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 <li>The inverse trig functions asin(), acos(), and atan() are not implemented
  49 <li>The gl_Color and gl_SecondaryColor varying vars are interpolated
  50     without perspective correction
  51 </ul>
  52
  53 <p>
  54 All other major features of the shading language should function.
  55 </p>
  56
  57
  58 <a name="notes">
  59 <h2>Implementation Notes</h2>
  60
  61 <ul>
  62 <li>Shading language programs are compiled into low-level programs
  63     very similar to those of GL_ARB_vertex/fragment_program.
  64 <li>All vector types (vec2, vec3, vec4, bvec2, etc) currently occupy full
  65     float[4] registers.
  66 <li>Float constants and variables are packed so that up to four floats
  67     can occupy one program parameter/register.
  68 <li>All function calls are inlined.
  69 <li>Shaders which use too many registers will not compile.
  70 <li>The quality of generated code is pretty good, register usage is fair.
  71 <li>Shader error detection and reporting of errors (InfoLog) is not
  72     very good yet.
  73 <li>The ftransform() function doesn't necessarily match the results of
  74     fixed-function transformation.
  75 </ul>
  76
  77 <p>
  78 These issues will be addressed/resolved in the future.
  79 </p>
  80
  81
  82 <a name="hints">
  83 <h2>Programming Hints</h2>
  84
  85 <ul>
  86 <li>Declare <em>in</em> function parameters as <em>const</em> whenever possible.
  87     This improves the efficiency of function inlining.
  88 </li>
  89 <br>
  90 <li>To reduce register usage, declare variables within smaller scopes.
  91     For example, the following code:
  92 <pre>
  93     void main()
  94     {
  95        vec4 a1, a2, b1, b2;
  96        gl_Position = expression using a1, a2.
  97        gl_Color = expression using b1, b2;
  98     }
  99 </pre>
 100     Can be rewritten as follows to use half as many registers:
 101 <pre>
 102     void main()
 103     {
 104        {
 105           vec4 a1, a2;
 106           gl_Position = expression using a1, a2.
 107        }
 108        {
 109           vec4 b1, b2;
 110           gl_Color = expression using b1, b2;
 111        }
 112     }
 113 </pre>
 114     Alternately, rather than using several float variables, use
 115     a vec4 instead.  Use swizzling and writemasks to access the
 116     components of the vec4 as floats.
 117 </li>
 118 <br>
 119 <li>Use the built-in library functions whenever possible.
 120     For example, instead of writing this:
 121 <pre>
 122         float x = 1.0 / sqrt(y);
 123 </pre>
 124     Write this:
 125 <pre>
 126         float x = inversesqrt(y);
 127 </pre>
 128 <li>
 129    Use ++i when possible as it's more efficient than i++
 130 </li>
 131 </ul>
 132
 133
 134 <a name="standalone">
 135 <h2>Stand-alone Compiler</h2>
 136
 137 <p>
 138 A unique stand-alone GLSL compiler driver has been added to Mesa.
 139 <p>
 140
 141 <p>
 142 The stand-alone compiler (like a conventional command-line compiler)
 143 is a tool that accepts Shading Language programs and emits low-level
 144 GPU programs.
 145 </p>
 146
 147 <p>
 148 This tool is useful for:
 149 <p>
 150 <ul>
 151 <li>Inspecting GPU code to gain insight into compilation
 152 <li>Generating initial GPU code for subsequent hand-tuning
 153 <li>Debugging the GLSL compiler itself
 154 </ul>
 155
 156 <p>
 157 To build the glslcompiler program (this will be improved someday):
 158 </p>
 159 <pre>
 160     cd src/mesa
 161     make libmesa.a
 162     cd drivers/glslcompiler
 163     make
 164 </pre>
 165
 166
 167 <p>
 168 Here's an example of using the compiler to compile a vertex shader and
 169 emit GL_ARB_vertex_program-style instructions:
 170 </p>
 171 <pre>
 172     glslcompiler --arb --linenumbers --vs vertshader.txt
 173 </pre>
 174 <p>
 175 The output may look similar to this:
 176 </p>
 177 <pre>
 178 !!ARBvp1.0
 179   0: MOV result.texcoord[0], vertex.texcoord[0];
 180   1: DP4 temp0.x, state.matrix.mvp.row[0], vertex.position;
 181   2: DP4 temp0.y, state.matrix.mvp.row[1], vertex.position;
 182   3: DP4 temp0.z, state.matrix.mvp.row[2], vertex.position;
 183   4: DP4 temp0.w, state.matrix.mvp.row[3], vertex.position;
 184   5: MOV result.position, temp0;
 185   6: END
 186 </pre>
 187
 188 <p>
 189 Note that some shading language constructs (such as uniform and varying
 190 variables) aren't expressible in ARB or NV-style programs.
 191 Therefore, the resulting output is not always legal by definition of
 192 those program languages.
 193 </p>
 194 <p>
 195 Also note that this compiler driver is still under development.
 196 Over time, the correctness of the GPU programs, with respect to the ARB
 197 and NV languagues, should improve.
 198 </p>
 199
 200
 201
 202 <a name="implementation">
 203 <h2>Compiler Implementation</h2>
 204
 205 <p>
 206 The source code for Mesa's shading language compiler is in the
 207 <code>src/mesa/shader/slang/</code> directory.
 208 </p>
 209
 210 <p>
 211 The compiler follows a fairly standard design and basically works as follows:
 212 </p>
 213 <ul>
 214 <li>The input string is tokenized (see grammar.c) and parsed
 215 (see slang_compiler_*.c) to produce an Abstract Syntax Tree (AST).
 216 The nodes in this tree are slang_operation structures
 217 (see slang_compile_operation.h).
 218 The nodes are decorated with symbol table, scoping and datatype information.
 219 <li>The AST is converted into an Intermediate representation (IR) tree
 220 (see the slang_codegen.c file).
 221 The IR nodes represent basic GPU instructions, like add, dot product,
 222 move, etc.
 223 The IR tree is mostly a binary tree, but a few nodes have three or four
 224 children.
 225 In principle, the IR tree could be executed by doing an in-order traversal.
 226 <li>The IR tree is traversed in-order to emit code (see slang_emit.c).
 227 This is also when registers are allocated to store variables and temps.
 228 <li>In the future, a pattern-matching code generator-generator may be
 229 used for code generation.
 230 Programs such as L-BURG (Bottom-Up Rewrite Generator) and Twig look for
 231 patterns in IR trees, compute weights for subtrees and use the weights
 232 to select the best instructions to represent the sub-tree.
 233 <li>The emitted GPU instructions (see prog_instruction.h) are stored in a
 234 gl_program object (see mtypes.h).
 235 <li>When a fragment shader and vertex shader are linked (see slang_link.c)
 236 the varying vars are matched up, uniforms are merged, and vertex
 237 attributes are resolved (rewriting instructions as needed).
 238 </ul>
 239
 240 <p>
 241 The final vertex and fragment programs may be interpreted in software
 242 (see prog_execute.c) or translated into a specific hardware architecture
 243 (see drivers/dri/i915/i915_fragprog.c for example).
 244 </p>
 245
 246 <h3>Code Generation Options</h3>
 247
 248 <p>
 249 Internally, there are several options that control the compiler's code
 250 generation and instruction selection.
 251 These options are seen in the gl_shader_state struct and may be set
 252 by the device driver to indicate its preferences:
 253
 254 <pre>
 255 struct gl_shader_state
 256 {
 257    ...
 258    /** Driver-selectable options: */
 259    GLboolean EmitHighLevelInstructions;
 260    GLboolean EmitCondCodes;
 261    GLboolean EmitComments;
 262 };
 263 </pre>
 264
 265 <ul>
 266 <li>EmitHighLevelInstructions
 267 <br>
 268 This option controls instruction selection for loops and conditionals.
 269 If the option is set high-level IF/ELSE/ENDIF, LOOP/ENDLOOP, CONT/BRK
 270 instructions will be emitted.
 271 Otherwise, those constructs will be implemented with BRA instructions.
 272 </li>
 273
 274 <li>EmitCondCodes
 275 <br>
 276 If set, condition codes (ala GL_NV_fragment_program) will be used for
 277 branching and looping.
 278 Otherwise, ordinary registers will be used (the IF instruction will
 279 examine the first operand's X component and do the if-part if non-zero).
 280 This option is only relevant if EmitHighLevelInstructions is set.
 281 </li>
 282
 283 <li>EmitComments
 284 <br>
 285 If set, instructions will be annoted with comments to help with debugging.
 286 Extra NOP instructions will also be inserted.
 287 </br>
 288
 289 </ul>
 290
 291
 292 <a name="validation">
 293 <h2>Compiler Validation</h2>
 294
 295 <p>
 296 A new <a href="http://glean.sf.net" target="_parent">Glean</a> test has
 297 been create to exercise the GLSL compiler.
 298 </p>
 299 <p>
 300 The <em>glsl1</em> test runs over 150 sub-tests to check that the language
 301 features and built-in functions work properly.
 302 This test should be run frequently while working on the compiler to catch
 303 regressions.
 304 </p>
 305 <p>
 306 The test coverage is reasonably broad and complete but additional tests
 307 should be added.
 308 </p>
 309
 310
 311 </BODY>
 312 </HTML>