[llvm] 179e15d - [SystemZ] Optimize bcmp calls (PR47420)

[Dávid Bolvanský via llvm-commits]

Author: Dávid Bolvanský
Date: 2020-09-25T17:55:39+02:00
New Revision: 179e15d53acdb1fc27e9e2c8f7e1d1e08dcf4a89

URL: https://github.com/llvm/llvm-project/commit/179e15d53acdb1fc27e9e2c8f7e1d1e08dcf4a89
DIFF: https://github.com/llvm/llvm-project/commit/179e15d53acdb1fc27e9e2c8f7e1d1e08dcf4a89.diff

LOG: [SystemZ] Optimize bcmp calls (PR47420)

Solves https://bugs.llvm.org/show_bug.cgi?id=47420

Reviewed By: uweigand

Differential Revision: https://reviews.llvm.org/D87988

Added: 
    llvm/test/CodeGen/SystemZ/bcmp.ll

Modified: 
    llvm/include/llvm/CodeGen/SelectionDAGTargetInfo.h
    llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
    llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h

Removed: 
    


################################################################################
diff  --git a/llvm/include/llvm/CodeGen/SelectionDAGTargetInfo.h b/llvm/include/llvm/CodeGen/SelectionDAGTargetInfo.h
index 014523f1af6a..78f6fc6656fa 100644
--- a/llvm/include/llvm/CodeGen/SelectionDAGTargetInfo.h
+++ b/llvm/include/llvm/CodeGen/SelectionDAGTargetInfo.h
@@ -85,7 +85,7 @@ class SelectionDAGTargetInfo {
     return SDValue();
   }
 
-  /// Emit target-specific code that performs a memcmp, in cases where that is
+  /// Emit target-specific code that performs a memcmp/bcmp, in cases where that is
   /// faster than a libcall. The first returned SDValue is the result of the
   /// memcmp and the second is the chain. Both SDValues can be null if a normal
   /// libcall should be used.

diff  --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index dcfa6e405c74..eef6a314a21f 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -7245,12 +7245,12 @@ void SelectionDAGBuilder::processIntegerCallValue(const Instruction &I,
   setValue(&I, Value);
 }
 
-/// See if we can lower a memcmp call into an optimized form. If so, return
+/// See if we can lower a memcmp/bcmp call into an optimized form. If so, return
 /// true and lower it. Otherwise return false, and it will be lowered like a
 /// normal call.
 /// The caller already checked that \p I calls the appropriate LibFunc with a
 /// correct prototype.
-bool SelectionDAGBuilder::visitMemCmpCall(const CallInst &I) {
+bool SelectionDAGBuilder::visitMemCmpBCmpCall(const CallInst &I) {
   const Value *LHS = I.getArgOperand(0), *RHS = I.getArgOperand(1);
   const Value *Size = I.getArgOperand(2);
   const ConstantInt *CSize = dyn_cast<ConstantInt>(Size);
@@ -7561,6 +7561,10 @@ void SelectionDAGBuilder::visitCall(const CallInst &I) {
         LibInfo->hasOptimizedCodeGen(Func)) {
       switch (Func) {
       default: break;
+      case LibFunc_bcmp:
+        if (visitMemCmpBCmpCall(I))
+          return;
+        break;
       case LibFunc_copysign:
       case LibFunc_copysignf:
       case LibFunc_copysignl:
@@ -7662,7 +7666,7 @@ void SelectionDAGBuilder::visitCall(const CallInst &I) {
           return;
         break;
       case LibFunc_memcmp:
-        if (visitMemCmpCall(I))
+        if (visitMemCmpBCmpCall(I))
           return;
         break;
       case LibFunc_mempcpy:

diff  --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
index 4904134a7d40..89b5de0a9f21 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
@@ -740,7 +740,7 @@ class SelectionDAGBuilder {
   void visitFence(const FenceInst &I);
   void visitPHI(const PHINode &I);
   void visitCall(const CallInst &I);
-  bool visitMemCmpCall(const CallInst &I);
+  bool visitMemCmpBCmpCall(const CallInst &I);
   bool visitMemPCpyCall(const CallInst &I);
   bool visitMemChrCall(const CallInst &I);
   bool visitStrCpyCall(const CallInst &I, bool isStpcpy);

diff  --git a/llvm/test/CodeGen/SystemZ/bcmp.ll b/llvm/test/CodeGen/SystemZ/bcmp.ll
new file mode 100644
index 000000000000..fc922abe44a1
--- /dev/null
+++ b/llvm/test/CodeGen/SystemZ/bcmp.ll
@@ -0,0 +1,29 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+
+declare signext i32 @bcmp(i8* nocapture, i8* nocapture, i64)
+
+define zeroext i1 @test_bcmp_eq_0(i8* nocapture readonly %A, i8* nocapture readonly %B) {
+; CHECK-LABEL: test_bcmp_eq_0:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    clc 0(2,%r3), 0(%r2)
+; CHECK-NEXT:    ipm %r0
+; CHECK-NEXT:    afi %r0, -268435456
+; CHECK-NEXT:    risbg %r2, %r0, 63, 191, 33
+; CHECK-NEXT:    br %r14
+  %c = tail call signext i32 @bcmp(i8* %A, i8* %B, i64 2)
+  %res  = icmp eq i32 %c, 0
+  ret i1 %res
+}
+
+define signext i32 @test_bcmp(i8* nocapture readonly %A, i8* nocapture readonly %B) {
+; CHECK-LABEL: test_bcmp:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    clc 0(2,%r3), 0(%r2)
+; CHECK-NEXT:    ipm %r0
+; CHECK-NEXT:    sllg %r0, %r0, 34
+; CHECK-NEXT:    srag %r2, %r0, 62
+; CHECK-NEXT:    br %r14
+  %res = tail call signext i32 @bcmp(i8* %A, i8* %B, i64 2)
+  ret i32 %res
+}