[PATCH] D134769: [ASan] Workaround for a performance problem in StackSafetyAnalysis pass.
Vitaly Buka via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Tue Sep 27 16:57:26 PDT 2022
vitalybuka added inline comments.
================
Comment at: llvm/test/Analysis/StackSafetyAnalysis/stack-safety-max-allocas.ll:1
+; REQUIRES: x86-registered-target
+
----------------
not needed?
================
Comment at: llvm/test/Analysis/StackSafetyAnalysis/stack-safety-max-allocas.ll:3
+
+; RUN: opt < %s -S -asan-instrumentation-with-call-threshold=0 -passes='asan-pipeline' -asan-use-stack-safety=1 \
+; RUN: -stack-safety-max-allocas=0 -o - | FileCheck %s --check-prefixes=ALLOCAS
----------------
broken lines hard to read when you have many of them, all RUNs looks the same them
long lines not nice in some editors, but very convenient to compare RUNs
so I prefer long RUN line exactly for readability
================
Comment at: llvm/test/Analysis/StackSafetyAnalysis/stack-safety-max-allocas.ll:3
+
+; RUN: opt < %s -S -asan-instrumentation-with-call-threshold=0 -passes='asan-pipeline' -asan-use-stack-safety=1 \
+; RUN: -stack-safety-max-allocas=0 -o - | FileCheck %s --check-prefixes=ALLOCAS
----------------
vitalybuka wrote:
> broken lines hard to read when you have many of them, all RUNs looks the same them
> long lines not nice in some editors, but very convenient to compare RUNs
>
> so I prefer long RUN line exactly for readability
>
instead of asan please directly use the pass: -passes="print-stack-safety"
example local.ll
================
Comment at: llvm/test/Analysis/StackSafetyAnalysis/stack-safety-max-allocas.ll:4
+; RUN: opt < %s -S -asan-instrumentation-with-call-threshold=0 -passes='asan-pipeline' -asan-use-stack-safety=1 \
+; RUN: -stack-safety-max-allocas=0 -o - | FileCheck %s --check-prefixes=ALLOCAS
+
----------------
can you do at least =1 and have one function with 1 and one function with 2 allocas
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D134769/new/
https://reviews.llvm.org/D134769
More information about the llvm-commits
mailing list