Page MenuHomePhabricator

dpeixott (David Peixotto)
User

Projects

User does not belong to any projects.

User Details

User Since
Jan 24 2014, 9:55 AM (282 w, 3 d)

Recent Activity

Jan 13 2015

dpeixott committed rL225812: Fix maxLoopDepth computation in ScopInfo.
Fix maxLoopDepth computation in ScopInfo
Jan 13 2015, 10:33 AM
dpeixott closed D6913: Fix maxLoopDepth computation in ScopInfo by committing rL225812: Fix maxLoopDepth computation in ScopInfo.
Jan 13 2015, 10:33 AM

Jan 12 2015

dpeixott added a comment to D6913: Fix maxLoopDepth computation in ScopInfo.

I added a test case and also an explanation in the commit message on when we can hit this problem.

Jan 12 2015, 5:19 PM
dpeixott updated the diff for D6913: Fix maxLoopDepth computation in ScopInfo.

Added test case and fixed commit message.

Jan 12 2015, 5:17 PM

Jan 9 2015

dpeixott retitled D6913: Fix maxLoopDepth computation in ScopInfo from to Fix maxLoopDepth computation in ScopInfo.
Jan 9 2015, 5:56 PM

Nov 7 2014

dpeixott closed D6174: Safely generate new loop metadata node.

Closed by commit rL221550 (authored by @dpeixott).

Nov 7 2014, 1:55 PM
dpeixott retitled D6174: Safely generate new loop metadata node from to Safely generate new loop metadata node.
Nov 7 2014, 11:19 AM

Oct 30 2014

dpeixott accepted D6035: [Refactor][NFC] Remove unused argument..

LGTM

Oct 30 2014, 9:38 AM

Oct 22 2014

dpeixott closed D5897: Change the RegionSet type to a SetVector.

Closed by commit rL220423 (authored by @dpeixott).

Oct 22 2014, 1:49 PM

Oct 21 2014

dpeixott retitled D5897: Change the RegionSet type to a SetVector from to Change the RegionSet type to a SetVector.
Oct 21 2014, 5:17 PM

Oct 15 2014

dpeixott added a comment to D5762: [Polly][Unfinished][NFC] Restructure the command line options.
In D5762#10, @jdoerfert wrote:
In D5762#9, @dpeixott wrote:

This looks like a nice cleanup to me. I think centralizing the available options for polly makes sense. The major downside in my opinion is that we now have a bunch of global variables. Previously you could look at a command line option that is declared static (without external storage) and know it is only used in that file. Now we are exposing these all as global variables.

That is actually one of the reasons I came up with this patch. I want the options to be available to other LLVM pasess/projects, thus I have to expose them somehow. While I agree that global variables are not the best solution to this problem it is/was the easiest solution to allow me to configure Polly from within another project. Do you think we should wait/look for a different solution without the globals?

Oct 15 2014, 6:52 PM

Oct 14 2014

dpeixott added a comment to D5728: [Polly][Fix] Handle wrap rounds in the loop bounds given by ScalarEvolution..

This patch makes me nervous because I don't see why it is obviously safe to always promote to the next size without changing semantics. For example, with this loop:

Oct 14 2014, 10:26 AM
dpeixott added a comment to D5762: [Polly][Unfinished][NFC] Restructure the command line options.

This looks like a nice cleanup to me. I think centralizing the available options for polly makes sense. The major downside in my opinion is that we now have a bunch of global variables. Previously you could look at a command line option that is declared static (without external storage) and know it is only used in that file. Now we are exposing these all as global variables.

Oct 14 2014, 10:11 AM

Oct 9 2014

dpeixott accepted D5707: [Polly][Refactor][NFC] Simplify and clean the handling of (new) access relations.
Oct 9 2014, 12:10 PM

Oct 2 2014

dpeixott accepted D5582: [Polly] Put the parallel context into the function entry block.

Needs test case. Otherwise LGTM.

Oct 2 2014, 10:51 AM
dpeixott accepted D5581: [Polly] Add option to limit the maximal number of parallel threads..

LGTM with small test case demonstrating the command line option is respected.

Oct 2 2014, 10:37 AM

Oct 1 2014

dpeixott added a comment to D5559: [Polly] Align copied load/store instructions as the original..

LGTM. Could also fold the alignment setting into the creation.

Oct 1 2014, 9:51 AM

Sep 26 2014

dpeixott closed D5335: Ignore annotation function calls in cost computation.

Closed by commit rL218525 (authored by @dpeixott).

Sep 26 2014, 10:58 AM

Sep 25 2014

dpeixott added a comment to D5335: Ignore annotation function calls in cost computation.

ping

Sep 25 2014, 10:45 AM

Sep 24 2014

dpeixott closed D5305: Fix assertion in LICM doFinalization().

Closed by commit rL218387 (authored by @dpeixott).

Sep 24 2014, 9:58 AM

Sep 23 2014

dpeixott added a comment to D5305: Fix assertion in LICM doFinalization().

ping

Sep 23 2014, 3:44 PM

Sep 17 2014

dpeixott accepted D5386: Restore "[ARM, Fix] Fix emitLeading/TrailingFence on old ARM processors".
Sep 17 2014, 2:23 PM

Sep 15 2014

dpeixott added a comment to D5305: Fix assertion in LICM doFinalization().

ping

Sep 15 2014, 10:11 AM

Sep 12 2014

dpeixott added inline comments to D5329: [Polly] Delinearize _all_ accesses to a multi-dimensional array.
Sep 12 2014, 1:49 PM
dpeixott added inline comments to D5333: [IndVarSimplify] Widen signed loop compare instructions to enable additional optimizations..
Sep 12 2014, 12:29 PM
dpeixott retitled D5335: Ignore annotation function calls in cost computation from to Ignore annotation function calls in cost computation.
Sep 12 2014, 10:43 AM

Sep 11 2014

dpeixott updated the diff for D5305: Fix assertion in LICM doFinalization().

Updated test case to a simplified loop so that is easier to see why it causes a failure.

Sep 11 2014, 1:47 PM

Sep 10 2014

dpeixott retitled D5305: Fix assertion in LICM doFinalization() from to Fix assertion in LICM doFinalization().
Sep 10 2014, 5:46 PM
dpeixott added inline comments to D5227: [Polly] Support ModRef function behaviour.
Sep 10 2014, 10:55 AM
dpeixott added a comment to D5227: [Polly] Support ModRef function behaviour.
In D5227#10, @jdoerfert wrote:
In D5227#7, @dpeixott wrote:

As far as I know detailed mod-ref-behaviour is currently only returned in the case of intrinsic functions. These functions are defined with the appropriate values and tablegen generates some code to return the correct values. You could use an intrinsic function in your test (e.g. llvm.prefetch).

Unfortunatly I cannot export the "onlyReadsArgumentPointees" of llvm.prefetch to IR... it gets overaproximated to "onlyReadsMemory" for now.

Sep 10 2014, 10:47 AM
dpeixott added a comment to D5297: [Polly] Allow to generate a loop without the GuardBB.

Needs a test case to demonstrate the omission of the guard bb. Otherwise, LGTM.

Sep 10 2014, 10:32 AM

Sep 9 2014

dpeixott retitled D5283: [Polly] Map isl_ast for nodes to LLVM Loops from to [Polly] Map isl_ast for nodes to LLVM Loops.
Sep 9 2014, 6:21 PM
dpeixott closed D5276: [Polly] Add -e to test generation script.

Closed by commit rL217463 (authored by @dpeixott).

Sep 9 2014, 3:24 PM
dpeixott retitled D5276: [Polly] Add -e to test generation script from to [Polly] Add -e to test generation script.
Sep 9 2014, 1:35 PM
dpeixott added inline comments to D5275: [WIP] Print Loop* associated with isl_ast_for.
Sep 9 2014, 1:30 PM
dpeixott added a reviewer for D5275: [WIP] Print Loop* associated with isl_ast_for: jdoerfert.
Sep 9 2014, 12:20 PM
dpeixott retitled D5275: [WIP] Print Loop* associated with isl_ast_for from to [WIP] Print Loop* associated with isl_ast_for.
Sep 9 2014, 12:18 PM

Sep 8 2014

dpeixott added a comment to D5227: [Polly] Support ModRef function behaviour.

As far as I know detailed mod-ref-behaviour is currently only returned in the case of intrinsic functions. These functions are defined with the appropriate values and tablegen generates some code to return the correct values. You could use an intrinsic function in your test (e.g. llvm.prefetch).

Sep 8 2014, 2:19 PM
dpeixott added a comment to D5076: [Polly][Refactor] Cleanup runtime code generation.

Minor typo. Otherwise the patch looks reasonable to me.

Sep 8 2014, 10:32 AM

Jul 18 2014

dpeixott closed D4279: MC: support different sized constants in constant pools.

Closed by commit rL213387 (authored by @dpeixott).

Jul 18 2014, 9:13 AM · deleted

Jul 17 2014

dpeixott accepted D4279: MC: support different sized constants in constant pools.

For completeness I would add one more test for loading a smaller constant to an x register to show that we generate a 64-bit entry in the constant pool even if the value does not require the full 64-bits. You don't need another review for that change.

Jul 17 2014, 1:59 PM · deleted
dpeixott added a comment to D4279: MC: support different sized constants in constant pools.

Thanks, this is looking much better. I had one minor suggestion and a question about the APInt usage.

Jul 17 2014, 10:50 AM · deleted

Jul 14 2014

dpeixott added a comment to D4279: MC: support different sized constants in constant pools.

Thanks for the patch. Overall I like the direction you are going. Please see specific comments below.

Jul 14 2014, 11:32 AM · deleted

Jun 24 2014

dpeixott added a comment to D4278: AArch64: use 32bit registers in ldr-pseudo.s tests.

It's not clear to me how this change improves the test. Rather than just renaming, I think it would be better to

Jun 24 2014, 3:45 PM · deleted

Jun 23 2014

dpeixott added a comment to D4163: [AArch64] Implement pseudo LDR <reg>, =<literal/label>.

LGTM.

Jun 23 2014, 1:33 PM

Jun 20 2014

dpeixott added a comment to D4163: [AArch64] Implement pseudo LDR <reg>, =<literal/label>.

Minor cleanup, otherwise LGTM.

Jun 20 2014, 11:30 AM
dpeixott updated subscribers of D4171: [polly] Support the new DiagnosticRemarks.

Adding llvm-commits to CC.

Jun 20 2014, 10:46 AM
dpeixott accepted D4171: [polly] Support the new DiagnosticRemarks.

Some minor c++11 suggestions. LGTM with those changes.

Jun 20 2014, 10:17 AM

Jun 17 2014

dpeixott accepted D4163: [AArch64] Implement pseudo LDR <reg>, =<literal/label>.

LGTM with a suitable commit message explaining why we want to split out the implementation.

Jun 17 2014, 6:55 PM
dpeixott updated subscribers of D4163: [AArch64] Implement pseudo LDR <reg>, =<literal/label>.

Adding llvm-commits to CC.

Jun 17 2014, 5:17 PM
dpeixott added a comment to D4163: [AArch64] Implement pseudo LDR <reg>, =<literal/label>.

Overall this looks good to me, but I think we should split this into two commits: 1 factoring commit that moves the constant pools to a place that can be accessed by both AArch64 and ARM backends, and a second commit to implement the ldr= psuedo for AArch64.

Jun 17 2014, 4:58 PM