Fix T41471 Cycles Bake: Setting small tile size results in wrong bake with stripes rather than the expected noise pattern

This problem was introduced in 983cbafd1877f8dbaae60b064a14e27b5b640f18
Basically the issue is that we were not getting a unique index in the
baking routine for the RNG (random number generator).

Reviewers: sergey

Differential Revision: https://developer.blender.org/D749

3 files changed, 16 insertions, 6 deletions

diff --git a/intern/cycles/device/device_cpu.cpp b/intern/cycles/device/device_cpu.cpp
index 4fdeef6bdcb..fd5ae1d7828 100644
--- a/intern/cycles/device/device_cpu.cpp
+++ b/intern/cycles/device/device_cpu.cpp
@@ -435,7 +435,8 @@ public:
 		if(system_cpu_support_avx2()) {
 			for(int sample = 0; sample < task.num_samples; sample++) {
 				for(int x = task.shader_x; x < task.shader_x + task.shader_w; x++)
-					kernel_cpu_avx2_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output, task.shader_eval_type, x, sample);
+					kernel_cpu_avx2_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output,
+					    task.shader_eval_type, x, task.offset, sample);
 
 				if(task.get_cancel() || task_pool.canceled())
 					break;
@@ -449,7 +450,8 @@ public:
 		if(system_cpu_support_avx()) {
 			for(int sample = 0; sample < task.num_samples; sample++) {
 				for(int x = task.shader_x; x < task.shader_x + task.shader_w; x++)
-					kernel_cpu_avx_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output, task.shader_eval_type, x, sample);
+					kernel_cpu_avx_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output,
+					    task.shader_eval_type, x, task.offset, sample);
 
 				if(task.get_cancel() || task_pool.canceled())
 					break;
@@ -463,7 +465,8 @@ public:
 		if(system_cpu_support_sse41()) {
 			for(int sample = 0; sample < task.num_samples; sample++) {
 				for(int x = task.shader_x; x < task.shader_x + task.shader_w; x++)
-					kernel_cpu_sse41_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output, task.shader_eval_type, x, sample);
+					kernel_cpu_sse41_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output,
+					    task.shader_eval_type, x, task.offset, sample);
 
 				if(task.get_cancel() || task_pool.canceled())
 					break;
@@ -477,7 +480,8 @@ public:
 		if(system_cpu_support_sse3()) {
 			for(int sample = 0; sample < task.num_samples; sample++) {
 				for(int x = task.shader_x; x < task.shader_x + task.shader_w; x++)
-					kernel_cpu_sse3_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output, task.shader_eval_type, x, sample);
+					kernel_cpu_sse3_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output,
+					    task.shader_eval_type, x, task.offset, sample);
 
 				if(task.get_cancel() || task_pool.canceled())
 					break;
@@ -491,7 +495,8 @@ public:
 		if(system_cpu_support_sse2()) {
 			for(int sample = 0; sample < task.num_samples; sample++) {
 				for(int x = task.shader_x; x < task.shader_x + task.shader_w; x++)
-					kernel_cpu_sse2_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output, task.shader_eval_type, x, sample);
+					kernel_cpu_sse2_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output,
+					    task.shader_eval_type, x, task.offset, sample);
 
 				if(task.get_cancel() || task_pool.canceled())
 					break;
@@ -504,7 +509,8 @@ public:
 		{
 			for(int sample = 0; sample < task.num_samples; sample++) {
 				for(int x = task.shader_x; x < task.shader_x + task.shader_w; x++)
-					kernel_cpu_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output, task.shader_eval_type, x, sample);
+					kernel_cpu_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output,
+					    task.shader_eval_type, x, task.offset, sample);
 
 				if(task.get_cancel() || task_pool.canceled())
 					break;
diff --git a/intern/cycles/device/device_cuda.cpp b/intern/cycles/device/device_cuda.cpp
index 6629069c6c6..9e3d703f5d9 100644
--- a/intern/cycles/device/device_cuda.cpp
+++ b/intern/cycles/device/device_cuda.cpp
@@ -676,6 +676,7 @@ public:
 		const int shader_chunk_size = 65536;
 		const int start = task.shader_x;
 		const int end = task.shader_x + task.shader_w;
+		int offset = task.offset;
 
 		bool canceled = false;
 		for(int sample = 0; sample < task.num_samples && !canceled; sample++) {
@@ -688,6 +689,7 @@ public:
 								 &task.shader_eval_type,
 								 &shader_x,
 								 &shader_w,
+								 &offset,
 								 &sample};
 
 				/* launch kernel */
diff --git a/intern/cycles/device/device_opencl.cpp b/intern/cycles/device/device_opencl.cpp
index 077ff9df51e..82419cd62b1 100644
--- a/intern/cycles/device/device_opencl.cpp
+++ b/intern/cycles/device/device_opencl.cpp
@@ -1004,6 +1004,7 @@ public:
 		cl_int d_shader_eval_type = task.shader_eval_type;
 		cl_int d_shader_x = task.shader_x;
 		cl_int d_shader_w = task.shader_w;
+		cl_int d_offset = task.offset;
 
 		/* sample arguments */
 		cl_uint narg = 0;
@@ -1033,6 +1034,7 @@ public:
 			opencl_assert(clSetKernelArg(kernel, narg++, sizeof(d_shader_eval_type), (void*)&d_shader_eval_type));
 			opencl_assert(clSetKernelArg(kernel, narg++, sizeof(d_shader_x), (void*)&d_shader_x));
 			opencl_assert(clSetKernelArg(kernel, narg++, sizeof(d_shader_w), (void*)&d_shader_w));
+			opencl_assert(clSetKernelArg(kernel, narg++, sizeof(d_offset), (void*)&d_offset));
 			opencl_assert(clSetKernelArg(kernel, narg++, sizeof(d_sample), (void*)&d_sample));
 
 			enqueue_kernel(kernel, task.shader_w, 1);