From: "Ondřej Bílka" <neleai@seznam.cz>
To: marcus.shawcroft@linaro.org
Cc: libc-ports@sourceware.org
Subject: memset on aarch64
Date: Mon, 25 Mar 2013 15:33:00 -0000 [thread overview]
Message-ID: <20130325152822.GA23513@domone.kolej.mff.cuni.cz> (raw)
Hi Marcus,
could you try how following memset implementation works on
aarch64.
It could be faster for small n, I found that on x64 overlapping stores
are best way how handle end conditions. I do not know how arm could handle that.
I do not know how good loop will gcc generate, if this is faster header
you can replace loop in assembly.
#include <stdint.h>
#include <stdlib.h>
/* Align VALUE down by ALIGN bytes. */
#define ALIGN_DOWN(value, align) \
ALIGN_DOWN_M1(value, align - 1)
/* Align VALUE down by ALIGN_M1 + 1 bytes.
Useful if you have precomputed ALIGN - 1. */
#define ALIGN_DOWN_M1(value, align_m1) \
(void *)((uintptr_t)(value) \
& ~(uintptr_t)(align_m1))
/* Align VALUE up by ALIGN bytes. */
#define ALIGN_UP(value, align) \
ALIGN_UP_M1(value, align - 1)
/* Align VALUE up by ALIGN_M1 + 1 bytes.
Useful if you have precomputed ALIGN - 1. */
#define ALIGN_UP_M1(value, align_m1) \
(void *)(((uintptr_t)(value) + (uintptr_t)(align_m1)) \
& ~(uintptr_t)(align_m1))
#define STOREU(x,y) STORE(x,y)
#define STORE(x,y) ((uint64_t*)(x))[0]=y; ((uint64_t*)(x))[1]=y;
static char *memset_small (char *dest, uint64_t c, size_t no, char *ret);
void *memset_new(char *dest, int _c, size_t n)
{
int i;
unsigned char c = _c;
uint64_t vc = 0x0101010101010101ULL*c;
if (n < 16)
{
return memset_small(dest, vc, n, dest);
}
else
{
STOREU(dest, vc);
STOREU(dest + n - 16, vc);
char *to = ALIGN_DOWN(dest + n, 16);
dest = ALIGN_DOWN(dest + 16, 16);
while (dest != to)
{
STORE(dest,vc);
dest += 16;
}
}
return dest;
}
static char *memset_small (char *dest, uint64_t c, size_t no, char *ret)
{
if (no & (8))
{
((uint64_t *) dest)[0] = c;
((uint64_t *)(dest + no - 8))[0] = c;
return ret;
}
if (no & 4)
{
((uint32_t *) dest)[0] = c;
((uint32_t *)(dest + no - 4))[0] = c;
return ret;
}
if (no & 1)
{
dest[0] = c;
}
if (no & 2)
{
((uint16_t *)(dest + no - 2))[0] = c;
}
return ret;
}
reply other threads:[~2013-03-25 15:33 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20130325152822.GA23513@domone.kolej.mff.cuni.cz \
--to=neleai@seznam.cz \
--cc=libc-ports@sourceware.org \
--cc=marcus.shawcroft@linaro.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).