I would expect gcc to generate comparable code for both functions below, or
perhaps even better code for foo() than for bar() since the code in foo() is
likely to be more common than the equivalent code in bar(). However, the code
produced for foo() is suboptimal in comparison to the code for bar(). In my
timings on x86 with gcc 4.3.0 at -O2, foo() appears to run about 5% slower than
bar().
$ cat t.c && gcc -S -O2 t.c && cat t.s
int foo (int *a, int *b) { return a && b || !a && !b; }
int bar (int *a, int *b) { return !!a == !!b; }
.file "t.c"
.text
.p2align 4,,15
.globl foo
.type foo, @function
foo:
.LFB2:
testq %rdi, %rdi
je .L2
testq %rsi, %rsi
movl $1, %eax
je .L2
rep
ret
.p2align 4,,10
.p2align 3
.L2:
testq %rdi, %rdi
sete %al
testq %rsi, %rsi
sete %dl
andl %edx, %eax
movzbl %al, %eax
ret
.LFE2:
.size foo, .-foo
.p2align 4,,15
.globl bar
.type bar, @function
bar:
.LFB3:
testq %rdi, %rdi
sete %al
testq %rsi, %rsi
setne %dl
xorl %edx, %eax
movzbl %al, %eax
ret
.LFE3:
.size bar, .-bar
.section .eh_frame,"a",@progbits
.Lframe1:
.long .LECIE1-.LSCIE1
.LSCIE1:
.long 0x0
.byte 0x1
.string "zR"
.uleb128 0x1
.sleb128 -8
.byte 0x10
.uleb128 0x1
.byte 0x3
.byte 0xc
.uleb128 0x7
.uleb128 0x8
.byte 0x90
.uleb128 0x1
.align 8
.LECIE1:
.LSFDE1:
.long .LEFDE1-.LASFDE1
.LASFDE1:
.long .LASFDE1-.Lframe1
.long .LFB2
.long .LFE2-.LFB2
.uleb128 0x0
.align 8
.LEFDE1:
.LSFDE3:
.long .LEFDE3-.LASFDE3
.LASFDE3:
.long .LASFDE3-.Lframe1
.long .LFB3
.long .LFE3-.LFB3
.uleb128 0x0
.align 8
.LEFDE3:
.ident "GCC: (GNU) 4.3.0 20080428 (Red Hat 4.3.0-8)"
.section .note.GNU-stack,"",@progbits
--
Summary: suboptimal code for (a && b || !a && !b)
Product: gcc
Version: 4.3.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: sebor at roguewave dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38126
