diff options
Diffstat (limited to 'intern/cycles/kernel/kernels/cpu')
-rw-r--r-- | intern/cycles/kernel/kernels/cpu/kernel.cpp | 133 | ||||
-rw-r--r-- | intern/cycles/kernel/kernels/cpu/kernel_avx.cpp | 87 | ||||
-rw-r--r-- | intern/cycles/kernel/kernels/cpu/kernel_avx2.cpp | 88 | ||||
-rw-r--r-- | intern/cycles/kernel/kernels/cpu/kernel_sse2.cpp | 84 | ||||
-rw-r--r-- | intern/cycles/kernel/kernels/cpu/kernel_sse3.cpp | 85 | ||||
-rw-r--r-- | intern/cycles/kernel/kernels/cpu/kernel_sse41.cpp | 86 |
6 files changed, 563 insertions, 0 deletions
diff --git a/intern/cycles/kernel/kernels/cpu/kernel.cpp b/intern/cycles/kernel/kernels/cpu/kernel.cpp new file mode 100644 index 00000000000..37a73ab2f04 --- /dev/null +++ b/intern/cycles/kernel/kernels/cpu/kernel.cpp @@ -0,0 +1,133 @@ +/* + * Copyright 2011-2013 Blender Foundation + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* CPU kernel entry points */ + +#include "kernel_compat_cpu.h" +#include "kernel.h" +#include "kernel_math.h" +#include "kernel_types.h" +#include "kernel_globals.h" +#include "kernel_film.h" +#include "kernel_path.h" +#include "kernel_path_branched.h" +#include "kernel_bake.h" + +CCL_NAMESPACE_BEGIN + +/* Memory Copy */ + +void kernel_const_copy(KernelGlobals *kg, const char *name, void *host, size_t size) +{ + if(strcmp(name, "__data") == 0) + memcpy(&kg->__data, host, size); + else + assert(0); +} + +void kernel_tex_copy(KernelGlobals *kg, const char *name, device_ptr mem, size_t width, size_t height, size_t depth, InterpolationType interpolation) +{ + if(0) { + } + +#define KERNEL_TEX(type, ttype, tname) \ + else if(strcmp(name, #tname) == 0) { \ + kg->tname.data = (type*)mem; \ + kg->tname.width = width; \ + } +#define KERNEL_IMAGE_TEX(type, ttype, tname) +#include "kernel_textures.h" + + else if(strstr(name, "__tex_image_float")) { + texture_image_float4 *tex = NULL; + int id = atoi(name + strlen("__tex_image_float_")); + int array_index = id; + + if(array_index >= 0 && array_index < MAX_FLOAT_IMAGES) { + tex = &kg->texture_float_images[array_index]; + } + + if(tex) { + tex->data = (float4*)mem; + tex->dimensions_set(width, height, depth); + tex->interpolation = interpolation; + } + } + else if(strstr(name, "__tex_image")) { + texture_image_uchar4 *tex = NULL; + int id = atoi(name + strlen("__tex_image_")); + int array_index = id - MAX_FLOAT_IMAGES; + + if(array_index >= 0 && array_index < MAX_BYTE_IMAGES) { + tex = &kg->texture_byte_images[array_index]; + } + + if(tex) { + tex->data = (uchar4*)mem; + tex->dimensions_set(width, height, depth); + tex->interpolation = interpolation; + } + } + else + assert(0); +} + +/* On x86-64, we can assume SSE2, so avoid the extra kernel and compile this one with SSE2 intrinsics */ +#if defined(__x86_64__) || defined(_M_X64) +#define __KERNEL_SSE2__ +#endif + +/* quiet unused define warnings */ +#if defined(__KERNEL_SSE2__) + /* do nothing */ +#endif + +/* Path Tracing */ + +void kernel_cpu_path_trace(KernelGlobals *kg, float *buffer, unsigned int *rng_state, int sample, int x, int y, int offset, int stride) +{ +#ifdef __BRANCHED_PATH__ + if(kernel_data.integrator.branched) + kernel_branched_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); + else +#endif + kernel_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); +} + +/* Film */ + +void kernel_cpu_convert_to_byte(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_byte(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +void kernel_cpu_convert_to_half_float(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_half_float(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +/* Shader Evaluation */ + +void kernel_cpu_shader(KernelGlobals *kg, uint4 *input, float4 *output, int type, int i, int offset, int sample) +{ + if(type >= SHADER_EVAL_BAKE) + kernel_bake_evaluate(kg, input, output, (ShaderEvalType)type, i, offset, sample); + else + kernel_shader_evaluate(kg, input, output, (ShaderEvalType)type, i, sample); +} + +CCL_NAMESPACE_END + diff --git a/intern/cycles/kernel/kernels/cpu/kernel_avx.cpp b/intern/cycles/kernel/kernels/cpu/kernel_avx.cpp new file mode 100644 index 00000000000..df77bedc729 --- /dev/null +++ b/intern/cycles/kernel/kernels/cpu/kernel_avx.cpp @@ -0,0 +1,87 @@ +/* + * Copyright 2011-2013 Blender Foundation + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* Optimized CPU kernel entry points. This file is compiled with AVX + * optimization flags and nearly all functions inlined, while kernel.cpp + * is compiled without for other CPU's. */ + +/* SSE optimization disabled for now on 32 bit, see bug #36316 */ +#if !(defined(__GNUC__) && (defined(i386) || defined(_M_IX86))) +#define __KERNEL_SSE2__ +#define __KERNEL_SSE3__ +#define __KERNEL_SSSE3__ +#define __KERNEL_SSE41__ +#define __KERNEL_AVX__ +#endif + +#include "util_optimization.h" + +#ifdef WITH_CYCLES_OPTIMIZED_KERNEL_AVX + +#include "kernel_compat_cpu.h" +#include "kernel.h" +#include "kernel_math.h" +#include "kernel_types.h" +#include "kernel_globals.h" +#include "kernel_film.h" +#include "kernel_path.h" +#include "kernel_path_branched.h" +#include "kernel_bake.h" + +CCL_NAMESPACE_BEGIN + +/* Path Tracing */ + +void kernel_cpu_avx_path_trace(KernelGlobals *kg, float *buffer, unsigned int *rng_state, int sample, int x, int y, int offset, int stride) +{ +#ifdef __BRANCHED_PATH__ + if(kernel_data.integrator.branched) + kernel_branched_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); + else +#endif + kernel_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); +} + +/* Film */ + +void kernel_cpu_avx_convert_to_byte(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_byte(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +void kernel_cpu_avx_convert_to_half_float(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_half_float(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +/* Shader Evaluate */ + +void kernel_cpu_avx_shader(KernelGlobals *kg, uint4 *input, float4 *output, int type, int i, int offset, int sample) +{ + if(type >= SHADER_EVAL_BAKE) + kernel_bake_evaluate(kg, input, output, (ShaderEvalType)type, i, offset, sample); + else + kernel_shader_evaluate(kg, input, output, (ShaderEvalType)type, i, sample); +} + +CCL_NAMESPACE_END +#else + +/* needed for some linkers in combination with scons making empty compilation unit in a library */ +void __dummy_function_cycles_avx(void); +void __dummy_function_cycles_avx(void) {} + +#endif diff --git a/intern/cycles/kernel/kernels/cpu/kernel_avx2.cpp b/intern/cycles/kernel/kernels/cpu/kernel_avx2.cpp new file mode 100644 index 00000000000..b3192369794 --- /dev/null +++ b/intern/cycles/kernel/kernels/cpu/kernel_avx2.cpp @@ -0,0 +1,88 @@ +/* + * Copyright 2011-2014 Blender Foundation + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* Optimized CPU kernel entry points. This file is compiled with AVX2 + * optimization flags and nearly all functions inlined, while kernel.cpp + * is compiled without for other CPU's. */ + +/* SSE optimization disabled for now on 32 bit, see bug #36316 */ +#if !(defined(__GNUC__) && (defined(i386) || defined(_M_IX86))) +#define __KERNEL_SSE2__ +#define __KERNEL_SSE3__ +#define __KERNEL_SSSE3__ +#define __KERNEL_SSE41__ +#define __KERNEL_AVX__ +#define __KERNEL_AVX2__ +#endif + +#include "util_optimization.h" + +#ifdef WITH_CYCLES_OPTIMIZED_KERNEL_AVX2 + +#include "kernel_compat_cpu.h" +#include "kernel.h" +#include "kernel_math.h" +#include "kernel_types.h" +#include "kernel_globals.h" +#include "kernel_film.h" +#include "kernel_path.h" +#include "kernel_path_branched.h" +#include "kernel_bake.h" + +CCL_NAMESPACE_BEGIN + +/* Path Tracing */ + +void kernel_cpu_avx2_path_trace(KernelGlobals *kg, float *buffer, unsigned int *rng_state, int sample, int x, int y, int offset, int stride) +{ +#ifdef __BRANCHED_PATH__ + if(kernel_data.integrator.branched) + kernel_branched_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); + else +#endif + kernel_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); +} + +/* Film */ + +void kernel_cpu_avx2_convert_to_byte(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_byte(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +void kernel_cpu_avx2_convert_to_half_float(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_half_float(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +/* Shader Evaluate */ + +void kernel_cpu_avx2_shader(KernelGlobals *kg, uint4 *input, float4 *output, int type, int i, int offset, int sample) +{ + if(type >= SHADER_EVAL_BAKE) + kernel_bake_evaluate(kg, input, output, (ShaderEvalType)type, i, offset, sample); + else + kernel_shader_evaluate(kg, input, output, (ShaderEvalType)type, i, sample); +} + +CCL_NAMESPACE_END +#else + +/* needed for some linkers in combination with scons making empty compilation unit in a library */ +void __dummy_function_cycles_avx2(void); +void __dummy_function_cycles_avx2(void) {} + +#endif diff --git a/intern/cycles/kernel/kernels/cpu/kernel_sse2.cpp b/intern/cycles/kernel/kernels/cpu/kernel_sse2.cpp new file mode 100644 index 00000000000..f9c5134e442 --- /dev/null +++ b/intern/cycles/kernel/kernels/cpu/kernel_sse2.cpp @@ -0,0 +1,84 @@ +/* + * Copyright 2011-2013 Blender Foundation + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* Optimized CPU kernel entry points. This file is compiled with SSE2 + * optimization flags and nearly all functions inlined, while kernel.cpp + * is compiled without for other CPU's. */ + +/* SSE optimization disabled for now on 32 bit, see bug #36316 */ +#if !(defined(__GNUC__) && (defined(i386) || defined(_M_IX86))) +#define __KERNEL_SSE2__ +#endif + +#include "util_optimization.h" + +#ifdef WITH_CYCLES_OPTIMIZED_KERNEL_SSE2 + +#include "kernel_compat_cpu.h" +#include "kernel.h" +#include "kernel_math.h" +#include "kernel_types.h" +#include "kernel_globals.h" +#include "kernel_film.h" +#include "kernel_path.h" +#include "kernel_path_branched.h" +#include "kernel_bake.h" + +CCL_NAMESPACE_BEGIN + +/* Path Tracing */ + +void kernel_cpu_sse2_path_trace(KernelGlobals *kg, float *buffer, unsigned int *rng_state, int sample, int x, int y, int offset, int stride) +{ +#ifdef __BRANCHED_PATH__ + if(kernel_data.integrator.branched) + kernel_branched_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); + else +#endif + kernel_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); +} + +/* Film */ + +void kernel_cpu_sse2_convert_to_byte(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_byte(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +void kernel_cpu_sse2_convert_to_half_float(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_half_float(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +/* Shader Evaluate */ + +void kernel_cpu_sse2_shader(KernelGlobals *kg, uint4 *input, float4 *output, int type, int i, int offset, int sample) +{ + if(type >= SHADER_EVAL_BAKE) + kernel_bake_evaluate(kg, input, output, (ShaderEvalType)type, i, offset, sample); + else + kernel_shader_evaluate(kg, input, output, (ShaderEvalType)type, i, sample); +} + +CCL_NAMESPACE_END + +#else + +/* needed for some linkers in combination with scons making empty compilation unit in a library */ +void __dummy_function_cycles_sse2(void); +void __dummy_function_cycles_sse2(void) {} + +#endif diff --git a/intern/cycles/kernel/kernels/cpu/kernel_sse3.cpp b/intern/cycles/kernel/kernels/cpu/kernel_sse3.cpp new file mode 100644 index 00000000000..2dbe4b81821 --- /dev/null +++ b/intern/cycles/kernel/kernels/cpu/kernel_sse3.cpp @@ -0,0 +1,85 @@ +/* + * Copyright 2011-2013 Blender Foundation + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* Optimized CPU kernel entry points. This file is compiled with SSE3/SSSE3 + * optimization flags and nearly all functions inlined, while kernel.cpp + * is compiled without for other CPU's. */ + +/* SSE optimization disabled for now on 32 bit, see bug #36316 */ +#if !(defined(__GNUC__) && (defined(i386) || defined(_M_IX86))) +#define __KERNEL_SSE2__ +#define __KERNEL_SSE3__ +#define __KERNEL_SSSE3__ +#endif + +#include "util_optimization.h" + +#ifdef WITH_CYCLES_OPTIMIZED_KERNEL_SSE3 + +#include "kernel_compat_cpu.h" +#include "kernel.h" +#include "kernel_math.h" +#include "kernel_types.h" +#include "kernel_globals.h" +#include "kernel_film.h" +#include "kernel_path.h" +#include "kernel_path_branched.h" +#include "kernel_bake.h" + +CCL_NAMESPACE_BEGIN + +/* Path Tracing */ + +void kernel_cpu_sse3_path_trace(KernelGlobals *kg, float *buffer, unsigned int *rng_state, int sample, int x, int y, int offset, int stride) +{ +#ifdef __BRANCHED_PATH__ + if(kernel_data.integrator.branched) + kernel_branched_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); + else +#endif + kernel_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); +} + +/* Film */ + +void kernel_cpu_sse3_convert_to_byte(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_byte(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +void kernel_cpu_sse3_convert_to_half_float(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_half_float(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +/* Shader Evaluate */ + +void kernel_cpu_sse3_shader(KernelGlobals *kg, uint4 *input, float4 *output, int type, int i, int offset, int sample) +{ + if(type >= SHADER_EVAL_BAKE) + kernel_bake_evaluate(kg, input, output, (ShaderEvalType)type, i, offset, sample); + else + kernel_shader_evaluate(kg, input, output, (ShaderEvalType)type, i, sample); +} + +CCL_NAMESPACE_END +#else + +/* needed for some linkers in combination with scons making empty compilation unit in a library */ +void __dummy_function_cycles_sse3(void); +void __dummy_function_cycles_sse3(void) {} + +#endif diff --git a/intern/cycles/kernel/kernels/cpu/kernel_sse41.cpp b/intern/cycles/kernel/kernels/cpu/kernel_sse41.cpp new file mode 100644 index 00000000000..5c57ad01181 --- /dev/null +++ b/intern/cycles/kernel/kernels/cpu/kernel_sse41.cpp @@ -0,0 +1,86 @@ +/* + * Copyright 2011-2013 Blender Foundation + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* Optimized CPU kernel entry points. This file is compiled with SSE3/SSSE3 + * optimization flags and nearly all functions inlined, while kernel.cpp + * is compiled without for other CPU's. */ + +/* SSE optimization disabled for now on 32 bit, see bug #36316 */ +#if !(defined(__GNUC__) && (defined(i386) || defined(_M_IX86))) +#define __KERNEL_SSE2__ +#define __KERNEL_SSE3__ +#define __KERNEL_SSSE3__ +#define __KERNEL_SSE41__ +#endif + +#include "util_optimization.h" + +#ifdef WITH_CYCLES_OPTIMIZED_KERNEL_SSE41 + +#include "kernel_compat_cpu.h" +#include "kernel.h" +#include "kernel_math.h" +#include "kernel_types.h" +#include "kernel_globals.h" +#include "kernel_film.h" +#include "kernel_path.h" +#include "kernel_path_branched.h" +#include "kernel_bake.h" + +CCL_NAMESPACE_BEGIN + +/* Path Tracing */ + +void kernel_cpu_sse41_path_trace(KernelGlobals *kg, float *buffer, unsigned int *rng_state, int sample, int x, int y, int offset, int stride) +{ +#ifdef __BRANCHED_PATH__ + if(kernel_data.integrator.branched) + kernel_branched_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); + else +#endif + kernel_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); +} + +/* Film */ + +void kernel_cpu_sse41_convert_to_byte(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_byte(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +void kernel_cpu_sse41_convert_to_half_float(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_half_float(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +/* Shader Evaluate */ + +void kernel_cpu_sse41_shader(KernelGlobals *kg, uint4 *input, float4 *output, int type, int i, int offset, int sample) +{ + if(type >= SHADER_EVAL_BAKE) + kernel_bake_evaluate(kg, input, output, (ShaderEvalType)type, i, offset, sample); + else + kernel_shader_evaluate(kg, input, output, (ShaderEvalType)type, i, sample); +} + +CCL_NAMESPACE_END +#else + +/* needed for some linkers in combination with scons making empty compilation unit in a library */ +void __dummy_function_cycles_sse41(void); +void __dummy_function_cycles_sse41(void) {} + +#endif |