[PATCH] D13442: [AArch64] Fold vector fmul by pow2 into floating-point to fixed-point conversion.

[Chad Rosier via llvm-commits]

mcrosier created this revision.
mcrosier added reviewers: t.p.northover, jmolloy, Gerolf.
mcrosier added subscribers: llvm-commits, mssimpso, haicheng, junbuml, gberry, bmakam.
mcrosier set the repository for this revision to rL LLVM.
Herald added subscribers: rengolin, aemerson.

FCVT (floating-point to signed/unsigned fixed-point conversion) can replace
combinations of FMUL and FCVT (floating-point to integer) when the FMUL has a
constant operand that is a power of 2.

For example, assuming v1 contains 16.000000e+00:

      fmul.2s v0, v1, v0[0]
      fcvtzs.2s v0, v0

becomes

      fcvtzs.2s v0, v0, #4

Please take a look.

 Thanks,
   Chad

Repository:
  rL LLVM

http://reviews.llvm.org/D13442

Files:
  lib/Target/AArch64/AArch64ISelLowering.cpp
  test/CodeGen/AArch64/fcvt_combine.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D13442.36534.patch
Type: text/x-patch
Size: 8758 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151005/73c09fe6/attachment.bin>