[Mesa-dev] [PATCH 2/8] anv: Use library mtime for cache UUID.

[Kenneth Graunke]

On Thursday, November 24, 2016 8:30:39 PM PST Emil Velikov wrote:
> From: Emil Velikov <emil.velikov at collabora.com>
> 
> Inspired by a similar commit for radv.
> 
> Rather than recomputing the timestamp on each make invocation, just
> fetch it at runtime.
> 
> Thus we no longer get the constant rebuild of anv_device.c and the
> follow-up libvulkan_intel.so link, when nothing has changed.
> 
> I.e. using make && make install is a little bit faster.
> 
> Signed-off-by: Emil Velikov <emil.velikov at collabora.com>
> ---
>  src/intel/vulkan/anv_device.c | 37 +++++++++++++++++++++++++++++++++----
>  1 file changed, 33 insertions(+), 4 deletions(-)
> 
> diff --git a/src/intel/vulkan/anv_device.c b/src/intel/vulkan/anv_device.c
> index 58c6b3f..4711501 100644
> --- a/src/intel/vulkan/anv_device.c
> +++ b/src/intel/vulkan/anv_device.c
> @@ -21,15 +21,16 @@
>   * IN THE SOFTWARE.
>   */
>  
> +#include <dlfcn.h>
>  #include <assert.h>
>  #include <stdbool.h>
>  #include <string.h>
>  #include <sys/mman.h>
> +#include <sys/stat.h>
>  #include <unistd.h>
>  #include <fcntl.h>
>  
>  #include "anv_private.h"
> -#include "anv_timestamp.h"
>  #include "util/strtod.h"
>  #include "util/debug.h"
>  
> @@ -53,11 +54,32 @@ compiler_perf_log(void *data, const char *fmt, ...)
>     va_end(args);
>  }
>  
> -static void
> +static int

I'd prefer to have us return a bool on success - we don't tend to use
the "0 means success" idiom much in Mesa.

> +anv_get_function_timestamp(void *ptr, uint32_t* timestamp)
> +{
> +   Dl_info info;
> +   struct stat st;
> +   if (!dladdr(ptr, &info) || !info.dli_fname)
> +      return -1;
> +
> +   if (stat(info.dli_fname, &st))
> +      return -1;
> +
> +   *timestamp = st.st_mtim.tv_sec;
> +   return 0;
> +}
> +
> +static int

Ditto.

>  anv_device_get_cache_uuid(void *uuid)
>  {
> +   uint32_t timestamp;
> +
>     memset(uuid, 0, VK_UUID_SIZE);
> -   snprintf(uuid, VK_UUID_SIZE, "anv-%s", ANV_TIMESTAMP);
> +   if (anv_get_function_timestamp(anv_device_get_cache_uuid, &timestamp))
> +         return -1;
> +
> +   snprintf(uuid, VK_UUID_SIZE, "anv-%d", timestamp);
> +   return 0;
>  }
>  
>  static VkResult
> @@ -186,7 +208,14 @@ anv_physical_device_init(struct anv_physical_device *device,
>     if (result != VK_SUCCESS)
>         goto fail;
>  
> -   anv_device_get_cache_uuid(device->uuid);
> +   if (anv_device_get_cache_uuid(device->uuid)) {
> +      anv_finish_wsi(device);
> +      ralloc_free(device->compiler);
> +      result = vk_errorf(VK_ERROR_INITIALIZATION_FAILED,
> +                         "cannot generate UUID");
> +      goto fail;
> +   }
> +

I don't see why this error case is special - all the others just have
"goto fail".  This stuff may need to get cleaned up, but shouldn't it
happen for the other error paths too?

I wrote these same patches today, not realizing you'd already done it,
so with those comments addressed, consider it a series-wide:

Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>

But please wait for Jason - it sounds like he has some concerns and
I'm not clear what those are.

>     /* XXX: Actually detect bit6 swizzling */
>     isl_device_init(&device->isl_dev, &device->info, swizzled);
> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: This is a digitally signed message part.
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20161126/0e2f23f1/attachment.sig>