Page MenuHomePhabricator

chatur01 (Charlie Turner)Email Not Verified
User

Projects

User does not belong to any projects.

User Details

User Since
Oct 7 2014, 8:15 AM (258 w, 1 d)

Recent Activity

Jan 4 2016

chatur01 added a comment to D15774: [PEI] Calculate stack protector frame offset before local stack allocation..

Ping after holiday season :)

Jan 4 2016, 1:54 AM

Dec 24 2015

chatur01 added reviewers for D15774: [PEI] Calculate stack protector frame offset before local stack allocation.: andrew, kristof.beyls.
Dec 24 2015, 11:00 AM
chatur01 updated D15774: [PEI] Calculate stack protector frame offset before local stack allocation..
Dec 24 2015, 10:56 AM
chatur01 retitled D15774: [PEI] Calculate stack protector frame offset before local stack allocation. from to [PEI] Calculate stack protector frame offset before local stack allocation..
Dec 24 2015, 10:56 AM
chatur01 added a comment to D15559: [SCEVExpander] Make findExistingExpansion smarter.

Current algorithm doesn't care about where does trip count's division come from.
That means, if TripCount is divided by some value in loop's preheader, compiler will give up doing unrolling. (Even if IV's step is one or minus one.)

If IV's step is constant likes one or minus one or multiple of 2, we don't need to generate division for computing trip count.

Dec 24 2015, 2:33 AM

Dec 22 2015

chatur01 added a comment to D15702: [AArch64] Optimize some simple TBZ/TBNZ cases..

Hi Geoff, looks good. :-)

Dec 22 2015, 10:57 AM

Dec 18 2015

chatur01 added a comment to D15559: [SCEVExpander] Make findExistingExpansion smarter.

I forgot to mention that I did see the following improvements in A57-T32 for spec2006:

Dec 18 2015, 1:52 AM
chatur01 added a comment to D15559: [SCEVExpander] Make findExistingExpansion smarter.

I've just seen the following regressions for spec2006 on the A57 in T32 mode,

Dec 18 2015, 1:43 AM

Dec 17 2015

chatur01 added a comment to D15559: [SCEVExpander] Make findExistingExpansion smarter.

Hi Junmo,

Dec 17 2015, 11:06 AM

Dec 16 2015

chatur01 added a comment to D15580: [SLPVectorizer] Ensure dominated reduction values..

Thanks for the review Nadav!

Dec 16 2015, 10:27 AM
chatur01 committed rL255792: [SLPVectorizer] Ensure dominated reduction values..
[SLPVectorizer] Ensure dominated reduction values.
Dec 16 2015, 10:27 AM
chatur01 closed D15580: [SLPVectorizer] Ensure dominated reduction values. by committing rL255792: [SLPVectorizer] Ensure dominated reduction values..
Dec 16 2015, 10:27 AM
chatur01 retitled D15580: [SLPVectorizer] Ensure dominated reduction values. from to [SLPVectorizer] Ensure dominated reduction values..
Dec 16 2015, 9:43 AM
chatur01 committed rL255783: [NFC] Update horizontal reduction test cases..
[NFC] Update horizontal reduction test cases.
Dec 16 2015, 9:25 AM

Nov 27 2015

chatur01 added a comment to D14866: [zorg] Add support for uploading artifacts to the 'llvmlab bisect' bucket and enable this for clang-cmake-mips..

Hi Daniel,

Nov 27 2015, 3:23 AM

Nov 26 2015

chatur01 committed rL254179: [LoopVectorize] Use MapVector rather than DenseMap for MinBWs..
[LoopVectorize] Use MapVector rather than DenseMap for MinBWs.
Nov 26 2015, 12:42 PM
chatur01 closed D14981: [LoopVectorize] Use MapVector rather than DenseMap for MinBWs. by committing rL254179: [LoopVectorize] Use MapVector rather than DenseMap for MinBWs..
Nov 26 2015, 12:42 PM
chatur01 updated the diff for D14981: [LoopVectorize] Use MapVector rather than DenseMap for MinBWs..

Added a test-case. This might violate the test guidelines of only comitting robust test cases. By definition I can't do that for a non-determinism bug, but I'm running the test case 5 times, which at least locally was always enough to catch the non-determinism on my computer, of course I can't guarantee that.

Nov 26 2015, 2:57 AM

Nov 25 2015

chatur01 updated the diff for D14981: [LoopVectorize] Use MapVector rather than DenseMap for MinBWs..

Nice one David! This saved me from publishing a nasty hack to maintain my own vector in LoopVectorize.

Nov 25 2015, 11:00 AM
chatur01 abandoned D14981: [LoopVectorize] Use MapVector rather than DenseMap for MinBWs..

Yep, thanks Silviu. I messed this up. I fixed the small testcase in the PR (by luck) and not the larger ones I have access to. Apologies.

Nov 25 2015, 4:06 AM
chatur01 retitled D14981: [LoopVectorize] Use MapVector rather than DenseMap for MinBWs. from to [LoopVectorize] Use std::map rather than DenseMap for MinBWs..
Nov 25 2015, 3:39 AM

Nov 18 2015

chatur01 added a comment to D14775: [FunctionAttrs] Provide a mechanism for adding function attributes from the command line.

Think you'll want a bit of extra checking here. What if I pass -force-attribute foo: by accident? Will get a segfault.

Nov 18 2015, 7:26 AM
chatur01 added a comment to D14775: [FunctionAttrs] Provide a mechanism for adding function attributes from the command line.

Hi James, this LGTM.
One nit inline. It would be nice if you could use a StringRef in in cl::list, but that doesn't seem possible.

Nov 18 2015, 7:12 AM

Nov 17 2015

chatur01 added a comment to D14743: [ARM] Don't pessimize i32 vselect..

I missed some mechanical changes in test/Analysis/CostModel/ARM/select.ll in this review. I committed the changes along with what's in this revision, which I felt were appropriate for post-commit review if necessary. Sorry about that!

Nov 17 2015, 9:30 AM
chatur01 committed rL253349: [ARM] Don't pessimize i32 vselect..
[ARM] Don't pessimize i32 vselect.
Nov 17 2015, 9:28 AM
chatur01 closed D14743: [ARM] Don't pessimize i32 vselect. by committing rL253349: [ARM] Don't pessimize i32 vselect..
Nov 17 2015, 9:28 AM
chatur01 added a comment to D14743: [ARM] Don't pessimize i32 vselect..

Thanks Renato. I double-checked A53 T32 and found no regressions in SPEC 2000, 2006 or LNT.

Nov 17 2015, 9:16 AM
chatur01 committed rL253334: [ARM] Match VABDL from log2 shuffles..
[ARM] Match VABDL from log2 shuffles.
Nov 17 2015, 5:24 AM
chatur01 closed D14664: [ARM] Match VABDL from log2 shuffles. by committing rL253334: [ARM] Match VABDL from log2 shuffles..
Nov 17 2015, 5:24 AM
chatur01 retitled D14743: [ARM] Don't pessimize i32 vselect. from to [ARM] Don't pessimize i32 vselect..
Nov 17 2015, 1:10 AM

Nov 16 2015

chatur01 updated the diff for D14664: [ARM] Match VABDL from log2 shuffles..

Made the patterns a bit nicer with the help of a PatFrag

Nov 16 2015, 3:20 AM

Nov 13 2015

chatur01 retitled D14664: [ARM] Match VABDL from log2 shuffles. from to [ARM] Match VABDL from log2 shuffles..
Nov 13 2015, 10:42 AM

Nov 11 2015

chatur01 committed rL252733: [SLP] Enable -slp-vectorize-hor by default..
[SLP] Enable -slp-vectorize-hor by default.
Nov 11 2015, 7:06 AM

Nov 9 2015

chatur01 committed rL252465: [AArch64] Add UABDL patterns for log2 shuffle..
[AArch64] Add UABDL patterns for log2 shuffle.
Nov 9 2015, 5:13 AM
chatur01 closed D14208: [AArch64] Add UABDL patterns for log2 shuffle..
Nov 9 2015, 5:13 AM
chatur01 committed rL252464: [AArch64] Handle extract_subvector(..., 0) in ISel..
[AArch64] Handle extract_subvector(..., 0) in ISel.
Nov 9 2015, 4:47 AM
chatur01 closed D14207: [AArch64] Handle extract_subvector(..., 0) in ISel..
Nov 9 2015, 4:47 AM
chatur01 added a comment to D14207: [AArch64] Handle extract_subvector(..., 0) in ISel..

Ping

Nov 9 2015, 2:26 AM
chatur01 added a reviewer for D14207: [AArch64] Handle extract_subvector(..., 0) in ISel.: jmolloy.
Nov 9 2015, 2:25 AM

Oct 30 2015

chatur01 updated D14208: [AArch64] Add UABDL patterns for log2 shuffle..
Oct 30 2015, 9:36 AM
chatur01 retitled D14208: [AArch64] Add UABDL patterns for log2 shuffle. from to [AArch64] Add UABDL patterns for log2 shuffle..
Oct 30 2015, 9:36 AM
chatur01 retitled D14207: [AArch64] Handle extract_subvector(..., 0) in ISel. from to [AArch64] Handle extract_subvector(..., 0) in ISel..
Oct 30 2015, 9:31 AM

Oct 27 2015

chatur01 committed rL251428: [SLP] Be more aggressive about reduction width selection..
[SLP] Be more aggressive about reduction width selection.
Oct 27 2015, 11:01 AM
chatur01 closed D14116: [SLP] Be more aggressive about reduction width selection. by committing rL251428: [SLP] Be more aggressive about reduction width selection..
Oct 27 2015, 11:01 AM
chatur01 committed rL251425: [SLP] Try a bit harder to find reduction PHIs.
[SLP] Try a bit harder to find reduction PHIs
Oct 27 2015, 10:56 AM
chatur01 closed D14063: [SLP] Try a bit harder to find reduction PHIs by committing rL251425: [SLP] Try a bit harder to find reduction PHIs.
Oct 27 2015, 10:56 AM
chatur01 committed rL251424: [SLP] Treat SelectInsts as reduction values..
[SLP] Treat SelectInsts as reduction values.
Oct 27 2015, 10:51 AM
chatur01 closed D13949: [SLP] Treat SelectInsts as reduction values. by committing rL251424: [SLP] Treat SelectInsts as reduction values..
Oct 27 2015, 10:51 AM
chatur01 retitled D14116: [SLP] Be more aggressive about reduction width selection. from to [SLP] Be more aggressive about reduction width selection..
Oct 27 2015, 6:52 AM
chatur01 updated the diff for D14063: [SLP] Try a bit harder to find reduction PHIs.

Address review comments.
Thanks again Nadav!

Oct 27 2015, 4:43 AM
chatur01 added a comment to D14082: [ARM] Expand ROTL and ROTR of vector value types.

Charlie - thanks for dealing with this - I'm surprised it didn't turn up in the fuzz testing that I've been doing recently (the XOP rotation lowering definitely caught several times). I agree with Tim that making Expand the default is probably the way forward.

Oct 27 2015, 3:29 AM
chatur01 committed rL251401: [ARM] Expand ROTL and ROTR of vector value types.
[ARM] Expand ROTL and ROTR of vector value types
Oct 27 2015, 3:28 AM
chatur01 closed D14082: [ARM] Expand ROTL and ROTR of vector value types.
Oct 27 2015, 3:27 AM

Oct 26 2015

chatur01 updated the diff for D13949: [SLP] Treat SelectInsts as reduction values..

Address review comments.

Oct 26 2015, 11:30 AM
chatur01 updated the diff for D14082: [ARM] Expand ROTL and ROTR of vector value types.

Add the same expansions on AArch64 as well. I missed that it was also bumping into this.

Oct 26 2015, 11:02 AM
chatur01 added a comment to D14082: [ARM] Expand ROTL and ROTR of vector value types.

Hi Tim,

Oct 26 2015, 10:47 AM
chatur01 retitled D14082: [ARM] Expand ROTL and ROTR of vector value types from to [ARM] Expand ROTL and ROTR of vector value types.
Oct 26 2015, 10:03 AM
chatur01 retitled D14063: [SLP] Try a bit harder to find reduction PHIs from to [SLP] Try a bit harder to find reduction PHIs.
Oct 26 2015, 6:35 AM
chatur01 added a comment to D13949: [SLP] Treat SelectInsts as reduction values..

Hi Nadav, I've anaysed my patches in more detail now. Sorry for how long it took me to do it!

Oct 26 2015, 6:15 AM

Oct 21 2015

chatur01 added a comment to D13949: [SLP] Treat SelectInsts as reduction values..

Do you know how effective is this pattern? Do you know how many times it hits when you compile the llvm test suite?

My testing has been on ARM and AArch64 targets, this patch has no effect on performance yet for the benchmarks I have access to. To vectorize the more general pattern, there are few more changes I need to propose, including searching for reductions in multi-block loops, and trying different reduction widths. I was going to go through them one at a time.

Oct 21 2015, 10:47 AM
chatur01 retitled D13949: [SLP] Treat SelectInsts as reduction values. from to [SLP] Treat SelectInsts as reduction values..
Oct 21 2015, 10:12 AM

Oct 16 2015

chatur01 committed rL250520: [AArch64] Implement vector splitting on UADDV..
[AArch64] Implement vector splitting on UADDV.
Oct 16 2015, 8:40 AM
chatur01 closed D13466: [AArch64] Implement vector splitting on UADDV..
Oct 16 2015, 8:40 AM

Oct 12 2015

chatur01 added a comment to D13121: Improve ISel across lane float min/max reduction.

Hi Jun, I've just got around to looking at this, and you need to update the test names, please see below.

Oct 12 2015, 5:21 AM

Oct 8 2015

chatur01 updated the diff for D13466: [AArch64] Implement vector splitting on UADDV..

Thanks Jun for your test cases. I have added some more to my patch.

Oct 8 2015, 2:10 AM

Oct 7 2015

chatur01 updated the diff for D13466: [AArch64] Implement vector splitting on UADDV..

Thanks Jun & James for the reviews!

Oct 7 2015, 9:05 AM
chatur01 added a comment to D13466: [AArch64] Implement vector splitting on UADDV..

Charlie, are you planing to support SADDV, [SU]{MAX,MIN}V as well ? Please let me know if you are not, then, I may extend it

Oct 7 2015, 7:00 AM

Oct 6 2015

chatur01 added a comment to D13466: [AArch64] Implement vector splitting on UADDV..

Indeed, I wanted to get review on this approach first before adding the other operations. I'll update the patch tomorrow. Thank you very much for your review!

Oct 6 2015, 10:03 AM
chatur01 updated the diff for D13466: [AArch64] Implement vector splitting on UADDV..

Thanks Jun, that was a mistake.

Oct 6 2015, 9:55 AM
chatur01 retitled D13466: [AArch64] Implement vector splitting on UADDV. from to [AArch64] Implement vector splitting on UADDV..
Oct 6 2015, 4:49 AM

Sep 30 2015

chatur01 committed rL248898: [test-suite] Don't allow FP_TOLERANCE and HASH_PROGRAM_OUTPUT to occur….
[test-suite] Don't allow FP_TOLERANCE and HASH_PROGRAM_OUTPUT to occur…
Sep 30 2015, 6:27 AM
chatur01 closed D13283: [test-suite] Don't allow FP_TOLERANCE and HASH_PROGRAM_OUTPUT to occur toghether..
Sep 30 2015, 6:26 AM
chatur01 retitled D13287: [ARM,AArch64] NFC. Add extra test cases for bswap lowering. from to [ARM,AArch64] NFC. Add extra test cases for bswap lowering..
Sep 30 2015, 6:23 AM
chatur01 retitled D13283: [test-suite] Don't allow FP_TOLERANCE and HASH_PROGRAM_OUTPUT to occur toghether. from to [test-suite] Don't allow FP_TOLERANCE and HASH_PROGRAM_OUTPUT to occur toghether..
Sep 30 2015, 5:24 AM

Sep 24 2015

chatur01 committed rL248482: [InstCombine] Recognize another bswap idiom..
[InstCombine] Recognize another bswap idiom.
Sep 24 2015, 3:26 AM
chatur01 closed D12637: [InstCombine] Recognize another bswap idiom..
Sep 24 2015, 3:26 AM
chatur01 updated the diff for D12637: [InstCombine] Recognize another bswap idiom..

Roll back to previous revision. I'll add the extra backend tests
in a separate revision.

Sep 24 2015, 3:08 AM
chatur01 added a comment to D12637: [InstCombine] Recognize another bswap idiom..

OK, sorry for the misunderstanding. Can I commit my original revision, and land the extra backend tests as part of a separate review?

Sep 24 2015, 1:23 AM

Sep 23 2015

chatur01 added a comment to D12637: [InstCombine] Recognize another bswap idiom..

I'm confused as to why this needs any target specific tests. This is a change to IR canonicalization.

Sep 23 2015, 8:09 AM
chatur01 updated the diff for D12637: [InstCombine] Recognize another bswap idiom..

Add tests for ARM backend lowering of the BSWAP instrinsic.

Sep 23 2015, 7:38 AM

Sep 16 2015

chatur01 accepted D9151: Loop Versioning for LICM.

Hi Ashutosh,

Sep 16 2015, 9:19 AM

Sep 8 2015

chatur01 added a comment to D12637: [InstCombine] Recognize another bswap idiom..

Hi Renato! Thanks for looking at my patch.

Sep 8 2015, 10:53 AM

Sep 4 2015

chatur01 retitled D12637: [InstCombine] Recognize another bswap idiom. from to [InstCombine] Recognize another bswap idiom..
Sep 4 2015, 9:09 AM

Aug 27 2015

chatur01 requested changes to D9151: Loop Versioning for LICM.

Hi Ashutosh,

Aug 27 2015, 3:53 PM

Aug 13 2015

chatur01 closed D11648: InstCombinePHI: Partial simplification of identity operations.

Landed r244887.

Aug 13 2015, 5:39 AM
chatur01 committed rL244887: [InstCombinePHI] Partial simplification of identity operations..
[InstCombinePHI] Partial simplification of identity operations.
Aug 13 2015, 5:39 AM

Jul 20 2015

chatur01 closed D9913: Mark libm builtins as such..

Thanks Renato, r242675 landed.

Jul 20 2015, 7:37 AM
chatur01 committed rL242675: Add some missing 'F' flags to libm/libc builtins..
Add some missing 'F' flags to libm/libc builtins.
Jul 20 2015, 7:37 AM
chatur01 updated the diff for D9913: Mark libm builtins as such..

Thanks very much Renato and Hal for giving this one another look and apologies for my delay getting back, I've been on holiday.

Jul 20 2015, 6:22 AM

Jul 2 2015

chatur01 added a comment to D10873: [GraphWriter] Don't wait on xdg-open when not on Apple.

In my experience I also often have cases where n graphs are generated but I am only interested in one of them which forces me to close n-1 graph viewers.

Jul 2 2015, 10:36 AM
chatur01 closed D10873: [GraphWriter] Don't wait on xdg-open when not on Apple.

As we are on the topic: I think the current approach of having llvm
spawn the viewer and wait/not wait for it should be changed: I'd
really prefer llvm just writing a bunch of .dot files to the current
directory (or /tmp) giving them some sensible names+numbers and maybe
print those names.

This would avoid llvm having to find and launch the correct viewer. In
my experience I also often have cases where n graphs are generated but
I am only interested in one of them which forces me to close n-1 graph
viewers.

Jul 2 2015, 2:39 AM
chatur01 committed rL241250: [GraphWriter] Don't wait on xdg-open when not on Apple..
[GraphWriter] Don't wait on xdg-open when not on Apple.
Jul 2 2015, 2:32 AM
chatur01 committed rL241249: [NFC] Minor editorial fixes to the CodeGen docs..
[NFC] Minor editorial fixes to the CodeGen docs.
Jul 2 2015, 2:32 AM

Jul 1 2015

chatur01 retitled D10873: [GraphWriter] Don't wait on xdg-open when not on Apple from to [GraphWriter] Don't wait on xdg-open when not on Apple.
Jul 1 2015, 10:34 AM

Jun 30 2015

chatur01 added a comment to D9912: PR20958 Allow redeclaration of type-generic builtins.

Weekly ping.

Jun 30 2015, 10:02 AM

Jun 23 2015

chatur01 added a comment to D9913: Mark libm builtins as such..

Hey Renato, thanks very much for taking another look.

Jun 23 2015, 6:30 AM
chatur01 added a comment to D9912: PR20958 Allow redeclaration of type-generic builtins.

Ping.

Jun 23 2015, 3:15 AM
chatur01 added a comment to D9913: Mark libm builtins as such..

Ping.

Jun 23 2015, 3:15 AM

Jun 22 2015

chatur01 added inline comments to D9151: Loop Versioning for LICM.
Jun 22 2015, 4:08 PM
chatur01 added a comment to D9151: Loop Versioning for LICM.

In addition to some dups you already noticed in LoopVectorize, I spotted a few more on my initial scan.

Jun 22 2015, 3:45 PM