diff options
-rw-r--r-- | intern/cycles/device/device.cpp | 1 | ||||
-rw-r--r-- | intern/cycles/device/device_memory.h | 1 | ||||
-rw-r--r-- | intern/cycles/kernel/CMakeLists.txt | 1 | ||||
-rw-r--r-- | intern/cycles/kernel/kernel_compat_cpu.h | 1 | ||||
-rw-r--r-- | intern/cycles/kernel/kernel_compat_cuda.h | 1 | ||||
-rw-r--r-- | intern/cycles/kernel/kernel_compat_opencl.h | 1 | ||||
-rw-r--r-- | intern/cycles/render/buffers.h | 1 | ||||
-rw-r--r-- | intern/cycles/util/CMakeLists.txt | 1 | ||||
-rw-r--r-- | intern/cycles/util/util_half.h | 91 | ||||
-rw-r--r-- | intern/cycles/util/util_types.h | 64 |
10 files changed, 99 insertions, 64 deletions
diff --git a/intern/cycles/device/device.cpp b/intern/cycles/device/device.cpp index 6283e34f563..a3264f0c516 100644 --- a/intern/cycles/device/device.cpp +++ b/intern/cycles/device/device.cpp @@ -23,6 +23,7 @@ #include "util_cuda.h" #include "util_debug.h" #include "util_foreach.h" +#include "util_half.h" #include "util_math.h" #include "util_opencl.h" #include "util_opengl.h" diff --git a/intern/cycles/device/device_memory.h b/intern/cycles/device/device_memory.h index 352a46f8c36..1427d12cba2 100644 --- a/intern/cycles/device/device_memory.h +++ b/intern/cycles/device/device_memory.h @@ -29,6 +29,7 @@ * to and from. */ #include "util_debug.h" +#include "util_half.h" #include "util_types.h" #include "util_vector.h" diff --git a/intern/cycles/kernel/CMakeLists.txt b/intern/cycles/kernel/CMakeLists.txt index 780aa417828..08427b17879 100644 --- a/intern/cycles/kernel/CMakeLists.txt +++ b/intern/cycles/kernel/CMakeLists.txt @@ -119,6 +119,7 @@ set(SRC_SVM_HEADERS set(SRC_UTIL_HEADERS ../util/util_color.h + ../util/util_half.h ../util/util_math.h ../util/util_transform.h ../util/util_types.h diff --git a/intern/cycles/kernel/kernel_compat_cpu.h b/intern/cycles/kernel/kernel_compat_cpu.h index 2b786bce0de..ce597a87315 100644 --- a/intern/cycles/kernel/kernel_compat_cpu.h +++ b/intern/cycles/kernel/kernel_compat_cpu.h @@ -20,6 +20,7 @@ #define __KERNEL_CPU__ #include "util_debug.h" +#include "util_half.h" #include "util_math.h" #include "util_simd.h" #include "util_types.h" diff --git a/intern/cycles/kernel/kernel_compat_cuda.h b/intern/cycles/kernel/kernel_compat_cuda.h index 76f885aefe0..15e7353ec38 100644 --- a/intern/cycles/kernel/kernel_compat_cuda.h +++ b/intern/cycles/kernel/kernel_compat_cuda.h @@ -40,6 +40,7 @@ /* Types */ +#include "util_half.h" #include "util_types.h" /* Textures */ diff --git a/intern/cycles/kernel/kernel_compat_opencl.h b/intern/cycles/kernel/kernel_compat_opencl.h index 1ff3615e448..4f4414cc298 100644 --- a/intern/cycles/kernel/kernel_compat_opencl.h +++ b/intern/cycles/kernel/kernel_compat_opencl.h @@ -113,6 +113,7 @@ /* define NULL */ #define NULL 0 +#include "util_half.h" #include "util_types.h" #endif /* __KERNEL_COMPAT_OPENCL_H__ */ diff --git a/intern/cycles/render/buffers.h b/intern/cycles/render/buffers.h index 0b1f9010e75..81eaf41077f 100644 --- a/intern/cycles/render/buffers.h +++ b/intern/cycles/render/buffers.h @@ -23,6 +23,7 @@ #include "kernel_types.h" +#include "util_half.h" #include "util_string.h" #include "util_thread.h" #include "util_types.h" diff --git a/intern/cycles/util/CMakeLists.txt b/intern/cycles/util/CMakeLists.txt index df188f8236a..c1150d226ae 100644 --- a/intern/cycles/util/CMakeLists.txt +++ b/intern/cycles/util/CMakeLists.txt @@ -38,6 +38,7 @@ set(SRC_HEADERS util_dynlib.h util_foreach.h util_function.h + util_half.h util_hash.h util_image.h util_list.h diff --git a/intern/cycles/util/util_half.h b/intern/cycles/util/util_half.h new file mode 100644 index 00000000000..21192024f7f --- /dev/null +++ b/intern/cycles/util/util_half.h @@ -0,0 +1,91 @@ +/* + * 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 + */ + +#ifndef __UTIL_HALF_H__ +#define __UTIL_HALF_H__ + +#include "util_types.h" + +CCL_NAMESPACE_BEGIN + +/* Half Floats */ + +#ifdef __KERNEL_OPENCL__ + +#define float4_store_half(h, f, scale) vstore_half4(*(f) * (scale), 0, h); + +#else + +typedef unsigned short half; +struct half4 { half x, y, z, w; }; + +#ifdef __KERNEL_CUDA__ + +ccl_device_inline void float4_store_half(half *h, const float4 *f, float scale) +{ + h[0] = __float2half_rn(f->x * scale); + h[1] = __float2half_rn(f->y * scale); + h[2] = __float2half_rn(f->z * scale); + h[3] = __float2half_rn(f->w * scale); +} + +#else + +ccl_device_inline void float4_store_half(half *h, const float4 *f, float scale) +{ +#ifndef __KERNEL_SSE2__ + for(int i = 0; i < 4; i++) { + /* optimized float to half for pixels: + * assumes no negative, no nan, no inf, and sets denormal to 0 */ + union { uint i; float f; } in; + float fscale = (*f)[i] * scale; + in.f = (fscale > 0.0f)? ((fscale < 65500.0f)? fscale: 65500.0f): 0.0f; + int x = in.i; + + int absolute = x & 0x7FFFFFFF; + int Z = absolute + 0xC8000000; + int result = (absolute < 0x38800000)? 0: Z; + int rshift = (result >> 13); + + h[i] = (rshift & 0x7FFF); + } +#else + /* same as above with SSE */ + const __m128 mm_scale = _mm_set_ps1(scale); + const __m128i mm_38800000 = _mm_set1_epi32(0x38800000); + const __m128i mm_7FFF = _mm_set1_epi32(0x7FFF); + const __m128i mm_7FFFFFFF = _mm_set1_epi32(0x7FFFFFFF); + const __m128i mm_C8000000 = _mm_set1_epi32(0xC8000000); + + __m128 mm_fscale = _mm_mul_ps(*(__m128*)f, mm_scale); + __m128i x = _mm_castps_si128(_mm_min_ps(_mm_max_ps(mm_fscale, _mm_set_ps1(0.0f)), _mm_set_ps1(65500.0f))); + __m128i absolute = _mm_and_si128(x, mm_7FFFFFFF); + __m128i Z = _mm_add_epi32(absolute, mm_C8000000); + __m128i result = _mm_andnot_si128(_mm_cmplt_epi32(absolute, mm_38800000), Z); + __m128i rh = _mm_and_si128(_mm_srai_epi32(result, 13), mm_7FFF); + + _mm_storel_pi((__m64*)h, _mm_castsi128_ps(_mm_packs_epi32(rh, rh))); +#endif +} + +#endif + +#endif + +CCL_NAMESPACE_END + +#endif /* __UTIL_HALF_H__ */ + diff --git a/intern/cycles/util/util_types.h b/intern/cycles/util/util_types.h index ebfd8b6700c..45038488c14 100644 --- a/intern/cycles/util/util_types.h +++ b/intern/cycles/util/util_types.h @@ -440,70 +440,6 @@ ccl_device_inline int4 make_int4(const float3& f) #endif -/* Half Floats */ - -#ifdef __KERNEL_OPENCL__ - -#define float4_store_half(h, f, scale) vstore_half4(*(f) * (scale), 0, h); - -#else - -typedef unsigned short half; -struct half4 { half x, y, z, w; }; - -#ifdef __KERNEL_CUDA__ - -ccl_device_inline void float4_store_half(half *h, const float4 *f, float scale) -{ - h[0] = __float2half_rn(f->x * scale); - h[1] = __float2half_rn(f->y * scale); - h[2] = __float2half_rn(f->z * scale); - h[3] = __float2half_rn(f->w * scale); -} - -#else - -ccl_device_inline void float4_store_half(half *h, const float4 *f, float scale) -{ -#ifndef __KERNEL_SSE2__ - for(int i = 0; i < 4; i++) { - /* optimized float to half for pixels: - * assumes no negative, no nan, no inf, and sets denormal to 0 */ - union { uint i; float f; } in; - float fscale = (*f)[i] * scale; - in.f = (fscale > 0.0f)? ((fscale < 65500.0f)? fscale: 65500.0f): 0.0f; - int x = in.i; - - int absolute = x & 0x7FFFFFFF; - int Z = absolute + 0xC8000000; - int result = (absolute < 0x38800000)? 0: Z; - int rshift = (result >> 13); - - h[i] = (rshift & 0x7FFF); - } -#else - /* same as above with SSE */ - const __m128 mm_scale = _mm_set_ps1(scale); - const __m128i mm_38800000 = _mm_set1_epi32(0x38800000); - const __m128i mm_7FFF = _mm_set1_epi32(0x7FFF); - const __m128i mm_7FFFFFFF = _mm_set1_epi32(0x7FFFFFFF); - const __m128i mm_C8000000 = _mm_set1_epi32(0xC8000000); - - __m128 mm_fscale = _mm_mul_ps(*(__m128*)f, mm_scale); - __m128i x = _mm_castps_si128(_mm_min_ps(_mm_max_ps(mm_fscale, _mm_set_ps1(0.0f)), _mm_set_ps1(65500.0f))); - __m128i absolute = _mm_and_si128(x, mm_7FFFFFFF); - __m128i Z = _mm_add_epi32(absolute, mm_C8000000); - __m128i result = _mm_andnot_si128(_mm_cmplt_epi32(absolute, mm_38800000), Z); - __m128i rh = _mm_and_si128(_mm_srai_epi32(result, 13), mm_7FFF); - - _mm_storel_pi((__m64*)h, _mm_castsi128_ps(_mm_packs_epi32(rh, rh))); -#endif -} - -#endif - -#endif - CCL_NAMESPACE_END #endif /* __UTIL_TYPES_H__ */ |