Lines Matching refs:from
398 * request that we have removed from the HW and put back on a run
478 * that we received an ACK from the HW, and so the context is not
522 * __notify_execute_cb from i915_request_submit() and we are not
563 * As this request likely depends on state from the lost
617 * requests that we have unsubmitted from HW, but left running
623 * We must remove the request from the caller's priority queue,
684 * busywaits from the GPU until their master is restarted. This is
691 /* We may be recursing from the signal callback of another i915 fence */
703 /* Will be called from irq-context when using foreign fences. */
725 * Before we remove this breadcrumb from the signal list, we have
754 /* Will be called from irq-context when using foreign fences. */
862 /* Ratelimit ourselves to prevent oom from malicious clients */
911 * race with the request being allocated from the slab freelist.
1073 * against it being retired and removed from the lists.
1133 * than one semaphore from each source, and not using a semaphore
1145 struct i915_request *from,
1157 err = intel_timeline_read_hwsp(from, to, &hwsp_offset);
1195 can_use_semaphore_wait(struct i915_request *to, struct i915_request *from)
1197 return to->engine->gt->ggtt == from->engine->gt->ggtt;
1202 struct i915_request *from,
1205 const intel_engine_mask_t mask = READ_ONCE(from->engine)->mask;
1208 if (!can_use_semaphore_wait(to, from))
1223 if (from->sched.flags & I915_SCHED_HAS_EXTERNAL_CHAIN)
1230 if (i915_request_await_start(to, from) < 0)
1234 if (__await_execution(to, from, gfp))
1237 if (__emit_semaphore_wait(to, from, from->fence.seqno))
1245 &from->fence, 0,
1265 struct i915_request *from)
1269 GEM_BUG_ON(intel_context_is_barrier(from->context));
1272 err = __await_execution(to, from, I915_FENCE_GFP);
1278 &from->fence))
1291 err = i915_request_await_start(to, from);
1301 * delayed by waiting on a semaphore from a remote engine. If our
1314 if (can_use_semaphore_wait(to, from) &&
1317 err = __emit_semaphore_wait(to, from, from->fence.seqno - 1);
1325 &from->sched,
1332 &from->fence);
1395 struct i915_request *from)
1398 return request_to_parent(to) == request_to_parent(from);
1450 await_request_submit(struct i915_request *to, struct i915_request *from)
1460 if (to->engine == READ_ONCE(from->engine))
1462 &from->submit,
1465 return __i915_request_await_execution(to, from);
1469 i915_request_await_request(struct i915_request *to, struct i915_request *from)
1473 GEM_BUG_ON(to == from);
1474 GEM_BUG_ON(to->timeline == from->timeline);
1476 if (i915_request_completed(from)) {
1477 i915_sw_fence_set_error_once(&to->submit, from->fence.error);
1483 &from->sched,
1490 is_power_of_2(to->execution_mask | READ_ONCE(from->execution_mask)))
1491 ret = await_request_submit(to, from);
1493 ret = emit_semaphore_wait(to, from, I915_FENCE_GFP);
1512 * amdgpu and we should not see any incoming fence-array from
1758 * order received from the execbuf IOCTL. So rather than using the
1874 * Cheaply and approximately convert from nanoseconds to microseconds.
1877 * error here is from the simple truncation.
2012 * switching away from the active task. Frequently, the client will
2023 * impacts from a context switch such as cache eviction.
2117 * NOTE: This function behaves differently from dma-fence wait semantics for
2214 * - the request has been transferred from the backend queue and
2219 * from the lists