

Use the -I<install path>/include . This directory <install path>/include should look something like: clang/ clang-c/ llvm/ llvm-c/ HTH ashok On 7/9/2012 3:15 PM, NY Knicks Fan wrote: > I downloaded the 3.1 LLVM and Clang sources. > > I followed the directions at: http://clang.llvm.org/get_started.html > > I am able to use Clang to compile stuff, but I could

On 7/9/2012 1:42 PM, Chad Rosier wrote: > Have you looked here: http://clang.llvm.org/get_started.html > > Chad > > > On Jul 9, 2012, at 11:33 AM, NY Knicks Fan wrote: > >> I'm trying to do the tutorials at: >> >> https://github.com/loarabia/Clang-tu...i/TutorialOrig >> <https://github.com/loarabia/Clang-tutorial/wiki/TutorialOrig> >>

The "make install" should collect everything into your <install path>. The <install path> will then have bin, lib and include dirs. On 7/9/2012 3:52 PM, NY Knicks Fan wrote: > Hi Ashok, > > The documentation suggests that I put clang inside of llvm/tools and so > I have two separate include directories. I tried both of them and > neither worked: > > $

Hello Duncan Sorry for the mistake. Actually that error occurred when I was compiling all the files at once, NOT in for loop. The for loop is working perfectly as it is dealing with individual files. I have now one new issue. Let me specify it briefly. If I compile the program using the following command line i.e. $ clang -O3 -lm *.c then $ time ./a.out real 0m2.606s user 0m2.584s sys

Owen/All, I've been working on adding support for reassociation with unsafe math (see: http://reviews.llvm.org/D4129). Do you know if this change, r156323, is still necessary? Specifically, do we need the reassociation pass to canonicalize FP operands for CSE to work effectively? This kinda scares me if it does! :( Side note: Without this canonicalization, I did see a 3% regression in Mesa

Thanks Duncan It was really helpful. Regards Abdul On Fri, Jun 8, 2012 at 7:23 PM, Duncan Sands <baldrick at free.fr> wrote: > Hi, > > >> If I compile the program using the following command line i.e. >> >> $ clang -O3 -lm *.c > > > this may be doing link time optimization. > > >> >> then >> >> $ time ./a.out >>