[llvm] 23a5073 - [RISCV] LUI used for address computation should not isAsCheapAsAMove

[Ben Shi via llvm-commits]

Author: Haocong.Lu
Date: 2022-02-12T07:14:38Z
New Revision: 23a50736004e94704a2393aa36a905d737f2b20f

URL: https://github.com/llvm/llvm-project/commit/23a50736004e94704a2393aa36a905d737f2b20f
DIFF: https://github.com/llvm/llvm-project/commit/23a50736004e94704a2393aa36a905d737f2b20f.diff

LOG: [RISCV] LUI used for address computation should not isAsCheapAsAMove

A LUI instruction with flag RISCVII::MO_HI is usually used in conjunction
with ADDI, and jointly complete address computation. To bind the cost
evaluation of address computation, the LUI should not be regarded as a cheap
 move separately, which is consistent with ADDI.

In this test case, it improves the unroll-loop code that the rematerialization
of array's base address miss MachineCSE with Heuristics #1 at isProfitableToCSE.

Reviewed By: asb, frasercrmck

Differential Revision: https://reviews.llvm.org/D118216

Added: 
    

Modified: 
    llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
    llvm/test/CodeGen/RISCV/unroll-loop-cse.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index f64717a6d88b8..4f39b9c8a9829 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -998,6 +998,8 @@ bool RISCVInstrInfo::isAsCheapAsAMove(const MachineInstr &MI) const {
     return (MI.getOperand(1).isReg() &&
             MI.getOperand(1).getReg() == RISCV::X0) ||
            (MI.getOperand(2).isImm() && MI.getOperand(2).getImm() == 0);
+  case RISCV::LUI:
+    return MI.getOperand(1).getTargetFlags() != RISCVII::MO_HI;
   }
   return MI.isAsCheapAsAMove();
 }

diff  --git a/llvm/test/CodeGen/RISCV/unroll-loop-cse.ll b/llvm/test/CodeGen/RISCV/unroll-loop-cse.ll
index 00b0d32e07d30..91aec53c47210 100644
--- a/llvm/test/CodeGen/RISCV/unroll-loop-cse.ll
+++ b/llvm/test/CodeGen/RISCV/unroll-loop-cse.ll
@@ -18,28 +18,20 @@ define signext i32 @unroll_loop_cse() {
 ; CHECK-NEXT:    bne a3, a4, .LBB0_6
 ; CHECK-NEXT:  # %bb.1:
 ; CHECK-NEXT:    addi a1, a1, %lo(x)
-; CHECK-NEXT:    lw a1, 4(a1)
+; CHECK-NEXT:    lw a3, 4(a1)
 ; CHECK-NEXT:    addi a2, a2, %lo(check)
-; CHECK-NEXT:    lw a2, 4(a2)
-; CHECK-NEXT:    bne a1, a2, .LBB0_6
+; CHECK-NEXT:    lw a4, 4(a2)
+; CHECK-NEXT:    bne a3, a4, .LBB0_6
 ; CHECK-NEXT:  # %bb.2:
-; CHECK-NEXT:    lui a1, %hi(x)
-; CHECK-NEXT:    addi a1, a1, %lo(x)
 ; CHECK-NEXT:    lw a3, 8(a1)
-; CHECK-NEXT:    lui a2, %hi(check)
-; CHECK-NEXT:    addi a2, a2, %lo(check)
 ; CHECK-NEXT:    lw a4, 8(a2)
 ; CHECK-NEXT:    bne a3, a4, .LBB0_6
 ; CHECK-NEXT:  # %bb.3:
-; CHECK-NEXT:    lw a1, 12(a1)
-; CHECK-NEXT:    lw a2, 12(a2)
-; CHECK-NEXT:    bne a1, a2, .LBB0_6
+; CHECK-NEXT:    lw a3, 12(a1)
+; CHECK-NEXT:    lw a4, 12(a2)
+; CHECK-NEXT:    bne a3, a4, .LBB0_6
 ; CHECK-NEXT:  # %bb.4:
-; CHECK-NEXT:    lui a1, %hi(x)
-; CHECK-NEXT:    addi a1, a1, %lo(x)
 ; CHECK-NEXT:    lw a3, 16(a1)
-; CHECK-NEXT:    lui a2, %hi(check)
-; CHECK-NEXT:    addi a2, a2, %lo(check)
 ; CHECK-NEXT:    lw a4, 16(a2)
 ; CHECK-NEXT:    bne a3, a4, .LBB0_6
 ; CHECK-NEXT:  # %bb.5: