[Bug tree-optimization/53346] [4.6/4.7/4.8 Regression] Bad vectorization in the proc cptrf2 of rnflow.f90

["ubizjak at gmail dot com" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53346

--- Comment #4 from Uros Bizjak <ubizjak at gmail dot com> 2012-05-17 20:09:42 UTC ---
Instead of this:

.L228:
    movl    $0, -4(%rdx,%rax,4)
    addq    $1, %rax
    cmpq    %rax, %rsi
    jge    .L228

vectorization generates following:

    movq    %rdx, %rax
    movq    %r9, %r8
    andl    $15, %eax
    shrq    $2, %rax
    negq    %rax
    andl    $3, %eax
    cmpq    %r9, %rax
    cmovbe    %rax, %r8
    cmpq    $6, %r9
    cmovbe    %r9, %r8
    testq    %r8, %r8
    je    .L233
    leaq    1(%r8), %rsi
    movl    $1, %eax
    .p2align 4,,10
    .p2align 3
.L176:
    movl    $0, -4(%rdx,%rax,4)
    addq    $1, %rax
    cmpq    %rsi, %rax
    jne    .L176
    cmpq    %r9, %r8
    je    .L182
.L174:
    movq    %r9, %rbp
    subq    %r8, %rbp
    movq    %rbp, %r11
    shrq    $2, %r11
    leaq    0(,%r11,4), %rbx
    testq    %rbx, %rbx
    je    .L181
    pxor    %xmm0, %xmm0
    leaq    (%rdx,%r8,4), %r8
    xorl    %esi, %esi
    .p2align 4,,10
    .p2align 3
.L183:
    addq    $1, %rsi
    movdqa    %xmm0, (%r8)
    addq    $16, %r8
    cmpq    %rsi, %r11
    ja    .L183
    addq    %rbx, %rax
    cmpq    %rbx, %rbp
    je    .L182
    .p2align 4,,10
    .p2align 3
.L181:
    movl    $0, -4(%rdx,%rax,4)
    addq    $1, %rax
    cmpq    %rax, %r9
    jge    .L181

Whoa.