[llvm-bugs] [Bug 47337] New: Slower sieve computation than gcc
via llvm-bugs
llvm-bugs at lists.llvm.org
Thu Aug 27 16:01:48 PDT 2020
https://bugs.llvm.org/show_bug.cgi?id=47337
Bug ID: 47337
Summary: Slower sieve computation than gcc
Product: libraries
Version: trunk
Hardware: PC
OS: Linux
Status: NEW
Severity: enhancement
Priority: P
Component: Loop Optimizer
Assignee: unassignedbugs at nondot.org
Reporter: david.bolvansky at gmail.com
CC: llvm-bugs at lists.llvm.org
#define Size 819000
static int sieve (int N) {
int i, k, prime, count, n; char flags[Size];
for (n = 0; n < N; n++) {
count = 0;
for (i = 0; i < Size; i++)
flags[i] = 1;
for (i = 0; i < Size; i++)
if (flags[i]) {
prime = i + i + 3;
for (k = i + prime; k < Size; k += prime)
flags[k] = 0;
count++;
}
}
return count;
}
int main (void) {
__builtin_printf ("sieve (100) = %d", sieve (100));
}
gcc -O3:
0m0,392s
gcc -O3 -march=haswell:
0m0,392s
clang -O3:
0m0,404s
clang -O3 -march=haswell:
0m0,393s
So -march=haswell enables loop unrolling which is profitable for this code.
But gcc does not unroll this loop and it is still faster.
Maybe clang's:
mov bl, 1 ***
xor ecx, ecx
mov edx, 3
xor esi, esi
test bl, bl ***
je .LBB0_7
.LBB0_3: # in Loop: Header=BB0_1 Depth=1
lea rdi, [rcx + 2*rcx]
add rdi, 3
cmp rdi, 818999
is slower than gcc's:
add rcx, 3
add rsi, 1
add rdx, 2
cmp rcx, 2457003
je .L12
.L5:
cmp BYTE PTR [rsi], 0
je .L2
cmp rcx, 818999
Godbolt: https://godbolt.org/z/6ExqE6
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20200827/5b899d92/attachment-0001.html>
More information about the llvm-bugs
mailing list