[all-commits] [llvm/llvm-project] 5840f7: [PowerPC] Respect rounding mode in the back end

[Nemanja Ivanovic via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 5840f7197d058371c01fd7ac9ad932a3ab2ced6a
      https://github.com/llvm/llvm-project/commit/5840f7197d058371c01fd7ac9ad932a3ab2ced6a
  Author: Nemanja Ivanovic <nemanja.i.ibm at gmail.com>
  Date:   2021-11-10 (Wed, 10 Nov 2021)

  Changed paths:
    M llvm/lib/Target/PowerPC/P10InstrResources.td
    M llvm/lib/Target/PowerPC/P9InstrResources.td
    M llvm/lib/Target/PowerPC/PPCISelLowering.cpp
    M llvm/lib/Target/PowerPC/PPCISelLowering.h
    M llvm/lib/Target/PowerPC/PPCInstr64Bit.td
    M llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
    M llvm/lib/Target/PowerPC/PPCInstrInfo.td
    M llvm/lib/Target/PowerPC/PPCInstrVSX.td
    A llvm/test/CodeGen/PowerPC/cse-despite-rounding-mode.ll
    A llvm/test/CodeGen/PowerPC/respect-rounding-mode.ll
    M llvm/test/CodeGen/PowerPC/vector-constrained-fp-intrinsics.ll

  Log Message:
  -----------
  [PowerPC] Respect rounding mode in the back end

Currently, the floating point instructions that depend on
rounding mode are correctly marked in the PPC back end with
an implicit use of the RM register. Similarly, instructions
that explicitly define the register are marked with an
implicit def of the same register. So for the most part,
RM-using code won't be moved across RM-setting instructions.

However, calls are not marked as RM-setting instructions so
code can be moved across calls. This is generally desired,
but so is the ability to turn off this behaviour with an
appropriate option - and -frounding-math really should be
that option.

This patch provides a set of call instructions (for direct
and indirect calls) that are marked with an implicit def of
the RM register. These will be used for calls that are marked
with the strictfp attribute.

Differential revision: https://reviews.llvm.org/D111433