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_SIZE (8 * 1024) 37 38 /* The number of batch slots in memory. 39 * 40 * One batch is being executed, one batch is being filled, the rest are 41 * waiting batches. There must be at least 1 slot for a waiting batch, 42 * so the minimum number of batches is 3. 43 */ 44 #define MARSHAL_MAX_BATCHES 8 45 46 /* Special value for glEnableClientState(GL_PRIMITIVE_RESTART_NV). */ 47 #define VERT_ATTRIB_PRIMITIVE_RESTART_NV -1 48 49 #include <inttypes.h> 50 #include <stdbool.h> 51 #include "util/u_queue.h" 52 #include "GL/gl.h" 53 #include "compiler/shader_enums.h" 54 #include "main/config.h" 55 56 struct gl_context; 57 struct gl_buffer_object; 58 struct _mesa_HashTable; 59 60 struct glthread_attrib_binding { 61 struct gl_buffer_object *buffer; /**< where non-VBO data was uploaded */ 62 int offset; /**< offset to uploaded non-VBO data */ 63 const void *original_pointer; /**< restore this pointer after the draw */ 64 }; 65 66 struct glthread_vao { 67 GLuint Name; 68 GLuint CurrentElementBufferName; 69 GLbitfield UserEnabled; /**< Vertex attribs enabled by the user. */ 70 GLbitfield Enabled; /**< UserEnabled with POS vs GENERIC0 aliasing resolved. */ 71 GLbitfield BufferEnabled; /**< "Enabled" converted to buffer bindings. */ 72 GLbitfield BufferInterleaved; /**< Bitmask of buffers used by multiple attribs. */ 73 GLbitfield UserPointerMask; /**< Bitmask of buffer bindings. */ 74 GLbitfield NonZeroDivisorMask; /**< Bitmask of buffer bindings. */ 75 76 struct { 77 /* Per attrib: */ 78 GLuint ElementSize; 79 GLuint RelativeOffset; 80 GLuint BufferIndex; /**< Referring to Attrib[BufferIndex]. */ 81 82 /* Per buffer binding: */ 83 GLsizei Stride; 84 GLuint Divisor; 85 int EnabledAttribCount; /**< Number of enabled attribs using this buffer. */ 86 const void *Pointer; 87 } Attrib[VERT_ATTRIB_MAX]; 88 }; 89 90 /** A single batch of commands queued up for execution. */ 91 struct glthread_batch 92 { 93 /** Batch fence for waiting for the execution to finish. */ 94 struct util_queue_fence fence; 95 96 /** The worker thread will access the context with this. */ 97 struct gl_context *ctx; 98 99 /** Amount of data used by batch commands, in bytes. */ 100 int used; 101 102 /** Data contained in the command buffer. */ 103 #ifdef _MSC_VER 104 __declspec(align(8)) 105 #else 106 __attribute__((aligned(8))) 107 #endif 108 uint8_t buffer[MARSHAL_MAX_CMD_SIZE]; 109 }; 110 111 struct glthread_client_attrib { 112 struct glthread_vao VAO; 113 GLuint CurrentArrayBufferName; 114 int ClientActiveTexture; 115 GLuint RestartIndex; 116 bool PrimitiveRestart; 117 bool PrimitiveRestartFixedIndex; 118 119 /** Whether this element of the client attrib stack contains saved state. */ 120 bool Valid; 121 }; 122 123 struct glthread_state 124 { 125 /** Multithreaded queue. */ 126 struct util_queue queue; 127 128 /** This is sent to the driver for framebuffer overlay / HUD. */ 129 struct util_queue_monitoring stats; 130 131 /** Whether GLThread is enabled. */ 132 bool enabled; 133 134 /** Whether GLThread is inside a display list generation. */ 135 bool inside_dlist; 136 137 /** For L3 cache pinning. */ 138 unsigned pin_thread_counter; 139 140 /** The ring of batches in memory. */ 141 struct glthread_batch batches[MARSHAL_MAX_BATCHES]; 142 143 /** Pointer to the batch currently being filled. */ 144 struct glthread_batch *next_batch; 145 146 /** Index of the last submitted batch. */ 147 unsigned last; 148 149 /** Index of the batch being filled and about to be submitted. */ 150 unsigned next; 151 152 /** Upload buffer. */ 153 struct gl_buffer_object *upload_buffer; 154 uint8_t *upload_ptr; 155 unsigned upload_offset; 156 int upload_buffer_private_refcount; 157 158 /** Caps. */ 159 GLboolean SupportsBufferUploads; 160 GLboolean SupportsNonVBOUploads; 161 162 /** Primitive restart state. */ 163 bool PrimitiveRestart; 164 bool PrimitiveRestartFixedIndex; 165 bool _PrimitiveRestart; 166 GLuint RestartIndex; 167 GLuint _RestartIndex[4]; /**< Restart index for index_size = 1,2,4. */ 168 169 /** Vertex Array objects tracked by glthread independently of Mesa. */ 170 struct _mesa_HashTable *VAOs; 171 struct glthread_vao *CurrentVAO; 172 struct glthread_vao *LastLookedUpVAO; 173 struct glthread_vao DefaultVAO; 174 struct glthread_client_attrib ClientAttribStack[MAX_CLIENT_ATTRIB_STACK_DEPTH]; 175 int ClientAttribStackTop; 176 int ClientActiveTexture; 177 178 /** Currently-bound buffer object IDs. */ 179 GLuint CurrentArrayBufferName; 180 GLuint CurrentDrawIndirectBufferName; 181 }; 182 183 void _mesa_glthread_init(struct gl_context *ctx); 184 void _mesa_glthread_destroy(struct gl_context *ctx); 185 186 void _mesa_glthread_restore_dispatch(struct gl_context *ctx, const char *func); 187 void _mesa_glthread_disable(struct gl_context *ctx, const char *func); 188 void _mesa_glthread_flush_batch(struct gl_context *ctx); 189 void _mesa_glthread_finish(struct gl_context *ctx); 190 void _mesa_glthread_finish_before(struct gl_context *ctx, const char *func); 191 void _mesa_glthread_upload(struct gl_context *ctx, const void *data, 192 GLsizeiptr size, unsigned *out_offset, 193 struct gl_buffer_object **out_buffer, 194 uint8_t **out_ptr); 195 void _mesa_glthread_reset_vao(struct glthread_vao *vao); 196 197 void _mesa_glthread_BindBuffer(struct gl_context *ctx, GLenum target, 198 GLuint buffer); 199 void _mesa_glthread_DeleteBuffers(struct gl_context *ctx, GLsizei n, 200 const GLuint *buffers); 201 202 void _mesa_glthread_BindVertexArray(struct gl_context *ctx, GLuint id); 203 void _mesa_glthread_DeleteVertexArrays(struct gl_context *ctx, 204 GLsizei n, const GLuint *ids); 205 void _mesa_glthread_GenVertexArrays(struct gl_context *ctx, 206 GLsizei n, GLuint *arrays); 207 void _mesa_glthread_set_prim_restart(struct gl_context *ctx, GLenum cap, 208 bool value); 209 void _mesa_glthread_PrimitiveRestartIndex(struct gl_context *ctx, GLuint index); 210 void _mesa_glthread_ClientState(struct gl_context *ctx, GLuint *vaobj, 211 gl_vert_attrib attrib, bool enable); 212 void _mesa_glthread_AttribDivisor(struct gl_context *ctx, const GLuint *vaobj, 213 gl_vert_attrib attrib, GLuint divisor); 214 void _mesa_glthread_AttribPointer(struct gl_context *ctx, gl_vert_attrib attrib, 215 GLint size, GLenum type, GLsizei stride, 216 const void *pointer); 217 void _mesa_glthread_DSAAttribPointer(struct gl_context *ctx, GLuint vao, 218 GLuint buffer, gl_vert_attrib attrib, 219 GLint size, GLenum type, GLsizei stride, 220 GLintptr offset); 221 void _mesa_glthread_AttribFormat(struct gl_context *ctx, GLuint attribindex, 222 GLint size, GLenum type, GLuint relativeoffset); 223 void _mesa_glthread_DSAAttribFormat(struct gl_context *ctx, GLuint vaobj, 224 GLuint attribindex, GLint size, GLenum type, 225 GLuint relativeoffset); 226 void _mesa_glthread_VertexBuffer(struct gl_context *ctx, GLuint bindingindex, 227 GLuint buffer, GLintptr offset, GLsizei stride); 228 void _mesa_glthread_DSAVertexBuffer(struct gl_context *ctx, GLuint vaobj, 229 GLuint bindingindex, GLuint buffer, 230 GLintptr offset, GLsizei stride); 231 void _mesa_glthread_DSAVertexBuffers(struct gl_context *ctx, GLuint vaobj, 232 GLuint first, GLsizei count, 233 const GLuint *buffers, 234 const GLintptr *offsets, 235 const GLsizei *strides); 236 void _mesa_glthread_BindingDivisor(struct gl_context *ctx, GLuint bindingindex, 237 GLuint divisor); 238 void _mesa_glthread_DSABindingDivisor(struct gl_context *ctx, GLuint vaobj, 239 GLuint bindingindex, GLuint divisor); 240 void _mesa_glthread_AttribBinding(struct gl_context *ctx, GLuint attribindex, 241 GLuint bindingindex); 242 void _mesa_glthread_DSAAttribBinding(struct gl_context *ctx, GLuint vaobj, 243 GLuint attribindex, GLuint bindingindex); 244 void _mesa_glthread_DSAElementBuffer(struct gl_context *ctx, GLuint vaobj, 245 GLuint buffer); 246 void _mesa_glthread_PushClientAttrib(struct gl_context *ctx, GLbitfield mask, 247 bool set_default); 248 void _mesa_glthread_PopClientAttrib(struct gl_context *ctx); 249 void _mesa_glthread_ClientAttribDefault(struct gl_context *ctx, GLbitfield mask); 250 void _mesa_glthread_InterleavedArrays(struct gl_context *ctx, GLenum format, 251 GLsizei stride, const GLvoid *pointer); 252 253 #endif /* _GLTHREAD_H*/ 254