[llvm-bugs] [Bug 51069] New: [X86] Failure to pull out common scaled address offset through select/cmov

via llvm-bugs llvm-bugs at lists.llvm.org
Mon Jul 12 14:37:16 PDT 2021


            Bug ID: 51069
           Summary: [X86] Failure to pull out common scaled address offset
                    through select/cmov
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: llvm-dev at redking.me.uk
                CC: craig.topper at gmail.com, llvm-bugs at lists.llvm.org,
                    llvm-dev at redking.me.uk, pengfei.wang at intel.com,
                    spatel+llvm at rotateright.com


void dec(int *base, long long offset, int sel) {
    int *ptr0 = base + offset + 0;
    int *ptr6 = base + offset + 6;
    int *ptr = sel ? ptr0 : ptr6;

define void @dec(i32* nocapture %0, i64 %1, i32 %2) {
  %4 = getelementptr inbounds i32, i32* %0, i64 %1
  %5 = getelementptr inbounds i32, i32* %4, i64 6
  %6 = icmp eq i32 %2, 0
  %7 = select i1 %6, i32* %5, i32* %4
  %8 = load i32, i32* %7, align 4, !tbaa !3
  %9 = add nsw i32 %8, -1
  store i32 %9, i32* %7, align 4, !tbaa !3
  ret void

        leaq    (%rdi,%rsi,4), %rax
        leaq    (%rdi,%rsi,4), %rcx
        addq    $24, %rcx
        testl   %edx, %edx
        cmovneq %rax, %rcx
        addl    $-1, (%rcx)

We can reduce the number of complex LEA ops by pulling the common "%rsi,4"
scaled offset through the cmov and into the addl address:

        leaq    24(%rdi), %rcx
        testl   %edx, %edx
        cmovneq %rdi, %rcx
        addl    $-1, (%rcx,%rsi,4)

This might be generally achievable by canonicalizing the gep-chain, but for now
I'm making this a X86 ticket.

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/20210712/1c4243e8/attachment.html>

More information about the llvm-bugs mailing list