bollu (Siddharth)
User

Projects

User does not belong to any projects.

User Details

User Since
Feb 20 2017, 1:20 PM (73 w, 3 d)

Recent Activity

Thu, Jul 12

bollu accepted D49154: Extend CFGPrinter and CallPrinter with Heat Colors: Polly changes.

So, how do we coordinate this? :) Or do you push the patches at the same time?

Thu, Jul 12, 9:01 AM
bollu added a comment to D49154: Extend CFGPrinter and CallPrinter with Heat Colors: Polly changes.

Thanks! However, I am a little confused: when do I merge this, precisely? It breaks on current LLVM HEAD.

Thu, Jul 12, 8:40 AM

Mon, Jul 9

bollu added inline comments to D49024: [Polly] [WIP] Introduce ShapeInfo into polly for sizes and strides..
Mon, Jul 9, 8:34 AM

Sun, Jul 8

bollu added inline comments to D49024: [Polly] [WIP] Introduce ShapeInfo into polly for sizes and strides..
Sun, Jul 8, 4:30 AM
bollu added inline comments to D49024: [Polly] [WIP] Introduce ShapeInfo into polly for sizes and strides..
Sun, Jul 8, 3:56 AM
bollu added inline comments to D49024: [Polly] [WIP] Introduce ShapeInfo into polly for sizes and strides..
Sun, Jul 8, 3:22 AM
bollu abandoned D48874: [Polly] [WIP] Introduce ShapeInfo into polly..

Abandoned in favour of D49024.

Sun, Jul 8, 3:17 AM · Restricted Project

Wed, Jul 4

bollu committed rL336288: [PPCGCodeGen] Change printf to outs() to prevent garbled output. [NFC].
[PPCGCodeGen] Change printf to outs() to prevent garbled output. [NFC]
Wed, Jul 4, 9:56 AM
bollu closed D40126: [Polly] [PCGCodeGen] Change printf to outs() to prevent garbled output. [NFC].
Wed, Jul 4, 9:56 AM
bollu committed rL336251: [Polly-ACC] Add isl_space.h to gpu_tree.c.
[Polly-ACC] Add isl_space.h to gpu_tree.c
Wed, Jul 4, 2:45 AM
bollu closed D48883: [Polly-ACC] Add isl_space.h to gpu_tree.c.
Wed, Jul 4, 2:45 AM
bollu accepted D48883: [Polly-ACC] Add isl_space.h to gpu_tree.c.
Wed, Jul 4, 2:12 AM

Tue, Jul 3

bollu added a comment to D48874: [Polly] [WIP] Introduce ShapeInfo into polly..

@Meinersbur in polly, we use sizes per dimension, so the sizes data structure actually stores the size of a dimension. In the "stride" version, we rather store the "block size" of each dimension, which is akin to the Fortran representation, and the representation (if I understand the chapel representation correctly).

Tue, Jul 3, 9:19 AM · Restricted Project
bollu updated the diff for D48874: [Polly] [WIP] Introduce ShapeInfo into polly..

Extract out codegen as well.

Tue, Jul 3, 9:13 AM · Restricted Project
bollu updated the diff for D48874: [Polly] [WIP] Introduce ShapeInfo into polly..
  • Backport code to detect chapel/fortran style indexing. TODO: Codegen.
Tue, Jul 3, 8:42 AM · Restricted Project
bollu added a project to D48874: [Polly] [WIP] Introduce ShapeInfo into polly.: Restricted Project.
Tue, Jul 3, 4:41 AM · Restricted Project
bollu updated the summary of D48874: [Polly] [WIP] Introduce ShapeInfo into polly..
Tue, Jul 3, 4:41 AM · Restricted Project
bollu created D48874: [Polly] [WIP] Introduce ShapeInfo into polly..
Tue, Jul 3, 4:39 AM · Restricted Project

Wed, Jun 27

bollu added a comment to D40778: [DebugIR] Revive the Debug IR pass. [Added llvm-commits].

As mentioned, I don't have the bandwidth :( If you could fix the issues that were raised during the review, it would be awesome.

Wed, Jun 27, 12:39 PM · debug-info
bollu accepted D48136: [Polly] Implement an iterator for isl maps, basic_maps, sets, basic_sets.

Sorry, I meant to accept this last week. I accept this as an acceptable "software engineering" compromise, while I still hold the philosophical reservations I have mentioned.

Wed, Jun 27, 12:33 PM
bollu added a comment to D48136: [Polly] Implement an iterator for isl maps, basic_maps, sets, basic_sets.

@Meinersbur If you are opposed to merging this, please request changes.

Wed, Jun 27, 12:33 PM

Tue, Jun 26

bollu added a comment to D48136: [Polly] Implement an iterator for isl maps, basic_maps, sets, basic_sets.

@Meinersbur even assuming that such a concept exists (since one could argue such a concept does exist for a range-based for loop to work), I still don't think it allows the compiler to optimise away iterators based on the interpretation of the spec I outlined (== is an equivalence relation whose domain is mutable).

Tue, Jun 26, 6:27 AM

Fri, Jun 22

bollu updated subscribers of D48026: [ScopHelper] Provide support for recognising collective invariant loads.

Also, please run the file through instnamer. That way, we will have "named" instructions and not "numbered" instructions. This is useful to modify the testcase in the future (it is less brittle): "numbered" instructions must be sequential, while named instructions need to be.

Fri, Jun 22, 9:04 AM
bollu added inline comments to D48026: [ScopHelper] Provide support for recognising collective invariant loads.
Fri, Jun 22, 6:55 AM
bollu added inline comments to D48026: [ScopHelper] Provide support for recognising collective invariant loads.
Fri, Jun 22, 4:59 AM
bollu added inline comments to D48136: [Polly] Implement an iterator for isl maps, basic_maps, sets, basic_sets.
Fri, Jun 22, 4:52 AM

Jun 17 2018

bollu requested changes to D48162: [GSoC] Schedule tree performance..

I'm going to request changes so this does not show up on the feed. I added two more comments regarding constructors :)

Jun 17 2018, 3:13 AM · Restricted Project

Jun 16 2018

bollu added a comment to D48162: [GSoC] Schedule tree performance..

Thanks for the patch! I've added some comments about general LLVM style, and some personal choices. Feel free to ignore the personal choices :)

Jun 16 2018, 12:15 PM · Restricted Project
bollu updated subscribers of D48162: [GSoC] Schedule tree performance..

Thanks for the patch! I've added some comments about general LLVM style, and some personal choices. Feel free to ignore the personal choices :)

Jun 16 2018, 12:15 PM · Restricted Project

Jun 4 2018

bollu added a comment to D47604: [Polly] Add isl C++ list iterators.

This might be too much to ask for, but is it possible to report_fatal_error() at the earliest when accessing out of bounds using the iterator? If ISL performs inbounds checks, then disregard this.

Jun 4 2018, 3:11 AM
bollu added a comment to D47604: [Polly] Add isl C++ list iterators.

This might be too much to ask for, but is it possible to report_fatal_error() at the earliest when accessing out of bounds using the iterator? If ISL performs inbounds checks, then disregard this.

Jun 4 2018, 3:11 AM

Apr 20 2018

bollu added inline comments to D45743: [Polly] Print executed statement instances at runtime..
Apr 20 2018, 8:47 AM · Restricted Project

Mar 21 2018

bollu added a comment to D40778: [DebugIR] Revive the Debug IR pass. [Added llvm-commits].

I'm sorry. I don't have the bandwidth for this right now. Please do take it up, it'd be great to see this upstreamed! If we can get the base version upstreamed, I don't mind quickly iterating on smaller features / tests / things like that.

Mar 21 2018, 7:47 AM · debug-info

Jan 17 2018

bollu added a comment to D42142: [Polly] [WIP] Adding Live-Range Reordering for Polly.

@Meinersbur I agree. However, I don't think that this is the right place for the code either. It is more a DependenceInfo thing, and maybe it makes sense to teach DependenceInfo to generate this information for us?

Jan 17 2018, 9:00 AM

Jan 16 2018

bollu added a comment to D42142: [Polly] [WIP] Adding Live-Range Reordering for Polly.

Thanks a lot for the patch! I took a shot at a first round of review :) I can be nitpicky, so we can hash out the details as we go along with the review.

Jan 16 2018, 9:37 PM

Dec 5 2017

bollu added a comment to D40778: [DebugIR] Revive the Debug IR pass. [Added llvm-commits].

Regarding StripDebugInfo, I did not know it existed. I should refactor the code to use it, I assume? I'll check this out.

Dec 5 2017, 1:50 PM · debug-info

Dec 4 2017

bollu committed rL319718: Update to latest clang-format. [NFC].
Update to latest clang-format. [NFC]
Dec 4 2017, 4:06 PM
bollu closed D40791: [Polly] update to latest clang-format. [NFC] by committing rL319718: Update to latest clang-format. [NFC].
Dec 4 2017, 4:06 PM · Restricted Project
bollu added a comment to D40791: [Polly] update to latest clang-format. [NFC].

No, unfortunately, I'm not sure. I didn't want to guess, so I didn't write a particular clang revision.

Dec 4 2017, 4:04 PM · Restricted Project
bollu added reviewers for D40791: [Polly] update to latest clang-format. [NFC]: grosser, philip.pfaffe, Meinersbur.
Dec 4 2017, 8:32 AM · Restricted Project
bollu created D40791: [Polly] update to latest clang-format. [NFC].
Dec 4 2017, 8:32 AM · Restricted Project
bollu updated the diff for D40778: [DebugIR] Revive the Debug IR pass. [Added llvm-commits].

Removed expletives from patch that arose due to frustration with debug metadata. Run clang-format on file.

Dec 4 2017, 5:47 AM · debug-info

Dec 3 2017

bollu abandoned D40207: [WIP] [DebugIR] Revive the old DebugIR pass for tip-of-tree LLVM..

Sorry for the delay, I resubmitted the patch as D40778. I'm closing this in favor of continuing the discussion there. Ping, @vsk, and thanks for all the help so far!

Dec 3 2017, 1:47 PM
bollu created D40778: [DebugIR] Revive the Debug IR pass. [Added llvm-commits].
Dec 3 2017, 1:46 PM · debug-info

Nov 28 2017

bollu accepted D40492: Handle Top-Level-Regions in polly::isHoistableLoad.
Nov 28 2017, 5:35 AM · Restricted Project
bollu added a comment to D40492: Handle Top-Level-Regions in polly::isHoistableLoad.

Other than request for documentation, LGTM. I did not test it, but I reduced the test case with bugpoint, so it should be the right fix. If not, I can complain again ;)

Nov 28 2017, 5:35 AM · Restricted Project
bollu added a comment to D40207: [WIP] [DebugIR] Revive the old DebugIR pass for tip-of-tree LLVM..

@vsk In that case, I think we agree. Bugpoint is useful, but a debugger is really helpful in certain contexts. It's easier for me (and I suspect others) to debug on the IR level when they would like to do so. Does that provide enough of a motivation to go ahead with the review of the patch? You mentioned that there are technical issues, I'd like to sort them out :)

Nov 28 2017, 5:01 AM

Nov 27 2017

bollu added a comment to D40207: [WIP] [DebugIR] Revive the old DebugIR pass for tip-of-tree LLVM..

@vsk Bugpoint does has four weaknesses that I have noticed not work (as far as I know, please do let me know if bugpoint can do this)

Nov 27 2017, 1:49 PM
bollu added a comment to D40207: [WIP] [DebugIR] Revive the old DebugIR pass for tip-of-tree LLVM..

@vsk I do use the same technique for testing my frontend. However, this is useful when stuff _breaks_ in the frontend.g I'm not as good at reading asm as I'd like to be, so being able to debug at the IR level is a boon.

Nov 27 2017, 11:07 AM

Nov 18 2017

bollu added reviewers for D40207: [WIP] [DebugIR] Revive the old DebugIR pass for tip-of-tree LLVM.: grosser, philip.pfaffe.
Nov 18 2017, 5:54 AM
bollu added reviewers for D40207: [WIP] [DebugIR] Revive the old DebugIR pass for tip-of-tree LLVM.: dexonsmith, echristo, aprantl, vsk.
Nov 18 2017, 5:34 AM
bollu created D40207: [WIP] [DebugIR] Revive the old DebugIR pass for tip-of-tree LLVM..
Nov 18 2017, 5:30 AM

Nov 16 2017

bollu updated the diff for D40126: [Polly] [PCGCodeGen] Change printf to outs() to prevent garbled output. [NFC].
  • remove unrelated files.
Nov 16 2017, 6:21 AM
bollu created D40126: [Polly] [PCGCodeGen] Change printf to outs() to prevent garbled output. [NFC].
Nov 16 2017, 6:14 AM
bollu updated the diff for D39904: [Polly] [PPCGCodeGeneration] Teach `SubtreeValues` to include possibly required base pointers, not remove all base pointers blindly..
  • update code and test case
Nov 16 2017, 6:02 AM

Nov 13 2017

bollu added inline comments to D37135: [Polly] [PPCGCodeGeneration] Allow PPCGCodeGeneration to run without libdevice..
Nov 13 2017, 5:01 PM

Nov 11 2017

bollu updated the diff for D39904: [Polly] [PPCGCodeGeneration] Teach `SubtreeValues` to include possibly required base pointers, not remove all base pointers blindly..
  • [NFC wrt patch] remove debug code
Nov 11 2017, 5:07 AM
bollu updated the diff for D39904: [Polly] [PPCGCodeGeneration] Teach `SubtreeValues` to include possibly required base pointers, not remove all base pointers blindly..
  • add changes in PPCGCodeGen that were missed due to incorrect rebase.
Nov 11 2017, 5:07 AM
bollu added a comment to D39904: [Polly] [PPCGCodeGeneration] Teach `SubtreeValues` to include possibly required base pointers, not remove all base pointers blindly..

@efriedma Thanks, I totally screwed up the rebase :)

Nov 11 2017, 5:07 AM

Nov 10 2017

bollu created D39904: [Polly] [PPCGCodeGeneration] Teach `SubtreeValues` to include possibly required base pointers, not remove all base pointers blindly..
Nov 10 2017, 8:45 AM

Oct 30 2017

bollu added inline comments to D39333: [Polly][ZoneAlgo/ForwardOpTree] Normalize PHIs to their known incoming values..
Oct 30 2017, 3:29 AM · Restricted Project

Oct 16 2017

bollu accepted D38953: Updating the TipsAndTricks.rst documentation of Polly related to bugpoint.

LGTM! Thanks, I didn't realize that the options could interact with bugs in llc.

Oct 16 2017, 7:51 AM · Restricted Project

Oct 4 2017

bollu added a comment to D38403: [Polly][ScopBuilder] Introduce -polly-stmt-granularity=scalar-indep option..

Thanks for the very clean patch :) I really like reviewing your code, it's always really well written.

Oct 4 2017, 5:15 PM · Restricted Project
bollu added inline comments to D38403: [Polly][ScopBuilder] Introduce -polly-stmt-granularity=scalar-indep option..
Oct 4 2017, 7:13 AM · Restricted Project

Sep 29 2017

bollu added inline comments to D35471: [Polly] [RFC] Calculate AST expression type.
Sep 29 2017, 8:33 AM
bollu added inline comments to D35471: [Polly] [RFC] Calculate AST expression type.
Sep 29 2017, 8:27 AM
bollu added inline comments to D38403: [Polly][ScopBuilder] Introduce -polly-stmt-granularity=scalar-indep option..
Sep 29 2017, 8:00 AM · Restricted Project

Sep 28 2017

bollu committed rL314422: [Docs] fix rendering of alpha and beta.
[Docs] fix rendering of alpha and beta
Sep 28 2017, 8:34 AM
bollu committed rL314423: [Docs] Replace 0-byte incorrect GEMM_double image with the one from www/images.
[Docs] Replace 0-byte incorrect GEMM_double image with the one from www/images
Sep 28 2017, 8:34 AM
bollu committed rL314419: [Docs] Add a performance document..
[Docs] Add a performance document.
Sep 28 2017, 8:12 AM
bollu closed D38330: [docs] Add a performance doc by committing rL314419: [Docs] Add a performance document..
Sep 28 2017, 8:12 AM · Restricted Project
bollu created D38353: [Polly] [PPCGCodeGeneration] Fix offset computation for managed arrays..
Sep 28 2017, 5:09 AM

Sep 25 2017

bollu added a comment to D38255: [WIP] Polyhedral Value Analysis.

@jdoerfert: Thanks for putting this up. I'm trying to read and understand the ideas behind this. I've left a few questions, could you please clarify them for me?

Sep 25 2017, 3:37 PM
bollu added a comment to D38255: [WIP] Polyhedral Value Analysis.

@jdoerfert: Thanks for putting this up. I'm trying to read and understand the ideas behind this. I've left a few questions, could you please clarify them for me?

Sep 25 2017, 3:36 PM

Sep 15 2017

bollu created D37901: [Polly] [Strided arrays, multidimensional indexing, fortran support] Teach Polly multidimensional strided array indexing. [WIP].
Sep 15 2017, 7:06 AM

Sep 7 2017

bollu committed rL312715: [PPCGCodeGen] Document pre-composition with Zero in getExtent. [NFC].
[PPCGCodeGen] Document pre-composition with Zero in getExtent. [NFC]
Sep 7 2017, 5:00 AM

Sep 1 2017

bollu added a comment to D37379: [Polly] [ISLNodeBuilder] [Fortran Arrays] Materialize Fortran array sizes of arrays without memory accesses..

Review please, this was a quick bugfix.

Sep 1 2017, 7:33 AM
bollu created D37379: [Polly] [ISLNodeBuilder] [Fortran Arrays] Materialize Fortran array sizes of arrays without memory accesses..
Sep 1 2017, 7:31 AM

Aug 31 2017

bollu closed D37056: [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible. by committing rL312239: [PPCGCodeGen] Convert intrinsics to libdevice functions whenever possible..
Aug 31 2017, 6:06 AM

Aug 29 2017

bollu committed rL311995: [ScopInfo] Fix comment grammar. "..to be build" -> "..to be built". [NFC].
[ScopInfo] Fix comment grammar. "..to be build" -> "..to be built". [NFC]
Aug 29 2017, 4:47 AM

Aug 28 2017

bollu added a comment to D37056: [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible..

@efriedma I made the changes you requested. Could you have a final look, please?

Aug 28 2017, 3:31 AM
bollu updated the diff for D37056: [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible..
  • [Diff update] Update diff with latest master.
Aug 28 2017, 2:37 AM
bollu updated the diff for D37056: [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible..
  • Don't parse the function name, keep a static map from intrinsic names to libdevice names
Aug 28 2017, 2:19 AM

Aug 27 2017

bollu accepted D37191: [Polly][GPGPU] Fixed undefined reference for CUDA's managed memory in Runtime library..

LGTM, thanks for generalising it :) I'm sorry, I'll keep the mixed CUDA/OpenCL thing in mind next time I mess with GPUJIt.

Aug 27 2017, 4:58 AM

Aug 25 2017

bollu added a comment to D37056: [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible..

@grosser ping, review please.

Aug 25 2017, 8:18 AM
bollu added a comment to D37061: [Polly] Third Updated translating the function Scop::foldSizeConstantsToRight() to isl C++.

Also, for future reference, note that an old review can be updated: There is an update diff option to the right hand side at the top of the web page. That way, we won't have multiple revisions like D37061 and D36843. This reduces the number of places someone needs to look to see the total change :). Thanks!

Aug 25 2017, 2:52 AM
bollu created D37135: [Polly] [PPCGCodeGeneration] Allow PPCGCodeGeneration to run without libdevice..
Aug 25 2017, 2:24 AM

Aug 24 2017

bollu added a comment to D37056: [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible..

@efriedma - The NVPTX backend does not lower either llvm.exp.* or llvm.powi.* (Unless I am doing something wrong).

Aug 24 2017, 3:01 AM
bollu committed rL311648: [Polly] [PPCGCodeGeneration] Mild refactoring of checking validity of functions….
[Polly] [PPCGCodeGeneration] Mild refactoring of checking validity of functions…
Aug 24 2017, 2:55 AM
bollu closed D37058: [Polly] [PPCGCodeGeneration] Mild refactoring of checking validity of functions in a kernel. by committing rL311648: [Polly] [PPCGCodeGeneration] Mild refactoring of checking validity of functions….
Aug 24 2017, 2:55 AM

Aug 23 2017

bollu created D37067: [Polly] [WIP] [ManagedMemoryRewrite] Rewrite global structs to managed memory..
Aug 23 2017, 9:00 AM
bollu updated the diff for D37056: [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible..
  • [Test] upadte tests to refect state of exp, powi in PPCGCodeGen
Aug 23 2017, 5:21 AM
bollu added a comment to D37056: [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible..

I should probably split this into two patches, but I wanted to run the idea by. I can split it when I commit (or split it now). Review, please.

Aug 23 2017, 4:50 AM
bollu retitled D37056: [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible. from [Polly] [PPCGCodeGeneration] [WIP] Convert intrinsics to libdevice functions whenever possible. to [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible..
Aug 23 2017, 4:48 AM
bollu updated the diff for D37056: [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible..
  • [Bugfix] add powi to the list of instructions that the NVPTX backend cannot lower.
  • [NFC] remove debug code
Aug 23 2017, 4:47 AM
bollu created D37058: [Polly] [PPCGCodeGeneration] Mild refactoring of checking validity of functions in a kernel..
Aug 23 2017, 4:37 AM
bollu created D37056: [Polly] [PPCGCodeGeneration] Convert intrinsics to libdevice functions whenever possible..
Aug 23 2017, 3:11 AM

Aug 22 2017

bollu committed rL311471: [GPUJIT] Make max managed pointers an environment variable..
[GPUJIT] Make max managed pointers an environment variable.
Aug 22 2017, 10:33 AM
bollu closed D37012: [GPUJIT] Make max managed pointers an environment variable by committing rL311471: [GPUJIT] Make max managed pointers an environment variable..
Aug 22 2017, 10:33 AM
bollu added inline comments to D37010: [Polly][PM] Properly require and preservation of OptimizationRemarkEmitter..
Aug 22 2017, 9:36 AM · Restricted Project