Page MenuHomePhabricator

dcaballe (Diego Caballero)
User

Projects

User does not belong to any projects.

User Details

User Since
Sep 16 2016, 11:47 AM (297 w, 10 h)

Recent Activity

Yesterday

dcaballe accepted D126454: [mlir][Linalg] Relax vectorization condition to allow transposed output..
Thu, May 26, 11:31 AM · Restricted Project, Restricted Project

Tue, May 17

dcaballe accepted D125583: [mlir][tosa] Rework tosa.apply_scale lowering for 32-bit.

Thanks for addressing the changes. LGTM but, as Mahesh mentioned, it's probably worth getting someone with experience in TOSA to take a look.

Tue, May 17, 2:38 PM · Restricted Project, Restricted Project

Mon, May 16

dcaballe added a comment to D125543: [mlir] support isa/cast/dyn_cast<Operation *>(operation) again.

I can confirm that D125590 is not fixing the problem. D123901 broke our mlir-based project (IREE) due to an Operation * self-cast. Only with the proposed changes in this diff things get back to work.

Mon, May 16, 5:24 PM · Restricted Project, Restricted Project
dcaballe added a comment to D125583: [mlir][tosa] Rework tosa.apply_scale lowering for 32-bit.

Thanks for working on this, Rob! I really can't comment on the actual implementation from the TOSA perspective. However, I would suggest not enabling this by default. Architectures with full 64-bit support should do better with the 64-bit lowering. I can follow up with a patch to enable this for the RISC-V specific cases.

Mon, May 16, 12:29 PM · Restricted Project, Restricted Project

Wed, May 11

dcaballe accepted D125101: [mlir][NFC] Fix `GpuKernelOutliningPass` copy constructor warnings.
Wed, May 11, 9:44 AM · Restricted Project, Restricted Project

Apr 20 2022

dcaballe accepted D124047: [mlir] Modify SuperVectorize to generate select op->combiner op .

Thanks!

Apr 20 2022, 11:38 AM · Restricted Project, Restricted Project

Apr 19 2022

dcaballe accepted D124047: [mlir] Modify SuperVectorize to generate select op->combiner op .

Thanks, Amy! In general, moving the select op before the operation makes more sense to me since the op might have side effects that would make the execution incorrect otherwise.
I hope we can replace all of this once we have the masking proposal in place.

Apr 19 2022, 3:38 PM · Restricted Project, Restricted Project

Apr 1 2022

dcaballe added a reviewer for D108536: [mlir][RISCVV][RFC] Initial RISCVV Dialect: dcaballe.
Apr 1 2022, 10:23 AM · Restricted Project, Restricted Project

Mar 24 2022

dcaballe accepted D118248: [mlir][Vector] Enable create_mask for scalable vectors.
Mar 24 2022, 12:01 PM · Restricted Project, Restricted Project
dcaballe accepted D122415: [mlir][vector][nfc] Rename index optimizations option.

Much better, thanks!

Mar 24 2022, 11:43 AM · Restricted Project, Restricted Project

Mar 23 2022

dcaballe added a comment to D118248: [mlir][Vector] Enable create_mask for scalable vectors.

Thanks for the contribution, Javier! Some comments inline.

Mar 23 2022, 5:53 PM · Restricted Project, Restricted Project
dcaballe accepted D122292: Update affine.load folding hook to fold global splat constant loads.

LGTM, thanks!

Mar 23 2022, 10:50 AM · Restricted Project, Restricted Project

Mar 10 2022

dcaballe committed rGf71f9958b984: [mlir][Vector] Modernize default lowering of vector transpose (authored by dcaballe).
[mlir][Vector] Modernize default lowering of vector transpose
Mar 10 2022, 2:35 PM · Restricted Project
dcaballe closed D121321: [mlir][Vector] Modernize default lowering of vector transpose.
Mar 10 2022, 2:35 PM · Restricted Project, Restricted Project

Mar 9 2022

dcaballe added a comment to D121321: [mlir][Vector] Modernize default lowering of vector transpose.

I'll commit it tomorrow if no more comments. Thanks!

Mar 9 2022, 6:23 PM · Restricted Project, Restricted Project
dcaballe updated the diff for D121321: [mlir][Vector] Modernize default lowering of vector transpose.

Addressed feedback.

Mar 9 2022, 6:22 PM · Restricted Project, Restricted Project
dcaballe accepted D121148: [mlir] Extend SimplifyTrivialLoops.
Mar 9 2022, 2:38 PM · Restricted Project, Restricted Project
dcaballe requested review of D121321: [mlir][Vector] Modernize default lowering of vector transpose.
Mar 9 2022, 11:48 AM · Restricted Project, Restricted Project

Mar 7 2022

dcaballe accepted D121170: NFC. Improve isInnermostAffineForOp - drop unnecessary check.

LGTM, thanks!

Mar 7 2022, 5:13 PM · Restricted Project, Restricted Project
dcaballe added a comment to D121148: [mlir] Extend SimplifyTrivialLoops.

I'll let @bondhugula clarify but my impression about his comment in D120776 is that we should be able to have a single implementation that works for both ScfForOp and AffineForOp by using/extending interfaces like LoopLikeInterface? Maybe I got it wrong. In any case, adding a scf specific implementation is also great. Thanks!

Mar 7 2022, 1:59 PM · Restricted Project, Restricted Project
dcaballe committed rG917d95fc8adb: [mlir][Vector] Improve default lowering of vector transpose operations (authored by dcaballe).
[mlir][Vector] Improve default lowering of vector transpose operations
Mar 7 2022, 9:58 AM · Restricted Project
dcaballe closed D120601: [mlir][Vector] Improve default lowering of vector transpose operations.
Mar 7 2022, 9:58 AM · Restricted Project, Restricted Project
dcaballe added inline comments to D120601: [mlir][Vector] Improve default lowering of vector transpose operations.
Mar 7 2022, 9:39 AM · Restricted Project, Restricted Project

Mar 4 2022

dcaballe added a comment to D120601: [mlir][Vector] Improve default lowering of vector transpose operations.

Thanks for the feedback! I addressed the trivial one. The cleanup/generalization would need more thought before we move forward. More info below.
I'll move forward with this if no more comments.

Mar 4 2022, 11:50 AM · Restricted Project, Restricted Project
dcaballe updated the diff for D120601: [mlir][Vector] Improve default lowering of vector transpose operations.

Addressed feedback.

Mar 4 2022, 11:49 AM · Restricted Project, Restricted Project

Mar 3 2022

dcaballe added a reviewer for D120601: [mlir][Vector] Improve default lowering of vector transpose operations: ThomasRaoux.
Mar 3 2022, 7:07 PM · Restricted Project, Restricted Project
dcaballe added reviewers for D120601: [mlir][Vector] Improve default lowering of vector transpose operations: gysit, pifon2a.
Mar 3 2022, 6:35 PM · Restricted Project, Restricted Project
dcaballe accepted D120617: [MLIR] Extend isLoopMemoryParallel to account for locally allocated memrefs.
Mar 3 2022, 12:11 PM · Restricted Project, Restricted Project
dcaballe accepted D120776: [mlir] Extend AffineForEmptyLoopFolder.

Thanks, Amy. LGTM once Uday's comments have been addressed.

Mar 3 2022, 10:35 AM · Restricted Project, Restricted Project

Mar 2 2022

dcaballe added a comment to D120776: [mlir] Extend AffineForEmptyLoopFolder.

Thanks, Amy! I think this is a very interesting extension! Some comments

Mar 2 2022, 11:07 AM · Restricted Project, Restricted Project
dcaballe added inline comments to D120612: [MLIR] Add affine.load fold hook on global constant memrefs.
Mar 2 2022, 10:10 AM · Restricted Project, Restricted Project
dcaballe accepted D120617: [MLIR] Extend isLoopMemoryParallel to account for locally allocated memrefs.
Mar 2 2022, 9:37 AM · Restricted Project, Restricted Project
Herald added a project to D120617: [MLIR] Extend isLoopMemoryParallel to account for locally allocated memrefs: Restricted Project.

Thanks, Uday. A couple of comments that would need attention. LGTM. You can move forward after addressing them.

Mar 2 2022, 9:37 AM · Restricted Project, Restricted Project

Feb 25 2022

dcaballe requested review of D120601: [mlir][Vector] Improve default lowering of vector transpose operations.
Feb 25 2022, 4:38 PM · Restricted Project, Restricted Project
dcaballe committed rG875bbce9f720: [mlir][Vector] Prevent AVX2 lowering for non-f32 transpose ops (authored by dcaballe).
[mlir][Vector] Prevent AVX2 lowering for non-f32 transpose ops
Feb 25 2022, 11:30 AM
dcaballe committed rGd7e0a0846b02: [mlir][Vector] Generalize AVX2 transpose lowering to n-D vectors (authored by dcaballe).
[mlir][Vector] Generalize AVX2 transpose lowering to n-D vectors
Feb 25 2022, 11:30 AM
dcaballe closed D120427: [mlir][Vector] Prevent AVX2 lowering for non-f32 transpose ops.
Feb 25 2022, 11:30 AM · Restricted Project
dcaballe closed D119505: [mlir][Vector] Generalize AVX2 transpose lowering to n-D vectors.
Feb 25 2022, 11:30 AM · Restricted Project

Feb 23 2022

dcaballe added a comment to D120427: [mlir][Vector] Prevent AVX2 lowering for non-f32 transpose ops.

Just curious if F32 data movement would not just work for any 32-bit entity, but change looks good to me.

Feb 23 2022, 6:09 PM · Restricted Project
dcaballe requested review of D120427: [mlir][Vector] Prevent AVX2 lowering for non-f32 transpose ops.
Feb 23 2022, 12:08 PM · Restricted Project
dcaballe updated the diff for D119505: [mlir][Vector] Generalize AVX2 transpose lowering to n-D vectors.

An internal test exposed a bug in the implementation. This new patch
fixes the issue (see 'areDimsTransposedIn2DSlice') and improves documentation.
Could you please take a quick look?

Feb 23 2022, 12:06 PM · Restricted Project

Feb 17 2022

dcaballe updated the diff for D119505: [mlir][Vector] Generalize AVX2 transpose lowering to n-D vectors.

Generalize the implementation even more by using shape_cast before and
after the transposition.

Feb 17 2022, 12:54 PM · Restricted Project
dcaballe retitled D119505: [mlir][Vector] Generalize AVX2 transpose lowering to n-D vectors from [mlir][Vector] Extend AVX2 transpose lowering to n-D vectors to [mlir][Vector] Generalize AVX2 transpose lowering to n-D vectors.
Feb 17 2022, 12:52 PM · Restricted Project

Feb 10 2022

dcaballe planned changes to D119505: [mlir][Vector] Generalize AVX2 transpose lowering to n-D vectors.
Feb 10 2022, 6:49 PM · Restricted Project
dcaballe requested review of D119505: [mlir][Vector] Generalize AVX2 transpose lowering to n-D vectors.
Feb 10 2022, 5:36 PM · Restricted Project
dcaballe accepted D119413: [MLIR] Add result status for normalizeAffineFor.

Thanks, Uday. LGTM!

Feb 10 2022, 9:26 AM · Restricted Project

Jan 26 2022

dcaballe accepted D118292: [mlir][amx] add a full tile matrix mult integral example to integration tests.

Thanks, Aart!

Jan 26 2022, 5:00 PM · Restricted Project
dcaballe accepted D118289: [mlir][amx] fix some whitespace issues in AMX tests.
Jan 26 2022, 3:31 PM · Restricted Project
dcaballe accepted D118277: [mlir][amx] add a full tile matrix mult example to integration tests.

Thanks, Aart! LGTM

Jan 26 2022, 1:44 PM · Restricted Project

Jan 14 2022

dcaballe added a comment to D116069: [mlir][vector] Allow values outside of [0; dim-size] in create_mask.

Thanks, Sergei! Much clearer now. I think your proposal makes more sense to me now. The problem is very specific to the way you are generating the code but it's clearing exposing some corner cases of the create_mask operation. I'll leave it to Aart, though, since he has much more context on these ops (I'm not even sure I understand why we have the constant and non-constant variant of this op so I can't fully understand the implications of this change).

Jan 14 2022, 1:52 AM · Restricted Project

Jan 13 2022

dcaballe added a comment to D116069: [mlir][vector] Allow values outside of [0; dim-size] in create_mask.

Thanks for the contribution, Sergei! I think I don't have enough experience with this op so I'll leave this to @nicolasvasilache.
How do you end up generating values that are out of the expected bounds of the mask?
Folding invalid values into valid ones could be surprising and maybe lead to silent bugs (?) but maybe it makes sense for this op. An alternative would be generating code in your use case to handle these out-of-bounds values and truncate them accordingly before they are passed to the create_mask op. I would useful if you could share a bit more about your use case.

Jan 13 2022, 7:13 AM · Restricted Project
dcaballe accepted D117209: [MLIR] NFC. affine data copy generate utility return value cleanup.

LGTM. Just a minor comment.

Jan 13 2022, 6:56 AM · Restricted Project
dcaballe added a comment to D116230: [MLIR] Introduce generic visitors..

Thanks, Rahul for contributing this! It's great that it can be integreated with the existing walk methods. LGTM. I'll leave the final approval to @rriddle.

Jan 13 2022, 3:51 AM · Restricted Project

Jan 11 2022

dcaballe committed rGe2b658cd5d0b: [mlir][GPU] Fix attribute name of DL specification (authored by dcaballe).
[mlir][GPU] Fix attribute name of DL specification
Jan 11 2022, 12:36 AM
dcaballe closed D116956: [mlir][GPU] Fix attribute name of DL specification.
Jan 11 2022, 12:36 AM · Restricted Project

Jan 10 2022

dcaballe requested review of D116956: [mlir][GPU] Fix attribute name of DL specification.
Jan 10 2022, 9:39 AM · Restricted Project

Dec 22 2021

dcaballe added a comment to D116151: [mlir][arith] Fix CmpIOP folding for vector types..

LGTM. Just a comment about the test.

Dec 22 2021, 6:49 AM · Restricted Project

Dec 16 2021

dcaballe committed rGe0514fc20047: [mlir] Fix Bazel build after 32fe1a8a2594 (authored by dcaballe).
[mlir] Fix Bazel build after 32fe1a8a2594
Dec 16 2021, 4:16 AM
dcaballe committed rG32fe1a8a2594: [mlir][GPU] Extend GPU kernel outlining to generate DL specification (authored by dcaballe).
[mlir][GPU] Extend GPU kernel outlining to generate DL specification
Dec 16 2021, 3:38 AM
dcaballe closed D115722: [mlir][GPU] Extend GPU kernel outlining to generate DL specification.
Dec 16 2021, 3:38 AM · Restricted Project
dcaballe committed rG0ca00c353823: [mlir][vector] Remove default value in populateVectorMultiReductionLoweringPatt… (authored by dcaballe).
[mlir][vector] Remove default value in populateVectorMultiReductionLoweringPatt…
Dec 16 2021, 1:46 AM
dcaballe closed D115805: [mlir][vector] Remove default value in populateVectorMultiReductionLoweringPatterns.
Dec 16 2021, 1:46 AM · Restricted Project

Dec 15 2021

dcaballe requested review of D115805: [mlir][vector] Remove default value in populateVectorMultiReductionLoweringPatterns.
Dec 15 2021, 8:18 AM · Restricted Project
dcaballe updated the diff for D115722: [mlir][GPU] Extend GPU kernel outlining to generate DL specification.

Addressed the feecback. The data layout spec is now passed using a string.
Much better. Thanks for the feedback!

Dec 15 2021, 7:49 AM · Restricted Project
dcaballe added inline comments to D115722: [mlir][GPU] Extend GPU kernel outlining to generate DL specification.
Dec 15 2021, 12:53 AM · Restricted Project

Dec 14 2021

dcaballe added inline comments to D115722: [mlir][GPU] Extend GPU kernel outlining to generate DL specification.
Dec 14 2021, 1:39 PM · Restricted Project
dcaballe requested review of D115722: [mlir][GPU] Extend GPU kernel outlining to generate DL specification.
Dec 14 2021, 4:42 AM · Restricted Project

Dec 10 2021

dcaballe added a comment to D112117: [llvm][LV] Drop poison-generating flags from FP instructions.

The change I'm working on is D115460. To my knowledge, there's no interaction with the vectorizer here, it's simply a code cleanup. e.g. Do the same thing for all types of poison generating flags. We may be too conservative in some places, but consistency has a strong value of it's own. At least per the tests being changed, the vectorizer is not impacted.

Dec 10 2021, 2:05 PM · Restricted Project
dcaballe accepted D115495: [MLIR][NFC] Move out affine scalar replacement utility to affine utils.

Thanks!

Dec 10 2021, 3:24 AM · Restricted Project
dcaballe updated subscribers of D112117: [llvm][LV] Drop poison-generating flags from FP instructions.

JFYI, I've been taking a look at trying to get FMF flags into the generic dropPoisonGeneratingFlags. So far, the test diff (which looks a lot like this one for the same reasons) has exposed a bunch of cases where the existing code had problems. I'm working through fixing them individually, with the goal of getting a reasonable diff.

The most relevant one which got some discussion on this review is the reduction expansion. That appears to be entirely an artifact of code which a) does nothing without the FMF handling in the generic code, and b) serves no purpose even with it. 0d13f94c and b24db85 simply remove most of it, and fix up some comments.

Dec 10 2021, 3:20 AM · Restricted Project

Dec 1 2021

dcaballe added a comment to D114803: [mlir][Vector] Thread 0-d vectors through vector.transfer ops.

Thanks for helping with the 0-d vector problem! Added some minor comments. I have a question about the lowering of 0-d vectors to the scalar world. See comments inline.

Dec 1 2021, 4:21 AM · Restricted Project

Nov 29 2021

dcaballe added a comment to D114662: [MLIR] Fix affine.for unroll for multi-result upper bound maps.

Thanks, Uday. LGTM but I'm not super familiar with the loop unrolling pass. I'll defer the approval to someone with more expertise in the pass.

Nov 29 2021, 5:56 AM · Restricted Project

Nov 23 2021

dcaballe added inline comments to D114387: [mlir][Vector] Thread 0-d vectors through ExtractElementOp..
Nov 23 2021, 4:15 AM · Restricted Project
dcaballe accepted D114388: [mlir][Vector] Thread 0-d vectors through InsertElementOp..

LGTM. Just one minor question. Thanks!

Nov 23 2021, 1:43 AM · Restricted Project

Nov 22 2021

dcaballe abandoned D112117: [llvm][LV] Drop poison-generating flags from FP instructions.

Dropping FMF doesn't seem to be needed after all. See D111846

Nov 22 2021, 3:06 AM · Restricted Project
dcaballe abandoned D112335: [LV] Pre-commit tests.

Dropping FMF doesn't seem to be needed after all. See D111846

Nov 22 2021, 3:06 AM · Restricted Project
dcaballe added a comment to D111846: [LV] Drop integer poison-generating flags from instructions that need predication.

Thanks for all the feedback! I think this will cover a good part of the cases that could actually go wrong in practice. We can accommodate more cases as we see fit.

Nov 22 2021, 3:04 AM · Restricted Project
dcaballe committed rG4348cd42c385: [LV] Drop integer poison-generating flags from instructions that need… (authored by dcaballe).
[LV] Drop integer poison-generating flags from instructions that need…
Nov 22 2021, 2:59 AM
dcaballe closed D111846: [LV] Drop integer poison-generating flags from instructions that need predication.
Nov 22 2021, 2:59 AM · Restricted Project
dcaballe committed rGa7027bb79971: [LV] Pre-commit test for D111846 (authored by dcaballe).
[LV] Pre-commit test for D111846
Nov 22 2021, 2:15 AM
dcaballe closed D112054: [LV] Pre-commit test for D111846.
Nov 22 2021, 2:15 AM · Restricted Project
dcaballe accepted D114335: [mlir][Vector] Add a vblendps-based impl for transpose8x8 (both intrin and inline_asm).
Nov 22 2021, 2:00 AM · Restricted Project
dcaballe added a comment to D114335: [mlir][Vector] Add a vblendps-based impl for transpose8x8 (both intrin and inline_asm).
Nov 22 2021, 2:00 AM · Restricted Project
dcaballe accepted D114164: [MLIR] Fix incorrect removal of source loop in loop fusion.

Thanks for addressing the feedback. LGTM. Just one minor comment. Feel free to fix it a commit it directly.

Nov 22 2021, 1:33 AM · Restricted Project

Nov 19 2021

dcaballe accepted D112985: [MLIR] Simplify Semi-affine expressions by rule based matching and replacing "expr - q * (expr floordiv q)" with "expr mod q" expression..

LGTM, thanks!

Nov 19 2021, 2:18 PM · Restricted Project
dcaballe added a comment to D111846: [LV] Drop integer poison-generating flags from instructions that need predication.

I talked to Nuno in private and he mentioned that I could go ahead and commit the changes and address any minor feedback in a separate commit since he is very busy right now.
I'll commit this on Monday if no other comments.

Nov 19 2021, 11:21 AM · Restricted Project

Nov 18 2021

dcaballe added a comment to D114164: [MLIR] Fix incorrect removal of source loop in loop fusion.

Thanks a lot for the fix! LGTM. Just a minor comment.

Nov 18 2021, 12:14 PM · Restricted Project

Nov 15 2021

dcaballe added a comment to D111846: [LV] Drop integer poison-generating flags from instructions that need predication.

Thanks, Florian! I addressed the feedback.

Nov 15 2021, 2:14 PM · Restricted Project
dcaballe updated the diff for D111846: [LV] Drop integer poison-generating flags from instructions that need predication.
  • Updated and added more comments.
  • Simplified redundant condition.
Nov 15 2021, 2:13 PM · Restricted Project
dcaballe added a comment to D112808: [MLIR] Simplify Semi Affine Expressions Using flattening.

ping @springerm and @dcaballe , can you please let me know whether the revision is ok or if any more changes are required to be made? Thanks in advance!

Nov 15 2021, 9:07 AM · Restricted Project

Nov 12 2021

dcaballe added a comment to D111846: [LV] Drop integer poison-generating flags from instructions that need predication.

Any other comments? :)

Nov 12 2021, 3:54 AM · Restricted Project

Nov 11 2021

dcaballe added inline comments to D112808: [MLIR] Simplify Semi Affine Expressions Using flattening.
Nov 11 2021, 2:02 PM · Restricted Project

Nov 9 2021

dcaballe updated the diff for D111846: [LV] Drop integer poison-generating flags from instructions that need predication.

Since it was already implemented in a previous version of the code, I restored the logic to drop flags
from Widen and WidenGEP recipes following Florian's suggestion. These changes only impact Florian's
test case. I modified the test case to also have a vector GEP feeding a masked load to also cover
that use case.

Nov 9 2021, 11:01 AM · Restricted Project
dcaballe updated the diff for D112054: [LV] Pre-commit test for D111846.

Added new test. Thanks, Florian!

Nov 9 2021, 10:44 AM · Restricted Project
dcaballe updated the diff for D111846: [LV] Drop integer poison-generating flags from instructions that need predication.

Fixed naming convention issues

Nov 9 2021, 8:27 AM · Restricted Project
dcaballe added inline comments to D111846: [LV] Drop integer poison-generating flags from instructions that need predication.
Nov 9 2021, 2:33 AM · Restricted Project

Nov 8 2021

dcaballe accepted D113347: [mlir][X86Vector] Add specialized vector.transpose lowering patterns for AVX2.

Thanks, Nicolas! LGTM.

Nov 8 2021, 12:52 PM · Restricted Project
dcaballe added a comment to D111846: [LV] Drop integer poison-generating flags from instructions that need predication.

Ok, I'll wait for Florian's approval to land this patch.

Nov 8 2021, 3:09 AM · Restricted Project
dcaballe updated the diff for D111846: [LV] Drop integer poison-generating flags from instructions that need predication.

Fixed failing test

Nov 8 2021, 3:08 AM · Restricted Project

Nov 7 2021

dcaballe added a comment to D112808: [MLIR] Simplify Semi Affine Expressions Using flattening.

Thanks a lot for the contribution! LGTM. Just some minor comments.

Nov 7 2021, 8:57 AM · Restricted Project