4
$\begingroup$

I am looking to run a Blender workload in Amazon Web Services (AWS). I am testing on a Windows Server image on a non-GPU compute resource. Instead of GPU I am attaching AWS Elastic Graphics to the instance. This provides OpenGL 4.3. This is working really well for another workload which uses Unity (in which case I need to force OpenGL mode).

When I open blender.exe (2.83), it immediately exits. Running it from the command line I am able to capture the following output, which indicates some sort of shader error on startup.

Edit: 2.79b launches just fine, so this seems to be a new bug introduced. We need to export to gltf so I'd like to be on the latest version.

I realize this may be an unusual case, but does anyone have any ideas on how I might get blender to run? Perhaps some command line settings? Thanks in advance!

C:\Program Files\Blender Foundation\Blender 2.83>blender
AL lib: (EE) ALCmmdevPlayback_open: Device init failed: 0x80070490
Warning: Could not find a matching GPU name. Things may not behave as expected.
Detected OpenGL configuration:
Vendor: Amazon Web Services, Inc.
Renderer: Amazon Elastic Graphics (eg1.medium)
found bundled python: C:\Program Files\Blender Foundation\Blender 2.83\2.83\python
GPUShader: compile error:
Vertex shader failed to compile with the following errors:
ERROR: 5:445: error(#438) Extension GL_ARB_shader_draw_parameters is requried' by gl_BaseInstanceARB'.
ERROR: 5:445: error(#102) #error: Builtin Name not supported
ERROR: 5:454: error(#438) Extension GL_ARB_shader_draw_parameters is requried' by gl_BaseInstanceARB'.
ERROR: 5:454: error(#102) #error: Builtin Name not supported
ERROR: 5:459: error(#438) Extension GL_ARB_shader_draw_parameters is requried' by gl_BaseInstanceARB'.
ERROR: 5:459: error(#102) #error: Builtin Name not supported
ERROR: 5:469: error(#438) Extension GL_ARB_shader_draw_parameters is requried' by gl_BaseInstanceARB'.
ERROR: 5:469: error(#102) #error: Builtin Name not supported
ERROR: error(#273) 8 compilation errors.  No code generated


Error   : EXCEPTION_ACCESS_VIOLATION
Address : 0x00007FF7CA4E8774
Module  : C:\Program Files\Blender Foundation\Blender 2.83\blender.exe

Here is out from using --debug-all in case that is helpful:

AL lib: (EE) ALCmmdevPlayback_open: Device init failed: 0x80070490
Switching to fully guarded memory allocator.
read file 
  Version 280 sub 39 date unknown hash unknown
DEG_relations_tag_update: Tagging relations for update.
DEG_relations_tag_update: Tagging relations for update.
DEG_relations_tag_update: Tagging relations for update.
DEG_relations_tag_update: Tagging relations for update.
DEG_relations_tag_update: Tagging relations for update.
graph_id_tag_update: id=SCScene flags=LEGACY_0 source=USER_EDIT
Warning: Could not find a matching GPU name. Things may not behave as expected.
Detected OpenGL configuration:
Vendor: Amazon Web Services, Inc.
Renderer: Amazon Elastic Graphics (eg1.medium)
ED_screen_refresh: set screen
[SCScene :: View Layer]: Removed 31 relations to no-op nodes
graph_id_tag_update: id=SCScene flags=TRANSFORM, GEOMETRY, SHADING, SELECT, BASE_FLAGS, POINT_CACHE, COPY_ON_WRITE, SEQUENCER_STRIPS, AUDIO_SEEK, AUDIO_FPS, AUDIO_VOLUME, AUDIO_MUTE, AUDIO_LISTENER, AUDIO, PARAMETERS source=RELATIONS
graph_id_tag_update: id=OBCube flags=TRANSFORM, GEOMETRY, COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=MECube flags=LEGACY_0 source=RELATIONS
graph_id_tag_update: id=MECube flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=MAMaterial flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=OBLight flags=TRANSFORM, GEOMETRY, COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=LALight flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=OBCamera flags=TRANSFORM, GEOMETRY, COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=CACamera flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=GRMaster Collection flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=GRCollection flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=WOWorld flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=LSLineStyle flags=COPY_ON_WRITE source=RELATIONS
graph_id_tag_update: id=SCScene flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=OBCube flags=TRANSFORM, GEOMETRY, COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=MECube flags=LEGACY_0 source=VISIBILITY
graph_id_tag_update: id=MECube flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=MAMaterial flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=OBLight flags=TRANSFORM, GEOMETRY, COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=LALight flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=OBCamera flags=TRANSFORM, GEOMETRY, COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=CACamera flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=GRMaster Collection flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=GRCollection flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=WOWorld flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=LSLineStyle flags=COPY_ON_WRITE source=VISIBILITY
Depsgraph built in 0.000456 seconds.
graph_id_tag_update: id=SCScene flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=OBCube flags=TRANSFORM, GEOMETRY, COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=MECube flags=LEGACY_0 source=VISIBILITY
graph_id_tag_update: id=MECube flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=MAMaterial flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=OBLight flags=TRANSFORM, GEOMETRY, COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=LALight flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=OBCamera flags=TRANSFORM, GEOMETRY, COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=CACamera flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=GRMaster Collection flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=GRCollection flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=WOWorld flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=NTShader Nodetree flags=COPY_ON_WRITE source=VISIBILITY
graph_id_tag_update: id=LSLineStyle flags=COPY_ON_WRITE source=VISIBILITY
[SCScene :: View Layer]: ... (output removed to save space)
I0723 23:12:13.248817  6420 blender_python.cpp:191] Debug flags initialized to:
CPU flags:
  AVX2       : True
  AVX        : True
  SSE4.1     : True
  SSE3       : True
  SSE2       : True
  BVH layout : BVH8
  Split      : False
CUDA flags:
  Adaptive Compile : False
OptiX flags:
  CUDA streams : 1
OpenCL flags:
  Device type    : ALL
  Debug          : False
  Memory limit   : 0
Depsgraph updated in 0.000455 seconds.
Depsgraph evaluation FPS: -nan(ind)
found bundled python: C:\Program Files\Blender Foundation\Blender 2.83\2.83\python
FRS_set_context: context 0x000002ECFED695F8 scene 0x000002EC821785B8
GPUShader: compile error:
===== shader string 1 ====
 1  #version 330
===== shader string 2 ====
 2  #define GPU_VERTEX_SHADER
===== shader string 3 ====
 3  #extension GL_ARB_texture_gather: enable
 4  #define GPU_ARB_texture_gather
 5  #extension GL_ARB_texture_query_lod: enable
 6  #extension GL_ARB_texture_cube_map_array : enable
 7  #define GPU_ARB_texture_cube_map_array
===== shader string 4 ====
 8  #define GPU_ATI
 9  #define OS_WIN
10  #define DFDX_SIGN 1.0
11  #define DFDY_SIGN 1.0
===== shader string 5 ====
12  #define blender_srgb_to_framebuffer_space(a) a
13  #define V3D_LIGHTING_STUDIO
14  #define WORKBENCH_ENCODE_NORMALS
15  #define OPAQUE_MATERIAL
===== shader string 6 ====
16  #define COMMON_VIEW_LIB
17  #define DRW_RESOURCE_CHUNK_LEN 512
18  
19  /* keep in sync with DRWManager.view_data */
20  layout(std140) uniform viewBlock
21  {
22    /* Same order as DRWViewportMatrixType */
23    mat4 ViewProjectionMatrix;
24    mat4 ViewProjectionMatrixInverse;
25    mat4 ViewMatrix;
26    mat4 ViewMatrixInverse;
27    mat4 ProjectionMatrix;
28    mat4 ProjectionMatrixInverse;
29  
30    vec4 clipPlanes[6];
31  
32    /* TODO move it elsewhere. */
33    vec4 CameraTexCoFactors;
34  };
35  
36  #ifdef world_clip_planes_calc_clip_distance
37  #  undef world_clip_planes_calc_clip_distance
38  #  define world_clip_planes_calc_clip_distance(p) \
39      _world_clip_planes_calc_clip_distance(p, clipPlanes)
40  #endif
41  
42  #ifdef COMMON_GLOBALS_LIB
43  float mul_project_m4_v3_zfac(in vec3 co)
44  {
45    return pixelFac * ((ViewProjectionMatrix[0][3] * co.x) + (ViewProjectionMatrix[1][3] * co.y) +
46                       (ViewProjectionMatrix[2][3] * co.z) + ViewProjectionMatrix[3][3]);
47  }
48  #endif
49  
50  /* Not the right place but need to be common to all overlay's.
51   * TODO Split to an overlay lib. */
52  mat4 extract_matrix_packed_data(mat4 mat, out vec4 dataA, out vec4 dataB)
53  {
54    const float div = 1.0 / 255.0;
55    int a = int(mat[0][3]);
56    int b = int(mat[1][3]);
57    int c = int(mat[2][3]);
58    int d = int(mat[3][3]);
59    dataA = vec4(a & 0xFF, a >> 8, b & 0xFF, b >> 8) * div;
60    dataB = vec4(c & 0xFF, c >> 8, d & 0xFF, d >> 8) * div;
61    mat[0][3] = mat[1][3] = mat[2][3] = 0.0;
62    mat[3][3] = 1.0;
63    return mat;
64  }
65  
66  /* Same here, Not the right place but need to be common to all overlay's.
67   * TODO Split to an overlay lib. */
68  /* edge_start and edge_pos needs to be in the range [0..sizeViewport]. */
69  vec4 pack_line_data(vec2 frag_co, vec2 edge_start, vec2 edge_pos)
70  {
71    vec2 edge = edge_start - edge_pos;
72    float len = length(edge);
73    if (len > 0.0) {
74      edge /= len;
75      vec2 perp = vec2(-edge.y, edge.x);
76      float dist = dot(perp, frag_co - edge_start);
77      /* Add 0.1 to diffenrentiate with cleared pixels. */
78      return vec4(perp * 0.5 + 0.5, dist * 0.25 + 0.5 + 0.1, 1.0);
79    }
80    else {
81      /* Default line if the origin is perfectly aligned with a pixel. */
82      return vec4(1.0, 0.0, 0.5 + 0.1, 1.0);
83    }
84  }
85  
86  uniform int resourceChunk;
87  
88  #ifdef GPU_VERTEX_SHADER
89  #  ifdef GL_ARB_shader_draw_parameters
90  #    define baseInstance gl_BaseInstanceARB
91  #  else /* no ARB_shader_draw_parameters */
92  uniform int baseInstance;
93  #  endif
94  
95  #  if defined(IN_PLACE_INSTANCES) || defined(INSTANCED_ATTR)
96  /* When drawing instances of an object at the same position. */
97  #    define instanceId 0
98  #  elif defined(GPU_DEPRECATED_AMD_DRIVER)
99  /* A driver bug make it so that when using an attribute with GL_INT_2_10_10_10_REV as format,
100   * the gl_InstanceID is incremented by the 2 bit component of the attribute.
101   * Ignore gl_InstanceID then. */
102  #    define instanceId 0
103  #  else
104  #    define instanceId gl_InstanceID
105  #  endif
106  
107  #  ifdef UNIFORM_RESOURCE_ID
108  /* This is in the case we want to do a special instance drawcall but still want to have the
109   * right resourceId and all the correct ubo datas. */
110  uniform int resourceId;
111  #    define resource_id resourceId
112  #  else
113  #    define resource_id (baseInstance + instanceId)
114  #  endif
115  
116  /* Use this to declare and pass the value if
117   * the fragment shader uses the resource_id. */
118  #  define RESOURCE_ID_VARYING flat out int resourceIDFrag;
119  #  define RESOURCE_ID_VARYING_GEOM flat out int resourceIDGeom;
120  #  define PASS_RESOURCE_ID resourceIDFrag = resource_id;
121  #  define PASS_RESOURCE_ID_GEOM resourceIDGeom = resource_id;
122  #endif
123  
124  /* If used in a fragment / geometry shader, we pass
125   * resource_id as varying. */
126  #ifdef GPU_GEOMETRY_SHADER
127  #  define RESOURCE_ID_VARYING \
128      flat out int resourceIDFrag; \
129      flat in int resourceIDGeom[];
130  
131  #  define resource_id resourceIDGeom
132  #  define PASS_RESOURCE_ID(i) resourceIDFrag = resource_id[i];
133  #endif
134  
135  #ifdef GPU_FRAGMENT_SHADER
136  flat in int resourceIDFrag;
137  #  define resource_id resourceIDFrag
138  #endif
139  
140  /* Breaking this across multiple lines causes issues for some older GLSL compilers. */
141  /* clang-format off */
142  #if !defined(GPU_INTEL) && !defined(GPU_DEPRECATED_AMD_DRIVER) && !defined(OS_MAC) && !defined(INSTANCED_ATTR)
143  /* clang-format on */
144  struct ObjectMatrices {
145    mat4 drw_modelMatrix;
146    mat4 drw_modelMatrixInverse;
147  };
148  
149  layout(std140) uniform modelBlock
150  {
151    ObjectMatrices drw_matrices[DRW_RESOURCE_CHUNK_LEN];
152  };
153  
154  #  define ModelMatrix (drw_matrices[resource_id].drw_modelMatrix)
155  #  define ModelMatrixInverse (drw_matrices[resource_id].drw_modelMatrixInverse)
156  
157  #else /* GPU_INTEL */
158  /* Intel GPU seems to suffer performance impact when the model matrix is in UBO storage.
159   * So for now we just force using the legacy path. */
160  /* Note that this is also a workaround of a problem on osx (amd or nvidia)
161   * and older amd driver on windows. */
162  uniform mat4 ModelMatrix;
163  uniform mat4 ModelMatrixInverse;
164  #endif
165  
166  #define resource_handle (resourceChunk * DRW_RESOURCE_CHUNK_LEN + resource_id)
167  
168  /** Transform shortcuts. */
169  /* Rule of thumb: Try to reuse world positions and normals because converting though viewspace
170   * will always be decomposed in at least 2 matrix operation. */
171  
172  /**
173   * Some clarification:
174   * Usually Normal matrix is transpose(inverse(ViewMatrix * ModelMatrix))
175   *
176   * But since it is slow to multiply matrices we decompose it. Decomposing
177   * inversion and transposition both invert the product order leaving us with
178   * the same original order:
179   * transpose(ViewMatrixInverse) * transpose(ModelMatrixInverse)
180   *
181   * Knowing that the view matrix is orthogonal, the transpose is also the inverse.
182   * Note: This is only valid because we are only using the mat3 of the ViewMatrixInverse.
183   * ViewMatrix * transpose(ModelMatrixInverse)
184   **/
185  #define normal_object_to_view(n) (mat3(ViewMatrix) * (transpose(mat3(ModelMatrixInverse)) * n))
186  #define normal_object_to_world(n) (transpose(mat3(ModelMatrixInverse)) * n)
187  #define normal_world_to_object(n) (transpose(mat3(ModelMatrix)) * n)
188  #define normal_world_to_view(n) (mat3(ViewMatrix) * n)
189  
190  #define point_object_to_ndc(p) (ViewProjectionMatrix * vec4((ModelMatrix * vec4(p, 1.0)).xyz, 1.0))
191  #define point_object_to_view(p) ((ViewMatrix * vec4((ModelMatrix * vec4(p, 1.0)).xyz, 1.0)).xyz)
192  #define point_object_to_world(p) ((ModelMatrix * vec4(p, 1.0)).xyz)
193  #define point_view_to_ndc(p) (ProjectionMatrix * vec4(p, 1.0))
194  #define point_view_to_object(p) ((ModelMatrixInverse * (ViewMatrixInverse * vec4(p, 1.0))).xyz)
195  #define point_view_to_world(p) ((ViewMatrixInverse * vec4(p, 1.0)).xyz)
196  #define point_world_to_ndc(p) (ViewProjectionMatrix * vec4(p, 1.0))
197  #define point_world_to_object(p) ((ModelMatrixInverse * vec4(p, 1.0)).xyz)
198  #define point_world_to_view(p) ((ViewMatrix * vec4(p, 1.0)).xyz)
199  
200  /* Due to some shader compiler bug, we somewhat need to access gl_VertexID
201   * to make vertex shaders work. even if it's actually dead code. */
202  #ifdef GPU_INTEL
203  #  define GPU_INTEL_VERTEX_SHADER_WORKAROUND gl_Position.x = float(gl_VertexID);
204  #else
205  #  define GPU_INTEL_VERTEX_SHADER_WORKAROUND
206  #endif
207  
208  #define DRW_BASE_SELECTED (1 << 1)
209  #define DRW_BASE_FROM_DUPLI (1 << 2)
210  #define DRW_BASE_FROM_SET (1 << 3)
211  #define DRW_BASE_ACTIVE (1 << 4)
212  
213  #ifdef GPU_VERTEX_SHADER
214  #  define IN_OUT out
215  #else
216  #  define IN_OUT in
217  #endif
218  
219  IN_OUT ShaderStageInterface
220  {
221    vec3 normal_interp;
222    vec3 color_interp;
223    float alpha_interp;
224    vec2 uv_interp;
225  #ifdef TRANSPARENT_MATERIAL
226    flat float roughness;
227    flat float metallic;
228  #else
229    flat float packed_rough_metal;
230  #endif
231    flat int object_id;
232  };
233  
234  #define EPSILON 0.00001
235  #define M_PI 3.14159265358979323846
236  
237  #define CAVITY_BUFFER_RANGE 4.0
238  
239  #ifdef WORKBENCH_ENCODE_NORMALS
240  
241  #  define WB_Normal vec2
242  
243  /* From http://aras-p.info/texts/CompactNormalStorage.html
244   * Using Method #4: Spheremap Transform */
245  vec3 workbench_normal_decode(vec4 enc)
246  {
247    vec2 fenc = enc.xy * 4.0 - 2.0;
248    float f = dot(fenc, fenc);
249    float g = sqrt(1.0 - f / 4.0);
250    vec3 n;
251    n.xy = fenc * g;
252    n.z = 1 - f / 2;
253    return n;
254  }
255  
256  /* From http://aras-p.info/texts/CompactNormalStorage.html
257   * Using Method #4: Spheremap Transform */
258  WB_Normal workbench_normal_encode(bool front_face, vec3 n)
259  {
260    n = normalize(front_face ? n : -n);
261    float p = sqrt(n.z * 8.0 + 8.0);
262    n.xy = clamp(n.xy / p + 0.5, 0.0, 1.0);
263    return n.xy;
264  }
265  
266  #else
267  #  define WB_Normal vec3
268  /* Well just do nothing... */
269  #  define workbench_normal_encode(f, a) (a)
270  #  define workbench_normal_decode(a) (a.xyz)
271  #endif /* WORKBENCH_ENCODE_NORMALS */
272  
273  /* Encoding into the alpha of a RGBA16F texture. (10bit mantissa) */
274  #define TARGET_BITCOUNT 8u
275  #define METALLIC_BITS 3u /* Metallic channel is less important. */
276  #define ROUGHNESS_BITS (TARGET_BITCOUNT - METALLIC_BITS)
277  
278  /* Encode 2 float into 1 with the desired precision. */
279  float workbench_float_pair_encode(float v1, float v2)
280  {
281    // const uint v1_mask = ~(0xFFFFFFFFu << ROUGHNESS_BITS);
282    // const uint v2_mask = ~(0xFFFFFFFFu << METALLIC_BITS);
283    /* Same as above because some compiler are dumb af. and think we use mediump int.  */
284    const int v1_mask = 0x1F;
285    const int v2_mask = 0x7;
286    int iv1 = int(v1 * float(v1_mask));
287    int iv2 = int(v2 * float(v2_mask)) << int(ROUGHNESS_BITS);
288    return float(iv1 | iv2);
289  }
290  
291  void workbench_float_pair_decode(float data, out float v1, out float v2)
292  {
293    // const uint v1_mask = ~(0xFFFFFFFFu << ROUGHNESS_BITS);
294    // const uint v2_mask = ~(0xFFFFFFFFu << METALLIC_BITS);
295    /* Same as above because some compiler are dumb af. and think we use mediump int.  */
296    const int v1_mask = 0x1F;
297    const int v2_mask = 0x7;
298    int idata = int(data);
299    v1 = float(idata & v1_mask) * (1.0 / float(v1_mask));
300    v2 = float(idata >> int(ROUGHNESS_BITS)) * (1.0 / float(v2_mask));
301  }
302  
303  vec3 view_vector_from_screen_uv(vec2 uv, vec4 viewvecs[3], mat4 proj_mat)
304  {
305    if (proj_mat[3][3] == 0.0) {
306      return normalize(viewvecs[0].xyz + vec3(uv, 0.0) * viewvecs[1].xyz);
307    }
308    else {
309      return vec3(0.0, 0.0, 1.0);
310    }
311  }
312  
313  vec3 view_position_from_depth(vec2 uvcoords, float depth, vec4 viewvecs[3], mat4 proj_mat)
314  {
315    if (proj_mat[3][3] == 0.0) {
316      /* Perspective */
317      float d = 2.0 * depth - 1.0;
318  
319      float zview = -proj_mat[3][2] / (d + proj_mat[2][2]);
320  
321      return zview * (viewvecs[0].xyz + vec3(uvcoords, 0.0) * viewvecs[1].xyz);
322    }
323    else {
324      /* Orthographic */
325      vec3 offset = vec3(uvcoords, depth);
326  
327      return viewvecs[0].xyz + offset * viewvecs[1].xyz;
328    }
329  }
330  
331  /* TODO(fclem) deduplicate code.  */
332  bool node_tex_tile_lookup(inout vec3 co, sampler2DArray ima, sampler1DArray map)
333  {
334    vec2 tile_pos = floor(co.xy);
335  
336    if (tile_pos.x < 0 || tile_pos.y < 0 || tile_pos.x >= 10)
337      return false;
338  
339    float tile = 10.0 * tile_pos.y + tile_pos.x;
340    if (tile >= textureSize(map, 0).x)
341      return false;
342  
343    /* Fetch tile information. */
344    float tile_layer = texelFetch(map, ivec2(tile, 0), 0).x;
345    if (tile_layer < 0.0)
346      return false;
347  
348    vec4 tile_info = texelFetch(map, ivec2(tile, 1), 0);
349  
350    co = vec3(((co.xy - tile_pos) * tile_info.zw) + tile_info.xy, tile_layer);
351    return true;
352  }
353  
354  vec4 workbench_sample_texture(sampler2D image, vec2 coord, bool nearest_sampling)
355  {
356    /* TODO(fclem) We could do the same with sampler objects.
357     * But this is a quick workaround instead of messing with the GPUTexture itself. */
358    if (nearest_sampling) {
359      /* Use texelFetch for nearest_sampling to reduce glitches. See: T73726 */
360      vec2 tex_size = vec2(textureSize(image, 0).xy);
361      ivec2 uv = ivec2(floor(coord * tex_size) + 0.5);
362      return texelFetch(image, uv, 0);
363    }
364    else {
365      return texture(image, coord);
366    }
367  }
368  
369  vec4 workbench_sample_texture_array(sampler2DArray tile_array,
370                                      sampler1DArray tile_data,
371                                      vec2 coord,
372                                      bool nearest_sampling)
373  {
374  
375    vec3 uv = vec3(coord, 0);
376    if (!node_tex_tile_lookup(uv, tile_array, tile_data))
377      return vec4(1.0, 0.0, 1.0, 1.0);
378  
379    /* TODO(fclem) We could do the same with sampler objects.
380     * But this is a quick workaround instead of messing with the GPUTexture itself. */
381    if (nearest_sampling) {
382      /* Use texelFetch for nearest_sampling to reduce glitches. See: T73726 */
383      vec3 tex_size = vec3(textureSize(tile_array, 0));
384      uv.xy = floor(uv.xy * tex_size.xy) + 0.5;
385      return texelFetch(tile_array, ivec3(uv), 0);
386    }
387    else {
388      return texture(tile_array, uv);
389    }
390  }
391  
392  uniform sampler2DArray imageTileArray;
393  uniform sampler1DArray imageTileData;
394  uniform sampler2D imageTexture;
395  
396  uniform float imageTransparencyCutoff = 0.1;
397  uniform bool imageNearest;
398  uniform bool imagePremult;
399  
400  vec3 workbench_image_color(vec2 uvs)
401  {
402  #ifdef V3D_SHADING_TEXTURE_COLOR
403  #  ifdef TEXTURE_IMAGE_ARRAY
404    vec4 color = workbench_sample_texture_array(imageTileArray, imageTileData, uvs, imageNearest);
405  #  else
406    vec4 color = workbench_sample_texture(imageTexture, uvs, imageNearest);
407  #  endif
408  
409    /* Unpremultiply if stored multiplied, since straight alpha is expected by shaders. */
410    if (imagePremult && !(color.a == 0.0 || color.a == 1.0)) {
411      color.rgb /= color.a;
412    }
413  
414  #  ifdef GPU_FRAGMENT_SHADER
415    if (color.a < imageTransparencyCutoff) {
416      discard;
417    }
418  #  endif
419  
420    return color.rgb;
421  #else
422    return vec3(1.0);
423  #endif
424  }
425  
426  layout(std140) uniform material_block
427  {
428    vec4 mat_data[4096];
429  };
430  
431  /* If set to -1, the resource handle is used instead. */
432  uniform int materialIndex;
433  
434  void workbench_material_data_get(
435      int handle, out vec3 color, out float alpha, out float roughness, out float metallic)
436  {
437    handle = (materialIndex != -1) ? materialIndex : handle;
438    vec4 data = mat_data[uint(handle) & 0xFFFu];
439    color = data.rgb;
440  
441    uint encoded_data = floatBitsToUint(data.w);
442    alpha = float((encoded_data >> 16u) & 0xFFu) * (1.0 / 255.0);
443    roughness = float((encoded_data >> 8u) & 0xFFu) * (1.0 / 255.0);
444    metallic = float(encoded_data & 0xFFu) * (1.0 / 255.0);
445  }
446  
447  #pragma BLENDER_REQUIRE(common_view_lib.glsl)
448  #pragma BLENDER_REQUIRE(workbench_shader_interface_lib.glsl)
449  #pragma BLENDER_REQUIRE(workbench_common_lib.glsl)
450  #pragma BLENDER_REQUIRE(workbench_material_lib.glsl)
451  #pragma BLENDER_REQUIRE(workbench_image_lib.glsl)
452  
453  in vec3 pos;
454  in vec3 nor;
455  in vec4 ac; /* active color */
456  in vec2 au; /* active texture layer */
457  
458  void main()
459  {
460    vec3 world_pos = point_object_to_world(pos);
461    gl_Position = point_world_to_ndc(world_pos);
462  
463  #ifdef USE_WORLD_CLIP_PLANES
464    world_clip_planes_calc_clip_distance(world_pos);
465  #endif
466  
467    uv_interp = au;
468  
469    normal_interp = normalize(normal_object_to_view(nor));
470  
471  #ifdef OPAQUE_MATERIAL
472    float metallic, roughness;
473  #endif
474    workbench_material_data_get(resource_handle, color_interp, alpha_interp, roughness, metallic);
475  
476    if (materialIndex == 0) {
477      color_interp = ac.rgb;
478    }
479  
480  #ifdef OPAQUE_MATERIAL
481    packed_rough_metal = workbench_float_pair_encode(roughness, metallic);
482  #endif
483  
484    object_id = int((uint(resource_id) + 1u) & 0xFFu);
485  }
Vertex shader failed to compile with the following errors:
ERROR: 5:445: error(#438) Extension GL_ARB_shader_draw_parameters is requried' by gl_BaseInstanceARB'.
ERROR: 5:445: error(#102) #error: Builtin Name not supported
ERROR: 5:454: error(#438) Extension GL_ARB_shader_draw_parameters is requried' by gl_BaseInstanceARB'.
ERROR: 5:454: error(#102) #error: Builtin Name not supported
ERROR: 5:459: error(#438) Extension GL_ARB_shader_draw_parameters is requried' by gl_BaseInstanceARB'.
ERROR: 5:459: error(#102) #error: Builtin Name not supported
ERROR: 5:469: error(#438) Extension GL_ARB_shader_draw_parameters is requried' by gl_BaseInstanceARB'.
ERROR: 5:469: error(#102) #error: Builtin Name not supported
ERROR: error(#273) 8 compilation errors.  No code generated


Error   : EXCEPTION_ACCESS_VIOLATION
Address : 0x00007FF7CA4E8774
Module  : C:\Program Files\Blender Foundation\Blender 2.83\blender.exe
$\endgroup$
1
  • 1
    $\begingroup$ This was also reported on the bug tracker in T79246 and fixed. It should work properly in current versions of Blender. $\endgroup$ – Robert Gützkow Dec 6 '20 at 12:36

Browse other questions tagged or ask your own question.