[InstCombine] reduce funnel-shift i16 X, X, 8 to bswap X

Prefer the more exact intrinsic to remove a use of the input value
and possibly make further transforms easier (we will still need
to match patterns with funnel-shift of wider types as pieces of
bswap, especially if we want to canonicalize to funnel-shift with
constant shift amount). Discussed in D46760.

llvm-svn: 364187

2 files changed, 16 insertions, 3 deletions

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index 1b247245c4c..63f313e1f7f 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -1983,6 +1983,13 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
       if (match(Op0, m_ZeroInt()) || match(Op0, m_Undef()))
         return BinaryOperator::CreateLShr(Op1,
                                           ConstantExpr::getSub(WidthC, ShAmtC));
+
+      // fshl i16 X, X, 8 --> bswap i16 X (reduce to more-specific form)
+      if (Op0 == Op1 && BitWidth == 16 && match(ShAmtC, m_SpecificInt(8))) {
+        Module *Mod = II->getModule();
+        Function *Bswap = Intrinsic::getDeclaration(Mod, Intrinsic::bswap, Ty);
+        return CallInst::Create(Bswap, { Op0 });
+      }
     }
 
     // Left or right might be masked.
diff --git a/llvm/test/Transforms/InstCombine/fsh.ll b/llvm/test/Transforms/InstCombine/fsh.ll
index 92be6d3795d..4300f041773 100644
--- a/llvm/test/Transforms/InstCombine/fsh.ll
+++ b/llvm/test/Transforms/InstCombine/fsh.ll
@@ -544,7 +544,7 @@ declare <3 x i16> @llvm.fshl.v3i16(<3 x i16>, <3 x i16>, <3 x i16>)
 
 define i16 @fshl_bswap(i16 %x) {
 ; CHECK-LABEL: @fshl_bswap(
-; CHECK-NEXT:    [[R:%.*]] = call i16 @llvm.fshl.i16(i16 [[X:%.*]], i16 [[X]], i16 8)
+; CHECK-NEXT:    [[R:%.*]] = call i16 @llvm.bswap.i16(i16 [[X:%.*]])
 ; CHECK-NEXT:    ret i16 [[R]]
 ;
   %r = call i16 @llvm.fshl.i16(i16 %x, i16 %x, i16 8)
@@ -553,7 +553,7 @@ define i16 @fshl_bswap(i16 %x) {
 
 define i16 @fshr_bswap(i16 %x) {
 ; CHECK-LABEL: @fshr_bswap(
-; CHECK-NEXT:    [[R:%.*]] = call i16 @llvm.fshl.i16(i16 [[X:%.*]], i16 [[X]], i16 8)
+; CHECK-NEXT:    [[R:%.*]] = call i16 @llvm.bswap.i16(i16 [[X:%.*]])
 ; CHECK-NEXT:    ret i16 [[R]]
 ;
   %r = call i16 @llvm.fshr.i16(i16 %x, i16 %x, i16 8)
@@ -562,13 +562,15 @@ define i16 @fshr_bswap(i16 %x) {
 
 define <3 x i16> @fshl_bswap_vector(<3 x i16> %x) {
 ; CHECK-LABEL: @fshl_bswap_vector(
-; CHECK-NEXT:    [[R:%.*]] = call <3 x i16> @llvm.fshl.v3i16(<3 x i16> [[X:%.*]], <3 x i16> [[X]], <3 x i16> <i16 8, i16 8, i16 8>)
+; CHECK-NEXT:    [[R:%.*]] = call <3 x i16> @llvm.bswap.v3i16(<3 x i16> [[X:%.*]])
 ; CHECK-NEXT:    ret <3 x i16> [[R]]
 ;
   %r = call <3 x i16> @llvm.fshl.v3i16(<3 x i16> %x, <3 x i16> %x, <3 x i16> <i16 8, i16 8, i16 8>)
   ret <3 x i16> %r
 }
 
+; Negative test
+
 define i16 @fshl_bswap_wrong_op(i16 %x, i16 %y) {
 ; CHECK-LABEL: @fshl_bswap_wrong_op(
 ; CHECK-NEXT:    [[R:%.*]] = call i16 @llvm.fshl.i16(i16 [[X:%.*]], i16 [[Y:%.*]], i16 8)
@@ -578,6 +580,8 @@ define i16 @fshl_bswap_wrong_op(i16 %x, i16 %y) {
   ret i16 %r
 }
 
+; Negative test
+
 define i16 @fshr_bswap_wrong_amount(i16 %x) {
 ; CHECK-LABEL: @fshr_bswap_wrong_amount(
 ; CHECK-NEXT:    [[R:%.*]] = call i16 @llvm.fshl.i16(i16 [[X:%.*]], i16 [[X]], i16 12)
@@ -587,6 +591,8 @@ define i16 @fshr_bswap_wrong_amount(i16 %x) {
   ret i16 %r
 }
 
+; Negative test
+
 define i32 @fshl_bswap_wrong_width(i32 %x) {
 ; CHECK-LABEL: @fshl_bswap_wrong_width(
 ; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 8)