From 9ac1432a5714f2c946d005dcdaa90dc5f738a6d8 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 19 Jul 2017 04:02:39 +0100 Subject: [PATCH] radv: port to new libdrm API. This bumps the libdrm requirement for amdgpu to the 2.4.82. Reviewed-by: Bas Nieuwenhuizen Signed-off-by: Dave Airlie --- configure.ac | 2 +- src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c | 121 +++++++++++++----- 2 files changed, 93 insertions(+), 30 deletions(-) diff --git a/configure.ac b/configure.ac index 38cd1271ecf..2689fc55e85 100644 --- a/configure.ac +++ b/configure.ac @@ -74,7 +74,7 @@ AC_SUBST([OPENCL_VERSION]) # in the first entry. LIBDRM_REQUIRED=2.4.75 LIBDRM_RADEON_REQUIRED=2.4.71 -LIBDRM_AMDGPU_REQUIRED=2.4.81 +LIBDRM_AMDGPU_REQUIRED=2.4.82 LIBDRM_INTEL_REQUIRED=2.4.75 LIBDRM_NVVIEUX_REQUIRED=2.4.66 LIBDRM_NOUVEAU_REQUIRED=2.4.66 diff --git a/src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c b/src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c index 91212d2c755..93243dfd135 100644 --- a/src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c +++ b/src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c @@ -96,10 +96,6 @@ static int ring_to_hw_ip(enum ring_type ring) } } -static void radv_amdgpu_wait_sems(struct radv_amdgpu_ctx *ctx, - uint32_t ip_type, - uint32_t ring, - struct radv_amdgpu_sem_info *sem_info); static int radv_amdgpu_signal_sems(struct radv_amdgpu_ctx *ctx, uint32_t ip_type, uint32_t ring, @@ -950,8 +946,6 @@ static int radv_amdgpu_winsys_cs_submit(struct radeon_winsys_ctx *_ctx, sem_info.signal_sems = signal_sem; sem_info.signal_sem_count = signal_sem_count; - radv_amdgpu_wait_sems(ctx, cs->hw_ip, queue_idx, &sem_info); - if (!cs->ws->use_ib_bos) { ret = radv_amdgpu_winsys_cs_submit_sysmem(_ctx, queue_idx, &sem_info, cs_array, cs_count, initial_preamble_cs, continue_preamble_cs, _fence); @@ -1062,31 +1056,17 @@ static bool radv_amdgpu_ctx_wait_idle(struct radeon_winsys_ctx *rwctx, static struct radeon_winsys_sem *radv_amdgpu_create_sem(struct radeon_winsys *_ws) { - int ret; - amdgpu_semaphore_handle sem; - - ret = amdgpu_cs_create_semaphore(&sem); - if (ret) + struct amdgpu_cs_fence *sem = CALLOC_STRUCT(amdgpu_cs_fence); + if (!sem) return NULL; + return (struct radeon_winsys_sem *)sem; } static void radv_amdgpu_destroy_sem(struct radeon_winsys_sem *_sem) { - amdgpu_semaphore_handle sem = (amdgpu_semaphore_handle)_sem; - amdgpu_cs_destroy_semaphore(sem); -} - -static void radv_amdgpu_wait_sems(struct radv_amdgpu_ctx *ctx, - uint32_t ip_type, - uint32_t ring, - struct radv_amdgpu_sem_info *sem_info) -{ - for (unsigned i = 0; i < sem_info->wait_sem_count; i++) { - amdgpu_semaphore_handle sem = (amdgpu_semaphore_handle)sem_info->wait_sems[i]; - amdgpu_cs_wait_semaphore(ctx->ctx, ip_type, 0, ring, - sem); - } + struct amdgpu_cs_fence *sem = (struct amdgpu_cs_fence *)_sem; + FREE(sem); } static int radv_amdgpu_signal_sems(struct radv_amdgpu_ctx *ctx, @@ -1095,9 +1075,12 @@ static int radv_amdgpu_signal_sems(struct radv_amdgpu_ctx *ctx, struct radv_amdgpu_sem_info *sem_info) { for (unsigned i = 0; i < sem_info->signal_sem_count; i++) { - amdgpu_semaphore_handle sem = (amdgpu_semaphore_handle)sem_info->signal_sems[i]; - amdgpu_cs_signal_semaphore(ctx->ctx, ip_type, 0, ring, - sem); + struct amdgpu_cs_fence *sem = (struct amdgpu_cs_fence *)sem_info->signal_sems[i]; + + if (sem->context) + return -EINVAL; + + *sem = ctx->last_submission[ip_type][ring].fence; } return 0; } @@ -1106,7 +1089,87 @@ static int radv_amdgpu_cs_submit(struct radv_amdgpu_ctx *ctx, struct amdgpu_cs_request *request, struct radv_amdgpu_sem_info *sem_info) { - return amdgpu_cs_submit(ctx->ctx, 0, request, 1); + int r; + int num_chunks; + int size; + bool user_fence; + struct drm_amdgpu_cs_chunk *chunks; + struct drm_amdgpu_cs_chunk_data *chunk_data; + struct drm_amdgpu_cs_chunk_dep *sem_dependencies = NULL; + int i; + struct amdgpu_cs_fence *sem; + user_fence = (request->fence_info.handle != NULL); + size = request->number_of_ibs + (user_fence ? 2 : 1) + 1; + + chunks = alloca(sizeof(struct drm_amdgpu_cs_chunk) * size); + + size = request->number_of_ibs + (user_fence ? 1 : 0); + + chunk_data = alloca(sizeof(struct drm_amdgpu_cs_chunk_data) * size); + + num_chunks = request->number_of_ibs; + for (i = 0; i < request->number_of_ibs; i++) { + struct amdgpu_cs_ib_info *ib; + chunks[i].chunk_id = AMDGPU_CHUNK_ID_IB; + chunks[i].length_dw = sizeof(struct drm_amdgpu_cs_chunk_ib) / 4; + chunks[i].chunk_data = (uint64_t)(uintptr_t)&chunk_data[i]; + + ib = &request->ibs[i]; + + chunk_data[i].ib_data._pad = 0; + chunk_data[i].ib_data.va_start = ib->ib_mc_address; + chunk_data[i].ib_data.ib_bytes = ib->size * 4; + chunk_data[i].ib_data.ip_type = request->ip_type; + chunk_data[i].ib_data.ip_instance = request->ip_instance; + chunk_data[i].ib_data.ring = request->ring; + chunk_data[i].ib_data.flags = ib->flags; + } + + if (user_fence) { + i = num_chunks++; + + chunks[i].chunk_id = AMDGPU_CHUNK_ID_FENCE; + chunks[i].length_dw = sizeof(struct drm_amdgpu_cs_chunk_fence) / 4; + chunks[i].chunk_data = (uint64_t)(uintptr_t)&chunk_data[i]; + + amdgpu_cs_chunk_fence_info_to_data(&request->fence_info, + &chunk_data[i]); + } + + if (sem_info->wait_sem_count) { + sem_dependencies = malloc(sizeof(struct drm_amdgpu_cs_chunk_dep) * sem_info->wait_sem_count); + if (!sem_dependencies) { + r = -ENOMEM; + goto error_out; + } + int sem_count = 0; + for (unsigned j = 0; j < sem_info->wait_sem_count; j++) { + sem = (struct amdgpu_cs_fence *)sem_info->wait_sems[j]; + if (!sem->context) + continue; + struct drm_amdgpu_cs_chunk_dep *dep = &sem_dependencies[sem_count++]; + + amdgpu_cs_chunk_fence_to_dep(sem, dep); + } + i = num_chunks++; + + /* dependencies chunk */ + chunks[i].chunk_id = AMDGPU_CHUNK_ID_DEPENDENCIES; + chunks[i].length_dw = sizeof(struct drm_amdgpu_cs_chunk_dep) / 4 * sem_count; + chunks[i].chunk_data = (uint64_t)(uintptr_t)sem_dependencies; + + sem_info->wait_sem_count = 0; + } + + r = amdgpu_cs_submit_raw(ctx->ws->dev, + ctx->ctx, + request->resources, + num_chunks, + chunks, + &request->seq_no); +error_out: + free(sem_dependencies); + return r; } void radv_amdgpu_cs_init_functions(struct radv_amdgpu_winsys *ws) -- 2.30.2