[llvm] 1a2bd44 - [RISCV] Make shouldConvertConstantLoadToIntImm return true unless enableUnalignedScalarMem is true.

[Craig Topper via llvm-commits]

Author: Craig Topper
Date: 2022-07-11T09:40:08-07:00
New Revision: 1a2bd44b77c2dd0c2aabf5e27e30eb17c4715832

URL: https://github.com/llvm/llvm-project/commit/1a2bd44b77c2dd0c2aabf5e27e30eb17c4715832
DIFF: https://github.com/llvm/llvm-project/commit/1a2bd44b77c2dd0c2aabf5e27e30eb17c4715832.diff

LOG: [RISCV] Make shouldConvertConstantLoadToIntImm return true unless enableUnalignedScalarMem is true.

This restores the old behavior before D129402 when
enableUnalignedScalarMem is false. This fixes a regression spotted
by @asb.

To fix this correctly, we need to consider alignment of the load
we'd be replacing, but that's not possible in the current interface.

Added: 
    

Modified: 
    llvm/lib/Target/RISCV/RISCVISelLowering.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 9c16da946e84..066c57614e34 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -1174,6 +1174,13 @@ bool RISCVTargetLowering::shouldConvertConstantLoadToIntImm(const APInt &Imm,
   if (isInt<32>(Val))
     return true;
 
+  // A constant pool entry may be more aligned thant he load we're trying to
+  // replace. If we don't support unaligned scalar mem, prefer the constant
+  // pool.
+  // TODO: Can the caller pass down the alignment?
+  if (!Subtarget.enableUnalignedScalarMem())
+    return true;
+
   // Prefer to keep the load if it would require many instructions.
   // This uses the same threshold we use for constant pools but doesn't
   // check useConstantPoolForLargeInts.