[Matrix] Propagate and use shape info for binary operators.

[fhahn]

Summary

This patch extends the current shape propagation and shape aware
lowering to also support binary operators. Those operators are uniform
with respect to their shape (shape of the input operands is the same as
the shape of their result).

[merge_guards_bot]

Build result: FAILURE - Could not check out parent git hash "7cce39f9a1eab3a98b761b84d09e100b0ce56650". It was not found in the repository. Did you configure the "Parent Revision" in Phabricator properly? Trying to apply the patch to the master branch instead...

ERROR: arc patch failed with error code 1. Check build log for details.
Log files: console-log.txt, CMakeCache.txt

[fhahn]

Update test check after address computation simplification.

[merge_guards_bot]

clang-format: pass.

Build artifacts: console-log.txt, diff.json

[fhahn]

ping

[anemet]

LGTM!

llvm/test/Transforms/LowerMatrixIntrinsics/bigger-expressions-double.ll

494

Rebase glitch?

llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp

bool setShapeInfo(Value *V, ShapeInfo Shape) {
     }
 
     ShapeMap.insert({V, Shape});
     LLVM_DEBUG(dbgs() << "  " << Shape.NumRows << " x " << Shape.NumColumns
                       << " for " << *V << "\n");
     return true;
   }
 
+  bool isUniformShape(Value *V) {
+    Instruction *I = dyn_cast<Instruction>(V);
+    if (!I)
+      return true;
+
+    switch (I->getOpcode()) {
+    case Instruction::FAdd:
+    case Instruction::FSub:
+    case Instruction::FMul: // Scalar multiply.
+    case Instruction::Add:
+    case Instruction::Mul:
+    case Instruction::Sub:
+      return true;
+    default:
+      return false;
+    }
+  }
+
   /// Returns true if shape information can be used for \p V. The supported
   /// instructions must match the instructions that can be lowered by this pass.
   bool supportsShapeInfo(Value *V) {
     Instruction *Inst = dyn_cast<Instruction>(V);
     if (!Inst)
       return false;
 
     IntrinsicInst *II = dyn_cast<IntrinsicInst>(Inst);
     if (II)
       switch (II->getIntrinsicID()) {
       case Intrinsic::matrix_multiply:
       case Intrinsic::matrix_transpose:
       case Intrinsic::matrix_columnwise_load:
       case Intrinsic::matrix_columnwise_store:
         return true;
       default:
         return false;
       }
-    return isa<StoreInst>(Inst);
+    return isUniformShape(V) || isa<StoreInst>(V);
   }
 
   /// Propagate the shape information of instructions to their users.
   void propagateShapeForward() {
     // The work list contains instructions for which we can compute the shape,
     // either based on the information provided by matrix intrinsics or known
     // shapes of operands.
     SmallVector<Instruction *, 8> WorkList;
while (!WorkList.empty()) {
                        m_Intrinsic<Intrinsic::matrix_columnwise_load>(
                            m_Value(), m_Value(), m_Value(M), m_Value(N)))) {
         Propagate = setShapeInfo(Inst, {M, N});
       } else if (match(Inst, m_Store(m_Value(MatrixA), m_Value()))) {
         auto OpShape = ShapeMap.find(MatrixA);
         if (OpShape != ShapeMap.end())
           setShapeInfo(Inst, OpShape->second);
         continue;
+      } else if (isUniformShape(Inst)) {
+        // Find the first operand that has a known shape and use that.
+        for (auto &Op : Inst->operands()) {
+          auto OpShape = ShapeMap.find(Op.get());
+          if (OpShape != ShapeMap.end()) {
+            Propagate |= setShapeInfo(Inst, OpShape->second);
+            break;
+          }
+        }
       }
 
       if (Propagate)
         for (auto *User : Inst->users())
           if (ShapeMap.count(User) == 0)
             WorkList.push_back(cast<Instruction>(User));
     }
   }
 
   bool Visit() {
     if (EnableShapePropagation)
       propagateShapeForward();
 
     ReversePostOrderTraversal<Function *> RPOT(&Func);
     bool Changed = false;
     for (auto *BB : RPOT) {
       for (Instruction &Inst : make_early_inc_range(*BB)) {
         IRBuilder<> Builder(&Inst);
 
         if (CallInst *CInst = dyn_cast<CallInst>(&Inst))
           Changed |= VisitCallInst(CInst);
 
         Value *Op1;
         Value *Op2;
-        if (match(&Inst, m_Store(m_Value(Op1), m_Value(Op2))))
+        if (auto *BinOp = dyn_cast<BinaryOperator>(&Inst))
+          Changed |= VisitBinaryOperator(BinOp);
+        else if (match(&Inst, m_Store(m_Value(Op1), m_Value(Op2))))
           Changed |= VisitStore(&Inst, Op1, Op2, Builder);
       }
     }
 
     for (Instruction *Inst : reverse(ToRemove))
       Inst->eraseFromParent();
 
     return Changed;
bool VisitStore(Instruction *Inst, Value *StoredVal, Value *Ptr,
                   IRBuilder<> &Builder) {
     auto I = ShapeMap.find(StoredVal);
     if (I == ShapeMap.end())
       return false;
 
     LowerStore(Inst, StoredVal, Ptr, Builder.getInt32(I->second.NumRows), I->second);
     return true;
   }
+
+  /// Lower binary operators, if shape information is available.
+  bool VisitBinaryOperator(BinaryOperator *Inst) {
+    auto I = ShapeMap.find(Inst);
+    if (I == ShapeMap.end())
+      return false;
+
+    Value *Lhs = Inst->getOperand(0);
+    Value *Rhs = Inst->getOperand(1);
+
+    IRBuilder<> Builder(Inst);
+    ShapeInfo &Shape = I->second;
+
+    ColumnMatrixTy LoweredLhs = getMatrix(Lhs, Shape, Builder);
+    ColumnMatrixTy LoweredRhs = getMatrix(Rhs, Shape, Builder);
+
+    // Add each column and store the result back into the opmapping
+    ColumnMatrixTy Result;
+    auto BuildColumnOp = [&Builder, Inst](Value *LHS, Value *RHS) {
+      switch (Inst->getOpcode()) {
+      case Instruction::Add:
+        return Builder.CreateAdd(LHS, RHS);
+      case Instruction::Mul:
+        return Builder.CreateMul(LHS, RHS);
+      case Instruction::Sub:
+        return Builder.CreateSub(LHS, RHS);
+      case Instruction::FAdd:
+        return Builder.CreateFAdd(LHS, RHS);
+      case Instruction::FMul:
+        return Builder.CreateFMul(LHS, RHS);
+      case Instruction::FSub:
+        return Builder.CreateFSub(LHS, RHS);
+      default:
+        llvm_unreachable("Unsupported binary operator for matrix");
+      }
+    };
+    for (unsigned C = 0; C < Shape.NumColumns; ++C)
+      Result.addColumn(
+          BuildColumnOp(LoweredLhs.getColumn(C), LoweredRhs.getColumn(C)));
+
+    finalizeLowering(Inst, Result, Builder);
+    return true;
+  }
 };
 } // namespace
 
 PreservedAnalyses LowerMatrixIntrinsicsPass::run(Function &F,
                                                  FunctionAnalysisManager &AM) {
   auto &TTI = AM.getResult<TargetIRAnalysis>(F);
   LowerMatrixIntrinsics LMT(F, TTI);
   if (LMT.Visit()) {

llvm/test/Transforms/LowerMatrixIntrinsics/bigger-expressions-double.ll

 ; CHECK-NEXT:    [[SPLAT_SPLAT83:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT82]], <1 x double> undef, <1 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP104:%.*]] = fmul <1 x double> [[BLOCK81]], [[SPLAT_SPLAT83]]
 ; CHECK-NEXT:    [[TMP105:%.*]] = fadd <1 x double> [[TMP102]], [[TMP104]]
 
 ;  Embed result of multiply into flat vector.
 
 ; CHECK-NEXT:    [[TMP106:%.*]] = shufflevector <1 x double> [[TMP105]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
 ; CHECK-NEXT:    [[TMP107:%.*]] = shufflevector <3 x double> [[TMP97]], <3 x double> [[TMP106]], <3 x i32> <i32 0, i32 1, i32 3>
-; CHECK-NEXT:    [[TMP108:%.*]] = shufflevector <3 x double> [[TMP47]], <3 x double> [[TMP77]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
-; CHECK-NEXT:    [[TMP109:%.*]] = shufflevector <3 x double> [[TMP107]], <3 x double> undef, <6 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef>
-; CHECK-NEXT:    [[TMP110:%.*]] = shufflevector <6 x double> [[TMP108]], <6 x double> [[TMP109]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
 
-; Load %C and add result of multiply.
+; Load %C.
 
 ; CHECK-NEXT:    [[C:%.*]] = load <9 x double>, <9 x double>* [[C_PTR:%.*]]
-; CHECK-NEXT:    [[RES:%.*]] = fadd <9 x double> [[C]], [[TMP110]]
-; CHECK-NEXT:    store <9 x double> [[RES]], <9 x double>* [[C_PTR]]
+
+; Extract columns from %C.
+
+; CHECK-NEXT:    [[SPLIT84:%.*]] = shufflevector <9 x double> [[C]], <9 x double> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT:    [[SPLIT85:%.*]] = shufflevector <9 x double> [[C]], <9 x double> undef, <3 x i32> <i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT86:%.*]] = shufflevector <9 x double> [[C]], <9 x double> undef, <3 x i32> <i32 6, i32 7, i32 8>
+
+; Add column vectors.
+
+; CHECK-NEXT:    [[TMP108:%.*]] = fadd <3 x double> [[SPLIT84]], [[TMP47]]
+; CHECK-NEXT:    [[TMP109:%.*]] = fadd <3 x double> [[SPLIT85]], [[TMP77]]
+; CHECK-NEXT:    [[TMP110:%.*]] = fadd <3 x double> [[SPLIT86]], [[TMP107]]
+
+; Store result columns.
+
+; CHECK-NEXT:    [[TMP111:%.*]] = bitcast <9 x double>* [[C_PTR]] to double*
+; CHECK-NEXT:    [[TMP112:%.*]] = bitcast double* [[TMP111]] to <3 x double>*
+; CHECK-NEXT:    store <3 x double> [[TMP108]], <3 x double>* [[TMP112]], align 8
+; CHECK-NEXT:    [[TMP113:%.*]] = getelementptr double, double* [[TMP111]], i32 3
+; CHECK-NEXT:    [[TMP114:%.*]] = bitcast double* [[TMP113]] to <3 x double>*
+; CHECK-NEXT:    store <3 x double> [[TMP109]], <3 x double>* [[TMP114]], align 8
+; CHECK-NEXT:    [[TMP115:%.*]] = getelementptr double, double* [[TMP111]], i32 6
+; CHECK-NEXT:    [[TMP116:%.*]] = bitcast double* [[TMP115]] to <3 x double>*
+; CHECK-NEXT:    store <3 x double> [[TMP110]], <3 x double>* [[TMP116]], align 8
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %a = load <9 x double>, <9 x double>* %A.Ptr
   %b = load <9 x double>, <9 x double>* %B.Ptr
   %a.trans  = call <9 x double> @llvm.matrix.transpose(<9 x double> %a, i32 3, i32 3)
   %mult = call <9 x double> @llvm.matrix.multiply.v9f64.v9f64.v9f64(<9 x double> %a.trans, <9 x double> %b, i32 3, i32 3, i32 3)
   %c = load <9 x double>, <9 x double>* %C.Ptr
   %res = fadd <9 x double> %c, %mult
 
   store <9 x double> %res, <9 x double>* %C.Ptr
   ret void
 }

llvm/test/Transforms/LowerMatrixIntrinsics/propagate-forward.ll

 ; SHAPE-NEXT:    ret void
 entry:
   %c  = call <8 x double> @llvm.matrix.transpose(<8 x double> %a, i32 2, i32 4)
   store <8 x double> %c, <8 x double>* %Ptr
   ret void
 }
 
 declare <8 x double> @llvm.matrix.transpose(<8 x double>, i32, i32)
+
+define <8 x double> @transpose_fadd(<8 x double> %a) {
+; CHECK-LABEL: @transpose_fadd(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <8 x double> [[A:%.*]], <8 x double> undef, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[SPLIT2:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <2 x i32> <i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT3:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <2 x i32> <i32 6, i32 7>
+; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <2 x double> [[SPLIT]], i64 0
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x double> undef, double [[TMP0]], i64 0
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[SPLIT1]], i64 0
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x double> [[TMP1]], double [[TMP2]], i64 1
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x double> [[TMP3]], double [[TMP4]], i64 2
+; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 0
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x double> [[TMP5]], double [[TMP6]], i64 3
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x double> [[SPLIT]], i64 1
+; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <4 x double> undef, double [[TMP8]], i64 0
+; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x double> [[SPLIT1]], i64 1
+; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <4 x double> [[TMP9]], double [[TMP10]], i64 1
+; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 1
+; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x double> [[TMP11]], double [[TMP12]], i64 2
+; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 1
+; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <4 x double> [[TMP13]], double [[TMP14]], i64 3
+; CHECK-NEXT:    [[SPLIT4:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[SPLIT5:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[TMP16:%.*]] = fadd <4 x double> [[TMP7]], [[SPLIT4]]
+; CHECK-NEXT:    [[TMP17:%.*]] = fadd <4 x double> [[TMP15]], [[SPLIT5]]
+; CHECK-NEXT:    [[TMP18:%.*]] = shufflevector <4 x double> [[TMP16]], <4 x double> [[TMP17]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    ret <8 x double> [[TMP18]]
+;
+entry:
+  %c  = call <8 x double> @llvm.matrix.transpose(<8 x double> %a, i32 2, i32 4)
+  %res = fadd <8 x double> %c, %a
+  ret <8 x double> %res
+}
+
+define <8 x double> @transpose_fmul(<8 x double> %a) {
+; CHECK-LABEL: @transpose_fmul(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <8 x double> [[A:%.*]], <8 x double> undef, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[SPLIT2:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <2 x i32> <i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT3:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <2 x i32> <i32 6, i32 7>
+; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <2 x double> [[SPLIT]], i64 0
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x double> undef, double [[TMP0]], i64 0
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[SPLIT1]], i64 0
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x double> [[TMP1]], double [[TMP2]], i64 1
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x double> [[TMP3]], double [[TMP4]], i64 2
+; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 0
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x double> [[TMP5]], double [[TMP6]], i64 3
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x double> [[SPLIT]], i64 1
+; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <4 x double> undef, double [[TMP8]], i64 0
+; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x double> [[SPLIT1]], i64 1
+; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <4 x double> [[TMP9]], double [[TMP10]], i64 1
+; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 1
+; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x double> [[TMP11]], double [[TMP12]], i64 2
+; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 1
+; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <4 x double> [[TMP13]], double [[TMP14]], i64 3
+; CHECK-NEXT:    [[SPLIT4:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[SPLIT5:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[TMP16:%.*]] = fmul <4 x double> [[TMP7]], [[SPLIT4]]
+; CHECK-NEXT:    [[TMP17:%.*]] = fmul <4 x double> [[TMP15]], [[SPLIT5]]
+; CHECK-NEXT:    [[TMP18:%.*]] = shufflevector <4 x double> [[TMP16]], <4 x double> [[TMP17]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    ret <8 x double> [[TMP18]]
+;
+entry:
+  %c  = call <8 x double> @llvm.matrix.transpose(<8 x double> %a, i32 2, i32 4)
+  %res = fmul <8 x double> %c, %a
+  ret <8 x double> %res
+}