To handle 40000 to 80000 interrupts per second running CPX mode with 4 streams/queues per KFD node, KFD interrupt handler becomes the performance bottleneck.
Remove the kfifo_out memcpy overhead by accessing ih_fifo data in-place and updating rptr with kfifo_skip_count. Signed-off-by: Philip Yang <[email protected]> Reviewed-by: Felix Kuehling <[email protected]> --- drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c | 35 +++++++++------------- 1 file changed, 14 insertions(+), 21 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c index 9b6b6e882593..e7412de9a0ac 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c @@ -114,50 +114,43 @@ void kfd_interrupt_exit(struct kfd_node *node) */ bool enqueue_ih_ring_entry(struct kfd_node *node, const void *ih_ring_entry) { - int count; - - count = kfifo_in(&node->ih_fifo, ih_ring_entry, - node->kfd->device_info.ih_ring_entry_size); - if (count != node->kfd->device_info.ih_ring_entry_size) { + if (kfifo_is_full(&node->ih_fifo)) { dev_dbg_ratelimited(node->adev->dev, - "Interrupt ring overflow, dropping interrupt %d\n", - count); + "Interrupt ring overflow, dropping interrupt\n"); return false; } + kfifo_in(&node->ih_fifo, ih_ring_entry, node->kfd->device_info.ih_ring_entry_size); return true; } /* * Assumption: single reader/writer. This function is not re-entrant */ -static bool dequeue_ih_ring_entry(struct kfd_node *node, void *ih_ring_entry) +static bool dequeue_ih_ring_entry(struct kfd_node *node, u32 **ih_ring_entry) { int count; - count = kfifo_out(&node->ih_fifo, ih_ring_entry, - node->kfd->device_info.ih_ring_entry_size); - - WARN_ON(count && count != node->kfd->device_info.ih_ring_entry_size); + if (kfifo_is_empty(&node->ih_fifo)) + return false; + count = kfifo_out_linear_ptr(&node->ih_fifo, ih_ring_entry, + node->kfd->device_info.ih_ring_entry_size); + WARN_ON(count != node->kfd->device_info.ih_ring_entry_size); return count == node->kfd->device_info.ih_ring_entry_size; } static void interrupt_wq(struct work_struct *work) { - struct kfd_node *dev = container_of(work, struct kfd_node, - interrupt_work); - uint32_t ih_ring_entry[KFD_MAX_RING_ENTRY_SIZE]; + struct kfd_node *dev = container_of(work, struct kfd_node, interrupt_work); + uint32_t *ih_ring_entry; unsigned long start_jiffies = jiffies; - if (dev->kfd->device_info.ih_ring_entry_size > sizeof(ih_ring_entry)) { - dev_err_once(dev->adev->dev, "Ring entry too small\n"); - return; - } - - while (dequeue_ih_ring_entry(dev, ih_ring_entry)) { + while (dequeue_ih_ring_entry(dev, &ih_ring_entry)) { dev->kfd->device_info.event_interrupt_class->interrupt_wq(dev, ih_ring_entry); + kfifo_skip_count(&dev->ih_fifo, dev->kfd->device_info.ih_ring_entry_size); + if (time_is_before_jiffies(start_jiffies + HZ)) { /* If we spent more than a second processing signals, * reschedule the worker to avoid soft-lockup warnings -- 2.47.1
