Clean up some of the shuffle decoding code for UNPCK instructions. Add instruction...

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@145924 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/InstPrinter/X86InstComments.cpp b/lib/Target/X86/InstPrinter/X86InstComments.cpp
index 59920ad9acf309abc98c0b89e3da1b7cf7ed0b9a..6e4b1b985adec5816d92cffed7961927d218150a 100644 (file)
--- a/lib/Target/X86/InstPrinter/X86InstComments.cpp
+++ b/lib/Target/X86/InstPrinter/X86InstComments.cpp
@@ -106,28 +106,92 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     // FALL THROUGH.
   case X86::PUNPCKHBWrm:
     Src1Name = getRegName(MI->getOperand(0).getReg());
-    DecodePUNPCKHMask(16, ShuffleMask);
+    DecodeUNPCKHMask(MVT::v16i8, ShuffleMask);
+    break;
+  case X86::VPUNPCKHBWrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKHBWrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKHMask(MVT::v16i8, ShuffleMask);
+    break;
+  case X86::VPUNPCKHBWYrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKHBWYrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKHMask(MVT::v32i8, ShuffleMask);
     break;
   case X86::PUNPCKHWDrr:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::PUNPCKHWDrm:
     Src1Name = getRegName(MI->getOperand(0).getReg());
-    DecodePUNPCKHMask(8, ShuffleMask);
+    DecodeUNPCKHMask(MVT::v8i16, ShuffleMask);
+    break;
+  case X86::VPUNPCKHWDrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKHWDrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKHMask(MVT::v8i16, ShuffleMask);
+    break;
+  case X86::VPUNPCKHWDYrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKHWDYrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKHMask(MVT::v16i16, ShuffleMask);
     break;
   case X86::PUNPCKHDQrr:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::PUNPCKHDQrm:
     Src1Name = getRegName(MI->getOperand(0).getReg());
-    DecodePUNPCKHMask(4, ShuffleMask);
+    DecodeUNPCKHMask(MVT::v4i32, ShuffleMask);
+    break;
+  case X86::VPUNPCKHDQrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKHDQrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKHMask(MVT::v4i32, ShuffleMask);
+    break;
+  case X86::VPUNPCKHDQYrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKHDQYrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKHMask(MVT::v8i32, ShuffleMask);
     break;
   case X86::PUNPCKHQDQrr:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::PUNPCKHQDQrm:
     Src1Name = getRegName(MI->getOperand(0).getReg());
-    DecodePUNPCKHMask(2, ShuffleMask);
+    DecodeUNPCKHMask(MVT::v2i64, ShuffleMask);
+    break;
+  case X86::VPUNPCKHQDQrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKHQDQrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKHMask(MVT::v2i64, ShuffleMask);
+    break;
+  case X86::VPUNPCKHQDQYrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKHQDQYrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKHMask(MVT::v4i64, ShuffleMask);
     break;
 
   case X86::PUNPCKLBWrr:
@@ -135,28 +199,92 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     // FALL THROUGH.
   case X86::PUNPCKLBWrm:
     Src1Name = getRegName(MI->getOperand(0).getReg());
-    DecodePUNPCKLBWMask(16, ShuffleMask);
+    DecodeUNPCKLMask(MVT::v16i8, ShuffleMask);
+    break;
+  case X86::VPUNPCKLBWrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKLBWrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKLMask(MVT::v16i8, ShuffleMask);
+    break;
+  case X86::VPUNPCKLBWYrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKLBWYrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKLMask(MVT::v32i8, ShuffleMask);
     break;
   case X86::PUNPCKLWDrr:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::PUNPCKLWDrm:
     Src1Name = getRegName(MI->getOperand(0).getReg());
-    DecodePUNPCKLWDMask(8, ShuffleMask);
+    DecodeUNPCKLMask(MVT::v8i16, ShuffleMask);
+    break;
+  case X86::VPUNPCKLWDrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKLWDrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKLMask(MVT::v8i16, ShuffleMask);
+    break;
+  case X86::VPUNPCKLWDYrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKLWDYrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKLMask(MVT::v16i16, ShuffleMask);
     break;
   case X86::PUNPCKLDQrr:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::PUNPCKLDQrm:
     Src1Name = getRegName(MI->getOperand(0).getReg());
-    DecodePUNPCKLDQMask(4, ShuffleMask);
+    DecodeUNPCKLMask(MVT::v4i32, ShuffleMask);
+    break;
+  case X86::VPUNPCKLDQrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKLDQrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKLMask(MVT::v4i32, ShuffleMask);
+    break;
+  case X86::VPUNPCKLDQYrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKLDQYrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKLMask(MVT::v8i32, ShuffleMask);
     break;
   case X86::PUNPCKLQDQrr:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::PUNPCKLQDQrm:
     Src1Name = getRegName(MI->getOperand(0).getReg());
-    DecodePUNPCKLQDQMask(2, ShuffleMask);
+    DecodeUNPCKLMask(MVT::v2i64, ShuffleMask);
+    break;
+  case X86::VPUNPCKLQDQrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKLQDQrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKLMask(MVT::v2i64, ShuffleMask);
+    break;
+  case X86::VPUNPCKLQDQYrr:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPUNPCKLQDQYrm:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeUNPCKLMask(MVT::v4i64, ShuffleMask);
     break;
 
   case X86::SHUFPDrri:
@@ -217,14 +345,14 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::UNPCKLPDrm:
-    DecodeUNPCKLPMask(MVT::v2f64, ShuffleMask);
+    DecodeUNPCKLMask(MVT::v2f64, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VUNPCKLPDrr:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VUNPCKLPDrm:
-    DecodeUNPCKLPMask(MVT::v2f64, ShuffleMask);
+    DecodeUNPCKLMask(MVT::v2f64, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -232,7 +360,7 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VUNPCKLPDYrm:
-    DecodeUNPCKLPMask(MVT::v4f64, ShuffleMask);
+    DecodeUNPCKLMask(MVT::v4f64, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -240,14 +368,14 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::UNPCKLPSrm:
-    DecodeUNPCKLPMask(MVT::v4f32, ShuffleMask);
+    DecodeUNPCKLMask(MVT::v4f32, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VUNPCKLPSrr:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VUNPCKLPSrm:
-    DecodeUNPCKLPMask(MVT::v4f32, ShuffleMask);
+    DecodeUNPCKLMask(MVT::v4f32, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -255,7 +383,7 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VUNPCKLPSYrm:
-    DecodeUNPCKLPMask(MVT::v8f32, ShuffleMask);
+    DecodeUNPCKLMask(MVT::v8f32, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -263,14 +391,14 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::UNPCKHPDrm:
-    DecodeUNPCKHPMask(MVT::v2f64, ShuffleMask);
+    DecodeUNPCKHMask(MVT::v2f64, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VUNPCKHPDrr:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VUNPCKHPDrm:
-    DecodeUNPCKHPMask(MVT::v2f64, ShuffleMask);
+    DecodeUNPCKHMask(MVT::v2f64, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -278,7 +406,7 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VUNPCKHPDYrm:
-    DecodeUNPCKLPMask(MVT::v4f64, ShuffleMask);
+    DecodeUNPCKHMask(MVT::v4f64, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -286,14 +414,14 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::UNPCKHPSrm:
-    DecodeUNPCKHPMask(MVT::v4f32, ShuffleMask);
+    DecodeUNPCKHMask(MVT::v4f32, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VUNPCKHPSrr:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VUNPCKHPSrm:
-    DecodeUNPCKHPMask(MVT::v4f32, ShuffleMask);
+    DecodeUNPCKHMask(MVT::v4f32, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -301,7 +429,7 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VUNPCKHPSYrm:
-    DecodeUNPCKHPMask(MVT::v8f32, ShuffleMask);
+    DecodeUNPCKHMask(MVT::v8f32, ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;

diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
index daca2401678df5262a76ca7bd944115f2f24626f..e7631b655777f92edbc45d93d3178ff873ecfddf 100644 (file)
--- a/lib/Target/X86/Utils/X86ShuffleDecode.cpp
+++ b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
@@ -95,39 +95,6 @@ void DecodePSHUFLWMask(unsigned Imm,
   ShuffleMask.push_back(7);
 }
 
-void DecodePUNPCKLBWMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask) {
-  DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i8, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLWDMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask) {
-  DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i16, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLDQMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask) {
-  DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i32, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLQDQMask(unsigned NElts,
-                          SmallVectorImpl<unsigned> &ShuffleMask) {
-  DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i64, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLMask(EVT VT,
-                       SmallVectorImpl<unsigned> &ShuffleMask) {
-  DecodeUNPCKLPMask(VT, ShuffleMask);
-}
-
-void DecodePUNPCKHMask(unsigned NElts,
-                       SmallVectorImpl<unsigned> &ShuffleMask) {
-  for (unsigned i = 0; i != NElts/2; ++i) {
-    ShuffleMask.push_back(i+NElts/2);
-    ShuffleMask.push_back(i+NElts+NElts/2);
-  }
-}
-
 void DecodeSHUFPMask(EVT VT, unsigned Imm,
                      SmallVectorImpl<unsigned> &ShuffleMask) {
   unsigned NumElts = VT.getVectorNumElements();
@@ -152,7 +119,7 @@ void DecodeSHUFPMask(EVT VT, unsigned Imm,
   }
 }
 
-void DecodeUNPCKHPMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask) {
+void DecodeUNPCKHMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask) {
   unsigned NumElts = VT.getVectorNumElements();
 
   // Handle 128 and 256-bit vector lengths. AVX defines UNPCK* to operate
@@ -171,10 +138,10 @@ void DecodeUNPCKHPMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask) {
   }
 }
 
-/// DecodeUNPCKLPMask - This decodes the shuffle masks for unpcklps/unpcklpd
+/// DecodeUNPCKLMask - This decodes the shuffle masks for unpcklps/unpcklpd
 /// etc.  VT indicates the type of the vector allowing it to handle different
 /// datatypes and vector widths.
-void DecodeUNPCKLPMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask) {
+void DecodeUNPCKLMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask) {
   unsigned NumElts = VT.getVectorNumElements();
 
   // Handle 128 and 256-bit vector lengths. AVX defines UNPCK* to operate

diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.h b/lib/Target/X86/Utils/X86ShuffleDecode.h
index 74c34b412bc09b914a1bdf0cb3d9a2acfe8df381..243728f81e3e23102d37c25d4efa8ee636c50378 100644 (file)
--- a/lib/Target/X86/Utils/X86ShuffleDecode.h
+++ b/lib/Target/X86/Utils/X86ShuffleDecode.h
@@ -46,36 +46,18 @@ void DecodePSHUFHWMask(unsigned Imm,
 void DecodePSHUFLWMask(unsigned Imm,
                        SmallVectorImpl<unsigned> &ShuffleMask);
 
-void DecodePUNPCKLBWMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask);
-
-void DecodePUNPCKLWDMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask);
-
-void DecodePUNPCKLDQMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask);
-
-void DecodePUNPCKLQDQMask(unsigned NElts,
-                          SmallVectorImpl<unsigned> &ShuffleMask);
-
-void DecodePUNPCKLMask(EVT VT,
-                       SmallVectorImpl<unsigned> &ShuffleMask);
-
-void DecodePUNPCKHMask(unsigned NElts,
-                       SmallVectorImpl<unsigned> &ShuffleMask);
-
 void DecodeSHUFPMask(EVT VT, unsigned Imm,
                      SmallVectorImpl<unsigned> &ShuffleMask);
 
-/// DecodeUNPCKHPMask - This decodes the shuffle masks for unpckhps/unpckhpd
+/// DecodeUNPCKHMask - This decodes the shuffle masks for unpckhps/unpckhpd
 /// etc.  VT indicates the type of the vector allowing it to handle different
 /// datatypes and vector widths.
-void DecodeUNPCKHPMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask);
+void DecodeUNPCKHMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask);
 
-/// DecodeUNPCKLPMask - This decodes the shuffle masks for unpcklps/unpcklpd
+/// DecodeUNPCKLMask - This decodes the shuffle masks for unpcklps/unpcklpd
 /// etc.  VT indicates the type of the vector allowing it to handle different
 /// datatypes and vector widths.
-void DecodeUNPCKLPMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask);
+void DecodeUNPCKLMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask);
 
 
 // DecodeVPERMILPMask - Decodes VPERMILPS/ VPERMILPD permutes for any 128-bit

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 7df931fdd4ec3d1432c2e8e11aa8f9ff38c7a55f..169e41b195b8dcbffa1a25860e873f197fb5d986 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -4461,16 +4461,12 @@ static SDValue getShuffleScalarElt(SDNode *N, int Index, SelectionDAG &DAG,
                       ShuffleMask);
       break;
     case X86ISD::PUNPCKH:
-      DecodePUNPCKHMask(NumElems, ShuffleMask);
-      break;
     case X86ISD::UNPCKHP:
-      DecodeUNPCKHPMask(VT, ShuffleMask);
+      DecodeUNPCKHMask(VT, ShuffleMask);
       break;
     case X86ISD::PUNPCKL:
-      DecodePUNPCKLMask(VT, ShuffleMask);
-      break;
     case X86ISD::UNPCKLP:
-      DecodeUNPCKLPMask(VT, ShuffleMask);
+      DecodeUNPCKLMask(VT, ShuffleMask);
       break;
     case X86ISD::MOVHLPS:
       DecodeMOVHLPSMask(NumElems, ShuffleMask);