The testcase:
int test (float a) {
return (a*a);
}
produces non-optimal asm code, when compiled with '-O2 -msse -mfpmath=387':
test:
pushl %ebp
movl %esp, %ebp
subl $4, %esp
flds 8(%ebp)
fmul %st(0), %st
fstps -4(%ebp)
(*) movss -4(%ebp), %xmm0
(*) cvttss2si %xmm0, %eax
leave
ret
A cvttss2si insn could use memory input operand, so movss to %xmm0 is not
needed. Following RTL sequence is generated in ._.23.greg:
(insn 31 11 32 0 (set (mem:SF (plus:SI (reg/f:SI 6 bp)
(const_int -4 [0xfffffffc])) [0 S4 A8])
(reg/v:SF 8 st [orig:59 a ] [59])) 61 {*movsf_1} (nil)
(nil))
(insn 32 31 12 0 (set (reg:SF 21 xmm0)
(mem:SF (plus:SI (reg/f:SI 6 bp)
(const_int -4 [0xfffffffc])) [0 S4 A8])) 61 {*movsf_1} (nil)
(nil))
(insn:HI 12 32 16 0 (set (reg:SI 0 ax [60])
(fix:SI (reg:SF 21 xmm0))) 112 {fix_truncsfsi_sse}
(insn_list:REG_DEP_TRUE 11 (nil))
(nil))
It looks like gcc does not take into account the fact that cvttss2si can use
memory input operand.
--
Summary: Non-optimal code with cvttss2si
Product: gcc
Version: 4.0.0
Status: UNCONFIRMED
Severity: normal
Priority: P2
Component: target
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: uros at kss-loka dot si
CC: gcc-bugs at gcc dot gnu dot org
GCC build triplet: i686-pc-linux-gnu
GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19398
