aengelke wrote:
> Adding this threshold check within isTrellis() feels somewhat unnatural. If
> compile time is a concern, could we simply check the size of functions (in
> terms of the number of blocks, as opposed to predecessor only) early in this
> pass and either skip it or switch to a fas
@@ -1592,6 +1603,11 @@ bool MachineBlockPlacement::hasBetterLayoutPredecessor(
BlockFrequency CandidateEdgeFreq = MBFI->getBlockFreq(BB) * RealSuccProb;
bool BadCFGConflict = false;
+ // Compile-time optimization: runtime is quadratic in the number of
aen
@@ -1030,6 +1036,11 @@ bool MachineBlockPlacement::isTrellis(
SmallPtrSet SeenPreds;
for (MachineBasicBlock *Succ : ViableSuccs) {
+// Compile-time optimization: runtime is quadratic in the number of
+// predecessors. For such uncommon cases, exit early.
+if (S
https://github.com/aengelke updated
https://github.com/llvm/llvm-project/pull/142584
>From 4cbc231699c11444cff73ff28b88dc0f3835c752 Mon Sep 17 00:00:00 2001
From: Alexis Engelke
Date: Wed, 4 Jun 2025 09:21:02 +
Subject: [PATCH] Move one check to beginning of function
Created using spr 1.3.
https://github.com/aengelke created
https://github.com/llvm/llvm-project/pull/142584
MachineBlockPlacement has quadratic runtime in the number of
predecessors: in some situation, for an edge, all predecessors of the
successor are considered.
Limit the number of considered predecessors to bound
aengelke wrote:
NB: I don't claim to fully understand what this code does, but it seems to be
safe to return a default value. [Example
generator](https://github.com/tpde2/tpde/blob/f6e87d2e97f49f403c12a27e7cf513a44f0f5dbc/tpde-llvm/test/filetest/many-preds.test)
to demonstrate the behavior, e.
aengelke wrote:
[c-t-t](http://llvm-compile-time-tracker.com/compare.php?from=2db9cb5fec35a7516b0e1d123d161ace78e14be6&to=c34780d18ac411ca2363eeff1cecd08aeb1d154a&stat=instructions:u)
-0.13% stage2-O3
https://github.com/llvm/llvm-project/pull/103400
_
https://github.com/aengelke created
https://github.com/llvm/llvm-project/pull/103400
Replace the DenseMap from blocks to their innermost loop a vector
indexed by block numbers, when possible.
This requires updating the loop info when blocks are renumbered. This
update is currently implemented b
https://github.com/aengelke updated
https://github.com/llvm/llvm-project/pull/103302
>From 6a2ac00a8424a4402475e2b7972bfb01330c3bf8 Mon Sep 17 00:00:00 2001
From: Alexis Engelke
Date: Tue, 13 Aug 2024 16:10:38 +
Subject: [PATCH 1/2] Only run instcombine in test case
Created using spr 1.3.5
https://github.com/aengelke updated
https://github.com/llvm/llvm-project/pull/103302
>From 6a2ac00a8424a4402475e2b7972bfb01330c3bf8 Mon Sep 17 00:00:00 2001
From: Alexis Engelke
Date: Tue, 13 Aug 2024 16:10:38 +
Subject: [PATCH 1/2] Only run instcombine in test case
Created using spr 1.3.5
aengelke wrote:
Done. Fun fact, llvm-reduce crashed on the input due to this bug.
https://github.com/llvm/llvm-project/pull/103302
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.
https://github.com/aengelke updated
https://github.com/llvm/llvm-project/pull/103302
>From 6a2ac00a8424a4402475e2b7972bfb01330c3bf8 Mon Sep 17 00:00:00 2001
From: Alexis Engelke
Date: Tue, 13 Aug 2024 16:10:38 +
Subject: [PATCH 1/2] Only run instcombine in test case
Created using spr 1.3.5
@@ -0,0 +1,576 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
UTC_ARGS: --version 5
+; RUN: opt -mtriple=arm64
-passes='inline,function(sroa,jump-threading,instcombine)' -S < %s
| FileCheck %s
aengelke wrote:
True, changed
https
https://github.com/aengelke updated
https://github.com/llvm/llvm-project/pull/103302
>From 6a2ac00a8424a4402475e2b7972bfb01330c3bf8 Mon Sep 17 00:00:00 2001
From: Alexis Engelke
Date: Tue, 13 Aug 2024 16:10:38 +
Subject: [PATCH] Only run instcombine in test case
Created using spr 1.3.5-bog
https://github.com/aengelke created
https://github.com/llvm/llvm-project/pull/103302
When looking at PHI operand for combining, only look at instructions and
arguments. The loop later iteraters over Arg's users, which is not
useful if Arg is a constant -- it's users are not meaningful and might
Author: Alexis Engelke
Date: 2024-08-13T07:50:05Z
New Revision: d0b1a582fd33e8c3605c027883c6deb35757f560
URL:
https://github.com/llvm/llvm-project/commit/d0b1a582fd33e8c3605c027883c6deb35757f560
DIFF:
https://github.com/llvm/llvm-project/commit/d0b1a582fd33e8c3605c027883c6deb35757f560.diff
LOG
Author: Alexis Engelke
Date: 2024-08-11T07:39:53Z
New Revision: 084d02577eb68dd2b6260b9b1d12a61631e8d799
URL:
https://github.com/llvm/llvm-project/commit/084d02577eb68dd2b6260b9b1d12a61631e8d799
DIFF:
https://github.com/llvm/llvm-project/commit/084d02577eb68dd2b6260b9b1d12a61631e8d799.diff
LOG
Author: Alexis Engelke
Date: 2024-08-11T05:55:56Z
New Revision: 1739628f12950e3ddbd80418750b93cdc11b48e8
URL:
https://github.com/llvm/llvm-project/commit/1739628f12950e3ddbd80418750b93cdc11b48e8
DIFF:
https://github.com/llvm/llvm-project/commit/1739628f12950e3ddbd80418750b93cdc11b48e8.diff
LOG
18 matches
Mail list logo
