shading.html revision b03e1712b2d06159c13564ad8f70f79cf586510e 
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>
12This page describes the features and status of Mesa's support for the
13<a href="http://opengl.org/documentation/glsl/" target="_parent">
14OpenGL Shading Language</a>.
15</p>
16
17<p>
18Last updated on 17 Feb 2007.
19</p>
20
21<p>
22Contents
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</ul>
31
32
33<a name="unsup">
34<h2>Unsupported Features</h2>
35
36<p>
37The following features of the shading language are not yet supported
38in Mesa:
39</p>
40
41<ul>
42<li>Dereferencing arrays with non-constant indexes
43<li>User-defined structs
44<li>Linking of multiple shaders is not supported
45<li>Integer operations are not fully implemented (most are implemented
46    as floating point).
47<li>gl_ClipVertex
48</ul>
49
50<p>
51All 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>There are massive memory leaks in the compiler.
71</ul>
72
73<p>
74These issues will be addressed/resolved in the future.
75</p>
76
77
78<a name="hints">
79<h2>Programming Hints</h2>
80
81<ul>
82<li>Declare <em>in</em> function parameters as <em>const</em> whenever possible.
83    This improves the efficiency of function inlining.
84</li>
85<br>
86<li>To reduce register usage, declare variables within smaller scopes.
87    For example, the following code:
88<pre>
89    void main()
90    {
91       vec4 a1, a2, b1, b2;
92       gl_Position = expression using a1, a2.
93       gl_Color = expression using b1, b2;
94    }
95</pre>
96    Can be rewritten as follows to use half as many registers:
97<pre>
98    void main()
99    {
100       {
101          vec4 a1, a2;
102          gl_Position = expression using a1, a2.
103       }
104       {
105          vec4 b1, b2;
106          gl_Color = expression using b1, b2;
107       }
108    }
109</pre>
110    Alternately, rather than using several float variables, use
111    a vec4 instead.  Use swizzling and writemasks to access the
112    components of the vec4 as floats.
113</li>
114<br>
115<li>Use the built-in library functions whenever possible.
116    For example, instead of writing this:
117<pre>
118        float x = 1.0 / sqrt(y);
119</pre>
120    Write this:
121<pre>
122        float x = inversesqrt(y);
123</pre>
124</ul>
125
126
127<a name="standalone">
128<h2>Stand-alone Compiler</h2>
129
130<p>
131A unique stand-alone GLSL compiler driver has been added to Mesa.
132<p>
133
134<p>
135The stand-alone compiler (like a conventional command-line compiler)
136is a tool that accepts Shading Language programs and emits low-level
137GPU programs.
138</p>
139
140<p>
141This tool is useful for:
142<p>
143<ul>
144<li>Inspecting GPU code to gain insight into compilation
145<li>Generating initial GPU code for subsequent hand-tuning
146<li>Debugging the GLSL compiler itself
147</ul>
148
149<p>
150To build the glslcompiler program (this will be improved someday):
151</p>
152<pre>
153    cd src/mesa
154    make libmesa.a
155    cd drivers/glslcompiler
156    make
157</pre>
158
159
160<p>
161Here's an example of using the compiler to compile a vertex shader and
162emit GL_ARB_vertex_program-style instructions:
163</p>
164<pre>
165    glslcompiler --arb --linenumbers --vs vertshader.txt
166</pre>
167<p>
168The output may look similar to this:
169</p>
170<pre>
171!!ARBvp1.0
172  0: MOV result.texcoord[0], vertex.texcoord[0];
173  1: DP4 temp0.x, state.matrix.mvp.row[0], vertex.position;
174  2: DP4 temp0.y, state.matrix.mvp.row[1], vertex.position;
175  3: DP4 temp0.z, state.matrix.mvp.row[2], vertex.position;
176  4: DP4 temp0.w, state.matrix.mvp.row[3], vertex.position;
177  5: MOV result.position, temp0;
178  6: END
179</pre>
180
181<p>
182Note that some shading language constructs (such as uniform and varying
183variables) aren't expressible in ARB or NV-style programs.
184Therefore, the resulting output is not always legal by definition of
185those program languages.
186</p>
187<p>
188Also note that this compiler driver is still under development.
189Over time, the correctness of the GPU programs, with respect to the ARB
190and NV languagues, should improve.
191</p>
192
193
194
195<a name="implementation">
196<h2>Compiler Implementation</h2>
197
198<p>
199The source code for Mesa's shading language compiler is in the
200<code>src/mesa/shader/slang/</code> directory.
201</p>
202
203<p>
204The compiler follows a fairly standard design and basically works as follows:
205</p>
206<ul>
207<li>The input string is tokenized (see grammar.c) and parsed
208(see slang_compiler_*.c) to produce an Abstract Syntax Tree (AST).
209The nodes in this tree are slang_operation structures
210(see slang_compile_operation.h).
211The nodes are decorated with symbol table, scoping and datatype information.
212<li>The AST is converted into an Intermediate representation (IR) tree
213(see the slang_codegen.c file).
214The IR nodes represent basic GPU instructions, like add, dot product,
215move, etc. 
216The IR tree is mostly a binary tree, but a few nodes have three or four
217children.
218In principle, the IR tree could be executed by doing an in-order traversal.
219<li>The IR tree is traversed in-order to emit code (see slang_emit.c).
220This is also when registers are allocated to store variables and temps.
221<li>In the future, a pattern-matching code generator-generator may be
222used for code generation.
223Programs such as L-BURG (Bottom-Up Rewrite Generator) and Twig look for
224patterns in IR trees, compute weights for subtrees and use the weights
225to select the best instructions to represent the sub-tree.
226<li>The emitted GPU instructions (see prog_instruction.h) are stored in a
227gl_program object (see mtypes.h).
228<li>When a fragment shader and vertex shader are linked (see slang_link.c)
229the varying vars are matched up, uniforms are merged, and vertex
230attributes are resolved (rewriting instructions as needed).
231</ul>
232
233<p>
234The final vertex and fragment programs may be interpreted in software
235(see prog_execute.c) or translated into a specific hardware architecture
236(see drivers/dri/i915/i915_fragprog.c for example).
237</p>
238
239
240
241</BODY>
242</HTML>
243