diff options
| author | Sergey Sharybin <sergey.vfx@gmail.com> | 2017-05-02 16:29:00 +0300 |
|---|---|---|
| committer | Sergey Sharybin <sergey.vfx@gmail.com> | 2017-05-02 16:29:00 +0300 |
| commit | 7f833c0da8848d424de9d2324412d8ca9d9c4e30 (patch) | |
| tree | 223a60022188132e5f7b64e9c649f80d72bfbcd7 /intern/cycles | |
| parent | ec9ce9e11864146dc6748f2388b56f95346a5053 (diff) | |
| parent | 4384a7cf463eedea83179da80bbe12ff7d55578a (diff) | |
Merge branch 'master' into blender2.8
Diffstat (limited to 'intern/cycles')
| -rw-r--r-- | intern/cycles/device/device_cuda.cpp | 16 |
1 files changed, 12 insertions, 4 deletions
diff --git a/intern/cycles/device/device_cuda.cpp b/intern/cycles/device/device_cuda.cpp index 3532e640eaa..e497ec6b0e1 100644 --- a/intern/cycles/device/device_cuda.cpp +++ b/intern/cycles/device/device_cuda.cpp @@ -119,6 +119,7 @@ public: int cuDevId; int cuDevArchitecture; bool first_error; + CUDASplitKernel *split_kernel; struct PixelMem { GLuint cuPBO; @@ -221,6 +222,8 @@ public: cuDevice = 0; cuContext = 0; + split_kernel = NULL; + need_bindless_mapping = false; /* intialize */ @@ -260,6 +263,8 @@ public: { task_pool.stop(); + delete split_kernel; + if(info.has_bindless_textures) { tex_free(bindless_mapping); } @@ -1357,12 +1362,14 @@ public: requested_features.max_closure = 64; } - CUDASplitKernel split_kernel(this); - split_kernel.load_kernels(requested_features); + if(split_kernel == NULL) { + split_kernel = new CUDASplitKernel(this); + split_kernel->load_kernels(requested_features); + } while(task->acquire_tile(this, tile)) { device_memory void_buffer; - split_kernel.path_trace(task, tile, void_buffer, void_buffer); + split_kernel->path_trace(task, tile, void_buffer, void_buffer); task->release_tile(tile); @@ -1648,7 +1655,8 @@ int2 CUDASplitKernel::split_kernel_global_size(device_memory& kg, device_memory& << string_human_readable_size(free) << ")."; size_t num_elements = max_elements_for_max_buffer_size(kg, data, free / 2); - int2 global_size = make_int2(round_down((int)sqrt(num_elements), 32), (int)sqrt(num_elements)); + size_t side = round_down((int)sqrt(num_elements), 32); + int2 global_size = make_int2(side, round_down(num_elements / side, 16)); VLOG(1) << "Global size: " << global_size << "."; return global_size; } |