| 1 : |
dbm |
637 |
|
| 2 : |
|
|
S M L / N J
|
| 3 : |
|
|
|
| 4 : |
|
|
1 1 0 . 2 8 N E W S
|
| 5 : |
|
|
|
| 6 : |
|
|
April 10, 2000
|
| 7 : |
|
|
|
| 8 : |
|
|
WARNING
|
| 9 : |
|
|
|
| 10 : |
|
|
This version is intended for compiler hackers.
|
| 11 : |
|
|
We are in the midst of substantial structural changes,
|
| 12 : |
|
|
and this is a snapshot.
|
| 13 : |
|
|
|
| 14 : |
|
|
http://cm.bell-labs.com/cm/cs/what/smlnj/index.html
|
| 15 : |
|
|
|
| 16 : |
|
|
------------------------------------------------------------------------
|
| 17 : |
|
|
Summary:
|
| 18 : |
|
|
|
| 19 : |
|
|
* This version has some tuning of phase ordering for FLINT, which
|
| 20 : |
|
|
mostly solves the excessive space problem during compilation. With
|
| 21 : |
|
|
this tuning, 110.28 is slightly better than 110.25 in performance,
|
| 22 : |
|
|
but not quite as good as 110.25+latest MLRISC. See the benchmarks
|
| 23 : |
|
|
below. Futher work on the register allocator in MLRISC should
|
| 24 : |
|
|
improve space performance further.
|
| 25 : |
|
|
|
| 26 : |
|
|
* CM
|
| 27 : |
|
|
Installation scripts have been tweaked. A new noweb tool has been
|
| 28 : |
|
|
added. ".fun" suffix has been added to those that imply sml compilation.
|
| 29 : |
|
|
New version numbering scheme added. Alternative syntax for anchored
|
| 30 : |
|
|
paths added. Parallel make made smarter. Support for "unsharing".
|
| 31 : |
|
|
Simple "makedepend" functionality added. Further details below.
|
| 32 : |
|
|
|
| 33 : |
|
|
See src/system/README and the latest version of the CM manual at
|
| 34 : |
|
|
|
| 35 : |
|
|
<//http://www.kurims.kyoto-u.ac.jp/~blume/SMLNJ-DEV/manual/index.html>
|
| 36 : |
|
|
<//http://www.kurims.kyoto-u.ac.jp/~blume/SMLNJ-DEV/manual.ps>
|
| 37 : |
|
|
|
| 38 : |
|
|
for further information about user-visible changes.
|
| 39 : |
|
|
|
| 40 : |
|
|
* A few MLRISC changes, most not SML/NJ related.
|
| 41 : |
|
|
|
| 42 : |
|
|
* ML-Yacc: updated manual and examples to be SML '97, SML/NJ 110+ compliant.
|
| 43 : |
|
|
Examples now compiled with CM.
|
| 44 : |
|
|
|
| 45 : |
|
|
* Reported bug fixes (details in dist/smlnj/working/bugs.1400-1600):
|
| 46 : |
|
|
1498. Specialized real arrays not pretty-printed correctly
|
| 47 : |
|
|
1510. Signature matching bug makes "casts" possible
|
| 48 : |
|
|
1562. CM complains about unrecognized file extension ".fun"
|
| 49 : |
|
|
1563. redundant pathconfig contents with multiple builds
|
| 50 : |
|
|
|
| 51 : |
|
|
|
| 52 : |
|
|
======================================================================
|
| 53 : |
|
|
Details of changes
|
| 54 : |
|
|
|
| 55 : |
|
|
|
| 56 : |
|
|
======================================================================
|
| 57 : |
|
|
FLINT:
|
| 58 : |
|
|
======================================================================
|
| 59 : |
|
|
|
| 60 : |
|
|
Tuned ordering of phases to improve time/space performance. See
|
| 61 : |
|
|
src/compiler/FLINT/opt/flintopt.txt for description of some of the
|
| 62 : |
|
|
issues.
|
| 63 : |
|
|
|
| 64 : |
|
|
|
| 65 : |
|
|
======================================================================
|
| 66 : |
|
|
MLRISC:
|
| 67 : |
|
|
======================================================================
|
| 68 : |
|
|
|
| 69 : |
|
|
This update synchronizes my repository with Yale's. Most of these
|
| 70 : |
|
|
changes, however, do not affect SML/NJ at all (the RA is an exception).
|
| 71 : |
|
|
|
| 72 : |
|
|
1. Register Allocator
|
| 73 : |
|
|
|
| 74 : |
|
|
a. An improvement in the interference graph construction:
|
| 75 : |
|
|
Given a copy
|
| 76 : |
|
|
|
| 77 : |
|
|
s <- t
|
| 78 : |
|
|
|
| 79 : |
|
|
no interference edge between s and t is added for this definition of s.
|
| 80 : |
|
|
|
| 81 : |
|
|
b. I've added two new spill heuristic modules that Fermin and I developed
|
| 82 : |
|
|
(in the new library RA.cm). These are unused in SML/NJ but maybe
|
| 83 : |
|
|
useful for others (Moby?)
|
| 84 : |
|
|
|
| 85 : |
|
|
c. Added a flag "ra-dump-size" to print out the size of the flowgraph
|
| 86 : |
|
|
and the interference graph.
|
| 87 : |
|
|
|
| 88 : |
|
|
2. X86
|
| 89 : |
|
|
|
| 90 : |
|
|
a. Various fixes in the backend provided by Fermin [C--] and Lal.
|
| 91 : |
|
|
|
| 92 : |
|
|
3. Alpha
|
| 93 : |
|
|
|
| 94 : |
|
|
a. Added the BSR instruction and code generation that goes with it [C--]
|
| 95 : |
|
|
|
| 96 : |
|
|
b. Other fixes too numerous to recount provided by Fermin [C--]
|
| 97 : |
|
|
|
| 98 : |
|
|
c. PSEUDOARITH was missing in AlphaRewrite. This causes an endless
|
| 99 : |
|
|
loop in C--.
|
| 100 : |
|
|
|
| 101 : |
|
|
4. Regmaps
|
| 102 : |
|
|
|
| 103 : |
|
|
a. The regmaps are not initialized with the identity physical bindings
|
| 104 : |
|
|
at creation time. This is unneeded.
|
| 105 : |
|
|
|
| 106 : |
|
|
5. MLRISC Optimizations
|
| 107 : |
|
|
|
| 108 : |
|
|
a. The DJ-Graph module can now compute the iterated dominance frontiers
|
| 109 : |
|
|
intersects with liveness incrementally in linear time! Woohoo!
|
| 110 : |
|
|
This is now used in my new SSA construction algorithm.
|
| 111 : |
|
|
|
| 112 : |
|
|
b. THe branch reorganization module is now smarter about linear chains of
|
| 113 : |
|
|
basic blocks.
|
| 114 : |
|
|
|
| 115 : |
|
|
|
| 116 : |
|
|
======================================================================
|
| 117 : |
|
|
CM
|
| 118 : |
|
|
======================================================================
|
| 119 : |
|
|
|
| 120 : |
|
|
Several manual updates
|
| 121 : |
|
|
|
| 122 : |
|
|
I always try to keep the manual in sync with CM's latest features.
|
| 123 : |
|
|
|
| 124 : |
|
|
New noweb tool
|
| 125 : |
|
|
|
| 126 : |
|
|
The existing system is entirely unaffected by this, but some CM users
|
| 127 : |
|
|
have asked for renewed noweb support. Everything is documented in the
|
| 128 : |
|
|
CM manual.
|
| 129 : |
|
|
|
| 130 : |
|
|
New (plugin) libraries:
|
| 131 : |
|
|
|
| 132 : |
|
|
noweb-tool.cm
|
| 133 : |
|
|
nw-ext.cm
|
| 134 : |
|
|
|
| 135 : |
|
|
CM version numbering added
|
| 136 : |
|
|
|
| 137 : |
|
|
This is an implementation of Lal's proposal for adding version
|
| 138 : |
|
|
numbers and version checking to .cm files. Lal said that his
|
| 139 : |
|
|
proposal was just that -- a proposal. For the time being I went
|
| 140 : |
|
|
ahead and implemented it so that people can comment on it.
|
| 141 : |
|
|
Everything is completely backward-compatible (except for the stable
|
| 142 : |
|
|
library format, i.e., new bootfiles!).
|
| 143 : |
|
|
|
| 144 : |
|
|
As usual, see the CM manual for details.
|
| 145 : |
|
|
|
| 146 : |
|
|
Alternative syntax for anchored paths
|
| 147 : |
|
|
|
| 148 : |
|
|
Dave has recently voiced the same concerns that I had when I did
|
| 149 : |
|
|
this, so there should be some support. My take is that eventually
|
| 150 : |
|
|
I will let support for the current syntax (where anchors are
|
| 151 : |
|
|
"implicit") fade out in favor of the new, explicit syntax.
|
| 152 : |
|
|
In order to be backward-compatible, both old and new syntax are
|
| 153 : |
|
|
currently supported.
|
| 154 : |
|
|
|
| 155 : |
|
|
Again, see the CM manual for details.
|
| 156 : |
|
|
|
| 157 : |
|
|
Parallel make is trying to be slightly smarter
|
| 158 : |
|
|
|
| 159 : |
|
|
When the master process finds a "bottleneck", i.e., when there is
|
| 160 : |
|
|
only one compilation unit that can be compiled and everybody else
|
| 161 : |
|
|
is waiting on it, then it will simply compile it directly instead
|
| 162 : |
|
|
of clumsily telling one of the slaves to do it.
|
| 163 : |
|
|
|
| 164 : |
|
|
Support for "unsharing" added
|
| 165 : |
|
|
|
| 166 : |
|
|
This is necessary in order to be able to have two different
|
| 167 : |
|
|
versions of the same library running at the same time (e.g., for
|
| 168 : |
|
|
trying out a new MLRISC while still having the old MLRISC linked
|
| 169 : |
|
|
into the current compiler, etc.) See the CM manual.
|
| 170 : |
|
|
|
| 171 : |
|
|
makedepend
|
| 172 : |
|
|
|
| 173 : |
|
|
Simple "makedepend" functionality added for generating Makefile
|
| 174 : |
|
|
dependency information. (This is rather crude at the moment.
|
| 175 : |
|
|
Expect some changes here in the future.)
|
| 176 : |
|
|
|
| 177 : |
|
|
New sml class prefix
|
| 178 : |
|
|
|
| 179 : |
|
|
".fun" added as a recognized suffix for ML files. Also documented
|
| 180 : |
|
|
explicitly in the manual that the fallback behavior (unknown suffix
|
| 181 : |
|
|
-> ML file) is not an official feature!
|
| 182 : |
|
|
|
| 183 : |
|
|
Pickler
|
| 184 : |
|
|
|
| 185 : |
|
|
Small changes to the pickler for stable libraries.
|
| 186 : |
|
|
|
| 187 : |
|
|
Internal cleanup
|
| 188 : |
|
|
|
| 189 : |
|
|
Several internal changes to CM (for cleanup/improvement).
|
| 190 : |
|
|
|
| 191 : |
|
|
install.sh changes
|
| 192 : |
|
|
|
| 193 : |
|
|
I changed config/install.sh to remove duplicate entries from the
|
| 194 : |
|
|
lib/pathconfig file at the end. Moreover, the final version of
|
| 195 : |
|
|
lib/pathconfig is sorted alphabetically. The same (sorting) is done
|
| 196 : |
|
|
in src/system/installml.
|
| 197 : |
|
|
|
| 198 : |
|
|
The config/install.sh script now consistently uses relative
|
| 199 : |
|
|
pathnames in lib/pathconfig whenever the anchor is in the lib
|
| 200 : |
|
|
directory. (So far this was true for the libraries that come
|
| 201 : |
|
|
pre-compiled and bundled as part of the bootfiles but not for
|
| 202 : |
|
|
libraries that are compiled by the script itself.)
|
| 203 : |
|
|
|
| 204 : |
|
|
Changed install.sh script to handle archive files without version number
|
| 205 : |
|
|
and to use "boot.<arch>-<os>" instead of "sml.boot.<arch>-<os>" for the
|
| 206 : |
|
|
name of the boot file archive.
|
| 207 : |
|
|
|
| 208 : |
|
|
|
| 209 : |
|
|
======================================================================
|
| 210 : |
|
|
BENCHMARKS
|
| 211 : |
|
|
======================================================================
|
| 212 : |
|
|
Allen Leung, 28 April 2000:
|
| 213 : |
|
|
|
| 214 : |
|
|
I've rerun the benchmarks to see if anything has slowed down in
|
| 215 : |
|
|
MLRISC recently, but found nothing. I compared 110.25, 110.25+latest
|
| 216 : |
|
|
MLRISC, and 110.27+ (which also has the latest MLRISC). 110.25 uses
|
| 217 : |
|
|
cpsopt, and 110.27+ uses flintopt. [110.27+ is repository state
|
| 218 : |
|
|
at the end of April, which is essentially 110.28 - dbm].
|
| 219 : |
|
|
|
| 220 : |
|
|
The results are:
|
| 221 : |
|
|
|
| 222 : |
|
|
110.25 versus 110.25+latest MLRISC
|
| 223 : |
|
|
----------------------------------
|
| 224 : |
|
|
Name Compilation Runtime Speedup
|
| 225 : |
|
|
barnesHut 5.560 5.368 3.57% 3.690 3.302 11.76%
|
| 226 : |
|
|
boyer 8.678 7.672 13.12% 0.313 0.312 0.53%
|
| 227 : |
|
|
count-graphs 2.048 1.915 6.96% 28.577 27.128 5.34%
|
| 228 : |
|
|
fft 1.325 1.233 7.43% 1.060 0.982 7.98%
|
| 229 : |
|
|
knuthBendix 5.337 4.680 14.03% 0.962 0.918 4.72%
|
| 230 : |
|
|
lexgen 10.310 9.950 3.62% 0.917 0.902 1.66%
|
| 231 : |
|
|
life 1.230 1.127 9.17% 0.128 0.100 28.33%
|
| 232 : |
|
|
logic 3.533 3.258 8.44% 5.750 5.413 6.22%
|
| 233 : |
|
|
mandelbrot 0.145 0.148 -2.25% 0.700 0.685 2.19%
|
| 234 : |
|
|
mlyacc 34.557 33.342 3.64% 0.553 0.533 3.75%
|
| 235 : |
|
|
nucleic 6.675 6.507 2.59% 0.173 0.167 4.00%
|
| 236 : |
|
|
ratio-regions 6.358 6.218 2.25% 119.753 120.772 -0.84%
|
| 237 : |
|
|
ray 2.280 2.247 1.48% 3.563 3.517 1.33%
|
| 238 : |
|
|
simple 9.798 9.650 1.54% 2.987 3.083 -3.14%
|
| 239 : |
|
|
tsp 1.763 1.528 15.38% 8.657 7.718 12.16%
|
| 240 : |
|
|
vliw 33.938 32.570 4.20% 1.982 2.000 -0.92%
|
| 241 : |
|
|
Average speedup: 5.32%
|
| 242 : |
|
|
Average compile time speedup: 5.95%
|
| 243 : |
|
|
|
| 244 : |
|
|
As you can see, MLRISC on x86 has improved slightly since the
|
| 245 : |
|
|
110.25. Now comparing 110.25 with 110.27+:
|
| 246 : |
|
|
|
| 247 : |
|
|
110.25 versus 110.27+
|
| 248 : |
|
|
---------------------
|
| 249 : |
|
|
Name Compilation Runtime Speedup
|
| 250 : |
|
|
barnesHut 5.560 5.568 -0.15% 3.690 3.347 10.26%
|
| 251 : |
|
|
boyer 8.678 10.058 -13.72% 0.313 0.322 -2.59%
|
| 252 : |
|
|
count-graphs 2.048 2.125 -3.61% 28.577 31.533 -9.38%
|
| 253 : |
|
|
fft 1.325 1.297 2.19% 1.060 1.052 0.79%
|
| 254 : |
|
|
knuthBendix 5.337 3.737 42.82% 0.962 1.175 -18.16%
|
| 255 : |
|
|
lexgen 10.310 9.985 3.25% 0.917 0.992 -7.56%
|
| 256 : |
|
|
life 1.230 1.105 11.31% 0.128 0.125 2.67%
|
| 257 : |
|
|
logic 3.533 3.302 7.02% 5.750 5.357 7.34%
|
| 258 : |
|
|
mandelbrot 0.145 0.162 -10.31% 0.700 0.753 -7.08%
|
| 259 : |
|
|
mlyacc 34.557 37.292 -7.33% 0.553 0.493 12.16%
|
| 260 : |
|
|
nucleic 6.675 7.717 -13.50% 0.173 0.165 5.05%
|
| 261 : |
|
|
ratio-regions 6.358 3.915 62.41% 119.753 126.153 -5.07%
|
| 262 : |
|
|
ray 2.280 2.142 6.46% 3.563 2.827 26.06%
|
| 263 : |
|
|
simple 9.798 10.067 -2.67% 2.987 2.812 6.22%
|
| 264 : |
|
|
tsp 1.763 1.605 9.87% 8.657 8.870 -2.41%
|
| 265 : |
|
|
vliw 33.938 46.907 -27.65% 1.982 1.802 9.99%
|
| 266 : |
|
|
Average speedup: 1.77%
|
| 267 : |
|
|
Average compile time speedup: 4.15%
|
| 268 : |
|
|
|
| 269 : |
|
|
110.25+latest MLRISC versus 110.27+
|
| 270 : |
|
|
-----------------------------------
|
| 271 : |
|
|
Name Compilation Runtime Speedup
|
| 272 : |
|
|
barnesHut 5.368 5.568 -3.59% 3.302 3.347 -1.34%
|
| 273 : |
|
|
boyer 7.672 10.058 -23.73% 0.312 0.322 -3.11%
|
| 274 : |
|
|
count-graphs 1.915 2.125 -9.88% 27.128 31.533 -13.97%
|
| 275 : |
|
|
fft 1.233 1.297 -4.88% 0.982 1.052 -6.66%
|
| 276 : |
|
|
knuthBendix 4.680 3.737 25.25% 0.918 1.175 -21.84%
|
| 277 : |
|
|
lexgen 9.950 9.985 -0.35% 0.902 0.992 -9.08%
|
| 278 : |
|
|
life 1.127 1.105 1.96% 0.100 0.125 -20.00%
|
| 279 : |
|
|
logic 3.258 3.302 -1.31% 5.413 5.357 1.06%
|
| 280 : |
|
|
mandelbrot 0.148 0.162 -8.25% 0.685 0.753 -9.07%
|
| 281 : |
|
|
mlyacc 33.342 37.292 -10.59% 0.533 0.493 8.11%
|
| 282 : |
|
|
nucleic 6.507 7.717 -15.68% 0.167 0.165 1.01%
|
| 283 : |
|
|
ratio-regions 6.218 3.915 58.83% 120.772 126.153 -4.27%
|
| 284 : |
|
|
ray 2.247 2.142 4.90% 3.517 2.827 24.41%
|
| 285 : |
|
|
simple 9.650 10.067 -4.14% 3.083 2.812 9.66%
|
| 286 : |
|
|
tsp 1.528 1.605 -4.78% 7.718 8.870 -12.98%
|
| 287 : |
|
|
vliw 32.570 46.907 -30.56% 2.000 1.802 11.01%
|
| 288 : |
|
|
Average speedup: -2.94%
|
| 289 : |
|
|
Average compile time speedup: -1.68%
|
| 290 : |
|
|
|
| 291 : |
|
|
Overall, I'd say 110.27+ is pretty competitive with 110.25. There are
|
| 292 : |
|
|
some big improvements (ray, vliw, mlyacc) which can only be attributed to FLINT
|
| 293 : |
|
|
changes. But something has slowed down (tsp, knuth-bendix, life, mandelbrot,
|
| 294 : |
|
|
count-graphs). [We know the reasons with knuth-bendix.]
|
| 295 : |
|
|
Also, compilation time in 110.27+ is generally slower (which is offset
|
| 296 : |
|
|
by huge improvements in knuth-bendix and ratio-regions).
|
Parent Directory
| 
Revision Log
