[llvm] 01acdea - [X86] Add test coverage for fshl/fshr with freeze

Simon Pilgrim via llvm-commits llvm-commits at lists.llvm.org
Sat Oct 22 10:40:07 PDT 2022


Author: Simon Pilgrim
Date: 2022-10-22T18:09:22+01:00
New Revision: 01acdea964959eea2787acd5081ca6acd1cfc3ca

URL: https://github.com/llvm/llvm-project/commit/01acdea964959eea2787acd5081ca6acd1cfc3ca
DIFF: https://github.com/llvm/llvm-project/commit/01acdea964959eea2787acd5081ca6acd1cfc3ca.diff

LOG: [X86] Add test coverage for fshl/fshr with freeze

Unlike most other freeze tests in the file, this is showing a missing SimplifyDemandedBits simplification instead of a merge of the ops

Added: 
    

Modified: 
    llvm/test/CodeGen/X86/freeze-binary.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/X86/freeze-binary.ll b/llvm/test/CodeGen/X86/freeze-binary.ll
index 552cd3325747..3f4dfb26f9b5 100644
--- a/llvm/test/CodeGen/X86/freeze-binary.ll
+++ b/llvm/test/CodeGen/X86/freeze-binary.ll
@@ -707,3 +707,51 @@ define <4 x i32> @freeze_rotr_vec(<4 x i32> %a0) nounwind {
   ret <4 x i32> %z
 }
 declare <4 x i32> @llvm.fshr.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
+
+define i32 @freeze_fshl(i32 %a0, i32 %a1, i32 %a2) nounwind {
+; X86-LABEL: freeze_fshl:
+; X86:       # %bb.0:
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    shrdl $27, %ecx, %eax
+; X86-NEXT:    shldl $27, %edx, %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: freeze_fshl:
+; X64:       # %bb.0:
+; X64-NEXT:    movl %edi, %eax
+; X64-NEXT:    shldl $5, %esi, %eax
+; X64-NEXT:    shldl $27, %edx, %eax
+; X64-NEXT:    retq
+  %f1 = freeze i32 %a1
+  %f2 = freeze i32 %a2
+  %x = call i32 @llvm.fshl.i32(i32 %a0, i32 %f1, i32 5)
+  %y = freeze i32 %x
+  %z = call i32 @llvm.fshl.i32(i32 %y, i32 %f2, i32 27)
+  ret i32 %z
+}
+
+define i32 @freeze_fshr(i32 %a0, i32 %a1, i32 %a2) nounwind {
+; X86-LABEL: freeze_fshr:
+; X86:       # %bb.0:
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    shrdl $1, %ecx, %eax
+; X86-NEXT:    shldl $1, %edx, %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: freeze_fshr:
+; X64:       # %bb.0:
+; X64-NEXT:    movl %edi, %eax
+; X64-NEXT:    shldl $31, %esi, %eax
+; X64-NEXT:    shldl $1, %edx, %eax
+; X64-NEXT:    retq
+  %f1 = freeze i32 %a1
+  %f2 = freeze i32 %a2
+  %x = call i32 @llvm.fshr.i32(i32 %a0, i32 %f1, i32 1)
+  %y = freeze i32 %x
+  %z = call i32 @llvm.fshr.i32(i32 %y, i32 %f2, i32 31)
+  ret i32 %z
+}


        


More information about the llvm-commits mailing list