1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
|
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html><head><title>Open LLVM Projects</title></head>
<body bgcolor=white>
<center><h1><font size=7>Open LLVM Projects</font></h1></center>
<ul>
<li><a href="#what">What is this?</a>
<li><a href="#improving">Improving the current system</a>
<ol>
<li><a href="#glibc">Port glibc to LLVM</a>
<li><a href="#NightlyTest">Improving the Nightly Tester</a>
<li><a href="#programs">Compile programs with the LLVM Compiler</a>
<li><a href="#llvm_ir">Extend the LLVM intermediate representation</a>
<li><a href="#misc_imp">Miscellaneous Improvements</a>
</ol>
<li><a href="#new">Adding new capabilities to LLVM</a>
<ol>
<li><a href="#pointeranalysis">Pointer and Alias Analysis</a>
<li><a href="#profileguided">Profile Guided Optimization</a>
<li><a href="#xforms">New Transformations and Analyses</a>
<li><a href="#x86be">X86 Back-end Improvements</a>
<li><a href="#misc_new">Miscellaneous Additions</a>
</ol>
</ul>
<br><br>
<!-- *********************************************************************** -->
<table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
<tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
<a name="what">What is this?
</b></font></td></tr></table><ul>
<!-- *********************************************************************** -->
This document is meant to be a sort of "big TODO list" for LLVM. Each project
in this document is something that would be useful for LLVM to have, and would
also be a great way to get familiar with the system. Some of these projects are
small and self-contained, which may be implemented in a couple of days, others
are larger. Several of these projects may lead to interesting research projects
in their own right. In any case, we welcome all contributions.<p>
If you are thinking about tackling one of these projects, please send a mail to
the <a href="http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM
Developer's</a> mailing list, so that we know the project is being worked on.
Additionally this is a good way to get more information about a specific project
or to suggest other projects to add to this page.<p>
<!-- *********************************************************************** -->
</ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
<tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
<a name="improving">Improving the current system
</b></font></td></tr></table><ul>
<!-- *********************************************************************** -->
Improvements to the current infrastructure are always very welcome and tend to
be fairly straight-forward to implement. Here are some of the key areas that
can use improvement...<p>
<!-- ======================================================================= -->
</ul><table width="50%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
<tr><td> </td><td width=100%>
<font color="#EEEEFF" size=+1 face="Georgia,Palatino"><b>
<a name="glibc">Port glibc to LLVM
</b></font></td></tr></table><ul>
It would be very useful to <a
href="http://www.gnu.org/software/libc/porting.html">port</a> <a
href="http://www.gnu.org/software/glibc/">glibc</a> to LLVM. This would allow a
variety of interprocedural algorithms to be much more effective in the face of
library calls. The most important pieces to port are things like the string
library and the <tt>stdio</tt> related functions... low-level system calls like
'<tt>read</tt>' should stay unimplemented in LLVM.<p>
<!-- ======================================================================= -->
</ul><table width="50%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
<tr><td> </td><td width=100%>
<font color="#EEEEFF" size=+1 face="Georgia,Palatino"><b>
<a name="NightlyTest">Improving the Nightly Tester
</b></font></td></tr></table><ul>
The <a href="/testresults/">Nightly Tester</a> is a simple perl script (located
in utils/NightlyTest.pl) which runs every night to generate a daily report. It
could use the following improvements:<p>
<ol>
<li>Olden timings - Time the compilation and execution times for the Olden
benchmark suite, keeping track of these values over time.
<li>Graphs - It would be great to have gnuplot graphs to keep track of how the
tree is changing over time. We already gather a several statistics, it
just neccesary to add the script-fu to gnuplotize it.
<li>Regression tests - We should run the regression tests in addition to the
program tests...
</ol><p>
<!-- ======================================================================= -->
</ul><table width="50%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
<tr><td> </td><td width=100%>
<font color="#EEEEFF" size=+1 face="Georgia,Palatino"><b>
<a name="programs">Compile programs with the LLVM Compiler
</b></font></td></tr></table><ul>
We are always looking for new testcases and benchmarks for use with LLVM. In
particular, it is useful to try compiling your favorite C source code with LLVM.
If it doesn't compile, try to figure out why or report it to the <a
href="http://mail.cs.uiuc.edu/pipermail/llvmbugs/">llvm-bugs</a> list. If you
get the program to compile, it would be extremely useful to convert the build
system to be compatible with the LLVM Programs testsuite so that we can check it
into CVS and the automated tester can use it to track progress of the
compiler.<p>
When testing a code, try running it with a variety of optimizations, and with
all the back-ends: CBE, llc, and lli.<p>
<!-- ======================================================================= -->
</ul><table width="50%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
<tr><td> </td><td width=100%>
<font color="#EEEEFF" size=+1 face="Georgia,Palatino"><b>
<a name="llvm_ir">Extend the LLVM intermediate representation
</b></font></td></tr></table><ul>
<ol>
<li>Transform <tt>setjmp</tt> and <tt>longjmp</tt> calls to use the LLVM
<tt>invoke</tt> mechanism.
<li>Add support for a volatile attribute on loads and stores
<li>Add a new conditional move instruction: <tt>X = select bool Cond, Y, Z</tt>
<li>Add support for platform independent prefetch support. The GCC <a
href="http://gcc.gnu.org/projects/prefetch.html">prefetch project</a> page
has a good survey of the prefetching capabilities of a variety of modern
processors.
</ol>
<!-- ======================================================================= -->
</ul><table width="50%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
<tr><td> </td><td width=100%>
<font color="#EEEEFF" size=+1 face="Georgia,Palatino"><b>
<a name="misc_imp">Miscellaneous Improvements
</b></font></td></tr></table><ul>
<ol>
<li>Someone needs to look into getting the <tt>ranlib</tt> tool to index LLVM
bytecode files, so that linking in .a files is not hideously slow. They
would also then have to implement the reader for this index in
<tt>gccld</tt>.<br>
<li>Improve the efficiency of the bytecode loader/writer, allow streaming lazy
loading of functions from the bytecode (for use by the JIT, for example)<br>
<li>Rework the PassManager to be more flexible
<li>Some transformations and analyses only work on reducible flow graphs. It
would be nice to have a transformation which could be "required" by these passes
which makes irreducible graphs reducible. This can easily be accomplished
through code duplication. See <a
href="http://citeseer.nj.nec.com/janssen97making.html">Making Graphs Reducible
with Controlled Node Splitting</a> and perhaps <a
href="http://doi.acm.org/10.1145/262004.262005">Nesting of Reducible and
Irreducible Loops</a>.
</ol>
<!-- *********************************************************************** -->
</ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
<tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
<a name="new">Adding new capabilities to LLVM
</b></font></td></tr></table><ul>
<!-- *********************************************************************** -->
Sometimes creating new things is more fun that improving existing things. These
projects tend to be more involved and perhaps require more work, but can also be
very rewarding.<p>
<!-- ======================================================================= -->
</ul><table width="50%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
<tr><td> </td><td width=100%>
<font color="#EEEEFF" size=+1 face="Georgia,Palatino"><b>
<a name="pointeranalysis">Pointer and Alias Analysis
</b></font></td></tr></table><ul>
We have a <a href="AliasAnalysis.html">strong base for development</a> of both
pointer analysis based optimizations as well as pointer analyses themselves. It
seems natural to want to take advantage of this...<p>
<ol>
<li>Implement a flow-sensitive context-sensitive alias analysis algorithm<br>
- Pick one of the somewhat efficient algorithms, but strive for maximum
precision
<li>Implement a flow-sensitive context-insensitive alias analysis algorithm<br>
- Just an efficient local algorithm perhaps?
<li>Implement an interface to update analyses in response to common code motion
transformations
<li>Implement alias analysis based optimizations:
<ul>
<li>Dead store elimination
</ul>
</ol>
<!-- ======================================================================= -->
</ul><table width="50%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
<tr><td> </td><td width=100%>
<font color="#EEEEFF" size=+1 face="Georgia,Palatino"><b>
<a name="profileguided">Profile Guided Optimization
</b></font></td></tr></table><ul>
We are getting to the point where we really need a unified infrastructure for
profile guided optimizations. It would be wonderful to be able to write profile
guided transformations which can be performed either at static compile time
(compile time or offline optimization time) or at runtime in a JIT type setup.
The LLVM transformation itself shouldn't need to know how it is being used.<p>
Ideas for profile guided transformations:<p>
<ol>
<li>Superblock formation (with many optimizations)
<li>Loop unrolling/peeling
<li>Profile directed inlining
<li>Code layout
<li>...
</ol><p>
<!-- ======================================================================= -->
</ul><table width="50%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
<tr><td> </td><td width=100%>
<font color="#EEEEFF" size=+1 face="Georgia,Palatino"><b>
<a name="xforms">New Transformations and Analyses
</b></font></td></tr></table><ul>
<ol>
<li>Implement a Dependence Analysis Infrastructure<br>
- Design some way to represent and query dep analysis
<li>Implement a faster Dominator Set Construction Algorithm<br>
- A linear time or nearly so algorithm
<li>Implement a strength reduction pass
<li>Value range propagation pass
<li>Implement a tail recursion elimination pass
<li>Implement an unswitching pass
<li>Write a loop unroller, with a simple heuristic for when to unroll
</ol>
<!-- ======================================================================= -->
</ul><table width="50%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
<tr><td> </td><td width=100%>
<font color="#EEEEFF" size=+1 face="Georgia,Palatino"><b>
<a name="x86be">X86 Back-end Improvements
</b></font></td></tr></table><ul>
<ol>
<li>Implement a global register allocator
<li>Implement a better instruction selector
<li>Implement support for the "switch" instruction without requiring the
lower-switches pass.
</ol>
<!-- ======================================================================= -->
</ul><table width="50%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
<tr><td> </td><td width=100%>
<font color="#EEEEFF" size=+1 face="Georgia,Palatino"><b>
<a name="misc_new">Miscellaneous Additions
</b></font></td></tr></table><ul>
<ol>
<li>Write a new frontend for some language (Java? OCaml? Forth?)
<li>Write a new backend for a target (IA64? MIPS? MMIX?)
</ol>
<!-- *********************************************************************** -->
</ul>
<!-- *********************************************************************** -->
<hr><font size-1>
<address><a href="mailto:sabre@nondot.org">Chris Lattner</a></address>
<!-- Created: Tue Aug 6 15:00:33 CDT 2002 -->
<!-- hhmts start -->
Last modified: Wed Aug 6 10:40:05 CDT 2003
<!-- hhmts end -->
</font></body></html>
|
