This is distilled from PR42500. The same testcase doesn't show a problem if
compiled with a C compiler only.
Why does the gimple for the following testcase contain ?
extern "C" void foo(int a[4]);
extern "C" void bar(int a, int b, int c, int d)
{
int v[4] = { a, b, c, d};
foo(v);
foo(v);
}
in .004t.gimple
void bar(int, int, int, int) (int a, int b, int c, int d)
{
int v[4];
v = {};
v[0] = a;
v[1] = b;
v[2] = c;
v[3] = d;
foo (&v[0]);
foo (&v[0]);
}
even when all members of the array have been initialized in your original
testcase. I wonder if this can be eliminated as a part of the lowering process
to GIMPLE in the C++ frontend rather than any other place.
At -O2 the RTL optimizers remove 2 of the stores but there are unnecessary
stores of 0 to memory locations at [sp+8] , [sp+12] and the code generated is
as follows. (Instructions marked A and B below).
bar:
stmfd sp!, {r4, lr}
.save {r4, lr}
.LCFI0:
.pad #16
sub sp, sp, #16
.LCFI1:
mov lr, #0
add ip, sp, #8
str lr, [ip], #4 // A
str r0, [sp, #0]
mov r0, sp
str lr, [ip, #0] // B
stmib sp, {r1, r2, r3} @ phole stm
bl foo
mov r0, sp
mov r4, sp
bl foo
add sp, sp, #16
ldmfd sp!, {r4, lr}
bx lr
At -Os because the zero initializer is in the form of a memset - the RTL
optimizers don't remove anything and the code generated is as follows.
bar:
.fnstart
.LFB0:
@ args = 0, pretend = 0, frame = 16
@ frame_needed = 0, uses_anonymous_args = 0
stmfd sp!, {r0, r1, r2, r3, r4, r5, r6, r7, r8, lr}
.save {r0, r1, r2, r3, r4, r5, r6, r7, r8, lr}
.LCFI0:
mov r6, r0
mov r5, r1
mov r4, r2
mov r1, #0
mov r2, #16
mov r0, sp
mov r8, r3
bl memset
mov r0, sp
str r6, [sp, #0]
str r5, [sp, #4]
mov r7, sp
str r4, [sp, #8]
str r8, [sp, #12]
bl foo
mov r0, sp
bl foo
ldmfd sp!, {r0, r1, r2, r3, r4, r5, r6, r7, r8, pc}
If I change the testcase to use the extended initializer list in c++0x and
gnu++0x . There is an unnecessary move from sp to r4 but that is the subject of
PR42500.
void foo(int a[4]);
void bar(int a, int b, int c, int d)
{
int v[4] ;
v = { a, b, c, d};
foo(v);
foo(v);
}
bar:
.fnstart
.LFB0:
@ args = 0, pretend = 0, frame = 16
@ frame_needed = 0, uses_anonymous_args = 0
stmfd sp!, {r0, r1, r2, r3, r4, lr}
.save {r0, r1, r2, r3, r4, lr}
.LCFI0:
str r0, [sp, #0]
mov r0, sp
stmib sp, {r1, r2, r3} @ phole stm
bl foo
mov r0, sp
mov r4, sp
bl foo
ldmfd sp!, {r0, r1, r2, r3, r4, pc}
.LFE0:
.fnend
cheers
Ramana
--
Summary: Unnecessary generation of a zero initializer for array
with C++
Product: gcc
Version: 4.5.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: ramana at gcc dot gnu dot org
GCC host triplet: x86_64 linux gnu
GCC target triplet: arm-eabi
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42556
