[PATCH] D127474: [SelectionDAG] fold Op0 - (X * MulC) --> Op0 + (X << log2(-MulC))

[Pretty-box via Phabricator via llvm-commits]

Pretty-box created this revision.
Pretty-box added reviewers: benshi001, craig.topper.
Pretty-box added a project: LLVM.
Herald added subscribers: StephenFan, frasercrmck, ecnelises, luismarques, apazos, sameer.abuasal, s.egerton, Jim, jocewei, PkmX, the_o, brucehoult, MartinMosbeck, rogfer01, edward-jones, zzheng, jrtc27, niosHD, sabuasal, simoncook, johnrusso, rbar, asb, hiraditya.
Herald added a project: All.
Pretty-box requested review of this revision.
Herald added subscribers: llvm-commits, pcwang-thead, MaskRay.

Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D127474

Files:
  llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
  llvm/test/CodeGen/RISCV/mul.ll


Index: llvm/test/CodeGen/RISCV/mul.ll
===================================================================
--- llvm/test/CodeGen/RISCV/mul.ll
+++ llvm/test/CodeGen/RISCV/mul.ll
@@ -1580,3 +1580,37 @@
   %r = and i8 %a, 15
   ret i8 %r
 }
+
+define i8 @mulsub(i8 %x, i8 %y) nounwind {
+; RV32I-LABEL: mulsub:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    slli a0, a0, 1
+; RV32I-NEXT:    add a0, a1, a0
+; RV32I-NEXT:    andi a0, a0, 15
+; RV32I-NEXT:    ret
+;
+; RV32IM-LABEL: mulsub:
+; RV32IM:       # %bb.0:
+; RV32IM-NEXT:    slli a0, a0, 1
+; RV32IM-NEXT:    add a0, a1, a0
+; RV32IM-NEXT:    andi a0, a0, 15
+; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: mulsub:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    slliw a0, a0, 1
+; RV64I-NEXT:    addw a0, a1, a0
+; RV64I-NEXT:    andi a0, a0, 15
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: mulsub:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    slliw a0, a0, 1
+; RV64IM-NEXT:    addw a0, a1, a0
+; RV64IM-NEXT:    andi a0, a0, 15
+; RV64IM-NEXT:    ret
+  %m = mul i8 %x, 14
+  %a = sub i8 %y, %m
+  %r = and i8 %a, 15
+  ret i8 %r
+}
Index: llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
===================================================================
--- llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -2516,24 +2516,29 @@
       return 0;
     };
 
-    auto foldMul = [&](SDValue X, SDValue Y, unsigned ShlAmt) {
+    auto foldMul = [&](SDValue Op, SDValue X, SDValue Y, unsigned ShlAmt) {
       EVT ShiftAmtTy = getShiftAmountTy(VT, TLO.DAG.getDataLayout());
       SDValue ShlAmtC = TLO.DAG.getConstant(ShlAmt, dl, ShiftAmtTy);
       SDValue Shl = TLO.DAG.getNode(ISD::SHL, dl, VT, X, ShlAmtC);
-      SDValue Sub = TLO.DAG.getNode(ISD::SUB, dl, VT, Y, Shl);
-      return TLO.CombineTo(Op, Sub);
+      SDValue Res = TLO.DAG.getNode(
+          Op.getOpcode() == ISD::ADD ? ISD::SUB : ISD::ADD, dl, VT, Y, Shl);
+      return TLO.CombineTo(Op, Res);
     };
 
     if (isOperationLegalOrCustom(ISD::SHL, VT)) {
       if (Op.getOpcode() == ISD::ADD) {
         // (X * MulC) + Op1 --> Op1 - (X << log2(-MulC))
         if (unsigned ShAmt = getShiftLeftAmt(Op0))
-          return foldMul(Op0.getOperand(0), Op1, ShAmt);
+          return foldMul(Op, Op0.getOperand(0), Op1, ShAmt);
         // Op0 + (X * MulC) --> Op0 - (X << log2(-MulC))
         if (unsigned ShAmt = getShiftLeftAmt(Op1))
-          return foldMul(Op1.getOperand(0), Op0, ShAmt);
-        // TODO:
+          return foldMul(Op, Op1.getOperand(0), Op0, ShAmt);
+      }
+      if (Op.getOpcode() == ISD::SUB) {
         // Op0 - (X * MulC) --> Op0 + (X << log2(-MulC))
+        if (unsigned ShAmt = getShiftLeftAmt(Op1)) {
+          return foldMul(Op, Op1.getOperand(0), Op0, ShAmt);
+        }
       }
     }
 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D127474.435805.patch
Type: text/x-patch
Size: 2795 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220610/4976059e/attachment.bin>