Re: [Xen-devel] [PATCH] perf: Replace malloc with alloca in hot path

[Santosh Jodh <Santosh.Jodh@xxxxxxxxxx>]

Yes - that is correct. This is used in the IO path for user mode block backend.

Thanks,
Santosh

-----Original Message-----
From: Ian Campbell 
Sent: Wednesday, February 01, 2012 8:15 AM
To: Santosh Jodh
Cc: xen-devel@xxxxxxxxxxxxxxxxxxx
Subject: Re: [PATCH] perf: Replace malloc with alloca in hot path

On Mon, 2012-01-30 at 19:07 +0000, Santosh Jodh wrote:
> Replace malloc with alloc in hot paths for improved performance.

This is a hot path in something like a userspace PV backend which is frequently 
mapping guest domain pages, or something like that?

> Signed-off-by: Santosh Jodh <santosh.jodh@xxxxxxxxxx>

Acked-by: Ian Campbell <ian.campbell@xxxxxxxxxx>

> 
> diff -r e2722b24dc09 -r 4952090d35e0 tools/libxc/xc_linux_osdep.c
> --- a/tools/libxc/xc_linux_osdep.c    Thu Jan 26 17:43:31 2012 +0000
> +++ b/tools/libxc/xc_linux_osdep.c    Mon Jan 30 11:02:32 2012 -0800
> @@ -242,7 +242,7 @@ static void *linux_privcmd_map_foreign_b
>           * IOCTL_PRIVCMD_MMAPBATCH_V2 is not supported - fall back to
>           * IOCTL_PRIVCMD_MMAPBATCH.
>           */
> -        xen_pfn_t *pfn = malloc(num * sizeof(*pfn));
> +        xen_pfn_t *pfn = alloca(num * sizeof(*pfn));
>  
>          if ( pfn )
>          {
> @@ -288,8 +288,6 @@ static void *linux_privcmd_map_foreign_b
>                  break;
>              }
>  
> -            free(pfn);
> -
>              if ( rc == -ENOENT && i == num )
>                  rc = 0;
>              else if ( rc )
> @@ -524,7 +522,7 @@ static void *linux_gnttab_grant_map(xc_g
>      if (flags & XC_GRANT_MAP_SINGLE_DOMAIN)
>          domids_stride = 0;
>  
> -    map = malloc(sizeof(*map) +
> +    map = alloca(sizeof(*map) +
>                   (count - 1) * sizeof(struct ioctl_gntdev_map_grant_ref));
>      if ( map == NULL )
>          return NULL;
> @@ -598,7 +596,6 @@ static void *linux_gnttab_grant_map(xc_g
>      }
>  
>   out:
> -    free(map);
>  
>      return addr;
>  }


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel