This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Malloc improvements
- From: Anton Blanchard <anton at au1 dot ibm dot com>
- To: "Carlos O'Donell" <carlos at redhat dot com>, Siddhesh Poyarekar <sid at reserved-bit dot com>, DJ Delorie <dj at redhat dot com>
- Cc: libc-alpha at sourceware dot org
- Date: Tue, 12 Jul 2016 10:10:10 +1000
- Subject: Malloc improvements
- Authentication-results: sourceware.org; auth=none
Hi,
It's great to see the current focus on improving malloc. One thing that
would really help POWER is reducing the number of locks and atomics in
the fast path. Right now we have 3 in the malloc fastpath and 2 in
free. These add up.
The relative performance of a simple malloc/free of 8 bytes on POWER8
shows just how much can be gained by having a lockless fast path:
glibc baseline 466%
dj/malloc 518%
jemalloc 147%
tcmalloc 100%
Anton
--
#include <malloc.h>
#define ITERATIONS 100000000
void *x;
int main(void)
{
unsigned long i;
for (i = 0; i < ITERATIONS; i++) {
x = malloc(8);
free(x);
}
}