Unexplained absolute addressing mode generated on amd64/FreeBSD. Bug?

Project Infrastructure LLVM Dev List Archives
1

Good evening!

I'm experiencing weird code generation on amd64 FreeBSD with the
system's clang 10.0.1. specifically, an absolute addressing mode is
unexpectedly chosen for ucomisd with a memory constant where I would
normally expect a rip-relative addressing mode. The behaviour has
proven to be hard to reproduce. Any attempt to simplify the code leads
to perfectly fine code. To reproduce, download my project:

    http://fuz.su/~fuz/dobutsu/dobutsu-2.tar.xz

and compile dobutsu.c with -O3 on amd64/FreeBSD (Linux should work,
too). The output has a bunch of instructions like this:

    ucomisd .LCPI0_0, %xmm1

with an absolute addressing mode where I would expect a relative
addressing mode like

    ucomisd .LCPI0_0(%rip), %xmm1

I wonder what the reason for this could be. I can reproduce this
error with clang 11.0.1 from FreeBSD but haven't been able to test
12.0. Will continue to try and reduce the example.

Yours,
Robert Clausecker

2

With some help from the IRC, I managed to use creduce to cut down the
test case. The result is this:

    double a;
    b;
    c() {
        while (b)
            if (a)
                d();
    }

This produces code like this (slightly cleaned up):

    .LCPI0_0:
            .quad 0
    b: .long 0
    c: pushq %rbp
            movq %rsp, %rbp
    .LBB0_1:
            movl b(%rip), %eax
    .LBB0_2:
            testl %eax, %eax
            je .LBB0_5
            movsd a(%rip), %xmm0
            ucomisd .LCPI0_0, %xmm0 ; ???
            jne .LBB0_4
            jnp .LBB0_2
    .LBB0_4:
            xorl %eax, %eax
            callq d
            jmp .LBB0_1
    .LBB0_5:
            popq %rbp
            retq

Why is an absolute addressing mode instead of a rip-relative addressing
mode chosen for the memory operand to ucomisd? And why is zero loaded
from memory at all? I might not be an expert, but

            xorsd %xmm0, %xmm0
            ucomisd a(%rip), %xmm0

looks a bit better than the movsd/ucomisd pair clang comes up with.

Yours,
Robert Clausecker