[llvm-bugs] [Bug 34558] New: use SDNodeFlags to enable reciprocal square root transform

via llvm-bugs llvm-bugs at lists.llvm.org
Mon Sep 11 10:21:44 PDT 2017


https://bugs.llvm.org/show_bug.cgi?id=34558

            Bug ID: 34558
           Summary: use SDNodeFlags to enable reciprocal square root
                    transform
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Common Code Generator Code
          Assignee: unassignedbugs at nondot.org
          Reporter: spatel+llvm at rotateright.com
                CC: llvm-bugs at lists.llvm.org

Given IR with fast-math-flags, those attributes should be passed on to the
corresponding FP nodes in the DAG:

define float @fast_recip_sqrt(float %x) {
  %y = call fast float @llvm.sqrt.f32(float %x)
  %z = fdiv fast float 1.0,  %y
  ret float %z
}
declare float @llvm.sqrt.f32(float) nounwind readonly

-------------------------------------------------------------------------------

For x86, this should trigger a reciprocal square root estimate optimization:
        rsqrtss %xmm0, %xmm1
        mulss   %xmm1, %xmm0
        mulss   %xmm1, %xmm0
        addss   LCPI0_0(%rip), %xmm0
        mulss   LCPI0_1(%rip), %xmm1
        mulss   %xmm1, %xmm0
        retq


There are currently 2 problems:
1. We don't apply FMF to intrinsics ( https://reviews.llvm.org/D37686 proposes
to fix this), so the sqrt node isn't marked with any relaxed FP properties.
2. The code in DAGCombiner::visitFDIV() doesn't check flags on nodes.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20170911/743198ea/attachment.html>


More information about the llvm-bugs mailing list