[llvm-dev] Legacy PM deprecation for optimization pipeline timeline

Arthur Eubanks via llvm-dev llvm-dev at lists.llvm.org
Tue Apr 27 13:10:43 PDT 2021


Yes we should make sure tests pass with the legacy PM as the default,
treating the bot like any other bot.
I've fixed the two func-attrs failures.
The two OpenMP failures look like real issues under the legacy PM that need
to be investigated.

diff --git a/llvm/test/Transforms/OpenMP/hide_mem_transfer_latency.ll
b/llvm/test/Transforms/OpenMP/hide_mem_transfer_latency.ll
index abc0d8769c24..fd2e499d9a28 100644
--- a/llvm/test/Transforms/OpenMP/hide_mem_transfer_latency.ll
+++ b/llvm/test/Transforms/OpenMP/hide_mem_transfer_latency.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
UTC_ARGS: -p --function-signature --scrub-attributes
-; RUN: opt -S -openmp-opt-cgscc -aa-pipeline=basic-aa
-openmp-hide-memory-transfer-latency < %s | FileCheck %s
+; RUN: opt -S -openmp-opt-cgscc -enable-new-pm=0
-openmp-hide-memory-transfer-latency < %s | FileCheck %s
 ; RUN: opt -S -passes=openmp-opt-cgscc -aa-pipeline=basic-aa
-openmp-hide-memory-transfer-latency < %s | FileCheck %s
 target datalayout =
"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"

to repro.

On Tue, Apr 27, 2021 at 12:40 PM Michael Kruse <llvmdev at meinersbur.de>
wrote:

> Am Di., 27. Apr. 2021 um 10:59 Uhr schrieb Arthur Eubanks via llvm-dev
> <llvm-dev at lists.llvm.org>:
> > We'd still have to have support in opt for the legacy PM to run IR
> passes that are part of the codegen pipeline. But we could still remove the
> -enable-new-pm flag after moving tests for passes in the optimization
> pipeline to use -passes.
>
> It might be worth noting that the regression tests are currently
> failing with ENABLE_EXPERIMENTAL_NEW_PASS_MANAGER=OFF. If it is still
> supported till the next release, they should be fixed.
>
> See http://meinersbur.de:8011/#/builders/148
>
> Michael
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20210427/7912c99f/attachment.html>


More information about the llvm-dev mailing list