diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c')
-rw-r--r-- | drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c | 115 |
1 files changed, 93 insertions, 22 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c index 37930629edc5..b1a6eb349bb3 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c @@ -28,6 +28,11 @@ #include "kfd_kernel_queue.h" #include "kfd_priv.h" +#define OVER_SUBSCRIPTION_PROCESS_COUNT (1 << 0) +#define OVER_SUBSCRIPTION_COMPUTE_QUEUE_COUNT (1 << 1) +#define OVER_SUBSCRIPTION_GWS_QUEUE_COUNT (1 << 2) +#define OVER_SUBSCRIPTION_XNACK_CONFLICT (1 << 3) + static inline void inc_wptr(unsigned int *wptr, unsigned int increment_bytes, unsigned int buffer_size_bytes) { @@ -40,7 +45,8 @@ static inline void inc_wptr(unsigned int *wptr, unsigned int increment_bytes, static void pm_calc_rlib_size(struct packet_manager *pm, unsigned int *rlib_size, - bool *over_subscription) + int *over_subscription, + int xnack_conflict) { unsigned int process_count, queue_count, compute_queue_count, gws_queue_count; unsigned int map_queue_size; @@ -58,17 +64,22 @@ static void pm_calc_rlib_size(struct packet_manager *pm, * hws_max_conc_proc has been done in * kgd2kfd_device_init(). */ - *over_subscription = false; + *over_subscription = 0; if (node->max_proc_per_quantum > 1) max_proc_per_quantum = node->max_proc_per_quantum; - if ((process_count > max_proc_per_quantum) || - compute_queue_count > get_cp_queues_num(pm->dqm) || - gws_queue_count > 1) { - *over_subscription = true; + if (process_count > max_proc_per_quantum) + *over_subscription |= OVER_SUBSCRIPTION_PROCESS_COUNT; + if (compute_queue_count > get_cp_queues_num(pm->dqm)) + *over_subscription |= OVER_SUBSCRIPTION_COMPUTE_QUEUE_COUNT; + if (gws_queue_count > 1) + *over_subscription |= OVER_SUBSCRIPTION_GWS_QUEUE_COUNT; + if (xnack_conflict && (node->adev->gmc.xnack_flags & AMDGPU_GMC_XNACK_FLAG_CHAIN)) + *over_subscription |= OVER_SUBSCRIPTION_XNACK_CONFLICT; + + if (*over_subscription) dev_dbg(dev, "Over subscribed runlist\n"); - } map_queue_size = pm->pmf->map_queues_size; /* calculate run list ib allocation size */ @@ -89,7 +100,8 @@ static int pm_allocate_runlist_ib(struct packet_manager *pm, unsigned int **rl_buffer, uint64_t *rl_gpu_buffer, unsigned int *rl_buffer_size, - bool *is_over_subscription) + int *is_over_subscription, + int xnack_conflict) { struct kfd_node *node = pm->dqm->dev; struct device *dev = node->adev->dev; @@ -98,7 +110,8 @@ static int pm_allocate_runlist_ib(struct packet_manager *pm, if (WARN_ON(pm->allocated)) return -EINVAL; - pm_calc_rlib_size(pm, rl_buffer_size, is_over_subscription); + pm_calc_rlib_size(pm, rl_buffer_size, is_over_subscription, + xnack_conflict); mutex_lock(&pm->lock); @@ -134,12 +147,28 @@ static int pm_create_runlist_ib(struct packet_manager *pm, struct qcm_process_device *qpd; struct queue *q; struct kernel_queue *kq; - bool is_over_subscription; + int is_over_subscription; + int xnack_enabled = -1; + bool xnack_conflict = 0; rl_wptr = retval = processes_mapped = 0; + /* Check if processes set different xnack modes */ + list_for_each_entry(cur, queues, list) { + qpd = cur->qpd; + if (xnack_enabled < 0) + /* First process */ + xnack_enabled = qpd->pqm->process->xnack_enabled; + else if (qpd->pqm->process->xnack_enabled != xnack_enabled) { + /* Found a process with a different xnack mode */ + xnack_conflict = 1; + break; + } + } + retval = pm_allocate_runlist_ib(pm, &rl_buffer, rl_gpu_addr, - &alloc_size_bytes, &is_over_subscription); + &alloc_size_bytes, &is_over_subscription, + xnack_conflict); if (retval) return retval; @@ -149,9 +178,13 @@ static int pm_create_runlist_ib(struct packet_manager *pm, dev_dbg(dev, "Building runlist ib process count: %d queues count %d\n", pm->dqm->processes_count, pm->dqm->active_queue_count); +build_runlist_ib: /* build the run list ib packet */ list_for_each_entry(cur, queues, list) { qpd = cur->qpd; + /* group processes with the same xnack mode together */ + if (qpd->pqm->process->xnack_enabled != xnack_enabled) + continue; /* build map process packet */ if (processes_mapped >= pm->dqm->processes_count) { dev_dbg(dev, "Not enough space left in runlist IB\n"); @@ -208,20 +241,33 @@ static int pm_create_runlist_ib(struct packet_manager *pm, alloc_size_bytes); } } + if (xnack_conflict) { + /* pick up processes with the other xnack mode */ + xnack_enabled = !xnack_enabled; + xnack_conflict = 0; + goto build_runlist_ib; + } dev_dbg(dev, "Finished map process and queues to runlist\n"); if (is_over_subscription) { if (!pm->is_over_subscription) - dev_warn( - dev, - "Runlist is getting oversubscribed. Expect reduced ROCm performance.\n"); + dev_warn(dev, "Runlist is getting oversubscribed due to%s%s%s%s. Expect reduced ROCm performance.\n", + is_over_subscription & OVER_SUBSCRIPTION_PROCESS_COUNT ? + " too many processes" : "", + is_over_subscription & OVER_SUBSCRIPTION_COMPUTE_QUEUE_COUNT ? + " too many queues" : "", + is_over_subscription & OVER_SUBSCRIPTION_GWS_QUEUE_COUNT ? + " multiple processes using cooperative launch" : "", + is_over_subscription & OVER_SUBSCRIPTION_XNACK_CONFLICT ? + " xnack on/off processes mixed on gfx9" : ""); + retval = pm->pmf->runlist(pm, &rl_buffer[rl_wptr], *rl_gpu_addr, alloc_size_bytes / sizeof(uint32_t), true); } - pm->is_over_subscription = is_over_subscription; + pm->is_over_subscription = !!is_over_subscription; for (i = 0; i < alloc_size_bytes / sizeof(uint32_t); i++) pr_debug("0x%2X ", rl_buffer[i]); @@ -248,7 +294,8 @@ int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm) default: if (KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 4, 2) || KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 4, 3) || - KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 4, 4)) + KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 4, 4) || + KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 5, 0)) pm->pmf = &kfd_aldebaran_pm_funcs; else if (KFD_GC_VERSION(dqm->dev) >= IP_VERSION(9, 0, 1)) pm->pmf = &kfd_v9_pm_funcs; @@ -383,14 +430,33 @@ out: return retval; } -int pm_update_grace_period(struct packet_manager *pm, uint32_t grace_period) +/* pm_config_dequeue_wait_counts: Configure dequeue timer Wait Counts + * by writing to CP_IQ_WAIT_TIME2 registers. + * + * @cmd: See emum kfd_config_dequeue_wait_counts_cmd definition + * @value: Depends on the cmd. This parameter is unused for + * KFD_DEQUEUE_WAIT_INIT and KFD_DEQUEUE_WAIT_RESET. For + * KFD_DEQUEUE_WAIT_SET_SCH_WAVE it holds value to be set + * + */ +int pm_config_dequeue_wait_counts(struct packet_manager *pm, + enum kfd_config_dequeue_wait_counts_cmd cmd, + uint32_t value) { struct kfd_node *node = pm->dqm->dev; struct device *dev = node->adev->dev; int retval = 0; uint32_t *buffer, size; - size = pm->pmf->set_grace_period_size; + if (!pm->pmf->config_dequeue_wait_counts || + !pm->pmf->config_dequeue_wait_counts_size) + return 0; + + if (cmd == KFD_DEQUEUE_WAIT_INIT && (KFD_GC_VERSION(pm->dqm->dev) < IP_VERSION(9, 4, 1) || + KFD_GC_VERSION(pm->dqm->dev) >= IP_VERSION(10, 0, 0))) + return 0; + + size = pm->pmf->config_dequeue_wait_counts_size; mutex_lock(&pm->lock); @@ -406,13 +472,18 @@ int pm_update_grace_period(struct packet_manager *pm, uint32_t grace_period) goto out; } - retval = pm->pmf->set_grace_period(pm, buffer, grace_period); - if (!retval) + retval = pm->pmf->config_dequeue_wait_counts(pm, buffer, + cmd, value); + if (!retval) { retval = kq_submit_packet(pm->priv_queue); - else + + /* If default value is modified, cache that in dqm->wait_times */ + if (!retval && cmd == KFD_DEQUEUE_WAIT_INIT) + update_dqm_wait_times(pm->dqm); + } else { kq_rollback_packet(pm->priv_queue); + } } - out: mutex_unlock(&pm->lock); return retval; |