diff options
author | Brecht Van Lommel <brecht@blender.org> | 2019-05-10 22:39:58 +0300 |
---|---|---|
committer | Lukas Stockner <lukas.stockner@freenet.de> | 2020-05-15 21:25:24 +0300 |
commit | d9773edaa394f61393f9c8b80275e62f74306097 (patch) | |
tree | 232b771b341e98a5403af16791bdcca133cb1edd /intern/cycles/device/opencl/device_opencl_impl.cpp | |
parent | 3ff8ca60e94db2584ca76e323a54c738e677d5f8 (diff) |
Cycles: code refactor to bake using regular render session and tiles
There should be no user visible change from this, except that tile size
now affects performance. The goal here is to simplify bake denoising in
D3099, letting it reuse more denoising tiles and pass code.
A lot of code is now shared with regular rendering, with the two main
differences being that we read some render result passes from the bake API
when starting to render a tile, and call the bake kernel instead of the
path trace kernel.
With this kind of design where Cycles asks for tiles from the bake API,
it should eventually be easier to reduce memory usage, show tiles as
they are baked, or bake multiple passes at once, though there's still
quite some work needed for that.
Reviewers: #cycles
Subscribers: monio, wmatyjewicz, lukasstockner97, michaelknubben
Differential Revision: https://developer.blender.org/D3108
Diffstat (limited to 'intern/cycles/device/opencl/device_opencl_impl.cpp')
-rw-r--r-- | intern/cycles/device/opencl/device_opencl_impl.cpp | 53 |
1 files changed, 49 insertions, 4 deletions
diff --git a/intern/cycles/device/opencl/device_opencl_impl.cpp b/intern/cycles/device/opencl/device_opencl_impl.cpp index 2766f85d17c..beb3174b111 100644 --- a/intern/cycles/device/opencl/device_opencl_impl.cpp +++ b/intern/cycles/device/opencl/device_opencl_impl.cpp @@ -1367,6 +1367,9 @@ void OpenCLDevice::thread_run(DeviceTask *task) */ clFinish(cqCommandQueue); } + else if (tile.task == RenderTile::BAKE) { + bake(*task, tile); + } else if (tile.task == RenderTile::DENOISE) { tile.sample = tile.start_sample + tile.num_samples; denoise(tile, denoising); @@ -1858,10 +1861,7 @@ void OpenCLDevice::shader(DeviceTask &task) cl_int d_offset = task.offset; OpenCLDevice::OpenCLProgram *program = &background_program; - if (task.shader_eval_type >= SHADER_EVAL_BAKE) { - program = &bake_program; - } - else if (task.shader_eval_type == SHADER_EVAL_DISPLACE) { + if (task.shader_eval_type == SHADER_EVAL_DISPLACE) { program = &displace_program; } program->wait_for_availability(); @@ -1892,6 +1892,51 @@ void OpenCLDevice::shader(DeviceTask &task) } } +void OpenCLDevice::bake(DeviceTask &task, RenderTile &rtile) +{ + scoped_timer timer(&rtile.buffers->render_time); + + /* Cast arguments to cl types. */ + cl_mem d_data = CL_MEM_PTR(const_mem_map["__data"]->device_pointer); + cl_mem d_buffer = CL_MEM_PTR(rtile.buffer); + cl_int d_x = rtile.x; + cl_int d_y = rtile.y; + cl_int d_w = rtile.w; + cl_int d_h = rtile.h; + cl_int d_offset = rtile.offset; + cl_int d_stride = rtile.stride; + + bake_program.wait_for_availability(); + cl_kernel kernel = bake_program(); + + cl_uint start_arg_index = kernel_set_args(kernel, 0, d_data, d_buffer); + + set_kernel_arg_buffers(kernel, &start_arg_index); + + start_arg_index += kernel_set_args( + kernel, start_arg_index, d_x, d_y, d_w, d_h, d_offset, d_stride); + + int start_sample = rtile.start_sample; + int end_sample = rtile.start_sample + rtile.num_samples; + + for (int sample = start_sample; sample < end_sample; sample++) { + if (task.get_cancel()) { + if (task.need_finish_queue == false) + break; + } + + kernel_set_args(kernel, start_arg_index, sample); + + enqueue_kernel(kernel, d_w, d_h); + + rtile.sample = sample + 1; + + task.update_progress(&rtile, rtile.w * rtile.h); + } + + clFinish(cqCommandQueue); +} + string OpenCLDevice::kernel_build_options(const string *debug_src) { string build_options = "-cl-no-signed-zeros -cl-mad-enable "; |