address comments

Author: AlexMaclean

llvm/docs/NVPTXUsage.rst

@@ -330,7 +330,7 @@ Syntax:
 
 .. code-block:: llvm
 
-    declare i32 @llvm.nvvm.fshl.clamp.i32(i32 %a, i32 %b, i32 %c)
+    declare i32 @llvm.nvvm.fshl.clamp.i32(i32 %hi, i32 %lo, i32 %n)
 
 Overview:
 """""""""
@@ -344,11 +344,11 @@ Semantics:
 """"""""""
 
 The '``llvm.nvvm.fshl.clamp``' family of intrinsic functions performs a clamped
-funnel shift left: the first two values are concatenated as { %a : %b } (%a is
-the most significant bits of the wide value), the combined value is shifted
+funnel shift left: the first two values are concatenated as { %hi : %lo } (%hi
+is the most significant bits of the wide value), the combined value is shifted
 left, and the most significant bits are extracted to produce a result that is
 the same size as the original arguments. The shift amount is the minimum of the
-value of %c and the bit width of the integer type.
+value of %n and the bit width of the integer type.
 
 '``llvm.nvvm.fshr.clamp.*``' Intrinsic
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
@@ -358,7 +358,7 @@ Syntax:
 
 .. code-block:: llvm
 
-    declare i32 @llvm.nvvm.fshr.clamp.i32(i32 %a, i32 %b, i32 %c)
+    declare i32 @llvm.nvvm.fshr.clamp.i32(i32 %hi, i32 %lo, i32 %n)
 
 Overview:
 """""""""
@@ -372,11 +372,11 @@ Semantics:
 """"""""""
 
 The '``llvm.nvvm.fshr.clamp``' family of intrinsic functions performs a clamped
-funnel shift right: the first two values are concatenated as { %a : %b } (%a is
-the most significant bits of the wide value), the combined value is shifted
+funnel shift right: the first two values are concatenated as { %hi : %lo } (%hi
+is the most significant bits of the wide value), the combined value is shifted
 right, and the least significant bits are extracted to produce a result that is
 the same size as the original arguments. The shift amount is the minimum of the
-value of %c and the bit width of the integer type.
+value of %n and the bit width of the integer type.
 
 
 Other Intrinsics

llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.cpp

@@ -395,13 +395,13 @@ static Instruction *simplifyNvvmIntrinsic(IntrinsicInst *II, InstCombiner &IC) {
         II->getArgOperand(0), II->getName());
 
   case SCP_FunnelShiftClamp: {
-    // Canoncialize a clamping funnel shift to the generic llvm funnel shift
+    // Canonicalize a clamping funnel shift to the generic llvm funnel shift
     // when possible, as this is easier for llvm to optimize further.
     if (const auto *ShiftConst = dyn_cast<ConstantInt>(II->getArgOperand(2))) {
+      const bool IsLeft = II->getIntrinsicID() == Intrinsic::nvvm_fshl_clamp;
       if (ShiftConst->getZExtValue() >= II->getType()->getIntegerBitWidth())
-        return IC.replaceInstUsesWith(*II, II->getArgOperand(1));
+        return IC.replaceInstUsesWith(*II, II->getArgOperand(IsLeft ? 1 : 0));
 
-      const bool IsLeft = II->getIntrinsicID() == Intrinsic::nvvm_fshl_clamp;
       const unsigned FshIID = IsLeft ? Intrinsic::fshl : Intrinsic::fshr;
       return CallInst::Create(Intrinsic::getOrInsertDeclaration(
                                   II->getModule(), FshIID, II->getType()),

llvm/test/Transforms/InstCombine/NVPTX/nvvm-intrins.ll

@@ -385,33 +385,47 @@ define float @test_sqrt_rn_f_ftz(float %a) #0 {
 }
 
 ; CHECK-LABEL: @test_fshl_clamp_1
-define i32 @test_fshl_clamp_1(i32 %a, i32 %b, i32 %c) {
+define i32 @test_fshl_clamp_1(i32 %a, i32 %b) {
 ; CHECK: call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 3)
   %call = call i32 @llvm.nvvm.fshl.clamp.i32(i32 %a, i32 %b, i32 3)
   ret i32 %call
 }
 
 ; CHECK-LABEL: @test_fshl_clamp_2
-define i32 @test_fshl_clamp_2(i32 %a, i32 %b, i32 %c) {
+define i32 @test_fshl_clamp_2(i32 %a, i32 %b) {
 ; CHECK: ret i32 %b
   %call = call i32 @llvm.nvvm.fshl.clamp.i32(i32 %a, i32 %b, i32 300)
   ret i32 %call
 }
 
+; CHECK-LABEL: @test_fshl_clamp_3
+define i32 @test_fshl_clamp_3(i32 %a, i32 %b, i32 %c) {
+; CHECK: call i32 @llvm.nvvm.fshl.clamp.i32(i32 %a, i32 %b, i32 %c)
+  %call = call i32 @llvm.nvvm.fshl.clamp.i32(i32 %a, i32 %b, i32 %c)
+  ret i32 %call
+}
+
 ; CHECK-LABEL: @test_fshr_clamp_1
-define i32 @test_fshr_clamp_1(i32 %a, i32 %b, i32 %c) {
+define i32 @test_fshr_clamp_1(i32 %a, i32 %b) {
 ; CHECK: call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 29)
   %call = call i32 @llvm.nvvm.fshr.clamp.i32(i32 %a, i32 %b, i32 3)
   ret i32 %call
 }
 
 ; CHECK-LABEL: @test_fshr_clamp_2
-define i32 @test_fshr_clamp_2(i32 %a, i32 %b, i32 %c) {
-; CHECK: ret i32 %b
+define i32 @test_fshr_clamp_2(i32 %a, i32 %b) {
+; CHECK: ret i32 %a
   %call = call i32 @llvm.nvvm.fshr.clamp.i32(i32 %a, i32 %b, i32 300)
   ret i32 %call
 }
 
+; CHECK-LABEL: @test_fshr_clamp_3
+define i32 @test_fshr_clamp_3(i32 %a, i32 %b, i32 %c) {
+; CHECK: call i32 @llvm.nvvm.fshr.clamp.i32(i32 %a, i32 %b, i32 %c)
+  %call = call i32 @llvm.nvvm.fshr.clamp.i32(i32 %a, i32 %b, i32 %c)
+  ret i32 %call
+}
+
 declare double @llvm.nvvm.add.rn.d(double, double)
 declare float @llvm.nvvm.add.rn.f(float, float)
 declare float @llvm.nvvm.add.rn.ftz.f(float, float)