[all-commits] [llvm/llvm-project] 3dd675: [AArch64] Add more complete support for BF16

[David Majnemer via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 3dd6750027cd168fce5fd9894b0bac0739652cf5
      https://github.com/llvm/llvm-project/commit/3dd6750027cd168fce5fd9894b0bac0739652cf5
  Author: David Majnemer <david.majnemer at gmail.com>
  Date:   2024-03-03 (Sun, 03 Mar 2024)

  Changed paths:
    M llvm/include/llvm/CodeGen/TargetLowering.h
    M llvm/include/llvm/CodeGen/ValueTypes.h
    M llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
    M llvm/lib/Target/AArch64/AArch64ISelLowering.h
    M llvm/lib/Target/AArch64/AArch64InstrFormats.td
    M llvm/lib/Target/AArch64/AArch64InstrInfo.td
    M llvm/lib/Target/AArch64/GISel/AArch64PostLegalizerLowering.cpp
    M llvm/test/Analysis/CostModel/AArch64/reduce-fadd.ll
    M llvm/test/CodeGen/AArch64/GlobalISel/lower-neon-vector-fcmp.mir
    M llvm/test/CodeGen/AArch64/implicitly-set-zero-high-64-bits.ll
    M llvm/test/CodeGen/AArch64/neon-compare-instructions.ll
    M llvm/test/CodeGen/AArch64/round-fptosi-sat-scalar.ll
    M llvm/test/CodeGen/AArch64/vector-fcopysign.ll

  Log Message:
  -----------
  [AArch64] Add more complete support for BF16

We can use a small amount of integer arithmetic to round FP32 to BF16
and extend BF16 to FP32.

While a number of operations still require promotion, this can be
reduced for some rather simple operations like abs, copysign, fneg but
these can be done in a follow-up.

A few neat optimizations are implemented:
- round-inexact-to-odd is used for F64 to BF16 rounding.
- quieting signaling NaNs for f32 -> bf16 tries to detect if a prior
  operation makes it unnecessary.



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications