dri-devel.lists.freedesktop.org archive mirror
 help / color / mirror / Atom feed
From: jeff.mcgee@intel.com
To: beignet@lists.freedesktop.org
Cc: intel-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org
Subject: [PATCH 2/2] Query the driver directly for compute units and subslice
Date: Mon,  2 Mar 2015 15:42:39 -0800	[thread overview]
Message-ID: <1425339759-19027-2-git-send-email-jeff.mcgee@intel.com> (raw)
In-Reply-To: <1425339759-19027-1-git-send-email-jeff.mcgee@intel.com>

From: Jeff McGee <jeff.mcgee@intel.com>

Values of device max compute units and max subslice obtained
directly from the driver should be more accurate than our own
ID-based lookup values. This is particularly important when a
single device ID may encompass more than one configuration. If
the driver cannot provide a valid value for the given device,
we fallback on the ID-based lookup value.

Signed-off-by: Jeff McGee <jeff.mcgee@intel.com>
---
 src/intel/intel_driver.c | 16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)

diff --git a/src/intel/intel_driver.c b/src/intel/intel_driver.c
index d61988c..d99fea9 100644
--- a/src/intel/intel_driver.c
+++ b/src/intel/intel_driver.c
@@ -757,10 +757,8 @@ static int intel_buffer_set_tiling(cl_buffer bo,
 static void
 intel_update_device_info(cl_device_id device)
 {
-#ifdef HAS_USERPTR
   intel_driver_t *driver;
-  const size_t sz = 4096;
-  void *host_ptr;
+  unsigned int eu_total, subslice_total;
 
   driver = intel_driver_new();
   assert(driver != NULL);
@@ -769,6 +767,10 @@ intel_update_device_info(cl_device_id device)
     return;
   }
 
+#ifdef HAS_USERPTR
+  const size_t sz = 4096;
+  void *host_ptr;
+
   host_ptr = cl_aligned_malloc(sz, 4096);
   if (host_ptr != NULL) {
     cl_buffer bo = intel_buffer_alloc_userptr((cl_buffer_mgr)driver->bufmgr,
@@ -781,12 +783,18 @@ intel_update_device_info(cl_device_id device)
   }
   else
     device->host_unified_memory = CL_FALSE;
+#endif
+
+  /* Prefer driver-queried value if supported */
+  if (!drm_intel_get_eu_total(driver->fd, &eu_total))
+    device->max_compute_unit = eu_total;
+  if (!drm_intel_get_subslice_total(driver->fd, &subslice_total))
+    device->sub_slice_count = subslice_total;
 
   intel_driver_context_destroy(driver);
   intel_driver_close(driver);
   intel_driver_terminate(driver);
   intel_driver_delete(driver);
-#endif
 }
 
 LOCAL void
-- 
2.3.0

_______________________________________________
Beignet mailing list
Beignet@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/beignet

  reply	other threads:[~2015-03-02 23:42 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-03-02 23:42 [PATCH 1/2] Add driver callback for updating device info jeff.mcgee
2015-03-02 23:42 ` jeff.mcgee [this message]
2015-03-09 23:35   ` [PATCH 2/2 v2] Query the driver directly for compute units and subslice jeff.mcgee
2015-03-12  2:08     ` [Beignet] " Zhigang Gong
2015-03-12 20:52       ` Jeff McGee
2015-03-09  0:21 ` [Beignet] [PATCH 1/2] Add driver callback for updating device info Zhigang Gong

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=1425339759-19027-2-git-send-email-jeff.mcgee@intel.com \
    --to=jeff.mcgee@intel.com \
    --cc=beignet@lists.freedesktop.org \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=intel-gfx@lists.freedesktop.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 NNTP newsgroup(s).