KIQ doesn't really use the GPU scheduler. The base
drivers generally use the KIQ ring directly rather than
submitting IBs. However, amdgpu_sched_hw_submission
(which defaults to 2) limits the number of outstanding
fences to 2. KFD uses the KIQ for TLB flushes and the
2 fence limit hurts performance when there are several KFD
processes running.
v2: move some expressions to one line
change KIQ sched_hw_submission to at least 16
v3: bump to 256
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If a ring is not initialized, it also should not be finished.
For example, in Vega10's SR-IOV environment, UVD's decode ring is not
initialized, but will be finnished in amdgpu_uvd_sw_fini, because UVD
driver put all the uvd decode ring's finish operation into
amdgpu_uvd_sw_fini function, while not uvd_vXXX_0_sw_fini. This will
lead to amdgpu module unloading failure.
Signed-off-by: Trigger Huang <trigger.huang@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
May waste a bit of memory, but simplifies the interface
significantly.
v2: convert internal accounting to use 256bit slots
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In review, Christian would like to keep the logic
inside amdgpu_vm.c with a cost of slightly slower.
The loop is still optimized out with this patch.
v2: remove the if statement. Now it is not slower.
Signed-off-by: Alex Xie <AlexBin.Xie@amd.com>
Reviewed-by: Christian König <christian.koeng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Depending on usage patterns, the current LRU policy may create a
non-injective mapping between userspace ring ids and kernel rings.
This behaviour is undesired as apps that attempt to fill all HW blocks
would be unable to reach some of them.
This change forces the LRU policy to create bijective mappings only.
v2: compress ring_blacklist
v3: simplify amdgpu_ring_is_blacklisted() logic
Signed-off-by: Andres Rodriguez <andresx7@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use an LRU policy to map usermode rings to HW compute queues.
Most compute clients use one queue, and usually the first queue
available. This results in poor pipe/queue work distribution when
multiple compute apps are running. In most cases pipe 0 queue 0 is
the only queue that gets used.
In order to better distribute work across multiple HW queues, we adopt
a policy to map the usermode ring ids to the LRU HW queue.
This fixes a large majority of multi-app compute workloads sharing the
same HW queue, even though 7 other queues are available.
v2: use ring->funcs->type instead of ring->hw_ip
v3: remove amdgpu_queue_mapper_funcs
v4: change ring_lru_list_lock to spinlock, grab only once in lru_get()
Signed-off-by: Andres Rodriguez <andresx7@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On gfx9 hardware the value is not wrapped and is a 64-bit value. So
we reduce it modulo the ring size.
Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
(v2) use buf_mask instead of computing on the fly
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
ring->buf_mask need be set prior to ring_clear_ring invoke
and fix ring_clear_ring as well which should use buf_mask
instead of ptr_mask
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Newer asics need 64 bit wptrs. If the wptr is now
smaller than the rptr that doesn't indicate a wrap-around
anymore.
v2: integrate Christian's comments.
Signed-off-by: Ken Wang <Qingqing.Wang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In resume routine, we need clr RB prior to the
ring test of engine, otherwise some engine hang
duplicated during GPU reset.
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Pull vfs updates from Al Viro:
- more ->d_init() stuff (work.dcache)
- pathname resolution cleanups (work.namei)
- a few missing iov_iter primitives - copy_from_iter_full() and
friends. Either copy the full requested amount, advance the iterator
and return true, or fail, return false and do _not_ advance the
iterator. Quite a few open-coded callers converted (and became more
readable and harder to fuck up that way) (work.iov_iter)
- several assorted patches, the big one being logfs removal
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
logfs: remove from tree
vfs: fix put_compat_statfs64() does not handle errors
namei: fold should_follow_link() with the step into not-followed link
namei: pass both WALK_GET and WALK_MORE to should_follow_link()
namei: invert WALK_PUT logics
namei: shift interpretation of LOOKUP_FOLLOW inside should_follow_link()
namei: saner calling conventions for mountpoint_last()
namei.c: get rid of user_path_parent()
switch getfrag callbacks to ..._full() primitives
make skb_add_data,{_nocache}() and skb_copy_to_page_nocache() advance only on success
[iov_iter] new primitives - copy_from_iter_full() and friends
don't open-code file_inode()
ceph: switch to use of ->d_init()
ceph: unify dentry_operations instances
lustre: switch to use of ->d_init()
It's constant, so it doesn't make to much sense to keep it
with the variable data.
v2: update vce and uvd phys mode ring structures as well
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
debugfs_create_file() returns NULL on error, it only returns error
pointers if debugfs isn't enabled in the config and we checked for that
earlier so it can't happen.
Fixes: 4f4824b556 ('drm/amd/amdgpu: Convert ring debugfs entries to binary')
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If a user tries to read a non-multiple of 4 bytes it would have
read until the end of the ring potentially crashing the user
task.
Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead of specifying the total ring size calculate that from the maximum
number of dw a submission can have and the number of concurrent submissions.
This fixes UVD with 8 concurrent submissions or more.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The padding depends on the firmware version and we need that for BO moves as
well, not only for VM updates.
v2: new approach of making pad_ib a ring function
v3: fix typo in macro name
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Rework the sdma structures in the driver to
consolidate all of the sdma info into a single
structure and allow for asics that may have
different numbers of sdma instances.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Embed the scheduler into the ring structure instead of allocating it.
Use the ring name directly instead of the id.
v2: rebased, whitespace cleanup
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Chunming Zhou<david1.zhou@amd.com>
Move the fence related stuff into amdgpu_fence.c
v2: rework commit message, cause this is actually not a bug
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou<david1.zhou@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
