[llvm-bugs] [Bug 32939] New: [DAGCombine] Missed A + -2.0*B*C -> A - (B+B)*C

via llvm-bugs llvm-bugs at lists.llvm.org
Fri May 5 07:44:50 PDT 2017


https://bugs.llvm.org/show_bug.cgi?id=32939

            Bug ID: 32939
           Summary: [DAGCombine] Missed A + -2.0*B*C -> A - (B+B)*C
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Common Code Generator Code
          Assignee: unassignedbugs at nondot.org
          Reporter: mcrosier at codeaurora.org
                CC: llvm-bugs at lists.llvm.org

Given the following C code

double test1(double a, double b, double c) {
  return a + 2.0*b*c;
}

Clang removes a multiply by transforming '2.0*b' to 'b+b' during DAGCombine. 
For example, we generate the following when targeting AArch64:

test1:                                  // @test1
        fadd    d1, d1, d1
        fmul    d1, d1, d2
        fadd    d0, d1, d0
        ret

Given the below code

double test2(double a, double b, double c) {
  return a + -2.0*b*c;
}

we should be able to perform a similar transform if we can remove the negative
constant by converting the addition into subtraction.

Currently we generate the below when targeting AArch64:
test2:                                  // @test2
        fmov    d3, #-2.00000000
        fmul    d1, d1, d3
        fmul    d1, d1, d2
        fadd    d0, d1, d0
        ret

but I'd expect

test2:                                  // @test2
        fadd    d1, d1, d1
        fmul    d1, d1, d2
        fsub    d0, d1, d0
        ret

AFAIK, this doesn't require fast-math.

-- 
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/20170505/34e7f582/attachment.html>


More information about the llvm-bugs mailing list