[PATCH] D30317: [X86] Clean up test/CodeGen/X86/2006-03-01-InstrSchedBug.ll

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added a reviewer: RKSimon.
RKSimon added inline comments.


================
Comment at: test/CodeGen/X86/2006-03-01-InstrSchedBug.ll:1
-; RUN: llc < %s -march=x86 | not grep "subl.*%esp"
+; RUN: llc < %s -march=x86 | FileCheck %s
+
----------------
Might as well regenerate the test file with the update script


================
Comment at: test/CodeGen/X86/2006-03-01-InstrSchedBug.ll:3
+
+target triple = "i686-unknown-unknown"
 
----------------
For a trivial test like this removing the -march and replacing with this triple (-mtriple=i686-unknown-unknown) will tidy things up more. 


================
Comment at: test/CodeGen/X86/2006-03-01-InstrSchedBug.ll:6
 define i32 @f(i32 %a, i32 %b) {
-        %tmp.2 = mul i32 %a, %a         ; <i32> [#uses=1]
-        %tmp.5 = shl i32 %a, 1          ; <i32> [#uses=1]
-        %tmp.6 = mul i32 %tmp.5, %b             ; <i32> [#uses=1]
-        %tmp.10 = mul i32 %b, %b                ; <i32> [#uses=1]
-        %tmp.7 = add i32 %tmp.10, %tmp.2                ; <i32> [#uses=1]
-        %tmp.11 = add i32 %tmp.7, %tmp.6                ; <i32> [#uses=1]
-        ret i32 %tmp.11
+entry:
+  %tmp.2 = mul i32 %a, %a
----------------
I'd recommend not adding this as it might affect the reproducibility of the exact codegen that caused the old bug (unlikely but maybe).


https://reviews.llvm.org/D30317