diff options
| author | Anton Sidorenko <anton.sidorenko@syntacore.com> | 2022-10-11 11:36:53 +0300 |
|---|---|---|
| committer | Anton Sidorenko <anton.sidorenko@syntacore.com> | 2022-11-17 13:24:04 +0300 |
| commit | b6c790736e77f79e819fa761fb13f2311296714a (patch) | |
| tree | 7ea2e5041f26e04e284b4c9bd5bbaa890a75c8d2 /llvm/lib/CodeGen/MachineCombiner.cpp | |
| parent | 50f8eb05af50904e740088385ec54aae017467e8 (diff) | |
| download | llvm-b6c790736e77f79e819fa761fb13f2311296714a.zip llvm-b6c790736e77f79e819fa761fb13f2311296714a.tar.gz llvm-b6c790736e77f79e819fa761fb13f2311296714a.tar.bz2 | |
[MachineCombiner][RISCV] Add fmadd/fmsub/fnmsub instructions patterns
This patch adds tranformation of fmul+fadd/fsub chains to fused multiply
instructions:
* fmul+fadd->fmadd
* fmul+fsub->fmsub/fnmsub
We also will try to combine these instructions if the fmul has more than one use
and cannot be deleted. However, removing the dependence between fmul and fadd can
still be profitable, and we rely on machine combiner approximations of scheduling.
Differential Revision: https://reviews.llvm.org/D136764
Diffstat (limited to 'llvm/lib/CodeGen/MachineCombiner.cpp')
| -rw-r--r-- | llvm/lib/CodeGen/MachineCombiner.cpp | 4 |
1 files changed, 4 insertions, 0 deletions
diff --git a/llvm/lib/CodeGen/MachineCombiner.cpp b/llvm/lib/CodeGen/MachineCombiner.cpp index ed0fe7d..bced97f7 100644 --- a/llvm/lib/CodeGen/MachineCombiner.cpp +++ b/llvm/lib/CodeGen/MachineCombiner.cpp @@ -319,6 +319,10 @@ static CombinerObjective getCombinerObjective(MachineCombinerPattern P) { case MachineCombinerPattern::REASSOC_XMM_AMM_BMM: case MachineCombinerPattern::SUBADD_OP1: case MachineCombinerPattern::SUBADD_OP2: + case MachineCombinerPattern::FMADD_AX: + case MachineCombinerPattern::FMADD_XA: + case MachineCombinerPattern::FMSUB: + case MachineCombinerPattern::FNMSUB: return CombinerObjective::MustReduceDepth; case MachineCombinerPattern::REASSOC_XY_BCA: case MachineCombinerPattern::REASSOC_XY_BAC: |