diff options
Diffstat (limited to 'intern/cycles/integrator/path_trace_work_gpu.cpp')
| -rw-r--r-- | intern/cycles/integrator/path_trace_work_gpu.cpp | 74 |
1 files changed, 54 insertions, 20 deletions
diff --git a/intern/cycles/integrator/path_trace_work_gpu.cpp b/intern/cycles/integrator/path_trace_work_gpu.cpp index 8306460d607..fa313f6460a 100644 --- a/intern/cycles/integrator/path_trace_work_gpu.cpp +++ b/intern/cycles/integrator/path_trace_work_gpu.cpp @@ -65,6 +65,8 @@ PathTraceWorkGPU::PathTraceWorkGPU(Device *device, integrator_shader_sort_counter_(device, "integrator_shader_sort_counter", MEM_READ_WRITE), integrator_shader_raytrace_sort_counter_( device, "integrator_shader_raytrace_sort_counter", MEM_READ_WRITE), + integrator_shader_mnee_sort_counter_( + device, "integrator_shader_mnee_sort_counter", MEM_READ_WRITE), integrator_shader_sort_prefix_sum_( device, "integrator_shader_sort_prefix_sum", MEM_READ_WRITE), integrator_next_main_path_index_(device, "integrator_next_main_path_index", MEM_READ_WRITE), @@ -150,7 +152,7 @@ void PathTraceWorkGPU::alloc_integrator_soa() total_soa_size += soa_memory->memory_size(); } - VLOG(3) << "GPU SoA state size: " << string_human_readable_size(total_soa_size); + VLOG_DEVICE_STATS << "GPU SoA state size: " << string_human_readable_size(total_soa_size); } } @@ -179,22 +181,45 @@ void PathTraceWorkGPU::alloc_integrator_queue() void PathTraceWorkGPU::alloc_integrator_sorting() { + /* Compute sort partitions, to balance between memory locality and coherence. + * Sort partitioning becomes less effective when more shaders are in the wavefront. In lieu of a + * more sophisticated heuristic we simply disable sort partitioning if the shader count is high. + */ + num_sort_partitions_ = 1; + if (device_scene_->data.max_shaders < 300) { + const int num_elements = queue_->num_sort_partition_elements(); + if (num_elements) { + num_sort_partitions_ = max(max_num_paths_ / num_elements, 1); + } + } + + integrator_state_gpu_.sort_partition_divisor = (int)divide_up(max_num_paths_, + num_sort_partitions_); + /* Allocate arrays for shader sorting. */ - const int max_shaders = device_scene_->data.max_shaders; - if (integrator_shader_sort_counter_.size() < max_shaders) { - integrator_shader_sort_counter_.alloc(max_shaders); + const int sort_buckets = device_scene_->data.max_shaders * num_sort_partitions_; + if (integrator_shader_sort_counter_.size() < sort_buckets) { + integrator_shader_sort_counter_.alloc(sort_buckets); integrator_shader_sort_counter_.zero_to_device(); + integrator_state_gpu_.sort_key_counter[DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE] = + (int *)integrator_shader_sort_counter_.device_pointer; - integrator_shader_raytrace_sort_counter_.alloc(max_shaders); - integrator_shader_raytrace_sort_counter_.zero_to_device(); + if (device_scene_->data.kernel_features & KERNEL_FEATURE_NODE_RAYTRACE) { + integrator_shader_raytrace_sort_counter_.alloc(sort_buckets); + integrator_shader_raytrace_sort_counter_.zero_to_device(); + integrator_state_gpu_.sort_key_counter[DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE] = + (int *)integrator_shader_raytrace_sort_counter_.device_pointer; + } - integrator_shader_sort_prefix_sum_.alloc(max_shaders); - integrator_shader_sort_prefix_sum_.zero_to_device(); + if (device_scene_->data.kernel_features & KERNEL_FEATURE_MNEE) { + integrator_shader_mnee_sort_counter_.alloc(sort_buckets); + integrator_shader_mnee_sort_counter_.zero_to_device(); + integrator_state_gpu_.sort_key_counter[DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_MNEE] = + (int *)integrator_shader_mnee_sort_counter_.device_pointer; + } - integrator_state_gpu_.sort_key_counter[DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE] = - (int *)integrator_shader_sort_counter_.device_pointer; - integrator_state_gpu_.sort_key_counter[DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE] = - (int *)integrator_shader_raytrace_sort_counter_.device_pointer; + integrator_shader_sort_prefix_sum_.alloc(sort_buckets); + integrator_shader_sort_prefix_sum_.zero_to_device(); } } @@ -232,7 +257,7 @@ void PathTraceWorkGPU::init_execution() /* Copy to device side struct in constant memory. */ device_->const_copy_to( - "__integrator_state", &integrator_state_gpu_, sizeof(integrator_state_gpu_)); + "integrator_state", &integrator_state_gpu_, sizeof(integrator_state_gpu_)); } void PathTraceWorkGPU::render_samples(RenderStatistics &statistics, @@ -326,7 +351,12 @@ void PathTraceWorkGPU::enqueue_reset() queue_->enqueue(DEVICE_KERNEL_INTEGRATOR_RESET, max_num_paths_, args); queue_->zero_to_device(integrator_queue_counter_); queue_->zero_to_device(integrator_shader_sort_counter_); - queue_->zero_to_device(integrator_shader_raytrace_sort_counter_); + if (device_scene_->data.kernel_features & KERNEL_FEATURE_NODE_RAYTRACE) { + queue_->zero_to_device(integrator_shader_raytrace_sort_counter_); + } + if (device_scene_->data.kernel_features & KERNEL_FEATURE_MNEE) { + queue_->zero_to_device(integrator_shader_mnee_sort_counter_); + } /* Tiles enqueue need to know number of active paths, which is based on this counter. Zero the * counter on the host side because `zero_to_device()` is not doing it. */ @@ -450,6 +480,7 @@ void PathTraceWorkGPU::enqueue_path_iteration(DeviceKernel kernel, const int num case DEVICE_KERNEL_INTEGRATOR_SHADE_SHADOW: case DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE: case DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE: + case DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_MNEE: case DEVICE_KERNEL_INTEGRATOR_SHADE_VOLUME: { /* Shading kernels with integrator state and render buffer. */ DeviceKernelArguments args(&d_path_index, &buffers_->buffer.device_pointer, &work_size); @@ -477,9 +508,9 @@ void PathTraceWorkGPU::compute_sorted_queued_paths(DeviceKernel kernel, /* Compute prefix sum of number of active paths with each shader. */ { const int work_size = 1; - int max_shaders = device_scene_->data.max_shaders; + int sort_buckets = device_scene_->data.max_shaders * num_sort_partitions_; - DeviceKernelArguments args(&d_counter, &d_prefix_sum, &max_shaders); + DeviceKernelArguments args(&d_counter, &d_prefix_sum, &sort_buckets); queue_->enqueue(DEVICE_KERNEL_PREFIX_SUM, work_size, args); } @@ -811,10 +842,10 @@ bool PathTraceWorkGPU::should_use_graphics_interop() interop_use_ = device->should_use_graphics_interop(); if (interop_use_) { - VLOG(2) << "Using graphics interop GPU display update."; + VLOG_INFO << "Using graphics interop GPU display update."; } else { - VLOG(2) << "Using naive GPU display update."; + VLOG_INFO << "Using naive GPU display update."; } interop_use_checked_ = true; @@ -1080,13 +1111,15 @@ int PathTraceWorkGPU::shadow_catcher_count_possible_splits() bool PathTraceWorkGPU::kernel_uses_sorting(DeviceKernel kernel) { return (kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE || - kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE); + kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE || + kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_MNEE); } bool PathTraceWorkGPU::kernel_creates_shadow_paths(DeviceKernel kernel) { return (kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE || kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE || + kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_MNEE || kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_VOLUME); } @@ -1094,7 +1127,8 @@ bool PathTraceWorkGPU::kernel_creates_ao_paths(DeviceKernel kernel) { return (device_scene_->data.kernel_features & KERNEL_FEATURE_AO) && (kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE || - kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE); + kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE || + kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_MNEE); } bool PathTraceWorkGPU::kernel_is_shadow_path(DeviceKernel kernel) |