[llvm] [RISCV] Restrict combineOp_VLToVWOp_VL w/ bf16 to vfwmadd_vl with zvfbfwma (PR #108798)

Luke Lau via llvm-commits llvm-commits at lists.llvm.org
Mon Sep 16 19:05:37 PDT 2024


lukel97 wrote:

> If we have an FMA and an FADD using the same bf16 fp_extend, will be turn the FMA into a widening extned, but leave the FADD using the original extend?

I think all of the users need to match successfully before anything is actually widened:

```c++
    // Right now we do an all or nothing approach.
    if (!Matched)
      return SDValue();
  }
```

The materialization step happens after we go through everything on the worklist. I added a test case to double check this

https://github.com/llvm/llvm-project/pull/108798


More information about the llvm-commits mailing list