docs/shading.html

   1 <HTML>
   2
   3 <TITLE>Shading Language Support</TITLE>
   4
   5 <link rel="stylesheet" type="text/css" href="mesa.css"></head>
   6
   7 <BODY>
   8
   9 <H1>Shading Language Support</H1>
  10
  11 <p>
  12 This page describes the features and status of Mesa's support for the
  13 <a href="http://opengl.org/documentation/glsl/" target="_parent">
  14 OpenGL Shading Language</a>.
  15 </p>
  16
  17 <p>
  18 Last updated on 28 March 2007.
  19 </p>
  20
  21 <p>
  22 Contents
  23 </p>
  24 <ul>
  25 <li><a href="#unsup">Unsupported Features</a>
  26 <li><a href="#notes">Implementation Notes</a>
  27 <li><a href="#hints">Programming Hints</a>
  28 <li><a href="#standalone">Stand-alone Compiler</a>
  29 <li><a href="#implementation">Compiler Implementation</a>
  30 <li><a href="#validation">Compiler Validation</a>
  31 </ul>
  32
  33
  34 <a name="unsup">
  35 <h2>Unsupported Features</h2>
  36
  37 <p>
  38 The following features of the shading language are not yet supported
  39 in Mesa:
  40 </p>
  41
  42 <ul>
  43 <li>Dereferencing arrays with non-constant indexes
  44 <li>Comparison of user-defined structs
  45 <li>Linking of multiple shaders is not supported
  46 <li>gl_ClipVertex
  47 <li>The derivative functions such as dFdx() are not implemented
  48 </ul>
  49
  50 <p>
  51 All other major features of the shading language should function.
  52 </p>
  53
  54
  55 <a name="notes">
  56 <h2>Implementation Notes</h2>
  57
  58 <ul>
  59 <li>Shading language programs are compiled into low-level programs
  60     very similar to those of GL_ARB_vertex/fragment_program.
  61 <li>All vector types (vec2, vec3, vec4, bvec2, etc) currently occupy full
  62     float[4] registers.
  63 <li>Float constants and variables are packed so that up to four floats
  64     can occupy one program parameter/register.
  65 <li>All function calls are inlined.
  66 <li>Shaders which use too many registers will not compile.
  67 <li>The quality of generated code is pretty good, register usage is fair.
  68 <li>Shader error detection and reporting of errors (InfoLog) is not
  69     very good yet.
  70 <li>The ftransform() function doesn't necessarily match the results of
  71     fixed-function transformation.
  72 </ul>
  73
  74 <p>
  75 These issues will be addressed/resolved in the future.
  76 </p>
  77
  78
  79 <a name="hints">
  80 <h2>Programming Hints</h2>
  81
  82 <ul>
  83 <li>Declare <em>in</em> function parameters as <em>const</em> whenever possible.
  84     This improves the efficiency of function inlining.
  85 </li>
  86 <br>
  87 <li>To reduce register usage, declare variables within smaller scopes.
  88     For example, the following code:
  89 <pre>
  90     void main()
  91     {
  92        vec4 a1, a2, b1, b2;
  93        gl_Position = expression using a1, a2.
  94        gl_Color = expression using b1, b2;
  95     }
  96 </pre>
  97     Can be rewritten as follows to use half as many registers:
  98 <pre>
  99     void main()
 100     {
 101        {
 102           vec4 a1, a2;
 103           gl_Position = expression using a1, a2.
 104        }
 105        {
 106           vec4 b1, b2;
 107           gl_Color = expression using b1, b2;
 108        }
 109     }
 110 </pre>
 111     Alternately, rather than using several float variables, use
 112     a vec4 instead.  Use swizzling and writemasks to access the
 113     components of the vec4 as floats.
 114 </li>
 115 <br>
 116 <li>Use the built-in library functions whenever possible.
 117     For example, instead of writing this:
 118 <pre>
 119         float x = 1.0 / sqrt(y);
 120 </pre>
 121     Write this:
 122 <pre>
 123         float x = inversesqrt(y);
 124 </pre>
 125 <li>
 126    Use ++i when possible as it's more efficient than i++
 127 </li>
 128 </ul>
 129
 130
 131 <a name="standalone">
 132 <h2>Stand-alone Compiler</h2>
 133
 134 <p>
 135 A unique stand-alone GLSL compiler driver has been added to Mesa.
 136 <p>
 137
 138 <p>
 139 The stand-alone compiler (like a conventional command-line compiler)
 140 is a tool that accepts Shading Language programs and emits low-level
 141 GPU programs.
 142 </p>
 143
 144 <p>
 145 This tool is useful for:
 146 <p>
 147 <ul>
 148 <li>Inspecting GPU code to gain insight into compilation
 149 <li>Generating initial GPU code for subsequent hand-tuning
 150 <li>Debugging the GLSL compiler itself
 151 </ul>
 152
 153 <p>
 154 To build the glslcompiler program (this will be improved someday):
 155 </p>
 156 <pre>
 157     cd src/mesa
 158     make libmesa.a
 159     cd drivers/glslcompiler
 160     make
 161 </pre>
 162
 163
 164 <p>
 165 Here's an example of using the compiler to compile a vertex shader and
 166 emit GL_ARB_vertex_program-style instructions:
 167 </p>
 168 <pre>
 169     glslcompiler --arb --linenumbers --vs vertshader.txt
 170 </pre>
 171 <p>
 172 The output may look similar to this:
 173 </p>
 174 <pre>
 175 !!ARBvp1.0
 176   0: MOV result.texcoord[0], vertex.texcoord[0];
 177   1: DP4 temp0.x, state.matrix.mvp.row[0], vertex.position;
 178   2: DP4 temp0.y, state.matrix.mvp.row[1], vertex.position;
 179   3: DP4 temp0.z, state.matrix.mvp.row[2], vertex.position;
 180   4: DP4 temp0.w, state.matrix.mvp.row[3], vertex.position;
 181   5: MOV result.position, temp0;
 182   6: END
 183 </pre>
 184
 185 <p>
 186 Note that some shading language constructs (such as uniform and varying
 187 variables) aren't expressible in ARB or NV-style programs.
 188 Therefore, the resulting output is not always legal by definition of
 189 those program languages.
 190 </p>
 191 <p>
 192 Also note that this compiler driver is still under development.
 193 Over time, the correctness of the GPU programs, with respect to the ARB
 194 and NV languagues, should improve.
 195 </p>
 196
 197
 198
 199 <a name="implementation">
 200 <h2>Compiler Implementation</h2>
 201
 202 <p>
 203 The source code for Mesa's shading language compiler is in the
 204 <code>src/mesa/shader/slang/</code> directory.
 205 </p>
 206
 207 <p>
 208 The compiler follows a fairly standard design and basically works as follows:
 209 </p>
 210 <ul>
 211 <li>The input string is tokenized (see grammar.c) and parsed
 212 (see slang_compiler_*.c) to produce an Abstract Syntax Tree (AST).
 213 The nodes in this tree are slang_operation structures
 214 (see slang_compile_operation.h).
 215 The nodes are decorated with symbol table, scoping and datatype information.
 216 <li>The AST is converted into an Intermediate representation (IR) tree
 217 (see the slang_codegen.c file).
 218 The IR nodes represent basic GPU instructions, like add, dot product,
 219 move, etc.
 220 The IR tree is mostly a binary tree, but a few nodes have three or four
 221 children.
 222 In principle, the IR tree could be executed by doing an in-order traversal.
 223 <li>The IR tree is traversed in-order to emit code (see slang_emit.c).
 224 This is also when registers are allocated to store variables and temps.
 225 <li>In the future, a pattern-matching code generator-generator may be
 226 used for code generation.
 227 Programs such as L-BURG (Bottom-Up Rewrite Generator) and Twig look for
 228 patterns in IR trees, compute weights for subtrees and use the weights
 229 to select the best instructions to represent the sub-tree.
 230 <li>The emitted GPU instructions (see prog_instruction.h) are stored in a
 231 gl_program object (see mtypes.h).
 232 <li>When a fragment shader and vertex shader are linked (see slang_link.c)
 233 the varying vars are matched up, uniforms are merged, and vertex
 234 attributes are resolved (rewriting instructions as needed).
 235 </ul>
 236
 237 <p>
 238 The final vertex and fragment programs may be interpreted in software
 239 (see prog_execute.c) or translated into a specific hardware architecture
 240 (see drivers/dri/i915/i915_fragprog.c for example).
 241 </p>
 242
 243 <h3>Code Generation Options</h3>
 244
 245 <p>
 246 Internally, there are several options that control the compiler's code
 247 generation and instruction selection.
 248 These options are seen in the gl_shader_state struct and may be set
 249 by the device driver to indicate its preferences:
 250
 251 <pre>
 252 struct gl_shader_state
 253 {
 254    ...
 255    /** Driver-selectable options: */
 256    GLboolean EmitHighLevelInstructions;
 257    GLboolean EmitCondCodes;
 258    GLboolean EmitComments;
 259 };
 260 </pre>
 261
 262 <ul>
 263 <li>EmitHighLevelInstructions
 264 <br>
 265 This option controls instruction selection for loops and conditionals.
 266 If the option is set high-level IF/ELSE/ENDIF, LOOP/ENDLOOP, CONT/BRK
 267 instructions will be emitted.
 268 Otherwise, those constructs will be implemented with BRA instructions.
 269 </li>
 270
 271 <li>EmitCondCodes
 272 <br>
 273 If set, condition codes (ala GL_NV_fragment_program) will be used for
 274 branching and looping.
 275 Otherwise, ordinary registers will be used (the IF instruction will
 276 examine the first operand's X component and do the if-part if non-zero).
 277 This option is only relevant if EmitHighLevelInstructions is set.
 278 </li>
 279
 280 <li>EmitComments
 281 <br>
 282 If set, instructions will be annoted with comments to help with debugging.
 283 Extra NOP instructions will also be inserted.
 284 </br>
 285
 286 </ul>
 287
 288
 289 <a name="validation">
 290 <h2>Compiler Validation</h2>
 291
 292 <p>
 293 A new <a href="http://glean.sf.net" target="_parent">Glean</a> test has
 294 been create to exercise the GLSL compiler.
 295 </p>
 296 <p>
 297 The <em>glsl1</em> test runs over 150 sub-tests to check that the language
 298 features and built-in functions work properly.
 299 This test should be run frequently while working on the compiler to catch
 300 regressions.
 301 </p>
 302 <p>
 303 The test coverage is reasonably broad and complete but additional tests
 304 should be added.
 305 </p>
 306
 307
 308 </BODY>
 309 </HTML>