calloc() implementation question

[Olivier Langlois]

Ondřej

On Thu, 2013-12-12 at 19:41 +0100, Ondřej Bílka wrote:
> To see how big is speedup/loss write a benchmark that compares variant
> with memset and one with inline expansion. 
> 
> void *
> calloc2 (size_t n)
> {
>   return memset (malloc (n), 0, n);
> }
> 
> void *
> calloc3 (size_t n)
> {
>   void *x = malloc (n);
>   if (n < 9 * 16)
>     ...
> }

that was an amazing suggestion as it did allow me to have some new
insight in glibc malloc/calloc performance.

first, the manual unrolled memset did turned out to be faster than the
unconditionnal memset() call.

The other unexpected result was that malloc + memset() (calloc2()) is
always faster than the real calloc() for allocation size <= ~ 64KB.

that was unexpected as I tought that because calloc may skip the memset
step, it would be faster and at worse be on par with malloc + memset().

I did my tests on Linux kernel 3.12.4 on a
Intel(R) Atom(TM) CPU N455   @ 1.66GHz

-------------- next part --------------
A non-text attachment was scrubbed...
Name: calloc_emul.c
Type: text/x-csrc
Size: 659 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20131213/cdec2163/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: tst-calloc.c
Type: text/x-csrc
Size: 1113 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20131213/cdec2163/attachment-0001.bin>