Page MenuHome
Differential D12367 Diff 41308 intern/cycles/kernel/kernel_write_passes.h

Changeset View

Standalone View

View Options

intern/cycles/kernel/kernel_write_passes.h

/* /*
* Copyright 2011-2013 Blender Foundation * Copyright 2011-2013 Blender Foundation
* *
* Licensed under the Apache License, Version 2.0 (the "License"); * Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License. * you may not use this file except in compliance with the License.
* You may obtain a copy of the License at * You may obtain a copy of the License at
* *
* http://www.apache.org/licenses/LICENSE-2.0 * http://www.apache.org/licenses/LICENSE-2.0
* *
* Unless required by applicable law or agreed to in writing, software * Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS, * distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and * See the License for the specific language governing permissions and
* limitations under the License. * limitations under the License.
*/ */
#if defined(__SPLIT_KERNEL__) || defined(__KERNEL_CUDA__) #pragma once
#ifdef __KERNEL_GPU__
# define __ATOMIC_PASS_WRITE__ # define __ATOMIC_PASS_WRITE__
#endif #endif
CCL_NAMESPACE_BEGIN CCL_NAMESPACE_BEGIN
ccl_device_inline void kernel_write_pass_float(ccl_global float *buffer, float value) ccl_device_inline void kernel_write_pass_float(ccl_global float *ccl_restrict buffer, float value)
{ {
ccl_global float *buf = buffer;
#ifdef __ATOMIC_PASS_WRITE__ #ifdef __ATOMIC_PASS_WRITE__
atomic_add_and_fetch_float(buf, value); atomic_add_and_fetch_float(buffer, value);
#else #else
*buf += value; *buffer += value;
#endif #endif
} }
ccl_device_inline void kernel_write_pass_float3(ccl_global float *buffer, float3 value) ccl_device_inline void kernel_write_pass_float3(ccl_global float *ccl_restrict buffer,
float3 value)
{ {
#ifdef __ATOMIC_PASS_WRITE__ #ifdef __ATOMIC_PASS_WRITE__
ccl_global float *buf_x = buffer + 0; ccl_global float *buf_x = buffer + 0;
ccl_global float *buf_y = buffer + 1; ccl_global float *buf_y = buffer + 1;
ccl_global float *buf_z = buffer + 2; ccl_global float *buf_z = buffer + 2;
atomic_add_and_fetch_float(buf_x, value.x); atomic_add_and_fetch_float(buf_x, value.x);
atomic_add_and_fetch_float(buf_y, value.y); atomic_add_and_fetch_float(buf_y, value.y);
atomic_add_and_fetch_float(buf_z, value.z); atomic_add_and_fetch_float(buf_z, value.z);
#else #else
ccl_global float3 *buf = (ccl_global float3 *)buffer; buffer[0] += value.x;
*buf += value; buffer[1] += value.y;
buffer[2] += value.z;
#endif #endif
} }
ccl_device_inline void kernel_write_pass_float4(ccl_global float *buffer, float4 value) ccl_device_inline void kernel_write_pass_float4(ccl_global float *ccl_restrict buffer,
float4 value)
{ {
#ifdef __ATOMIC_PASS_WRITE__ #ifdef __ATOMIC_PASS_WRITE__
ccl_global float *buf_x = buffer + 0; ccl_global float *buf_x = buffer + 0;
ccl_global float *buf_y = buffer + 1; ccl_global float *buf_y = buffer + 1;
ccl_global float *buf_z = buffer + 2; ccl_global float *buf_z = buffer + 2;
ccl_global float *buf_w = buffer + 3; ccl_global float *buf_w = buffer + 3;
atomic_add_and_fetch_float(buf_x, value.x); atomic_add_and_fetch_float(buf_x, value.x);
atomic_add_and_fetch_float(buf_y, value.y); atomic_add_and_fetch_float(buf_y, value.y);
atomic_add_and_fetch_float(buf_z, value.z); atomic_add_and_fetch_float(buf_z, value.z);
atomic_add_and_fetch_float(buf_w, value.w); atomic_add_and_fetch_float(buf_w, value.w);
#else #else
ccl_global float4 *buf = (ccl_global float4 *)buffer; buffer[0] += value.x;
*buf += value; buffer[1] += value.y;
buffer[2] += value.z;
buffer[3] += value.w;
#endif #endif
} }
#ifdef __DENOISING_FEATURES__ ccl_device_inline float kernel_read_pass_float(ccl_global float *ccl_restrict buffer)
ccl_device_inline void kernel_write_pass_float_variance(ccl_global float *buffer, float value)
{ {
kernel_write_pass_float(buffer, value); return *buffer;
/* The online one-pass variance update that's used for the megakernel can't easily be implemented
* with atomics, so for the split kernel the E[x^2] - 1/N * (E[x])^2 fallback is used. */
kernel_write_pass_float(buffer + 1, value * value);
} }
# ifdef __ATOMIC_PASS_WRITE__ ccl_device_inline float3 kernel_read_pass_float3(ccl_global float *ccl_restrict buffer)
# define kernel_write_pass_float3_unaligned kernel_write_pass_float3
# else
ccl_device_inline void kernel_write_pass_float3_unaligned(ccl_global float *buffer, float3 value)
{ {
buffer[0] += value.x; return make_float3(buffer[0], buffer[1], buffer[2]);
buffer[1] += value.y;
buffer[2] += value.z;
} }
# endif
ccl_device_inline void kernel_write_pass_float3_variance(ccl_global float *buffer, float3 value) ccl_device_inline float4 kernel_read_pass_float4(ccl_global float *ccl_restrict buffer)
{ {
kernel_write_pass_float3_unaligned(buffer, value); return make_float4(buffer[0], buffer[1], buffer[2], buffer[3]);
kernel_write_pass_float3_unaligned(buffer + 3, value * value);
} }
#endif /* __DENOISING_FEATURES__ */
CCL_NAMESPACE_END CCL_NAMESPACE_END