search for: llcbeta

...using TCO. I had to fixed a number of subtle bugs to get it working and even now I am not too happy with it. My focus was on finding non-ABI changing automatic tail call cases (aka gcc's sibcall). It's now done so I'll leave -tailcallopt alone for now. I'll run -tailcallopt as x86 llcbeta to see if JIT is indeed broken. Evan On Feb 5, 2010, at 7:32 PM, Albert Graef wrote: > Evan Cheng wrote: >> As far as I can tell only PPC and X86 targets are supporting this option. Does anyone actually using it? I'd prefer to just remove it to clean up the implementation if no on...

Ok. Like I said, I am not too concerned with one or two llcbeta failures. That's expected. Thx, Evan On Oct 5, 2007, at 3:18 PM, Arnold Schwaighofer <arnold.schwaighofer at gmail.com > wrote: > Another round ;) Sadly not beer. > > tested against r42630 darwin-x86: > SPASS llc-beta still fails > > thanks for the patience >...

...imization to x86. This is different from what -tailcallopt does, which forces fastcc function to be tail callable. My changes detect opportunities to do tail call without having to change the ABI. I've looked at the codegen of -tailcallopt and it doesn't look all that good. Running it as a llcbeta option shows it significantly pessimize code in most cases. As far as I can tell only PPC and X86 targets are supporting this option. Does anyone actually using it? I'd prefer to just remove it to clean up the implementation if no one has any objections. Evan

...the X86CallingConv.td file has: class CCIfSubtarget<string F, CCAction A> : CCIf<!strconcat("State.getTarget().getSubtarget<X86Subtarget>().", F), A>; It would be straight-forward to have a CCIf defined to check some command line argument. I think enabling this as llcbeta for a few nights makes sense before turning it on by default. -Chris -- http://nondot.org/sabre/ http://llvm.org/

...42486). The same number (46) of failing test cases on patched version and vanilla version. LCCBETA was enabled on both. I changed the LCCBETA option in Makefile.programs in the patched version to include the tail-call-opt flags so that the optimization code gets exercised. ifeq ($(ARCH),x86) LLCBETAOPTION := -regalloc=local -fast -tail-call-opt -tail-call-opt- align-stack On 26 Sep 2007, at 02:26, Chris Lattner wrote: > I think enabling this as llcbeta for a few nights makes > sense before turning it on by default. > > -Chris What does turning on by default mean? Does that mean...

Hi John, Unless anyone else has any comments. I recommend you check in the PEI patches so we can start testing it as llcbeta. Chris, should I commit the patch for John or can you grant him commit access (since part of his patch has already been committed)? Evan On Mar 16, 2009, at 3:23 PM, John Mosby wrote: > Here is the latest shrink wrapping patch, with fixes for issues > identified by Evan. > > I...

Evan Cheng wrote: > I'll run -tailcallopt as x86 llcbeta to see if JIT is indeed broken. Jeffrey apparently fixed that now, I'm going to test against trunk asap. Albert -- Dr. Albert Gr"af Dept. of Music-Informatics, University of Mainz, Germany Email: Dr.Graef at t-online.de, ag at muwiinfa.geschichte.uni-mainz.de WWW: http://www.musiki...

...llvm test suite? > No not yet. Ok, I don't expect everything would just work. But please run at least SingleSource and MultiSource and let us know what breaks. Once it's proven to pass a good chunk of the tests it can be checked in but turned off. This will allow us to test it as llcbeta. > > Expect to see a new patch after i have done the testing with the > whole test suite. Sounds good. Thanks. Evan > > > regards arnold > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm....

...st anymore (darwin-x86). if/when someone has the time to look at it - i would be happy to mail it. i think i should no longer clobber the mailing list with my patches :) regards arnold On 7 Oct 2007, at 00:04, Evan Cheng wrote: > Ok. Like I said, I am not too concerned with one or two llcbeta > failures. That's expected. > > Thx, > > Evan >

...to x86. This is different from what -tailcallopt does, which forces fastcc function to be tail callable. My changes detect opportunities to do tail call without having to change the ABI. > > I've looked at the codegen of -tailcallopt and it doesn't look all that good. Running it as a llcbeta option shows it significantly pessimize code in most cases. > > As far as I can tell only PPC and  X86 targets are supporting this option. Does anyone actually using it?  I'd prefer to just remove it to clean up the implementation if no one has any objections. > > Evan > ________...

Here is the latest shrink wrapping patch, with fixes for issues identified by Evan. I am including a few small additions/fixes to include/llvm/ADT/{SparseBitVector,DepthFirstIterator}.h. Files: include/llvm/ADT/DepthFirstIterator.h include/llvm/ADT/SparseBitVector.h lib/CodeGen/PrologEpilogInserter.cpp Evan, let me know how it looks when you get a chance. Thanks much, John >

> Ok, I don't expect everything would just work. But please run at least > SingleSource and MultiSource and let us know what breaks. Once it's > proven to pass a good chunk of the tests it can be checked in but > turned off. This will allow us to test it as llcbeta. > I just ran the test suite in llvm-test with a vanilla copy of the svn tree on darwin-x86. To perform the tests i downloaded the test-suite as instructed on the web. cd'ed into llvm-test. configured (--withllvmfoo etc) and make'd ;). What worries me is that 148 tests are failing. to de...

...vanilla version. LCCBETA was enabled on both. I changed the LCCBETA > option in Makefile.programs in the patched version to include the > tail-call-opt flags so that the optimization code gets exercised. Ok. I'll review the patch soon. Thanks. > > > ifeq ($(ARCH),x86) > LLCBETAOPTION := -regalloc=local -fast -tail-call-opt -tail-call-opt- > align-stack Please remove -regalloc=local -fast. We want to test this patch separately. Can you explain the advantages / disadvantages of -tail- call-opt-align-stack? > > > On 26 Sep 2007, at 02:26, Chris Lattner wrot...

...compile > time > for the test. The better way is to add a custom report for this. See TEST.llc.Makefile and TEST.llc.report under llvm-test as an example. You can do make TEST=llc to generate custom reports. If this change can somehow be optionally controlled, then you can set it as llcbeta option and write a custom report that compare the time spend in some particular passes (TEST.beta-compare is an example, but it comparesf number instructions). The biggest problem is right now time spent in scheduling is lumped into "DAG->DAG Instruction Selection" time becaus...

> > ifeq ($(ARCH),x86) > > LLCBETAOPTION := -regalloc=local -fast -tail-call-opt -tail-call-opt- > > align-stack okay i ll do another round of testing with LLCBETAOPTION := -tail-call-opt -tail-call-opt-align-stack > Please remove -regalloc=local -fast. We want to test this patch > separately. Can you explain the advant...

On Saturday 22 December 2007 02:33, Evan Cheng wrote: > > After some very simple testing, I see slowdowns of around 1.7%. I > > assume > > this is ok, but want to check. > > Can you clarify? Is this 1.7% slowdown of scheduling time or overall > codegen time? If it's the later, then it seems a bit too much. Also, > please test it with all the

...load/store sequence only for + // arguments which would be overwritten otherwise Please capitalize "copy", "this", etc. Also please end sentences with proper punctuation. :-) Overall, a very good patch. Please also incorporate the minor changes mentions in other emails (LLCBETAOPTION, -tail-call-opt-align-stack). I think it'll be ready for svn comit after one more iteration. :-) Evan On Oct 2, 2007, at 2:27 AM, Arnold Schwaighofer wrote: > Hi all, > > I changed the code that checks whether a tail call is really > eligible for optimization so that it...

...; class CCIfSubtarget<string F, CCAction A> > : CCIf<!strconcat("State.getTarget().getSubtarget<X86Subtarget> > ().", F), A>; > > It would be straight-forward to have a CCIf defined to check some > command > line argument. I think enabling this as llcbeta for a few nights > makes > sense before turning it on by default. No not directly. The code related to "caller/callee cleans arguments off the stack" is not controlled by the .td. It's controlled in code by the operands of CALLSEQ_END. for example in SDOperand X86TargetLo...

Begin forwarded message: > From: Evan Cheng <evan.cheng at apple.com> > Date: 11 September 2007 19:26:39 GMT+02:00 > To: LLVM Developers Mailing List <llvmdev at cs.uiuc.edu> > Subject: Re: [LLVMdev] RFC: Tail call optimization X86 > Reply-To: LLVM Developers Mailing List <llvmdev at cs.uiuc.edu> > > Hi Arnold, > > Thanks for the patch. Some questions

...This seem to be doing traversal at least one too many times? Can this be improved? 11. Can you explain a bit more about AnticIn, AvailIn, etc.? 12. Let's worry about edge splitting for a later time. :-) 13. After the code is cleaned up, we should consider checking it in and try it out as llcbeta. Do you have any idea of its compile time impact? Thanks, Evan On Mar 4, 2009, at 7:57 PM, John Mosby wrote: > Here is an updated patch for shrink wrapping with: > > - spills/restores done with stack slot stores/loads > - stack adjustment removed > - refactoring (but still in n...