https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94824
Bug ID: 94824
Summary: Failure to optimize with __builtin_bswap32 as well as
with a function recognized as such
Product: gcc
Version: 10.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: gabravier at gmail dot com
Target Milestone: ---
uint32_t swap32(uint32_t x)
{
return ((x << 24) | ((x << 8) & 0x00FF0000) | ((x >> 8) & 0x0000FF00) | (x
>> 24));
}
uint64_t swap64v1(uint64_t x)
{
uint64_t a = __builtin_bswap32(x);
x >>= 32;
a <<= 32;
return __builtin_bswap32(x) | a;
}
uint64_t swap64v2(uint64_t x)
{
uint64_t a = swap32(x);
x >>= 32;
a <<= 32;
return swap32(x) | a;
}
swap64v1 and swap64v2 are identical, since bswap32 is equivalent to
__builtin_bswap32. However, only swap64v2 is optimized to __builtin_bswap64.
swap64v1 is compiled to this by gcc -O3 :
swap64v1(unsigned long):
mov rdx, rdi
mov eax, edi
shr rdx, 32
bswap eax
sal rax, 32
bswap edx
mov edi, edx
or rax, rdi
ret
And swap64v2 gives this :
swap64v2(unsigned long):
mov rax, rdi
bswap rax
ret
