Welcome to mirror list, hosted at ThFree Co, Russian Federation.

git.blender.org/blender.git - Unnamed repository; edit this file 'description' to name the repository.
summaryrefslogtreecommitdiff
path: root/intern/cycles/kernel
diff options
context:
space:
mode:
Diffstat (limited to 'intern/cycles/kernel')
-rw-r--r--intern/cycles/kernel/geom/geom_volume.h13
-rw-r--r--intern/cycles/kernel/kernel_compat_cuda.h25
-rw-r--r--intern/cycles/kernel/kernel_textures.h68
-rw-r--r--intern/cycles/kernel/svm/svm_image.h82
-rw-r--r--intern/cycles/kernel/svm/svm_voxel.h17
5 files changed, 68 insertions, 137 deletions
diff --git a/intern/cycles/kernel/geom/geom_volume.h b/intern/cycles/kernel/geom/geom_volume.h
index ef02c01dec6..2044aafc877 100644
--- a/intern/cycles/kernel/geom/geom_volume.h
+++ b/intern/cycles/kernel/geom/geom_volume.h
@@ -29,7 +29,7 @@ CCL_NAMESPACE_BEGIN
/* Return position normalized to 0..1 in mesh bounds */
-#ifdef __KERNEL_GPU__
+#if defined(__KERNEL_GPU__) && __CUDA_ARCH__ < 300
ccl_device float4 volume_image_texture_3d(int id, float x, float y, float z)
{
float4 r;
@@ -65,7 +65,13 @@ ccl_device float volume_attribute_float(KernelGlobals *kg, const ShaderData *sd,
{
float3 P = volume_normalized_position(kg, sd, sd->P);
#ifdef __KERNEL_GPU__
+# if __CUDA_ARCH__ >= 300
+ CUtexObject tex = kernel_tex_fetch(__bindless_mapping, id);
+ float f = kernel_tex_image_interp_3d_float(tex, P.x, P.y, P.z);
+ float4 r = make_float4(f, f, f, 1.0);
+# else
float4 r = volume_image_texture_3d(id, P.x, P.y, P.z);
+# endif
#else
float4 r;
if(sd->flag & SD_VOLUME_CUBIC)
@@ -84,7 +90,12 @@ ccl_device float3 volume_attribute_float3(KernelGlobals *kg, const ShaderData *s
{
float3 P = volume_normalized_position(kg, sd, sd->P);
#ifdef __KERNEL_GPU__
+# if __CUDA_ARCH__ >= 300
+ CUtexObject tex = kernel_tex_fetch(__bindless_mapping, id);
+ float4 r = kernel_tex_image_interp_3d_float4(tex, P.x, P.y, P.z);
+# else
float4 r = volume_image_texture_3d(id, P.x, P.y, P.z);
+# endif
#else
float4 r;
if(sd->flag & SD_VOLUME_CUBIC)
diff --git a/intern/cycles/kernel/kernel_compat_cuda.h b/intern/cycles/kernel/kernel_compat_cuda.h
index d10d3255e1b..42314756f02 100644
--- a/intern/cycles/kernel/kernel_compat_cuda.h
+++ b/intern/cycles/kernel/kernel_compat_cuda.h
@@ -67,20 +67,29 @@ typedef texture<uchar4, 2, cudaReadModeNormalizedFloat> texture_image_uchar4;
/* Macros to handle different memory storage on different devices */
-/* In order to use full 6GB of memory on Titan cards, use arrays instead
- * of textures. On earlier cards this seems slower, but on Titan it is
- * actually slightly faster in tests. */
+/* On Fermi cards (4xx and 5xx), we use regular textures for both data and images.
+ * On Kepler (6xx) and above, we use Bindless Textures for images and arrays for data.
+ *
+ * Arrays are necessary in order to use the full VRAM on newer cards, and it's slightly faster.
+ * Using Arrays on Fermi turned out to be slower.*/
+
+/* Fermi */
#if __CUDA_ARCH__ < 300
# define __KERNEL_CUDA_TEX_STORAGE__
-#endif
-
-#ifdef __KERNEL_CUDA_TEX_STORAGE__
# define kernel_tex_fetch(t, index) tex1Dfetch(t, index)
+
+# define kernel_tex_image_interp(t, x, y) tex2D(t, x, y)
+# define kernel_tex_image_interp_3d(t, x, y, z) tex3D(t, x, y, z)
+
+/* Kepler */
#else
# define kernel_tex_fetch(t, index) t[(index)]
+
+# define kernel_tex_image_interp_float4(t, x, y) tex2D<float4>(t, x, y)
+# define kernel_tex_image_interp_float(t, x, y) tex2D<float>(t, x, y)
+# define kernel_tex_image_interp_3d_float4(t, x, y, z) tex3D<float4>(t, x, y, z)
+# define kernel_tex_image_interp_3d_float(t, x, y, z) tex3D<float>(t, x, y, z)
#endif
-#define kernel_tex_image_interp(t, x, y) tex2D(t, x, y)
-#define kernel_tex_image_interp_3d(t, x, y, z) tex3D(t, x, y, z)
#define kernel_data __data
diff --git a/intern/cycles/kernel/kernel_textures.h b/intern/cycles/kernel/kernel_textures.h
index 62b0a6f2923..245d236ff97 100644
--- a/intern/cycles/kernel/kernel_textures.h
+++ b/intern/cycles/kernel/kernel_textures.h
@@ -72,6 +72,8 @@ KERNEL_TEX(float, texture_float, __lookup_table)
/* sobol */
KERNEL_TEX(uint, texture_uint, __sobol_directions)
+#ifdef __KERNEL_CUDA__
+# if __CUDA_ARCH__ < 300
/* full-float image */
KERNEL_IMAGE_TEX(float4, texture_image_float4, __tex_image_float4_000)
KERNEL_IMAGE_TEX(float4, texture_image_float4, __tex_image_float4_001)
@@ -174,66 +176,12 @@ KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_089)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_090)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_091)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_092)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_093)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_094)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_095)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_096)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_097)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_098)
-
-/* Kepler and above */
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_099)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_100)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_101)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_102)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_103)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_104)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_105)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_106)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_107)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_108)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_109)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_110)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_111)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_112)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_113)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_114)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_115)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_116)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_117)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_118)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_119)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_120)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_121)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_122)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_123)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_124)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_125)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_126)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_127)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_128)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_129)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_130)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_131)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_132)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_133)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_134)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_135)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_136)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_137)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_138)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_139)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_140)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_141)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_142)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_143)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_144)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_145)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_146)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_147)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_148)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_149)
-KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_150)
+
+# else
+/* bindless textures */
+KERNEL_TEX(uint, texture_uint, __bindless_mapping)
+# endif
+#endif
/* packed image (opencl) */
KERNEL_TEX(uchar4, texture_uchar4, __tex_image_byte4_packed)
diff --git a/intern/cycles/kernel/svm/svm_image.h b/intern/cycles/kernel/svm/svm_image.h
index faff4ce3e6d..92d2b36bbb1 100644
--- a/intern/cycles/kernel/svm/svm_image.h
+++ b/intern/cycles/kernel/svm/svm_image.h
@@ -18,11 +18,15 @@ CCL_NAMESPACE_BEGIN
/* Float4 textures on various devices. */
#if defined(__KERNEL_CPU__)
- #define TEX_NUM_FLOAT4_IMAGES TEX_NUM_FLOAT4_IMAGES_CPU
+# define TEX_NUM_FLOAT4_IMAGES TEX_NUM_FLOAT4_IMAGES_CPU
#elif defined(__KERNEL_CUDA__)
- #define TEX_NUM_FLOAT4_IMAGES TEX_NUM_FLOAT4_IMAGES_CUDA
+# if __CUDA_ARCH__ < 300
+# define TEX_NUM_FLOAT4_IMAGES TEX_NUM_FLOAT4_IMAGES_CUDA
+# else
+# define TEX_NUM_FLOAT4_IMAGES TEX_NUM_FLOAT4_IMAGES_CUDA_KEPLER
+# endif
#else
- #define TEX_NUM_FLOAT4_IMAGES TEX_NUM_FLOAT4_IMAGES_OPENCL
+# define TEX_NUM_FLOAT4_IMAGES TEX_NUM_FLOAT4_IMAGES_OPENCL
#endif
#ifdef __KERNEL_OPENCL__
@@ -151,6 +155,7 @@ ccl_device float4 svm_image_texture(KernelGlobals *kg, int id, float x, float y,
#else
float4 r;
+# if __CUDA_ARCH__ < 300
/* not particularly proud of this massive switch, what are the
* alternatives?
* - use a single big 1D texture, and do our own lookup/filtering
@@ -254,72 +259,19 @@ ccl_device float4 svm_image_texture(KernelGlobals *kg, int id, float x, float y,
case 90: r = kernel_tex_image_interp(__tex_image_byte4_090, x, y); break;
case 91: r = kernel_tex_image_interp(__tex_image_byte4_091, x, y); break;
case 92: r = kernel_tex_image_interp(__tex_image_byte4_092, x, y); break;
-
-# if defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 300)
- case 93: r = kernel_tex_image_interp(__tex_image_byte4_093, x, y); break;
- case 94: r = kernel_tex_image_interp(__tex_image_byte4_094, x, y); break;
- case 95: r = kernel_tex_image_interp(__tex_image_byte4_095, x, y); break;
- case 96: r = kernel_tex_image_interp(__tex_image_byte4_096, x, y); break;
- case 97: r = kernel_tex_image_interp(__tex_image_byte4_097, x, y); break;
- case 98: r = kernel_tex_image_interp(__tex_image_byte4_098, x, y); break;
- case 99: r = kernel_tex_image_interp(__tex_image_byte4_099, x, y); break;
- case 100: r = kernel_tex_image_interp(__tex_image_byte4_100, x, y); break;
- case 101: r = kernel_tex_image_interp(__tex_image_byte4_101, x, y); break;
- case 102: r = kernel_tex_image_interp(__tex_image_byte4_102, x, y); break;
- case 103: r = kernel_tex_image_interp(__tex_image_byte4_103, x, y); break;
- case 104: r = kernel_tex_image_interp(__tex_image_byte4_104, x, y); break;
- case 105: r = kernel_tex_image_interp(__tex_image_byte4_105, x, y); break;
- case 106: r = kernel_tex_image_interp(__tex_image_byte4_106, x, y); break;
- case 107: r = kernel_tex_image_interp(__tex_image_byte4_107, x, y); break;
- case 108: r = kernel_tex_image_interp(__tex_image_byte4_108, x, y); break;
- case 109: r = kernel_tex_image_interp(__tex_image_byte4_109, x, y); break;
- case 110: r = kernel_tex_image_interp(__tex_image_byte4_110, x, y); break;
- case 111: r = kernel_tex_image_interp(__tex_image_byte4_111, x, y); break;
- case 112: r = kernel_tex_image_interp(__tex_image_byte4_112, x, y); break;
- case 113: r = kernel_tex_image_interp(__tex_image_byte4_113, x, y); break;
- case 114: r = kernel_tex_image_interp(__tex_image_byte4_114, x, y); break;
- case 115: r = kernel_tex_image_interp(__tex_image_byte4_115, x, y); break;
- case 116: r = kernel_tex_image_interp(__tex_image_byte4_116, x, y); break;
- case 117: r = kernel_tex_image_interp(__tex_image_byte4_117, x, y); break;
- case 118: r = kernel_tex_image_interp(__tex_image_byte4_118, x, y); break;
- case 119: r = kernel_tex_image_interp(__tex_image_byte4_119, x, y); break;
- case 120: r = kernel_tex_image_interp(__tex_image_byte4_120, x, y); break;
- case 121: r = kernel_tex_image_interp(__tex_image_byte4_121, x, y); break;
- case 122: r = kernel_tex_image_interp(__tex_image_byte4_122, x, y); break;
- case 123: r = kernel_tex_image_interp(__tex_image_byte4_123, x, y); break;
- case 124: r = kernel_tex_image_interp(__tex_image_byte4_124, x, y); break;
- case 125: r = kernel_tex_image_interp(__tex_image_byte4_125, x, y); break;
- case 126: r = kernel_tex_image_interp(__tex_image_byte4_126, x, y); break;
- case 127: r = kernel_tex_image_interp(__tex_image_byte4_127, x, y); break;
- case 128: r = kernel_tex_image_interp(__tex_image_byte4_128, x, y); break;
- case 129: r = kernel_tex_image_interp(__tex_image_byte4_129, x, y); break;
- case 130: r = kernel_tex_image_interp(__tex_image_byte4_130, x, y); break;
- case 131: r = kernel_tex_image_interp(__tex_image_byte4_131, x, y); break;
- case 132: r = kernel_tex_image_interp(__tex_image_byte4_132, x, y); break;
- case 133: r = kernel_tex_image_interp(__tex_image_byte4_133, x, y); break;
- case 134: r = kernel_tex_image_interp(__tex_image_byte4_134, x, y); break;
- case 135: r = kernel_tex_image_interp(__tex_image_byte4_135, x, y); break;
- case 136: r = kernel_tex_image_interp(__tex_image_byte4_136, x, y); break;
- case 137: r = kernel_tex_image_interp(__tex_image_byte4_137, x, y); break;
- case 138: r = kernel_tex_image_interp(__tex_image_byte4_138, x, y); break;
- case 139: r = kernel_tex_image_interp(__tex_image_byte4_139, x, y); break;
- case 140: r = kernel_tex_image_interp(__tex_image_byte4_140, x, y); break;
- case 141: r = kernel_tex_image_interp(__tex_image_byte4_141, x, y); break;
- case 142: r = kernel_tex_image_interp(__tex_image_byte4_142, x, y); break;
- case 143: r = kernel_tex_image_interp(__tex_image_byte4_143, x, y); break;
- case 144: r = kernel_tex_image_interp(__tex_image_byte4_144, x, y); break;
- case 145: r = kernel_tex_image_interp(__tex_image_byte4_145, x, y); break;
- case 146: r = kernel_tex_image_interp(__tex_image_byte4_146, x, y); break;
- case 147: r = kernel_tex_image_interp(__tex_image_byte4_147, x, y); break;
- case 148: r = kernel_tex_image_interp(__tex_image_byte4_148, x, y); break;
- case 149: r = kernel_tex_image_interp(__tex_image_byte4_149, x, y); break;
- case 150: r = kernel_tex_image_interp(__tex_image_byte4_150, x, y); break;
-# endif
-
default:
kernel_assert(0);
return make_float4(0.0f, 0.0f, 0.0f, 0.0f);
}
+# else
+ CUtexObject tex = kernel_tex_fetch(__bindless_mapping, id);
+ if(id < 2048) /* TODO(dingto): Make this a variable */
+ r = kernel_tex_image_interp_float4(tex, x, y);
+ else {
+ float f = kernel_tex_image_interp_float(tex, x, y);
+ r = make_float4(f, f, f, 1.0);
+ }
+# endif
#endif
#ifdef __KERNEL_SSE2__
diff --git a/intern/cycles/kernel/svm/svm_voxel.h b/intern/cycles/kernel/svm/svm_voxel.h
index 85ba2f906fa..d2cc2c3730e 100644
--- a/intern/cycles/kernel/svm/svm_voxel.h
+++ b/intern/cycles/kernel/svm/svm_voxel.h
@@ -42,10 +42,21 @@ ccl_device void svm_node_tex_voxel(KernelGlobals *kg,
tfm.w = read_node_float(kg, offset);
co = transform_point(&tfm, co);
}
+ float4 r;
# if defined(__KERNEL_GPU__)
- float4 r = volume_image_texture_3d(id, co.x, co.y, co.z);
-# else
- float4 r = kernel_tex_image_interp_3d(id, co.x, co.y, co.z);
+# if __CUDA_ARCH__ >= 300
+ CUtexObject tex = kernel_tex_fetch(__bindless_mapping, id);
+ if(id < 2048) /* TODO(dingto): Make this a variable */
+ r = kernel_tex_image_interp_3d_float4(tex, co.x, co.y, co.z);
+ else {
+ float f = kernel_tex_image_interp_3d_float(tex, co.x, co.y, co.z);
+ r = make_float4(f, f, f, 1.0);
+ }
+# else /* __CUDA_ARCH__ >= 300 */
+ r = volume_image_texture_3d(id, co.x, co.y, co.z);
+# endif
+# else /* __KERNEL_GPU__ */
+ r = kernel_tex_image_interp_3d(id, co.x, co.y, co.z);
# endif
#else
float4 r = make_float4(0.0f, 0.0f, 0.0f, 0.0f);
generated by cgit v1.2.3 (git 2.25.1) at 2024-07-17 23:28:42 +0300