1 /* 2 * Copyright © 2012 Intel Corporation 3 * 4 * Permission is hereby granted, free of charge, to any person obtaining a 5 * copy of this software and associated documentation files (the "Software"), 6 * to deal in the Software without restriction, including without limitation 7 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 8 * and/or sell copies of the Software, and to permit persons to whom the 9 * Software is furnished to do so, subject to the following conditions: 10 * 11 * The above copyright notice and this permission notice (including the next 12 * paragraph) shall be included in all copies or substantial portions of the 13 * Software. 14 * 15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 21 * IN THE SOFTWARE. 22 */ 23 24 #ifndef _GLTHREAD_H 25 #define _GLTHREAD_H 26 27 /* The size of one batch and the maximum size of one call. 28 * 29 * This should be as low as possible, so that: 30 * - multiple synchronizations within a frame don't slow us down much 31 * - a smaller number of calls per frame can still get decent parallelism 32 * - the memory footprint of the queue is low, and with that comes a lower 33 * chance of experiencing CPU cache thrashing 34 * but it should be high enough so that u_queue overhead remains negligible. 35 */ 36 #define MARSHAL_MAX_CMD_BUFFER_SIZE (8 * 1024) 37 38 /* We need to leave 1 slot at the end to insert the END marker for unmarshal 39 * calls that look ahead to know where the batch ends. 40 */ 41 #define MARSHAL_MAX_CMD_SIZE (MARSHAL_MAX_CMD_BUFFER_SIZE - 8) 42 43 /* The number of batch slots in memory. 44 * 45 * One batch is being executed, one batch is being filled, the rest are 46 * waiting batches. There must be at least 1 slot for a waiting batch, 47 * so the minimum number of batches is 3. 48 */ 49 #define MARSHAL_MAX_BATCHES 8 50 51 /* Special value for glEnableClientState(GL_PRIMITIVE_RESTART_NV). */ 52 #define VERT_ATTRIB_PRIMITIVE_RESTART_NV -1 53 54 #include <inttypes.h> 55 #include <stdbool.h> 56 #include "util/u_queue.h" 57 #include "compiler/shader_enums.h" 58 #include "main/config.h" 59 #include "main/hash.h" 60 #include "util/glheader.h" 61 62 #ifdef __cplusplus 63 extern "C" { 64 #endif 65 66 struct gl_context; 67 struct gl_buffer_object; 68 struct _glapi_table; 69 70 /* Used by both glthread and gl_context. */ 71 union gl_vertex_format_user { 72 struct { 73 GLenum16 Type; /**< datatype: GL_FLOAT, GL_INT, etc */ 74 bool Bgra; /**< true if GL_BGRA, else GL_RGBA */ 75 uint8_t Size:5; /**< components per element (1,2,3,4) */ 76 bool Normalized:1; /**< GL_ARB_vertex_program */ 77 bool Integer:1; /**< Integer-valued? */ 78 bool Doubles:1; /**< double values are not converted to floats */ 79 }; 80 uint32_t All; 81 }; 82 83 #define MESA_PACK_VFORMAT(type, size, normalized, integer, doubles) \ 84 (union gl_vertex_format_user){{ \ 85 .Type = MIN2(type, 0xffff), /* 0xffff means invalid value */ \ 86 .Bgra = size == GL_BGRA, \ 87 .Size = size == GL_BGRA ? 4 : MIN2(size, 5), /* 5 means invalid value */ \ 88 .Normalized = normalized, \ 89 .Integer = integer, \ 90 .Doubles = doubles \ 91 }} 92 93 struct glthread_attrib { 94 /* Per attrib: */ 95 uint8_t ElementSize; /**< max 32 */ 96 uint8_t BufferIndex; /**< Referring to Attrib[BufferIndex]. */ 97 uint16_t RelativeOffset; /**< max 0xffff in Mesa */ 98 union gl_vertex_format_user Format; 99 100 /* Per buffer binding: */ 101 GLuint Divisor; 102 int16_t Stride; /**< max 2048 */ 103 int8_t EnabledAttribCount; /**< Number of enabled attribs using this buffer. */ 104 const void *Pointer; 105 }; 106 107 struct glthread_vao { 108 GLuint Name; 109 GLuint CurrentElementBufferName; 110 GLbitfield UserEnabled; /**< Vertex attribs enabled by the user. */ 111 GLbitfield Enabled; /**< UserEnabled with POS vs GENERIC0 aliasing resolved. */ 112 GLbitfield BufferEnabled; /**< "Enabled" converted to buffer bindings. */ 113 GLbitfield BufferInterleaved; /**< Bitmask of buffers used by multiple attribs. */ 114 GLbitfield UserPointerMask; /**< Bitmask of buffer bindings. */ 115 GLbitfield NonNullPointerMask; /**< Bitmask of buffer bindings with non-NULL user pointers. */ 116 GLbitfield NonZeroDivisorMask; /**< Bitmask of buffer bindings. */ 117 118 struct glthread_attrib Attrib[VERT_ATTRIB_MAX]; 119 }; 120 121 /** A single batch of commands queued up for execution. */ 122 struct glthread_batch 123 { 124 /** Batch fence for waiting for the execution to finish. */ 125 struct util_queue_fence fence; 126 127 /** The worker thread will access the context with this. */ 128 struct gl_context *ctx; 129 130 /** 131 * Number of uint64_t elements filled already. 132 * This is 0 when it's being filled because glthread::used holds the real 133 * value temporarily, and glthread::used is copied to this variable when 134 * the batch is submitted. 135 */ 136 unsigned used; 137 138 /** Data contained in the command buffer. */ 139 uint64_t buffer[MARSHAL_MAX_CMD_BUFFER_SIZE / 8]; 140 }; 141 142 struct glthread_client_attrib { 143 struct glthread_vao VAO; 144 GLuint CurrentArrayBufferName; 145 int ClientActiveTexture; 146 GLuint RestartIndex; 147 bool PrimitiveRestart; 148 bool PrimitiveRestartFixedIndex; 149 150 /** Whether this element of the client attrib stack contains saved state. */ 151 bool Valid; 152 }; 153 154 /* For glPushAttrib / glPopAttrib. */ 155 struct glthread_attrib_node { 156 GLbitfield Mask; 157 int ActiveTexture; 158 GLenum16 MatrixMode; 159 bool Blend; 160 bool CullFace; 161 bool DepthTest; 162 bool Lighting; 163 bool PolygonStipple; 164 }; 165 166 typedef enum { 167 M_MODELVIEW, 168 M_PROJECTION, 169 M_PROGRAM0, 170 M_PROGRAM_LAST = M_PROGRAM0 + MAX_PROGRAM_MATRICES - 1, 171 M_TEXTURE0, 172 M_TEXTURE_LAST = M_TEXTURE0 + MAX_TEXTURE_UNITS - 1, 173 M_DUMMY, /* used instead of reporting errors */ 174 M_NUM_MATRIX_STACKS, 175 } gl_matrix_index; 176 177 struct glthread_state 178 { 179 /** Multithreaded queue. */ 180 struct util_queue queue; 181 182 /** This is sent to the driver for framebuffer overlay / HUD. */ 183 struct util_queue_monitoring stats; 184 185 /** Whether GLThread is enabled. */ 186 bool enabled; 187 bool inside_begin_end; 188 bool thread_sched_enabled; 189 190 /** Display lists. */ 191 GLenum16 ListMode; /**< Zero if not inside display list, else list mode. */ 192 unsigned ListBase; 193 unsigned ListCallDepth; 194 195 /** For L3 cache pinning. */ 196 unsigned pin_thread_counter; 197 unsigned thread_sched_state; 198 199 /** The ring of batches in memory. */ 200 struct glthread_batch batches[MARSHAL_MAX_BATCHES]; 201 202 /** Pointer to the batch currently being filled. */ 203 struct glthread_batch *next_batch; 204 205 /** Index of the last submitted batch. */ 206 unsigned last; 207 208 /** Index of the batch being filled and about to be submitted. */ 209 unsigned next; 210 211 /** Number of uint64_t elements filled already. */ 212 unsigned used; 213 214 /** Upload buffer. */ 215 struct gl_buffer_object *upload_buffer; 216 uint8_t *upload_ptr; 217 unsigned upload_offset; 218 int upload_buffer_private_refcount; 219 220 /** Primitive restart state. */ 221 bool PrimitiveRestart; 222 bool PrimitiveRestartFixedIndex; 223 bool _PrimitiveRestart; 224 GLuint RestartIndex; 225 GLuint _RestartIndex[4]; /**< Restart index for index_size = 1,2,4. */ 226 227 /** Vertex Array objects tracked by glthread independently of Mesa. */ 228 struct _mesa_HashTable VAOs; 229 struct glthread_vao *CurrentVAO; 230 struct glthread_vao *LastLookedUpVAO; 231 struct glthread_vao DefaultVAO; 232 struct glthread_client_attrib ClientAttribStack[MAX_CLIENT_ATTRIB_STACK_DEPTH]; 233 int ClientAttribStackTop; 234 int ClientActiveTexture; 235 236 /** Currently-bound buffer object IDs. */ 237 GLuint CurrentArrayBufferName; 238 GLuint CurrentDrawIndirectBufferName; 239 GLuint CurrentPixelPackBufferName; 240 GLuint CurrentPixelUnpackBufferName; 241 GLuint CurrentQueryBufferName; 242 243 /** 244 * The batch index of the last occurence of glLinkProgram or 245 * glDeleteProgram or -1 if there is no such enqueued call. 246 */ 247 int LastProgramChangeBatch; 248 249 /** 250 * The batch index of the last occurence of glEndList or 251 * glDeleteLists or -1 if there is no such enqueued call. 252 */ 253 int LastDListChangeBatchIndex; 254 255 /** Basic matrix state tracking. */ 256 int ActiveTexture; 257 GLenum16 MatrixMode; 258 gl_matrix_index MatrixIndex; 259 struct glthread_attrib_node AttribStack[MAX_ATTRIB_STACK_DEPTH]; 260 int AttribStackDepth; 261 int MatrixStackDepth[M_NUM_MATRIX_STACKS]; 262 263 /** Enable states. */ 264 bool Blend; 265 bool DepthTest; 266 bool CullFace; 267 bool DebugOutputSynchronous; 268 bool Lighting; 269 bool PolygonStipple; 270 271 GLuint CurrentDrawFramebuffer; 272 GLuint CurrentReadFramebuffer; 273 GLuint CurrentProgram; 274 275 /** The last added call of the given function. */ 276 struct marshal_cmd_CallList *LastCallList; 277 struct marshal_cmd_BindBuffer *LastBindBuffer1; 278 struct marshal_cmd_BindBuffer *LastBindBuffer2; 279 280 /** Global mutex update info. */ 281 unsigned GlobalLockUpdateBatchCounter; 282 bool LockGlobalMutexes; 283 }; 284 285 void _mesa_glthread_init(struct gl_context *ctx); 286 void _mesa_glthread_destroy(struct gl_context *ctx); 287 288 void _mesa_glthread_init_dispatch0(struct gl_context *ctx, 289 struct _glapi_table *table); 290 void _mesa_glthread_init_dispatch1(struct gl_context *ctx, 291 struct _glapi_table *table); 292 void _mesa_glthread_init_dispatch2(struct gl_context *ctx, 293 struct _glapi_table *table); 294 void _mesa_glthread_init_dispatch3(struct gl_context *ctx, 295 struct _glapi_table *table); 296 void _mesa_glthread_init_dispatch4(struct gl_context *ctx, 297 struct _glapi_table *table); 298 void _mesa_glthread_init_dispatch5(struct gl_context *ctx, 299 struct _glapi_table *table); 300 void _mesa_glthread_init_dispatch6(struct gl_context *ctx, 301 struct _glapi_table *table); 302 void _mesa_glthread_init_dispatch7(struct gl_context *ctx, 303 struct _glapi_table *table); 304 305 void _mesa_glthread_enable(struct gl_context *ctx); 306 void _mesa_glthread_disable(struct gl_context *ctx); 307 void _mesa_glthread_flush_batch(struct gl_context *ctx); 308 void _mesa_glthread_finish(struct gl_context *ctx); 309 void _mesa_glthread_finish_before(struct gl_context *ctx, const char *func); 310 bool _mesa_glthread_invalidate_zsbuf(struct gl_context *ctx); 311 void _mesa_glthread_release_upload_buffer(struct gl_context *ctx); 312 void _mesa_glthread_upload(struct gl_context *ctx, const void *data, 313 GLsizeiptr size, unsigned *out_offset, 314 struct gl_buffer_object **out_buffer, 315 uint8_t **out_ptr, 316 unsigned start_offset); 317 void _mesa_glthread_reset_vao(struct glthread_vao *vao); 318 void _mesa_error_glthread_safe(struct gl_context *ctx, GLenum error, 319 bool glthread, const char *format, ...); 320 void _mesa_glthread_execute_list(struct gl_context *ctx, GLuint list); 321 322 void _mesa_glthread_DeleteBuffers(struct gl_context *ctx, GLsizei n, 323 const GLuint *buffers); 324 325 void _mesa_glthread_BindVertexArray(struct gl_context *ctx, GLuint id); 326 void _mesa_glthread_DeleteVertexArrays(struct gl_context *ctx, 327 GLsizei n, const GLuint *ids); 328 void _mesa_glthread_GenVertexArrays(struct gl_context *ctx, 329 GLsizei n, GLuint *arrays); 330 void _mesa_glthread_set_prim_restart(struct gl_context *ctx, GLenum cap, 331 bool value); 332 void _mesa_glthread_PrimitiveRestartIndex(struct gl_context *ctx, GLuint index); 333 void _mesa_glthread_ClientState(struct gl_context *ctx, GLuint *vaobj, 334 gl_vert_attrib attrib, bool enable); 335 void _mesa_glthread_AttribDivisor(struct gl_context *ctx, const GLuint *vaobj, 336 gl_vert_attrib attrib, GLuint divisor); 337 void _mesa_glthread_AttribPointer(struct gl_context *ctx, gl_vert_attrib attrib, 338 union gl_vertex_format_user format, 339 GLsizei stride, const void *pointer); 340 void _mesa_glthread_DSAAttribPointer(struct gl_context *ctx, GLuint vao, 341 GLuint buffer, gl_vert_attrib attrib, 342 union gl_vertex_format_user format, 343 GLsizei stride, GLintptr offset); 344 void _mesa_glthread_AttribFormat(struct gl_context *ctx, GLuint attribindex, 345 union gl_vertex_format_user format, 346 GLuint relativeoffset); 347 void _mesa_glthread_DSAAttribFormat(struct gl_context *ctx, GLuint vaobj, 348 GLuint attribindex, 349 union gl_vertex_format_user format, 350 GLuint relativeoffset); 351 void _mesa_glthread_VertexBuffer(struct gl_context *ctx, GLuint bindingindex, 352 GLuint buffer, GLintptr offset, GLsizei stride); 353 void _mesa_glthread_DSAVertexBuffer(struct gl_context *ctx, GLuint vaobj, 354 GLuint bindingindex, GLuint buffer, 355 GLintptr offset, GLsizei stride); 356 void _mesa_glthread_DSAVertexBuffers(struct gl_context *ctx, GLuint vaobj, 357 GLuint first, GLsizei count, 358 const GLuint *buffers, 359 const GLintptr *offsets, 360 const GLsizei *strides); 361 void _mesa_glthread_BindingDivisor(struct gl_context *ctx, GLuint bindingindex, 362 GLuint divisor); 363 void _mesa_glthread_DSABindingDivisor(struct gl_context *ctx, GLuint vaobj, 364 GLuint bindingindex, GLuint divisor); 365 void _mesa_glthread_AttribBinding(struct gl_context *ctx, GLuint attribindex, 366 GLuint bindingindex); 367 void _mesa_glthread_DSAAttribBinding(struct gl_context *ctx, GLuint vaobj, 368 GLuint attribindex, GLuint bindingindex); 369 void _mesa_glthread_DSAElementBuffer(struct gl_context *ctx, GLuint vaobj, 370 GLuint buffer); 371 void _mesa_glthread_PushClientAttrib(struct gl_context *ctx, GLbitfield mask, 372 bool set_default); 373 void _mesa_glthread_PopClientAttrib(struct gl_context *ctx); 374 void _mesa_glthread_ClientAttribDefault(struct gl_context *ctx, GLbitfield mask); 375 void _mesa_glthread_InterleavedArrays(struct gl_context *ctx, GLenum format, 376 GLsizei stride, const GLvoid *pointer); 377 void _mesa_glthread_ProgramChanged(struct gl_context *ctx); 378 void _mesa_glthread_UnrollDrawElements(struct gl_context *ctx, 379 GLenum mode, GLsizei count, GLenum type, 380 const GLvoid *indices, GLint basevertex); 381 void _mesa_glthread_unbind_uploaded_vbos(struct gl_context *ctx); 382 383 #ifdef __cplusplus 384 } 385 #endif 386 387 #endif /* _GLTHREAD_H*/ 388