On 13/10/2023 21:51, Rodrigo Vivi wrote:
On Thu, Sep 28, 2023 at 01:48:34PM +0100, Tvrtko Ursulin wrote:
On 27/09/2023 20:34, Belgaumkar, Vinay wrote:
On 9/21/2023 3:41 AM, Tvrtko Ursulin wrote:
On 20/09/2023 22:56, Vinay Belgaumkar wrote:
Provide a bit to disable waitboost while waiting on a gem object.
Waitboost results in increased power consumption by requesting RP0
while waiting for the request to complete. Add a bit in the gem_wait()
IOCTL where this can be disabled.
This is related to the libva API change here -
Link:
https://github.com/XinfengZhang/libva/commit/3d90d18c67609a73121bb71b20ee4776b54b61a7
This link does not appear to lead to userspace code using this uapi?
We have asked Carl (cc'd) to post a patch for the same.
Ack.
I'm glad to see that we will have the end-to-end flow of the high-level API.
Cc: Rodrigo Vivi <rodrigo.v...@intel.com>
Signed-off-by: Vinay Belgaumkar <vinay.belgaum...@intel.com>
---
drivers/gpu/drm/i915/gem/i915_gem_wait.c | 9 ++++++---
drivers/gpu/drm/i915/i915_request.c | 3 ++-
drivers/gpu/drm/i915/i915_request.h | 1 +
include/uapi/drm/i915_drm.h | 1 +
4 files changed, 10 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_wait.c
b/drivers/gpu/drm/i915/gem/i915_gem_wait.c
index d4b918fb11ce..955885ec859d 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_wait.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_wait.c
@@ -72,7 +72,8 @@ i915_gem_object_wait_reservation(struct
dma_resv *resv,
struct dma_fence *fence;
long ret = timeout ?: 1;
- i915_gem_object_boost(resv, flags);
+ if (!(flags & I915_WAITBOOST_DISABLE))
+ i915_gem_object_boost(resv, flags);
dma_resv_iter_begin(&cursor, resv,
dma_resv_usage_rw(flags & I915_WAIT_ALL));
@@ -236,7 +237,7 @@ i915_gem_wait_ioctl(struct drm_device *dev,
void *data, struct drm_file *file)
ktime_t start;
long ret;
- if (args->flags != 0)
+ if (args->flags != 0 || args->flags != I915_GEM_WAITBOOST_DISABLE)
return -EINVAL;
obj = i915_gem_object_lookup(file, args->bo_handle);
@@ -248,7 +249,9 @@ i915_gem_wait_ioctl(struct drm_device *dev,
void *data, struct drm_file *file)
ret = i915_gem_object_wait(obj,
I915_WAIT_INTERRUPTIBLE |
I915_WAIT_PRIORITY |
- I915_WAIT_ALL,
+ I915_WAIT_ALL |
+ (args->flags & I915_GEM_WAITBOOST_DISABLE ?
+ I915_WAITBOOST_DISABLE : 0),
to_wait_timeout(args->timeout_ns));
if (args->timeout_ns > 0) {
diff --git a/drivers/gpu/drm/i915/i915_request.c
b/drivers/gpu/drm/i915/i915_request.c
index f59081066a19..2957409b4b2a 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -2044,7 +2044,8 @@ long i915_request_wait_timeout(struct
i915_request *rq,
* but at a cost of spending more power processing the workload
* (bad for battery).
*/
- if (flags & I915_WAIT_PRIORITY && !i915_request_started(rq))
+ if (!(flags & I915_WAITBOOST_DISABLE) && (flags &
I915_WAIT_PRIORITY) &&
+ !i915_request_started(rq))
intel_rps_boost(rq);
wait.tsk = current;
diff --git a/drivers/gpu/drm/i915/i915_request.h
b/drivers/gpu/drm/i915/i915_request.h
index 0ac55b2e4223..3cc00e8254dc 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -445,6 +445,7 @@ long i915_request_wait(struct i915_request *rq,
#define I915_WAIT_INTERRUPTIBLE BIT(0)
#define I915_WAIT_PRIORITY BIT(1) /* small priority bump
for the request */
#define I915_WAIT_ALL BIT(2) /* used by
i915_gem_object_wait() */
+#define I915_WAITBOOST_DISABLE BIT(3) /* used by
maybe name it I915_WAIT_NO_BOOST to align a bit better with the existent ones?
I thought it would be better to not mention wait boost in the uapi, but
leave it as implementation detail.
My suggestion was along the lines of I915_GEM_WAIT_BACKGROUND/IDLE.
In essence saying allowing userspace to say this is not an important
wait. Yes, it implies that other waits are (more) important, but I think
this is still better than starting to mention wait boost in the uapi.
Since that would kind of cement it exists, while we always just viewed
it as an "go faster" driver internal heuristics and could freely decide
not to employ it even for default waits.
Historically even we had a period when waits were getting elevated
scheduling priority. We removed it, would have to dig through git and
email history to remember exactly why, but probably along the lines that
it is not always justified. Same as waitboost is not always justified
and can be harmful.
So I think a generic name for the uapi leaves more freedom for the
driver. Might be a wrong name that I am suggesting and should be
something else, not sure.
[Comes back later.]
eec39e441c29 ("drm/i915: Remove wait priority boosting")
So it seems we only removed it because corner cases with the current
scheduler were hard. Unfortunately improved deadline based scheduler
never got in despite being ready so we can not restore this now.
i915_gem_object_wait() */
void i915_request_show(struct drm_printer *m,
const struct i915_request *rq,
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 7000e5910a1d..4adee70e39cf 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -1928,6 +1928,7 @@ struct drm_i915_gem_wait {
/** Handle of BO we shall wait on */
__u32 bo_handle;
__u32 flags;
+#define I915_GEM_WAITBOOST_DISABLE (1u<<0)
Probably would be good to avoid mentioning waitboost in the uapi
since so far it wasn't an explicit feature/contract. Something like
I915_GEM_WAIT_BACKGROUND_PRIORITY? Low priority?
sure.
I also wonder if there could be a possible angle to help Rob (+cc)
upstream the syncobj/fence deadline code if our media driver might
make use of that somehow.
Like if either we could wire up the deadline into GEM_WAIT (in a
backward compatible manner), or if media could use sync fd wait
instead. Assuming they have an out fence already, which may not be
true.
Makes sense. We could add a SET_DEADLINE flag or something similar and
pass in the deadline when appropriate.
Rob - do you have time and motivation to think about this angle at all
currently? If not I guess we just proceed with the new flag for our
GEM_WAIT.
Well, this could be the first user for that uapi that Rob was proposing
indeed.
The downside is probably because we should implement the deadline in i915
and consider all the deadline as 0 (urgent) and boost, unless in this
case where before the gem_wait the UMD would use the set_deadline to
something higher (max?).
Well, if we have a clarity on how to proceed with the deadline we should
probably go there. But for simplicity I would be in favor of this proposed
gem_wait flag as is, because this already solves many real important cases.
Yes I don't think we had consensus on the semantics of when no deadline
is set, so it does sound better to proceed with i915 specific solution
for now. The two wouldn't be incompatible if deadlines were later added.
Regards,
Tvrtko
Regards,
Tvrtko
Thanks,
Vinay.
Regards,
Tvrtko
/** Number of nanoseconds to wait, Returns time remaining. */
__s64 timeout_ns;
};
