[X86] Teach 'getTargetShuffleMask' how to look through ISD::WrapperRIP when decoding...

The function 'getTargetShuffleMask' already knows how to deal with PSHUFB nodes
where the mask node is a load from constant pool, and the constant pool node
is wrapped by a X86ISD::Wrapper node. This patch extends that logic by teaching
it how to also look through X86ISD::WrapperRIP.

This helps function combineX86ShufflesRecusively to combine more shuffle
sequences containing PSHUFB nodes if we are in RIPRel PIC mode.

Before this change, llc (with -relocation-model=pic -march=x86-64) was unable
to decode a pshufb where the mask was loaded from a constant pool. For example,
the no-op shuffle from test 'x86-fold-pshufb.ll' was not folded into its
operand, so instead of generating a single 'movaps' the backend always
generated a sub-optimal 'movdqa + pshufb' sequence.

Added test x86-fold-pshufb.ll.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236863 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index cac5b54187b0ae1fdba8ca17077e0c3bd5acadd9..35517212acc01d677420d3dc488c4627897020c9 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -4352,7 +4352,8 @@ static bool getTargetShuffleMask(SDNode *N, MVT VT,
       return false;
 
     SDValue Ptr = MaskLoad->getBasePtr();
-    if (Ptr->getOpcode() == X86ISD::Wrapper)
+    if (Ptr->getOpcode() == X86ISD::Wrapper ||
+        Ptr->getOpcode() == X86ISD::WrapperRIP)
       Ptr = Ptr->getOperand(0);
 
     auto *MaskCP = dyn_cast<ConstantPoolSDNode>(Ptr);

diff --git a/test/CodeGen/X86/x86-fold-pshufb.ll b/test/CodeGen/X86/x86-fold-pshufb.ll
new file mode 100644 (file)
index 0000000..c29e592
--- /dev/null
+++ b/test/CodeGen/X86/x86-fold-pshufb.ll
@@ -0,0 +1,17 @@
+; RUN: llc -relocation-model=pic -march=x86-64 -mtriple=x86_64-unknown-unknown -mattr=+ssse3 < %s | FileCheck %s
+
+; Verify that the backend correctly folds the shuffle in function 'fold_pshufb'
+; into a simple load from constant pool.
+
+define <2 x i64> @fold_pshufb() {
+; CHECK-LABEL: fold_pshufb:
+; CHECK:       # BB#0:
+; CHECK-NEXT:    movaps {{.*#+}} xmm0 = [0,0,0,0,1,0,0,0,2,0,0,0,3,0,0,0]
+; CHECK-NEXT:    retq
+entry:
+  %0 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 1, i8 0, i8 0, i8 0, i8 2, i8 0, i8 0, i8 0, i8 3, i8 0, i8 0, i8 0>, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>)
+  %1 = bitcast <16 x i8> %0 to <2 x i64>
+  ret <2 x i64> %1
+}
+
+declare <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8>, <16 x i8>)