Cycles: Remove ccl_addr_space from RNG passed to functions

Simplifies code quite a bit, making it shorter and easier to extend.
Currently no functional changes for users, but is required for the
upcoming work of shadow catcher support with OpenCL.

9 files changed, 37 insertions, 27 deletions

diff --git a/intern/cycles/kernel/split/kernel_buffer_update.h b/intern/cycles/kernel/split/kernel_buffer_update.h
index f36899b884a..fc264734b7e 100644
--- a/intern/cycles/kernel/split/kernel_buffer_update.h
+++ b/intern/cycles/kernel/split/kernel_buffer_update.h
@@ -87,7 +87,7 @@ ccl_device void kernel_buffer_update(KernelGlobals *kg,
 	ccl_global Ray *ray = &kernel_split_state.ray[ray_index];
 	ccl_global float3 *throughput = &kernel_split_state.throughput[ray_index];
 	ccl_global float *L_transparent = &kernel_split_state.L_transparent[ray_index];
-	ccl_global uint *rng = &kernel_split_state.rng[ray_index];
+	RNG rng = kernel_split_state.rng[ray_index];
 	ccl_global float *buffer = kernel_split_params.buffer;
 
 	unsigned int work_index;
@@ -120,7 +120,7 @@ ccl_device void kernel_buffer_update(KernelGlobals *kg,
 
 		/* accumulate result in output buffer */
 		kernel_write_pass_float4(buffer, sample, L_rad);
-		path_rng_end(kg, rng_state, *rng);
+		path_rng_end(kg, rng_state, rng);
 
 		ASSIGN_RAY_STATE(ray_state, ray_index, RAY_TO_REGENERATE);
 	}
@@ -146,7 +146,7 @@ ccl_device void kernel_buffer_update(KernelGlobals *kg,
 			buffer += (kernel_split_params.offset + pixel_x + pixel_y*stride) * kernel_data.film.pass_stride;
 
 			/* Initialize random numbers and ray. */
-			kernel_path_trace_setup(kg, rng_state, sample, pixel_x, pixel_y, rng, ray);
+			kernel_path_trace_setup(kg, rng_state, sample, pixel_x, pixel_y, &rng, ray);
 
 			if(ray->t != 0.0f) {
 				/* Initialize throughput, L_transparent, Ray, PathState;
@@ -155,7 +155,7 @@ ccl_device void kernel_buffer_update(KernelGlobals *kg,
 				*throughput = make_float3(1.0f, 1.0f, 1.0f);
 				*L_transparent = 0.0f;
 				path_radiance_init(L, kernel_data.film.use_light_pass);
-				path_state_init(kg, &kernel_split_state.sd_DL_shadow[ray_index], state, rng, sample, ray);
+				path_state_init(kg, &kernel_split_state.sd_DL_shadow[ray_index], state, &rng, sample, ray);
 #ifdef __SUBSURFACE__
 				kernel_path_subsurface_init_indirect(&kernel_split_state.ss_rays[ray_index]);
 #endif
@@ -170,12 +170,13 @@ ccl_device void kernel_buffer_update(KernelGlobals *kg,
 				float4 L_rad = make_float4(0.0f, 0.0f, 0.0f, 0.0f);
 				/* Accumulate result in output buffer. */
 				kernel_write_pass_float4(buffer, sample, L_rad);
-				path_rng_end(kg, rng_state, *rng);
+				path_rng_end(kg, rng_state, rng);
 
 				ASSIGN_RAY_STATE(ray_state, ray_index, RAY_TO_REGENERATE);
 			}
 		}
 	}
+	kernel_split_state.rng[ray_index] = rng;
 
 #ifndef __COMPUTE_DEVICE_GPU__
 	}
diff --git a/intern/cycles/kernel/split/kernel_direct_lighting.h b/intern/cycles/kernel/split/kernel_direct_lighting.h
index 3d062cf0e2b..0a0733b9a9f 100644
--- a/intern/cycles/kernel/split/kernel_direct_lighting.h
+++ b/intern/cycles/kernel/split/kernel_direct_lighting.h
@@ -83,11 +83,12 @@ ccl_device void kernel_direct_lighting(KernelGlobals *kg,
 		    (sd->flag & SD_BSDF_HAS_EVAL)))
 		{
 			/* Sample illumination from lights to find path contribution. */
-			ccl_global RNG* rng = &kernel_split_state.rng[ray_index];
-			float light_t = path_state_rng_1D(kg, rng, state, PRNG_LIGHT);
+			RNG rng = kernel_split_state.rng[ray_index];
+			float light_t = path_state_rng_1D(kg, &rng, state, PRNG_LIGHT);
 			float light_u, light_v;
-			path_state_rng_2D(kg, rng, state, PRNG_LIGHT_U, &light_u, &light_v);
-			float terminate = path_state_rng_light_termination(kg, rng, state);
+			path_state_rng_2D(kg, &rng, state, PRNG_LIGHT_U, &light_u, &light_v);
+			float terminate = path_state_rng_light_termination(kg, &rng, state);
+			kernel_split_state.rng[ray_index] = rng;
 
 			LightSample ls;
 			if(light_sample(kg,
diff --git a/intern/cycles/kernel/split/kernel_do_volume.h b/intern/cycles/kernel/split/kernel_do_volume.h
index b1df45d6bb2..47d3c280831 100644
--- a/intern/cycles/kernel/split/kernel_do_volume.h
+++ b/intern/cycles/kernel/split/kernel_do_volume.h
@@ -50,7 +50,7 @@ ccl_device void kernel_do_volume(KernelGlobals *kg)
 
 		ccl_global float3 *throughput = &kernel_split_state.throughput[ray_index];
 		ccl_global Ray *ray = &kernel_split_state.ray[ray_index];
-		ccl_global RNG *rng = &kernel_split_state.rng[ray_index];
+		RNG rng = kernel_split_state.rng[ray_index];
 		ccl_global Intersection *isect = &kernel_split_state.isect[ray_index];
 		ShaderData *sd = &kernel_split_state.sd[ray_index];
 		ShaderData *sd_input = &kernel_split_state.sd_DL_shadow[ray_index];
@@ -69,15 +69,15 @@ ccl_device void kernel_do_volume(KernelGlobals *kg)
 			{
 				/* integrate along volume segment with distance sampling */
 				VolumeIntegrateResult result = kernel_volume_integrate(
-					kg, state, sd, &volume_ray, L, throughput, rng, heterogeneous);
+					kg, state, sd, &volume_ray, L, throughput, &rng, heterogeneous);
 
 #  ifdef __VOLUME_SCATTER__
 				if(result == VOLUME_PATH_SCATTERED) {
 					/* direct lighting */
-					kernel_path_volume_connect_light(kg, rng, sd, sd_input, *throughput, state, L);
+					kernel_path_volume_connect_light(kg, &rng, sd, sd_input, *throughput, state, L);
 
 					/* indirect light bounce */
-					if(kernel_path_volume_bounce(kg, rng, sd, throughput, state, L, ray))
+					if(kernel_path_volume_bounce(kg, &rng, sd, throughput, state, L, ray))
 						ASSIGN_RAY_STATE(kernel_split_state.ray_state, ray_index, RAY_REGENERATED);
 					else
 						ASSIGN_RAY_STATE(kernel_split_state.ray_state, ray_index, RAY_UPDATE_BUFFER);
@@ -85,6 +85,7 @@ ccl_device void kernel_do_volume(KernelGlobals *kg)
 #  endif
 			}
 		}
+		kernel_split_state.rng[ray_index] = rng;
 	}
 
 #endif
diff --git a/intern/cycles/kernel/split/kernel_holdout_emission_blurring_pathtermination_ao.h b/intern/cycles/kernel/split/kernel_holdout_emission_blurring_pathtermination_ao.h
index 1834a791b91..8e1f5472871 100644
--- a/intern/cycles/kernel/split/kernel_holdout_emission_blurring_pathtermination_ao.h
+++ b/intern/cycles/kernel/split/kernel_holdout_emission_blurring_pathtermination_ao.h
@@ -98,7 +98,7 @@ ccl_device void kernel_holdout_emission_blurring_pathtermination_ao(
 	unsigned int tile_y;
 	unsigned int sample;
 
-	ccl_global RNG *rng = 0x0;
+	RNG rng = kernel_split_state.rng[ray_index];
 	ccl_global PathState *state = 0x0;
 	float3 throughput;
 
@@ -110,7 +110,6 @@ ccl_device void kernel_holdout_emission_blurring_pathtermination_ao(
 
 		throughput = kernel_split_state.throughput[ray_index];
 		state = &kernel_split_state.path_state[ray_index];
-		rng = &kernel_split_state.rng[ray_index];
 
 		work_index = kernel_split_state.work_array[ray_index];
 		sample = get_work_sample(kg, work_index, ray_index) + kernel_split_params.start_sample;
@@ -194,7 +193,7 @@ ccl_device void kernel_holdout_emission_blurring_pathtermination_ao(
 
 		if(IS_STATE(ray_state, ray_index, RAY_ACTIVE)) {
 			if(probability != 1.0f) {
-				float terminate = path_state_rng_1D_for_decision(kg, rng, state, PRNG_TERMINATE);
+				float terminate = path_state_rng_1D_for_decision(kg, &rng, state, PRNG_TERMINATE);
 				if(terminate >= probability) {
 					ASSIGN_RAY_STATE(ray_state, ray_index, RAY_UPDATE_BUFFER);
 					enqueue_flag = 1;
@@ -214,7 +213,7 @@ ccl_device void kernel_holdout_emission_blurring_pathtermination_ao(
 		{
 			/* todo: solve correlation */
 			float bsdf_u, bsdf_v;
-			path_state_rng_2D(kg, rng, state, PRNG_BSDF_U, &bsdf_u, &bsdf_v);
+			path_state_rng_2D(kg, &rng, state, PRNG_BSDF_U, &bsdf_u, &bsdf_v);
 
 			float ao_factor = kernel_data.background.ao_factor;
 			float3 ao_N;
@@ -243,6 +242,8 @@ ccl_device void kernel_holdout_emission_blurring_pathtermination_ao(
 		}
 	}
 #endif  /* __AO__ */
+	kernel_split_state.rng[ray_index] = rng;
+
 
 #ifndef __COMPUTE_DEVICE_GPU__
 	}
diff --git a/intern/cycles/kernel/split/kernel_next_iteration_setup.h b/intern/cycles/kernel/split/kernel_next_iteration_setup.h
index 056fb1d8c08..a9957ee1d8a 100644
--- a/intern/cycles/kernel/split/kernel_next_iteration_setup.h
+++ b/intern/cycles/kernel/split/kernel_next_iteration_setup.h
@@ -141,15 +141,16 @@ ccl_device void kernel_next_iteration_setup(KernelGlobals *kg,
 	if(IS_STATE(ray_state, ray_index, RAY_ACTIVE)) {
 		ccl_global float3 *throughput = &kernel_split_state.throughput[ray_index];
 		ccl_global Ray *ray = &kernel_split_state.ray[ray_index];
-		ccl_global RNG *rng = &kernel_split_state.rng[ray_index];
+		RNG rng = kernel_split_state.rng[ray_index];
 		state = &kernel_split_state.path_state[ray_index];
 		L = &kernel_split_state.path_radiance[ray_index];
 
 		/* Compute direct lighting and next bounce. */
-		if(!kernel_path_surface_bounce(kg, rng, &kernel_split_state.sd[ray_index], throughput, state, L, ray)) {
+		if(!kernel_path_surface_bounce(kg, &rng, &kernel_split_state.sd[ray_index], throughput, state, L, ray)) {
 			ASSIGN_RAY_STATE(ray_state, ray_index, RAY_UPDATE_BUFFER);
 			enqueue_flag = 1;
 		}
+		kernel_split_state.rng[ray_index] = rng;
 	}
 
 #ifndef __COMPUTE_DEVICE_GPU__
diff --git a/intern/cycles/kernel/split/kernel_path_init.h b/intern/cycles/kernel/split/kernel_path_init.h
index f879fca5009..a7ecde7c80d 100644
--- a/intern/cycles/kernel/split/kernel_path_init.h
+++ b/intern/cycles/kernel/split/kernel_path_init.h
@@ -60,12 +60,14 @@ ccl_device void kernel_path_init(KernelGlobals *kg) {
 	ccl_global float *buffer = kernel_split_params.buffer;
 	buffer += (kernel_split_params.offset + pixel_x + pixel_y * kernel_split_params.stride) * kernel_data.film.pass_stride;
 
+	RNG rng = kernel_split_state.rng[ray_index];
+
 	/* Initialize random numbers and ray. */
 	kernel_path_trace_setup(kg,
 	                        rng_state,
 	                        my_sample,
 	                        pixel_x, pixel_y,
-	                        &kernel_split_state.rng[ray_index],
+	                        &rng,
 	                        &kernel_split_state.ray[ray_index]);
 
 	if(kernel_split_state.ray[ray_index].t != 0.0f) {
@@ -78,7 +80,7 @@ ccl_device void kernel_path_init(KernelGlobals *kg) {
 		path_state_init(kg,
 		                &kernel_split_state.sd_DL_shadow[ray_index],
 		                &kernel_split_state.path_state[ray_index],
-		                &kernel_split_state.rng[ray_index],
+		                &rng,
 		                my_sample,
 		                &kernel_split_state.ray[ray_index]);
 #ifdef __SUBSURFACE__
@@ -97,6 +99,7 @@ ccl_device void kernel_path_init(KernelGlobals *kg) {
 		path_rng_end(kg, rng_state, kernel_split_state.rng[ray_index]);
 		ASSIGN_RAY_STATE(kernel_split_state.ray_state, ray_index, RAY_TO_REGENERATE);
 	}
+	kernel_split_state.rng[ray_index] = rng;
 }
 
 CCL_NAMESPACE_END
diff --git a/intern/cycles/kernel/split/kernel_scene_intersect.h b/intern/cycles/kernel/split/kernel_scene_intersect.h
index 66f549f59b7..684760eedee 100644
--- a/intern/cycles/kernel/split/kernel_scene_intersect.h
+++ b/intern/cycles/kernel/split/kernel_scene_intersect.h
@@ -72,7 +72,7 @@ ccl_device void kernel_scene_intersect(KernelGlobals *kg)
 		}
 
 		extmax = kernel_data.curve.maximum_width;
-		lcg_state = lcg_state_init(&rng, &state, 0x51633e2d);
+		lcg_state = lcg_state_init(&rng, state.rng_offset, state.sample, 0x51633e2d);
 	}
 
 	if(state.bounce > kernel_data.integrator.ao_bounces) {
diff --git a/intern/cycles/kernel/split/kernel_shader_eval.h b/intern/cycles/kernel/split/kernel_shader_eval.h
index b739f86338c..0f1696e34a0 100644
--- a/intern/cycles/kernel/split/kernel_shader_eval.h
+++ b/intern/cycles/kernel/split/kernel_shader_eval.h
@@ -54,7 +54,7 @@ ccl_device void kernel_shader_eval(KernelGlobals *kg,
 	/* Continue on with shader evaluation. */
 	if((ray_index != QUEUE_EMPTY_SLOT) && IS_STATE(kernel_split_state.ray_state, ray_index, RAY_ACTIVE)) {
 		Intersection isect = kernel_split_state.isect[ray_index];
-		ccl_global uint *rng = &kernel_split_state.rng[ray_index];
+		RNG rng = kernel_split_state.rng[ray_index];
 		ccl_global PathState *state = &kernel_split_state.path_state[ray_index];
 		Ray ray = kernel_split_state.ray[ray_index];
 
@@ -62,8 +62,9 @@ ccl_device void kernel_shader_eval(KernelGlobals *kg,
 		                      &kernel_split_state.sd[ray_index],
 		                      &isect,
 		                      &ray);
-		float rbsdf = path_state_rng_1D_for_decision(kg, rng, state, PRNG_BSDF);
-		shader_eval_surface(kg, &kernel_split_state.sd[ray_index], rng, state, rbsdf, state->flag, SHADER_CONTEXT_MAIN);
+		float rbsdf = path_state_rng_1D_for_decision(kg, &rng, state, PRNG_BSDF);
+		shader_eval_surface(kg, &kernel_split_state.sd[ray_index], &rng, state, rbsdf, state->flag, SHADER_CONTEXT_MAIN);
+		kernel_split_state.rng[ray_index] = rng;
 	}
 }
 
diff --git a/intern/cycles/kernel/split/kernel_subsurface_scatter.h b/intern/cycles/kernel/split/kernel_subsurface_scatter.h
index 4eaa7f56332..0b4d50c70ee 100644
--- a/intern/cycles/kernel/split/kernel_subsurface_scatter.h
+++ b/intern/cycles/kernel/split/kernel_subsurface_scatter.h
@@ -55,7 +55,7 @@ ccl_device void kernel_subsurface_scatter(KernelGlobals *kg,
 	ccl_global char *ray_state = kernel_split_state.ray_state;
 	ccl_global PathState *state = &kernel_split_state.path_state[ray_index];
 	PathRadiance *L = &kernel_split_state.path_radiance[ray_index];
-	ccl_global RNG *rng = &kernel_split_state.rng[ray_index];
+	RNG rng = kernel_split_state.rng[ray_index];
 	ccl_global Ray *ray = &kernel_split_state.ray[ray_index];
 	ccl_global float3 *throughput = &kernel_split_state.throughput[ray_index];
 	ccl_global SubsurfaceIndirectRays *ss_indirect = &kernel_split_state.ss_rays[ray_index];
@@ -69,7 +69,7 @@ ccl_device void kernel_subsurface_scatter(KernelGlobals *kg,
 			                                  emission_sd,
 			                                  L,
 			                                  state,
-			                                  rng,
+			                                  &rng,
 			                                  ray,
 			                                  throughput,
 			                                  ss_indirect)) {
@@ -77,6 +77,7 @@ ccl_device void kernel_subsurface_scatter(KernelGlobals *kg,
 				enqueue_flag = 1;
 			}
 		}
+		kernel_split_state.rng[ray_index] = rng;
 	}
 
 #ifndef __COMPUTE_DEVICE_GPU__