Cr8-xyz/blender
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Cycles: Remove Fermi texture code.

Browse Source 
This should be the last Fermi removal commit, unless I missed something.
It's been a pleasure Fermi!
This commit is contained in:
Thomas Dinges
2018-02-17 22:56:58 +01:00
parent e1ef902058
commit 9e717c0495
9 changed files with 74 additions and 407 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
intern/cycles/device/device.cpp
View File

@@ -359,7 +359,6 @@ DeviceInfo Device::get_multi_device(const vector<DeviceInfo>& subdevices, int th
info.description = "Multi Device";
info.num = 0;
info.has_fermi_limits = false;
info.has_half_images = true;
info.has_volume_decoupled = true;
info.bvh_layout_mask = BVH_LAYOUT_ALL;
@@ -395,8 +394,6 @@ DeviceInfo Device::get_multi_device(const vector<DeviceInfo>& subdevices, int th
}
/* Accumulate device info. */
info.has_fermi_limits = info.has_fermi_limits ||
device.has_fermi_limits;
info.has_half_images &= device.has_half_images;
info.has_volume_decoupled &= device.has_volume_decoupled;
info.bvh_layout_mask = device.bvh_layout_mask & info.bvh_layout_mask;

2
intern/cycles/device/device.h
View File

@@ -56,7 +56,6 @@ public:
int num;
bool display_device; /* GPU is used as a display device. */
bool advanced_shading; /* Supports full shading system. */
bool has_fermi_limits; /* Fixed number of textures limit. */
bool has_half_images; /* Support half-float textures. */
bool has_volume_decoupled; /* Decoupled volume shading. */
BVHLayoutMask bvh_layout_mask; /* Bitmask of supported BVH layouts. */
@@ -73,7 +72,6 @@ public:
cpu_threads = 0;
display_device = false;
advanced_shading = true;
has_fermi_limits = false;
has_half_images = false;
has_volume_decoupled = false;
bvh_layout_mask = BVH_LAYOUT_NONE;

56
intern/cycles/device/device_cuda.cpp
View File

@@ -309,9 +309,7 @@ public:
delete split_kernel;
if(!info.has_fermi_limits) {
texture_info.free();
}
cuda_assert(cuCtxDestroy(cuContext));
}
@@ -680,7 +678,7 @@ public:
void load_texture_info()
{
if(!info.has_fermi_limits && need_texture_info) {
if(need_texture_info) {
texture_info.copy_to_device();
need_texture_info = false;
}
@@ -1018,9 +1016,6 @@ public:
{
CUDAContextScope scope(this);
/* Check if we are on sm_30 or above, for bindless textures. */
bool has_fermi_limits = info.has_fermi_limits;
/* General variables for both architectures */
string bind_name = mem.name;
size_t dsize = datatype_size(mem.data_type);
@@ -1076,25 +1071,6 @@ public:
/* Image Texture Storage */
CUtexref texref = NULL;
if(has_fermi_limits) {
if(mem.data_depth > 1) {
/* Kernel uses different bind names for 2d and 3d float textures,
* so we have to adjust couple of things here.
*/
vector<string> tokens;
string_split(tokens, mem.name, "_");
bind_name = string_printf("__tex_image_%s_3d_%s",
tokens[2].c_str(),
tokens[3].c_str());
}
cuda_assert(cuModuleGetTexRef(&texref, cuModule, bind_name.c_str()));
if(!texref) {
return;
}
}
CUarray_format_enum format;
switch(mem.data_type) {
case TYPE_UCHAR: format = CU_AD_FORMAT_UNSIGNED_INT8; break;
@@ -1187,7 +1163,6 @@ public:
cuda_assert(cuMemcpyHtoD(mem.device_pointer, mem.host_pointer, size));
}
if(!has_fermi_limits) {
/* Kepler+, bindless textures. */
int flat_slot = 0;
if(string_startswith(mem.name, "__tex_image")) {
@@ -1251,34 +1226,6 @@ public:
info.depth = mem.data_depth;
need_texture_info = true;
}
else {
/* Fermi, fixed texture slots. */
if(array_3d) {
cuda_assert(cuTexRefSetArray(texref, array_3d, CU_TRSA_OVERRIDE_FORMAT));
}
else if(mem.data_height > 0) {
CUDA_ARRAY_DESCRIPTOR array_desc;
array_desc.Format = format;
array_desc.Height = mem.data_height;
array_desc.Width = mem.data_width;
array_desc.NumChannels = mem.data_elements;
cuda_assert(cuTexRefSetAddress2D_v3(texref, &array_desc, mem.device_pointer, dst_pitch));
}
else {
cuda_assert(cuTexRefSetAddress(NULL, texref, cuda_device_ptr(mem.device_pointer), size));
}
/* Attach to texture reference. */
cuda_assert(cuTexRefSetFilterMode(texref, filter_mode));
cuda_assert(cuTexRefSetFlags(texref, CU_TRSF_NORMALIZED_COORDINATES));
cuda_assert(cuTexRefSetFormat(texref, format, mem.data_elements));
cuda_assert(cuTexRefSetAddressMode(texref, 0, address_mode));
cuda_assert(cuTexRefSetAddressMode(texref, 1, address_mode));
if(mem.data_depth > 1) {
cuda_assert(cuTexRefSetAddressMode(texref, 2, address_mode));
}
}
}
void tex_free(device_memory& mem)
{
@@ -2545,7 +2492,6 @@ void device_cuda_info(vector<DeviceInfo>& devices)
info.num = num;
info.advanced_shading = (major >= 3);
info.has_fermi_limits = !(major >= 3);
info.has_half_images = (major >= 3);
info.has_volume_decoupled = false;
info.bvh_layout_mask = BVH_LAYOUT_BVH2;

11
intern/cycles/kernel/kernel_compat_cuda.h
View File

@@ -135,18 +135,9 @@ ccl_device_inline uint ccl_num_groups(uint d)
/* Textures */
/* Use arrays for regular data. This is a little slower than textures on Fermi,
* but allows for cleaner code and we will stop supporting Fermi soon. */
/* Use arrays for regular data. */
#define kernel_tex_fetch(t, index) t[(index)]
/* On Kepler (6xx) and above, we use Bindless Textures for images.
* On Fermi cards (4xx and 5xx), we have to use regular textures. */
#if __CUDA_ARCH__ < 300
typedef texture<float4, 2> texture_image_float4;
typedef texture<float4, 3> texture_image3d_float4;
typedef texture<uchar4, 2, cudaReadModeNormalizedFloat> texture_image_uchar4;
#endif
#define kernel_data __data
/* Use fast math functions */

103
intern/cycles/kernel/kernel_textures.h
View File

@@ -78,111 +78,8 @@ KERNEL_TEX(float, __lookup_table)
/* sobol */
KERNEL_TEX(uint, __sobol_directions)
#if !defined(__KERNEL_CUDA__) || __CUDA_ARCH__ >= 300
/* image textures */
KERNEL_TEX(TextureInfo, __texture_info)
#else
/* full-float image */
KERNEL_IMAGE_TEX(float4, texture_image_float4, __tex_image_float4_000)
KERNEL_IMAGE_TEX(float4, texture_image_float4, __tex_image_float4_008)
KERNEL_IMAGE_TEX(float4, texture_image_float4, __tex_image_float4_016)
KERNEL_IMAGE_TEX(float4, texture_image_float4, __tex_image_float4_024)
KERNEL_IMAGE_TEX(float4, texture_image_float4, __tex_image_float4_032)
KERNEL_IMAGE_TEX(float4, texture_image3d_float4, __tex_image_float4_3d_000)
KERNEL_IMAGE_TEX(float4, texture_image3d_float4, __tex_image_float4_3d_008)
KERNEL_IMAGE_TEX(float4, texture_image3d_float4, __tex_image_float4_3d_016)
KERNEL_IMAGE_TEX(float4, texture_image3d_float4, __tex_image_float4_3d_024)
KERNEL_IMAGE_TEX(float4, texture_image3d_float4, __tex_image_float4_3d_032)
/* image
* These texture names are encoded to their flattened slots as
* ImageManager::type_index_to_flattened_slot() returns them. */
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_001)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_009)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_017)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_025)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_033)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_041)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_049)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_057)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_065)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_073)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_081)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_089)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_097)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_105)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_113)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_121)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_129)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_137)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_145)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_153)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_161)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_169)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_177)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_185)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_193)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_201)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_209)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_217)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_225)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_233)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_241)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_249)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_257)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_265)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_273)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_281)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_289)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_297)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_305)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_313)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_321)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_329)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_337)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_345)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_353)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_361)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_369)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_377)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_385)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_393)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_401)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_409)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_417)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_425)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_433)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_441)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_449)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_457)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_465)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_473)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_481)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_489)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_497)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_505)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_513)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_521)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_529)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_537)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_545)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_553)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_561)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_569)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_577)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_585)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_593)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_601)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_609)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_617)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_625)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_633)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_641)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_649)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_657)
KERNEL_IMAGE_TEX(uchar4, texture_image_uchar4, __tex_image_byte4_665)
#endif /* defined(__KERNEL_CUDA__) && __CUDA_ARCH__ < 300 */
#undef KERNEL_TEX
#undef KERNEL_IMAGE_TEX

121
intern/cycles/kernel/kernels/cuda/kernel_cuda_image.h
View File

@@ -14,10 +14,6 @@
* limitations under the License.
*/
#if __CUDA_ARCH__ >= 300
/* Kepler */
/* w0, w1, w2, and w3 are the four cubic B-spline basis functions. */
ccl_device float cubic_w0(float a)
{
@@ -191,120 +187,3 @@ ccl_device float4 kernel_tex_image_interp_3d(KernelGlobals *kg, int id, float x,
}
}
#else
/* Fermi */
ccl_device float4 kernel_tex_image_interp(KernelGlobals *kg, int id, float x, float y)
{
float4 r;
switch(id) {
case 0: r = tex2D(__tex_image_float4_000, x, y); break;
case 8: r = tex2D(__tex_image_float4_008, x, y); break;
case 16: r = tex2D(__tex_image_float4_016, x, y); break;
case 24: r = tex2D(__tex_image_float4_024, x, y); break;
case 32: r = tex2D(__tex_image_float4_032, x, y); break;
case 1: r = tex2D(__tex_image_byte4_001, x, y); break;
case 9: r = tex2D(__tex_image_byte4_009, x, y); break;
case 17: r = tex2D(__tex_image_byte4_017, x, y); break;
case 25: r = tex2D(__tex_image_byte4_025, x, y); break;
case 33: r = tex2D(__tex_image_byte4_033, x, y); break;
case 41: r = tex2D(__tex_image_byte4_041, x, y); break;
case 49: r = tex2D(__tex_image_byte4_049, x, y); break;
case 57: r = tex2D(__tex_image_byte4_057, x, y); break;
case 65: r = tex2D(__tex_image_byte4_065, x, y); break;
case 73: r = tex2D(__tex_image_byte4_073, x, y); break;
case 81: r = tex2D(__tex_image_byte4_081, x, y); break;
case 89: r = tex2D(__tex_image_byte4_089, x, y); break;
case 97: r = tex2D(__tex_image_byte4_097, x, y); break;
case 105: r = tex2D(__tex_image_byte4_105, x, y); break;
case 113: r = tex2D(__tex_image_byte4_113, x, y); break;
case 121: r = tex2D(__tex_image_byte4_121, x, y); break;
case 129: r = tex2D(__tex_image_byte4_129, x, y); break;
case 137: r = tex2D(__tex_image_byte4_137, x, y); break;
case 145: r = tex2D(__tex_image_byte4_145, x, y); break;
case 153: r = tex2D(__tex_image_byte4_153, x, y); break;
case 161: r = tex2D(__tex_image_byte4_161, x, y); break;
case 169: r = tex2D(__tex_image_byte4_169, x, y); break;
case 177: r = tex2D(__tex_image_byte4_177, x, y); break;
case 185: r = tex2D(__tex_image_byte4_185, x, y); break;
case 193: r = tex2D(__tex_image_byte4_193, x, y); break;
case 201: r = tex2D(__tex_image_byte4_201, x, y); break;
case 209: r = tex2D(__tex_image_byte4_209, x, y); break;
case 217: r = tex2D(__tex_image_byte4_217, x, y); break;
case 225: r = tex2D(__tex_image_byte4_225, x, y); break;
case 233: r = tex2D(__tex_image_byte4_233, x, y); break;
case 241: r = tex2D(__tex_image_byte4_241, x, y); break;
case 249: r = tex2D(__tex_image_byte4_249, x, y); break;
case 257: r = tex2D(__tex_image_byte4_257, x, y); break;
case 265: r = tex2D(__tex_image_byte4_265, x, y); break;
case 273: r = tex2D(__tex_image_byte4_273, x, y); break;
case 281: r = tex2D(__tex_image_byte4_281, x, y); break;
case 289: r = tex2D(__tex_image_byte4_289, x, y); break;
case 297: r = tex2D(__tex_image_byte4_297, x, y); break;
case 305: r = tex2D(__tex_image_byte4_305, x, y); break;
case 313: r = tex2D(__tex_image_byte4_313, x, y); break;
case 321: r = tex2D(__tex_image_byte4_321, x, y); break;
case 329: r = tex2D(__tex_image_byte4_329, x, y); break;
case 337: r = tex2D(__tex_image_byte4_337, x, y); break;
case 345: r = tex2D(__tex_image_byte4_345, x, y); break;
case 353: r = tex2D(__tex_image_byte4_353, x, y); break;
case 361: r = tex2D(__tex_image_byte4_361, x, y); break;
case 369: r = tex2D(__tex_image_byte4_369, x, y); break;
case 377: r = tex2D(__tex_image_byte4_377, x, y); break;
case 385: r = tex2D(__tex_image_byte4_385, x, y); break;
case 393: r = tex2D(__tex_image_byte4_393, x, y); break;
case 401: r = tex2D(__tex_image_byte4_401, x, y); break;
case 409: r = tex2D(__tex_image_byte4_409, x, y); break;
case 417: r = tex2D(__tex_image_byte4_417, x, y); break;
case 425: r = tex2D(__tex_image_byte4_425, x, y); break;
case 433: r = tex2D(__tex_image_byte4_433, x, y); break;
case 441: r = tex2D(__tex_image_byte4_441, x, y); break;
case 449: r = tex2D(__tex_image_byte4_449, x, y); break;
case 457: r = tex2D(__tex_image_byte4_457, x, y); break;
case 465: r = tex2D(__tex_image_byte4_465, x, y); break;
case 473: r = tex2D(__tex_image_byte4_473, x, y); break;
case 481: r = tex2D(__tex_image_byte4_481, x, y); break;
case 489: r = tex2D(__tex_image_byte4_489, x, y); break;
case 497: r = tex2D(__tex_image_byte4_497, x, y); break;
case 505: r = tex2D(__tex_image_byte4_505, x, y); break;
case 513: r = tex2D(__tex_image_byte4_513, x, y); break;
case 521: r = tex2D(__tex_image_byte4_521, x, y); break;
case 529: r = tex2D(__tex_image_byte4_529, x, y); break;
case 537: r = tex2D(__tex_image_byte4_537, x, y); break;
case 545: r = tex2D(__tex_image_byte4_545, x, y); break;
case 553: r = tex2D(__tex_image_byte4_553, x, y); break;
case 561: r = tex2D(__tex_image_byte4_561, x, y); break;
case 569: r = tex2D(__tex_image_byte4_569, x, y); break;
case 577: r = tex2D(__tex_image_byte4_577, x, y); break;
case 585: r = tex2D(__tex_image_byte4_585, x, y); break;
case 593: r = tex2D(__tex_image_byte4_593, x, y); break;
case 601: r = tex2D(__tex_image_byte4_601, x, y); break;
case 609: r = tex2D(__tex_image_byte4_609, x, y); break;
case 617: r = tex2D(__tex_image_byte4_617, x, y); break;
case 625: r = tex2D(__tex_image_byte4_625, x, y); break;
case 633: r = tex2D(__tex_image_byte4_633, x, y); break;
case 641: r = tex2D(__tex_image_byte4_641, x, y); break;
case 649: r = tex2D(__tex_image_byte4_649, x, y); break;
case 657: r = tex2D(__tex_image_byte4_657, x, y); break;
case 665: r = tex2D(__tex_image_byte4_665, x, y); break;
default: r = make_float4(0.0f, 0.0f, 0.0f, 0.0f);
}
return r;
}
ccl_device float4 kernel_tex_image_interp_3d(KernelGlobals *kg, int id, float x, float y, float z, InterpolationType interp)
{
float4 r;
switch(id) {
case 0: r = tex3D(__tex_image_float4_3d_000, x, y, z); break;
case 8: r = tex3D(__tex_image_float4_3d_008, x, y, z); break;
case 16: r = tex3D(__tex_image_float4_3d_016, x, y, z); break;
case 24: r = tex3D(__tex_image_float4_3d_024, x, y, z); break;
case 32: r = tex3D(__tex_image_float4_3d_032, x, y, z); break;
}
return r;
}
#endif

27
intern/cycles/render/image.cpp
View File

@@ -49,7 +49,6 @@ ImageManager::ImageManager(const DeviceInfo& info)
/* Set image limits */
max_num_images = TEX_NUM_MAX;
has_half_images = info.has_half_images;
cuda_fermi_limits = info.has_fermi_limits;
for(size_t type = 0; type < IMAGE_DATA_NUM_TYPES; type++) {
tex_num_images[type] = 0;
@@ -255,7 +254,7 @@ int ImageManager::add_image(const string& filename,
/* Check whether it's a float texture. */
is_float = (type == IMAGE_DATA_TYPE_FLOAT || type == IMAGE_DATA_TYPE_FLOAT4);
/* No single channel and half textures on CUDA (Fermi) and no half on OpenCL, use available slots */
/* No half textures on OpenCL, use available slots */
if(!has_half_images) {
if(type == IMAGE_DATA_TYPE_HALF4) {
type = IMAGE_DATA_TYPE_FLOAT4;
@@ -265,15 +264,6 @@ int ImageManager::add_image(const string& filename,
}
}
if(cuda_fermi_limits) {
if(type == IMAGE_DATA_TYPE_FLOAT) {
type = IMAGE_DATA_TYPE_FLOAT4;
}
else if(type == IMAGE_DATA_TYPE_BYTE) {
type = IMAGE_DATA_TYPE_BYTE4;
}
}
/* Fnd existing image. */
for(slot = 0; slot < images[type].size(); slot++) {
img = images[type][slot];
@@ -303,18 +293,8 @@ int ImageManager::add_image(const string& filename,
break;
}
/* Count if we're over the limit */
if(cuda_fermi_limits) {
if(tex_num_images[IMAGE_DATA_TYPE_BYTE4] == TEX_NUM_BYTE4_CUDA
|| tex_num_images[IMAGE_DATA_TYPE_FLOAT4] == TEX_NUM_FLOAT4_CUDA)
{
printf("ImageManager::add_image: Reached %s image limit (%d), skipping '%s'\n",
name_from_type(type).c_str(), tex_num_images[type], filename.c_str());
return -1;
}
}
else {
/* Very unlikely, since max_num_images is insanely big. But better safe than sorry. */
/* Count if we're over the limit.
* Very unlikely, since max_num_images is insanely big. But better safe than sorry. */
int tex_count = 0;
for(int type = 0; type < IMAGE_DATA_NUM_TYPES; type++) {
tex_count += tex_num_images[type];
@@ -324,7 +304,6 @@ int ImageManager::add_image(const string& filename,
max_num_images, filename.c_str());
return -1;
}
}
if(slot == images[type].size()) {
images[type].resize(images[type].size() + 1);

1
intern/cycles/render/image.h
View File

@@ -121,7 +121,6 @@ private:
int tex_num_images[IMAGE_DATA_NUM_TYPES];
int max_num_images;
bool has_half_images;
bool cuda_fermi_limits;
thread_mutex device_mutex;
int animation_frame;

21
intern/cycles/util/util_texture.h
View File

@@ -20,22 +20,6 @@
CCL_NAMESPACE_BEGIN
/* Texture limits on devices. */
/* CUDA (Geforce 4xx and 5xx) */
#define TEX_NUM_FLOAT4_CUDA 5
#define TEX_NUM_BYTE4_CUDA 84
#define TEX_NUM_HALF4_CUDA 0
#define TEX_NUM_FLOAT_CUDA 0
#define TEX_NUM_BYTE_CUDA 0
#define TEX_NUM_HALF_CUDA 0
#define TEX_START_FLOAT4_CUDA 0
#define TEX_START_BYTE4_CUDA TEX_NUM_FLOAT4_CUDA
#define TEX_START_HALF4_CUDA (TEX_NUM_FLOAT4_CUDA + TEX_NUM_BYTE4_CUDA)
#define TEX_START_FLOAT_CUDA (TEX_NUM_FLOAT4_CUDA + TEX_NUM_BYTE4_CUDA + TEX_NUM_HALF4_CUDA)
#define TEX_START_BYTE_CUDA (TEX_NUM_FLOAT4_CUDA + TEX_NUM_BYTE4_CUDA + TEX_NUM_HALF4_CUDA + TEX_NUM_FLOAT_CUDA)
#define TEX_START_HALF_CUDA (TEX_NUM_FLOAT4_CUDA + TEX_NUM_BYTE4_CUDA + TEX_NUM_HALF4_CUDA + TEX_NUM_FLOAT_CUDA + TEX_NUM_BYTE_CUDA)
/* Any architecture other than old CUDA cards */
#define TEX_NUM_MAX (INT_MAX >> 4)
/* Color to use when textures are not found. */
@@ -44,11 +28,8 @@ CCL_NAMESPACE_BEGIN
#define TEX_IMAGE_MISSING_B 1
#define TEX_IMAGE_MISSING_A 1
#if defined (__KERNEL_CUDA__) && (__CUDA_ARCH__ < 300)
# define kernel_tex_type(tex) (tex < TEX_START_BYTE4_CUDA ? IMAGE_DATA_TYPE_FLOAT4 : IMAGE_DATA_TYPE_BYTE4)
#else
/* Texture type. */
#define kernel_tex_type(tex) (tex & IMAGE_DATA_TYPE_MASK)
#endif
/* Interpolation types for textures
* cuda also use texture space to store other objects */