Cycles: Optionally output luminance from the shader evaluation kernel

This makes it possible to move some parts of evaluation from host to the device
and hopefully reduce memory usage by avoid having full RGBA buffer on the host.

Reviewers: juicyfruit, lukasstockner97, brecht

Reviewed By: lukasstockner97, brecht

Differential Revision: https://developer.blender.org/D1702

1 files changed, 13 insertions, 7 deletions

diff --git a/intern/cycles/device/device_cuda.cpp b/intern/cycles/device/device_cuda.cpp
index d9d6fd77ecb..5c9ca3454c6 100644
--- a/intern/cycles/device/device_cuda.cpp
+++ b/intern/cycles/device/device_cuda.cpp
@@ -726,6 +726,7 @@ public:
 		CUfunction cuShader;
 		CUdeviceptr d_input = cuda_device_ptr(task.shader_input);
 		CUdeviceptr d_output = cuda_device_ptr(task.shader_output);
+		CUdeviceptr d_output_luma = cuda_device_ptr(task.shader_output_luma);
 
 		/* get kernel function */
 		if(task.shader_eval_type >= SHADER_EVAL_BAKE) {
@@ -747,13 +748,18 @@ public:
 				int shader_w = min(shader_chunk_size, end - shader_x);
 
 				/* pass in parameters */
-				void *args[] = {&d_input,
-								 &d_output,
-								 &task.shader_eval_type,
-								 &shader_x,
-								 &shader_w,
-								 &offset,
-								 &sample};
+				void *args[8];
+				int arg = 0;
+				args[arg++] = &d_input;
+				args[arg++] = &d_output;
+				if(task.shader_eval_type < SHADER_EVAL_BAKE) {
+					args[arg++] = &d_output_luma;
+				}
+				args[arg++] = &task.shader_eval_type;
+				args[arg++] = &shader_x;
+				args[arg++] = &shader_w;
+				args[arg++] = &offset;
+				args[arg++] = &sample;
 
 				/* launch kernel */
 				int threads_per_block;