From: John.C.Harrison@Intel.com
To: Intel-GFX@lists.freedesktop.org
Subject: [RFC 40/44] drm/i915: REVERTME Hack to allow IGT to test pre-emption
Date: Thu, 26 Jun 2014 18:24:31 +0100 [thread overview]
Message-ID: <1403803475-16337-41-git-send-email-John.C.Harrison@Intel.com> (raw)
In-Reply-To: <1403803475-16337-1-git-send-email-John.C.Harrison@Intel.com>
From: John Harrison <John.C.Harrison@Intel.com>
In order to test pre-emption, a flag has been added to the execbuffer() API to
explicitly request that a given batch buffer is made pre-emptive. This is purely
a temporary measure to allow an IGT test to queue pre-emptive and non-preemptive
work loads.
Note that the final solution will be to add an IOCTL to set the priority of a
batch buffer (work in progress elsewhere). The scheduler will then decide to
pre-empt or not based on the assigned priority level.
---
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 12 ++++++++++++
include/uapi/drm/i915_drm.h | 5 +++++
2 files changed, 17 insertions(+)
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 81acdf2..b7d0737 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -1361,6 +1361,18 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
qe.params.mask = mask;
qe.params.mode = mode;
+ /* Hack for testing pre-empting prior to having an official priority API */
+ if (qe.params.args_flags & I915_EXEC_PREEMPT) {
+#ifdef CONFIG_DRM_I915_SCHEDULER_PREEMPTION
+ struct i915_scheduler *scheduler = dev_priv->scheduler;
+ qe.priority += scheduler->priority_level_preempt;
+#else
+ DRM_DEBUG("Buffer flags 0x%X includes PREEMPT!\n",
+ qe.params.args_flags);
+#endif
+ }
+ /* Hack for testing pre-empting prior to having an official priority API */
+
#ifdef CONFIG_DRM_I915_SCHEDULER
/*
* Save away the list of objects used by this batch buffer for the
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index de6f603..d391222 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -694,6 +694,11 @@ struct drm_i915_gem_execbuffer2 {
#define I915_EXEC_BLT (3<<0)
#define I915_EXEC_VEBOX (4<<0)
+/* Pre-emption flag
+ * If this flag is set, this batchbuffer preempts those already submitted
+ */
+#define I915_EXEC_PREEMPT (1<<5)
+
/* Used for switching the constants addressing mode on gen4+ RENDER ring.
* Gen6+ only supports relative addressing to dynamic state (default) and
* absolute addressing.
--
1.7.9.5
next prev parent reply other threads:[~2014-06-26 17:26 UTC|newest]
Thread overview: 90+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-06-26 17:23 [RFC 00/44] GPU scheduler for i915 driver John.C.Harrison
2014-06-26 17:23 ` [RFC 01/44] drm/i915: Corrected 'file_priv' to 'file' in 'i915_driver_preclose()' John.C.Harrison
2014-06-30 21:03 ` Jesse Barnes
2014-07-07 18:02 ` Daniel Vetter
2014-06-26 17:23 ` [RFC 02/44] drm/i915: Added getparam for native sync John.C.Harrison
2014-07-07 18:52 ` Daniel Vetter
2014-06-26 17:23 ` [RFC 03/44] drm/i915: Add extra add_request calls John.C.Harrison
2014-06-30 21:10 ` Jesse Barnes
2014-07-07 18:41 ` Daniel Vetter
2014-07-08 7:44 ` Chris Wilson
2014-06-26 17:23 ` [RFC 04/44] drm/i915: Fix null pointer dereference in error capture John.C.Harrison
2014-06-30 21:40 ` Jesse Barnes
2014-07-01 7:12 ` Chris Wilson
2014-07-07 18:49 ` Daniel Vetter
2014-07-01 7:20 ` [PATCH] drm/i915: Remove num_pages parameter to i915_error_object_create() Chris Wilson
2014-06-26 17:23 ` [RFC 05/44] drm/i915: Updating assorted register and status page definitions John.C.Harrison
2014-07-02 17:49 ` Jesse Barnes
2014-06-26 17:23 ` [RFC 06/44] drm/i915: Fixes for FIFO space queries John.C.Harrison
2014-07-02 17:50 ` Jesse Barnes
2014-06-26 17:23 ` [RFC 07/44] drm/i915: Disable 'get seqno' workaround for VLV John.C.Harrison
2014-07-02 17:51 ` Jesse Barnes
2014-07-07 18:56 ` Daniel Vetter
2014-06-26 17:23 ` [RFC 08/44] drm/i915: Added GPU scheduler config option John.C.Harrison
2014-07-07 18:58 ` Daniel Vetter
2014-06-26 17:24 ` [RFC 09/44] drm/i915: Start of GPU scheduler John.C.Harrison
2014-07-02 17:55 ` Jesse Barnes
2014-07-07 19:02 ` Daniel Vetter
2014-06-26 17:24 ` [RFC 10/44] drm/i915: Prepare retire_requests to handle out-of-order seqnos John.C.Harrison
2014-07-02 18:11 ` Jesse Barnes
2014-07-07 19:05 ` Daniel Vetter
2014-07-09 14:08 ` Daniel Vetter
2014-06-26 17:24 ` [RFC 11/44] drm/i915: Added scheduler hook into i915_seqno_passed() John.C.Harrison
2014-07-02 18:14 ` Jesse Barnes
2014-06-26 17:24 ` [RFC 12/44] drm/i915: Disable hardware semaphores when GPU scheduler is enabled John.C.Harrison
2014-07-02 18:16 ` Jesse Barnes
2014-06-26 17:24 ` [RFC 13/44] drm/i915: Added scheduler hook when closing DRM file handles John.C.Harrison
2014-07-02 18:20 ` Jesse Barnes
2014-07-23 15:10 ` John Harrison
2014-07-23 15:39 ` Jesse Barnes
2014-06-26 17:24 ` [RFC 14/44] drm/i915: Added getparam for GPU scheduler John.C.Harrison
2014-07-02 18:21 ` Jesse Barnes
2014-07-07 19:11 ` Daniel Vetter
2014-06-26 17:24 ` [RFC 15/44] drm/i915: Added deferred work handler for scheduler John.C.Harrison
2014-07-07 19:14 ` Daniel Vetter
2014-07-23 15:37 ` John Harrison
2014-07-23 18:50 ` Daniel Vetter
2014-07-24 15:42 ` John Harrison
2014-07-25 7:18 ` Daniel Vetter
2014-06-26 17:24 ` [RFC 16/44] drm/i915: Alloc early seqno John.C.Harrison
2014-07-02 18:29 ` Jesse Barnes
2014-07-23 15:11 ` John Harrison
2014-06-26 17:24 ` [RFC 17/44] drm/i915: Prelude to splitting i915_gem_do_execbuffer in two John.C.Harrison
2014-07-02 18:34 ` Jesse Barnes
2014-07-07 19:21 ` Daniel Vetter
2014-07-23 16:33 ` John Harrison
2014-07-23 18:14 ` Daniel Vetter
2014-06-26 17:24 ` [RFC 18/44] drm/i915: Added scheduler debug macro John.C.Harrison
2014-07-02 18:37 ` Jesse Barnes
2014-07-07 19:23 ` Daniel Vetter
2014-06-26 17:24 ` [RFC 19/44] drm/i915: Split i915_dem_do_execbuffer() in half John.C.Harrison
2014-06-26 17:24 ` [RFC 20/44] drm/i915: Redirect execbuffer_final() via scheduler John.C.Harrison
2014-06-26 17:24 ` [RFC 21/44] drm/i915: Added tracking/locking of batch buffer objects John.C.Harrison
2014-06-26 17:24 ` [RFC 22/44] drm/i915: Ensure OLS & PLR are always in sync John.C.Harrison
2014-06-26 17:24 ` [RFC 23/44] drm/i915: Added manipulation of OLS/PLR John.C.Harrison
2014-06-26 17:24 ` [RFC 24/44] drm/i915: Added scheduler interrupt handler hook John.C.Harrison
2014-06-26 17:24 ` [RFC 25/44] drm/i915: Added hook to catch 'unexpected' ring submissions John.C.Harrison
2014-06-26 17:24 ` [RFC 26/44] drm/i915: Added scheduler support to __wait_seqno() calls John.C.Harrison
2014-06-26 17:24 ` [RFC 27/44] drm/i915: Added scheduler support to page fault handler John.C.Harrison
2014-06-26 17:24 ` [RFC 28/44] drm/i915: Added scheduler flush calls to ring throttle and idle functions John.C.Harrison
2014-06-26 17:24 ` [RFC 29/44] drm/i915: Hook scheduler into intel_ring_idle() John.C.Harrison
2014-06-26 17:24 ` [RFC 30/44] drm/i915: Added a module parameter for allowing scheduler overrides John.C.Harrison
2014-06-26 17:24 ` [RFC 31/44] drm/i915: Implemented the GPU scheduler John.C.Harrison
2014-06-26 17:24 ` [RFC 32/44] drm/i915: Added immediate submission override to scheduler John.C.Harrison
2014-06-26 17:24 ` [RFC 33/44] drm/i915: Added trace points " John.C.Harrison
2014-06-26 17:24 ` [RFC 34/44] drm/i915: Added scheduler queue throttling by DRM file handle John.C.Harrison
2014-06-26 17:24 ` [RFC 35/44] drm/i915: Added debugfs interface to scheduler tuning parameters John.C.Harrison
2014-06-26 17:24 ` [RFC 36/44] drm/i915: Added debug state dump facilities to scheduler John.C.Harrison
2014-06-26 17:24 ` [RFC 37/44] drm/i915: Added facility for cancelling an outstanding request John.C.Harrison
2014-06-26 17:24 ` [RFC 38/44] drm/i915: Add early exit to execbuff_final() if insufficient ring space John.C.Harrison
2014-06-26 17:24 ` [RFC 39/44] drm/i915: Added support for pre-emptive scheduling John.C.Harrison
2014-06-26 17:24 ` John.C.Harrison [this message]
2014-06-26 17:24 ` [RFC 41/44] drm/i915: Added validation callback to trace points John.C.Harrison
2014-06-26 17:24 ` [RFC 42/44] drm/i915: Added scheduler statistic reporting to debugfs John.C.Harrison
2014-06-26 17:24 ` [RFC 43/44] drm/i915: Added support for submitting out-of-batch ring commands John.C.Harrison
2014-06-26 17:24 ` [RFC 44/44] drm/i915: Fake batch support for page flips John.C.Harrison
2014-07-07 19:25 ` Daniel Vetter
2014-06-26 20:44 ` [RFC 00/44] GPU scheduler for i915 driver Dave Airlie
2014-07-07 15:57 ` Daniel Vetter
2014-10-10 10:35 ` Steven Newbury
2014-10-20 10:31 ` John Harrison
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=1403803475-16337-41-git-send-email-John.C.Harrison@Intel.com \
--to=john.c.harrison@intel.com \
--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