bool has_exec_fence;
bool has_syncobj;
bool has_syncobj_wait;
+ bool has_syncobj_wait_available;
bool has_context_priority;
bool has_context_isolation;
+ bool has_thread_submit;
bool has_mem_available;
bool has_mmap_offset;
uint64_t gtt_size;
bool has_bindless_images;
/** True if we can use bindless access for samplers */
bool has_bindless_samplers;
+ /** True if we can use timeline semaphores through execbuf */
+ bool has_exec_timeline;
/** True if we can read the GPU timestamp register
*
uint32_t fence_count;
uint32_t fence_array_length;
struct drm_i915_gem_exec_fence * fences;
+ uint64_t * fence_values;
uint32_t temporary_semaphore_count;
uint32_t temporary_semaphore_array_length;
uint32_t sync_fd_semaphore_array_length;
/* Allocated only with non shareable timelines. */
- struct anv_timeline ** wait_timelines;
+ union {
+ struct anv_timeline ** wait_timelines;
+ uint32_t * wait_timeline_syncobjs;
+ };
uint32_t wait_timeline_count;
uint32_t wait_timeline_array_length;
uint64_t * wait_timeline_values;
struct anv_queue {
struct vk_object_base base;
- struct anv_device * device;
+ struct anv_device * device;
- /*
- * A list of struct anv_queue_submit to be submitted to i915.
- */
- struct list_head queued_submits;
+ VkDeviceQueueCreateFlags flags;
+
+ /* Set once from the device api calls. */
+ bool lost_signaled;
+
+ /* Only set once atomically by the queue */
+ int lost;
+ int error_line;
+ const char * error_file;
+ char error_msg[80];
+
+ /*
+ * This mutext protects the variables below.
+ */
+ pthread_mutex_t mutex;
+
+ pthread_t thread;
+ pthread_cond_t cond;
- VkDeviceQueueCreateFlags flags;
+ /*
+ * A list of struct anv_queue_submit to be submitted to i915.
+ */
+ struct list_head queued_submits;
+
+ /* Set to true to stop the submission thread */
+ bool quit;
};
struct anv_pipeline_cache {
int fd;
bool can_chain_batches;
bool robust_buffer_access;
+ bool has_thread_submit;
struct anv_device_extension_table enabled_extensions;
struct anv_device_dispatch_table dispatch;
pthread_mutex_t mutex;
pthread_cond_t queue_submit;
int _lost;
+ int lost_reported;
struct gen_batch_decode_ctx decoder_ctx;
/*
uint64_t perf_metric; /* 0 if unset */
struct gen_aux_map_context *aux_map_ctx;
+
+ struct gen_debug_block_frame *debug_frame_desc;
};
static inline struct anv_instance *
void anv_device_init_blorp(struct anv_device *device);
void anv_device_finish_blorp(struct anv_device *device);
-void _anv_device_set_all_queue_lost(struct anv_device *device);
+void _anv_device_report_lost(struct anv_device *device);
VkResult _anv_device_set_lost(struct anv_device *device,
const char *file, int line,
const char *msg, ...)
#define anv_device_set_lost(dev, ...) \
_anv_device_set_lost(dev, __FILE__, __LINE__, __VA_ARGS__)
#define anv_queue_set_lost(queue, ...) \
- _anv_queue_set_lost(queue, __FILE__, __LINE__, __VA_ARGS__)
+ (queue)->device->has_thread_submit ? \
+ _anv_queue_set_lost(queue, __FILE__, __LINE__, __VA_ARGS__) : \
+ _anv_device_set_lost(queue->device, __FILE__, __LINE__, __VA_ARGS__)
static inline bool
anv_device_is_lost(struct anv_device *device)
{
- return unlikely(p_atomic_read(&device->_lost));
+ int lost = p_atomic_read(&device->_lost);
+ if (unlikely(lost && !device->lost_reported))
+ _anv_device_report_lost(device);
+ return lost;
}
VkResult anv_device_query_status(struct anv_device *device);
int anv_gem_get_context_param(int fd, int context, uint32_t param,
uint64_t *value);
int anv_gem_get_param(int fd, uint32_t param);
+uint64_t anv_gem_get_drm_cap(int fd, uint32_t capability);
int anv_gem_get_tiling(struct anv_device *device, uint32_t gem_handle);
bool anv_gem_get_bit6_swizzle(int fd, uint32_t tiling);
int anv_gem_gpu_get_reset_stats(struct anv_device *device,
void anv_gem_syncobj_reset(struct anv_device *device, uint32_t handle);
bool anv_gem_supports_syncobj_wait(int fd);
int anv_gem_syncobj_wait(struct anv_device *device,
- uint32_t *handles, uint32_t num_handles,
+ const uint32_t *handles, uint32_t num_handles,
int64_t abs_timeout_ns, bool wait_all);
+int anv_gem_syncobj_timeline_wait(struct anv_device *device,
+ const uint32_t *handles, const uint64_t *points,
+ uint32_t num_items, int64_t abs_timeout_ns,
+ bool wait_all, bool wait_materialize);
+int anv_gem_syncobj_timeline_signal(struct anv_device *device,
+ const uint32_t *handles, const uint64_t *points,
+ uint32_t num_items);
+int anv_gem_syncobj_timeline_query(struct anv_device *device,
+ const uint32_t *handles, uint64_t *points,
+ uint32_t num_items);
uint64_t anv_vma_alloc(struct anv_device *device,
uint64_t size, uint64_t align,
/* Number of dynamic offsets used by this descriptor set */
uint16_t dynamic_offset_count;
- /* For each shader stage, which offsets apply to that stage */
- uint16_t stage_dynamic_offsets[MESA_SHADER_STAGES];
+ /* For each dynamic buffer, which VkShaderStageFlagBits stages are using
+ * this buffer
+ */
+ VkShaderStageFlags dynamic_offset_stages[MAX_DYNAMIC_BUFFERS];
/* Size of the descriptor buffer for this descriptor set */
uint32_t descriptor_buffer_size;
struct anv_vertex_binding {
struct anv_buffer * buffer;
VkDeviceSize offset;
+ VkDeviceSize stride;
+ VkDeviceSize size;
};
struct anv_xfb_binding {
/** Dynamic offsets for dynamic UBOs and SSBOs */
uint32_t dynamic_offsets[MAX_DYNAMIC_BUFFERS];
- uint64_t push_reg_mask;
+ /* Robust access pushed registers. */
+ uint64_t push_reg_mask[MESA_SHADER_STAGES];
/** Pad out to a multiple of 32 bytes */
uint32_t pad[2];
uint32_t back;
} stencil_reference;
+ struct {
+ struct {
+ VkStencilOp fail_op;
+ VkStencilOp pass_op;
+ VkStencilOp depth_fail_op;
+ VkCompareOp compare_op;
+ } front;
+ struct {
+ VkStencilOp fail_op;
+ VkStencilOp pass_op;
+ VkStencilOp depth_fail_op;
+ VkCompareOp compare_op;
+ } back;
+ } stencil_op;
+
struct {
uint32_t factor;
uint16_t pattern;
} line_stipple;
+
+ VkCullModeFlags cull_mode;
+ VkFrontFace front_face;
+ VkPrimitiveTopology primitive_topology;
+ bool depth_test_enable;
+ bool depth_write_enable;
+ VkCompareOp depth_compare_op;
+ bool depth_bounds_test_enable;
+ bool stencil_test_enable;
+ bool dyn_vbo_stride;
+ bool dyn_vbo_size;
};
extern const struct anv_dynamic_state default_dynamic_state;
struct anv_cmd_pipeline_state {
struct anv_descriptor_set *descriptors[MAX_SETS];
struct anv_push_descriptor_set *push_descriptors[MAX_SETS];
+
+ struct anv_push_constants push_constants;
+
+ /* Push constant state allocated when flushing push constants. */
+ struct anv_state push_constants_state;
};
/** State tracking for graphics pipeline
struct anv_vb_cache_range vb_bound_ranges[33];
struct anv_vb_cache_range vb_dirty_ranges[33];
+ VkShaderStageFlags push_constant_stages;
+
struct anv_dynamic_state dynamic;
+ uint32_t primitive_topology;
+
struct {
struct anv_buffer *index_buffer;
uint32_t index_type; /**< 3DSTATE_INDEX_BUFFER.IndexFormat */
struct anv_vertex_binding vertex_bindings[MAX_VBS];
bool xfb_enabled;
struct anv_xfb_binding xfb_bindings[MAX_XFB_BUFFERS];
- VkShaderStageFlags push_constant_stages;
- struct anv_push_constants push_constants[MESA_SHADER_STAGES];
struct anv_state binding_tables[MESA_SHADER_STAGES];
struct anv_state samplers[MESA_SHADER_STAGES];
void anv_cmd_buffer_emit_state_base_address(struct anv_cmd_buffer *cmd_buffer);
struct anv_state
-anv_cmd_buffer_push_constants(struct anv_cmd_buffer *cmd_buffer,
- gl_shader_stage stage);
+anv_cmd_buffer_gfx_push_constants(struct anv_cmd_buffer *cmd_buffer);
struct anv_state
anv_cmd_buffer_cs_push_constants(struct anv_cmd_buffer *cmd_buffer);
ANV_SEMAPHORE_TYPE_SYNC_FILE,
ANV_SEMAPHORE_TYPE_DRM_SYNCOBJ,
ANV_SEMAPHORE_TYPE_TIMELINE,
+ ANV_SEMAPHORE_TYPE_DRM_SYNCOBJ_TIMELINE,
};
struct anv_timeline_point {
struct {
uint32_t sf[7];
uint32_t depth_stencil_state[3];
+ uint32_t clip[4];
} gen7;
struct {
ANV_DUMP_FRAMEBUFFERS_BIT = 0x1,
};
+#ifdef DEBUG
+PUBLIC
+#endif
void anv_dump_start(struct anv_device *device, enum anv_dump_action actions);
+#ifdef DEBUG
+PUBLIC
+#endif
void anv_dump_finish(void);
void anv_dump_add_attachments(struct anv_cmd_buffer *cmd_buffer);
return subpass_id;
}
+struct anv_performance_configuration_intel {
+ struct vk_object_base base;
+
+ struct gen_perf_registers *register_config;
+
+ uint64_t config_id;
+};
+
struct gen_perf_config *anv_get_perf(const struct gen_device_info *devinfo, int fd);
void anv_device_perf_init(struct anv_device *device);
void anv_perf_write_pass_results(struct gen_perf_config *perf,
VK_DEFINE_NONDISP_HANDLE_CASTS(anv_ycbcr_conversion, base,
VkSamplerYcbcrConversion,
VK_OBJECT_TYPE_SAMPLER_YCBCR_CONVERSION)
+VK_DEFINE_NONDISP_HANDLE_CASTS(anv_performance_configuration_intel, base,
+ VkPerformanceConfigurationINTEL,
+ VK_OBJECT_TYPE_PERFORMANCE_CONFIGURATION_INTEL)
/* Gen-specific function declarations */
#ifdef genX
projects / mesa.git / blobdiff