similar to: [LLVMdev] Reassociating expressions involving GEPs

On Fri, Jan 30, 2009 at 3:03 PM, Stefanus Du Toit <stefanus.dutoit at rapidmind.com> wrote: > The computation of %base then becomes loop-invariant and can be lifted out. > > What's the best way to add this optimization to LLVM? Probably the best place is LICM itself... only loop transformations are aware whether something is loop-invariant. Although, I'm not completely

On 30-Jan-09, at 6:14 PM, Eli Friedman wrote: > On Fri, Jan 30, 2009 at 3:03 PM, Stefanus Du Toit > <stefanus.dutoit at rapidmind.com> wrote: >> The computation of %base then becomes loop-invariant and can be >> lifted out. >> >> What's the best way to add this optimization to LLVM? > > Probably the best place is LICM itself... only loop

On Feb 25, 2009, at 12:12 PM, Stefanus Du Toit wrote: >> Although, I'm not completely sure the transformation is safe, at >> least >> the way you're stating it; unlike add, GEP has undefined overflow, so >> this isn't right in cases like %call == %tmp4 == INT_MIN. > > Hmm, you raise a good point. There's a similar issue even without > overflow,

On 9-Mar-09, at 12:14 AM, Chris Lattner wrote: > GEP overflow is undefined, but this sentence means that *accesses* to > an array must be within its bounds. It is fine to GEP outside the > array as long as you readjust the pointer back before access. Thanks, that was my hope. Maybe this should be spelled out in the langref, especially since it's inconsistent with C, where merely

Our frontend can guarantee that loads from globals are rematerializable and do not alias with any stores in any function in the given module. We'd like the optimization passes (and ideally the register allocator as well) to be able to use this fact. The globals are not constant "forever" but are constant during the calling of any given function in the module. There seem to

Hi Stefanus, Thanks for verifying this. Could you patch this or should I open a new bug report and find a generic solution first? Cheers, Nicolas From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu] On Behalf Of Stefanus Du Toit Sent: woensdag 1 april 2009 18:59 To: LLVM Developers Mailing List Subject: Re: [LLVMdev] Shuffle combine On 1-Apr-09, at 12:42

For those curious about uses of LLVM, we just officially announced our adoption of LLVM in our products: http://www.rapidmind.com/News-Nov10-08-LLVM-OpenCL.php Thanks for all the support so far on here, we look forward to continuing to work with LLVM! -- Stefanus Du Toit <stefanus.dutoit at rapidmind.com> RapidMind Inc. phone: +1 519 885 5455 x116 -- fax: +1 519 885 1463

Hi Nicolas, On 2-Apr-09, at 6:04 PM, Nicolas Capens wrote: > Thanks for verifying this. Could you patch this or should I open a > new bug report and find a generic solution first? I don't have write access so the best I could do would be to submit a patch, and I'm crazy busy at the moment. I actually think the check I described below is fine and would fix this bug (but

Hi Stefanus, Thanks for the info. I still think it's a bug though. Take for example a case where the vectors each have four elements. The values in Mask[] can range from 0 to 7, while HLSMask only has 4 elements. So LHSMask[Mask[i]] can go out of bounds, no? Cheers, Nicolas From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu] On Behalf Of Stefanus Du

The llvm-dev archives (and other llvm/clang mailing list archives) on the web don't seem to have any new messages since some time Monday night. Stefanus -- Stefanus Du Toit <stefanus.dutoit at rapidmind.com> RapidMind Inc. phone: +1 519 885 5455 x116 -- fax: +1 519 885 1463

On 1-Apr-09, at 12:42 PM, Nicolas Capens wrote: > Hi Stefanus, > > Thanks for the info. I still think it’s a bug though. Take for > example a case where the vectors each have four elements. The values > in Mask[] can range from 0 to 7, while HLSMask only has 4 elements. > So LHSMask[Mask[i]] can go out of bounds, no? Good point! One easy way to fix this would be to use:

On Nov 10, 2008, at 12:01 PM, Stefanus Du Toit wrote: > For those curious about uses of LLVM, we just officially announced our > adoption of LLVM in our products: > > http://www.rapidmind.com/News-Nov10-08-LLVM-OpenCL.php I'm thrilled to read an official announcement of that! Do you use LLVM only for static code generation or are you also doing any late (e.g.,

On 31-Jul-08, at 2:38 PM, Dan Gohman wrote: > On Jul 31, 2008, at 7:22 AM, Nicolas Capens wrote: >> In the same breath I’d also like to kindly ask if someone could have >> a look at the reverse operations, namely trunk from 128 to 64 bit >> using movdq2q, and 128 to 32 and 64 to 32 using movd. This also >> seems related to Bug 2585. Thanks again. > > The operations

Does 'make check' allow just running on a particualar directory of tests ? Many thanks in advance, Aaron -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20090617/e0dc48e4/attachment.html>

On 21-Jul-08, at 7:33 PM, Eli Friedman wrote: > On Mon, Jul 21, 2008 at 1:21 PM, Stefanus Du Toit > <stefanus.dutoit at rapidmind.com> wrote: >> 1) Vector shl, lshr, ashr >> > I have a rough draft of a patch for this that works reasonably well > for simple cases... I don't think I really have the time to finish it > properly, but I'll clean it up a bit and

Hi all, I've located a regression that causes my project to crash. It's in revision 67979, where PointerIntPair is changed from storing the integer in the upper bits instead of the lower bits. My project is an experimental JIT-compiler in Windows. So I was wondering if anyone had any clue why the new PointerIntPair implementation might fail. It doesn't seem very safe to me to

Stefanus Du Toit wrote: > The llvm-dev archives (and other llvm/clang mailing list archives) on > the web don't seem to have any new messages since some time Monday > night. > It seems to be working for me. Does it work for you now? -- John T. > Stefanus > > -- > Stefanus Du Toit <stefanus.dutoit at rapidmind.com> > RapidMind Inc. > phone: +1

Hi all, I'm having some trouble understanding the following lines in InstructionCombining.cpp, which possibly contain a bug: if (Mask[i] >= 2*e) NewMask.push_back(2*e); else NewMask.push_back(LHSMask[Mask[i]]); When Mask[i] is bigger than the size of LHSMask it reads out of bounds on that last line. I believe the first line is there to try to prevent that but then it

Hi Stefanus, I'm not if using MMX instructions when doing operations on 64-bit vectors is so terrible? With x86-64 you have double the registers, but it comes at the cost of longer instruction encodings. So there's probably no benefit using SSE. Or am I missing something? Cheers, Nicolas -----Original Message----- From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at

From http://www.llvm.org/docs/TestingGuide.html#quickdejagnu : To run only a subdirectory of tests in llvm/test using DejaGNU (ie. Transforms), just set the TESTSUITE variable to the path of the subdirectory (relative to llvm/test): % gmake TESTSUITE=Transforms check On 17-Jun-09, at 1:33 PM, Aaron Gray wrote: > Does 'make check' allow just running on a particualar directory of