-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathREADME
193 lines (161 loc) · 7.22 KB
/
README
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
removed copying source to /aux directories
removed .c .s .text binaries
examples/backprop.vlad
examples/commute.vlad
examples/dn.vlad
examples/double-agent.vlad
examples/gaussian*.vlad
examples/marble.vlad
examples/particle.vlad
examples/saddle.vlad
examples/secant.vlad
examples/forward/debugging.vlad
examples/forward/double-agent.vlad
examples/forward/particle.vlad
examples/forward/saddle.vlad
examples/forward/secant.vlad
examples/flow-tests/
examples/must-alias/
examples/no-perturbations/
examples/p-examples/
examples/popl2008/
examples/struct/
examples/tests/
examples/union2struct/
examples/widen-lists/
debugging
needs work
here I am
writing (only in c)
belongs in QobiScheme
factor out sqrt like foo in exp
eliminated
examples/2saddle/compile
examples/5saddle/compile
examples/generator-tests/run-multiply
examples/generator-tests/run-flow-analysis
examples/generator-tests/run-flow-analysis2
examples/generator-tests/run-flow-analysis3
examples/popl2009-few-real/compile
examples/popl2009-few-real/post-compile
examples/popl2009-few-real/post-compile-link-all
examples/popl2009-few-real/run-short
examples/popl2009-lots-real/compile
examples/popl2009-lots-real/post-compile
examples/popl2009-lots-real/post-compile-link-all
examples/popl2009-lots-real/run-short
examples/popl2009-lots-real/compile-nsaddle
examples/popl2009-lots-real/compile-nsaddle-real
examples/probabilistic-programming/compile
examples/probabilistic-programming/post-compile-link-all
regression-test doesn't run:
examples/2saddle
examples/5saddle
examples/generator-tests/multiple{1,2,3,4,5,6}.vlad
examples/generator-tests/
t{20a,21a,22,25{a,b,c,d},26{a,b},28{a,b,c,d}}.vlad
t{15,20,21,22} are not run in certain variants of generator-tests
qobi@tlamachilistli>fgrep "#echo" run-*
run-1:#echo Compiling t15
run-1:#echo Compiling t20
run-1:#echo Compiling t21
run-1:#echo Compiling t22
run-1:#echo Compiling t15
run-1:#echo Compiling t20
run-1:#echo Compiling t22
run-1-2:#echo Compiling t15
run-1-2:#echo Compiling t20
run-1-2:#echo Compiling t22
run-1-2:#echo Compiling t15
run-1-2:#echo Compiling t20
run-1-2:#echo Compiling t22
run-1-3:#echo Compiling t15
run-1-3:#echo Compiling t15
run-0 doesn't run t23 but the others do
-order-limit for some of t{15,16,18,19} in run-{1,1-2,1-3,widen-lists}
examples/popl2009-few-real
examples/popl2009-lots-real/nsaddle-{real-,}{F,R}{F,R}.vlad
-------------------------------------------------------------------------------
with gcc 4.3.2 sometimes i686 is better and sometimes amd64 is better
this appears to be fixed with gcc 4.5.1
sometimes i686 is better
Compiling cannon-R
amd64:
4.3.2: 1015 lines 97 adds 26 subs 125 muls 19 divs 0 sqrts 26.3% density
4.5.1: 487 lines 27 adds 12 subs 34 muls 5 divs 3 sqrts 16.6% density
i686:
4.3.2: 445 lines 27 adds 12 subs 34 muls 5 divs 0 sqrts 17.5% density
4.5.1: 437 lines 27 adds 12 subs 34 muls 5 divs 3 sqrts 18.5% density
sometimes amd64 is better
Compiling nr-sqrt-FR
amd64:
4.3.2: 450 lines 16 adds 12 subs 20 muls 6 divs 0 sqrts 12.0% density
4.5.1: 427 lines 16 adds 11 subs 20 muls 6 divs 0 sqrts 12.4% density
i686:
4.3.2: 618 lines 40 adds 18 subs 59 muls 18 divs 0 sqrts 21.8% density
4.5.1: 403 lines 16 adds 11 subs 20 muls 6 divs 0 sqrts 13.1% density
-------------------------------------------------------------------------------
no diffs forward, interpreter, ad2008 a/b, backprop, popl2009-lots-real a/b
only diffs
generator-tests a/b
either b fails some tests or b does more aggressive code elimination
generator-tests b/c
either c fails some tests or c does more aggressive code elimination
c has some code generator issues
warning: excess elements in array initializer
t{20,22}
ad2008 b/c
b has some conformance warnings due to imprecision that c doesn't have
should be no diff in code generation but there are
< 257 lines 23 adds 14 subs 27 muls 4 divs 0 sqrts 26.4% density
> 271 lines 27 adds 14 subs 28 muls 4 divs 0 sqrts 26.9% density
< 255 lines 23 adds 14 subs 27 muls 4 divs 0 sqrts 26.6% density
> 276 lines 27 adds 14 subs 28 muls 4 divs 0 sqrts 26.4% density
< 1015 lines 97 adds 26 subs 125 muls 19 divs 0 sqrts 26.3% density
> 466 lines 26 adds 13 subs 34 muls 5 divs 0 sqrts 16.7% density
< 1023 lines 97 adds 26 subs 125 muls 19 divs 0 sqrts 26.0% density
> 484 lines 26 adds 13 subs 34 muls 5 divs 0 sqrts 16.1% density
should be no diff in result but there is for cannon{,1000}-R
< 221.114669566850978
> 221.114669566850949
popl2009-lots-real b/c
b has some conformance warnings due to imprecision that c doesn't have
should be no diff in code generation but there are
< 67 lines 2 adds 0 subs 3 muls 0 divs 0 sqrts 7.4% density
> 54 lines 1 adds 0 subs 3 muls 0 divs 0 sqrts 7.4% density
< 112 lines 2 adds 9 subs 6 muls 2 divs 0 sqrts 16.9% density
> 124 lines 6 adds 15 subs 11 muls 4 divs 0 sqrts 29.0% density
< 698 lines 29 adds 0 subs 30 muls 0 divs 0 sqrts 8.4% density
> 424 lines 11 adds 0 subs 12 muls 0 divs 0 sqrts 5.4% density
< 315 lines 8 adds 6 subs 7 muls 3 divs 0 sqrts 7.6% density
> 320 lines 6 adds 7 subs 6 muls 3 divs 0 sqrts 6.8% density
< 79 lines 4 adds 0 subs 4 muls 0 divs 0 sqrts 10.1% density
> 139 lines 13 adds 0 subs 13 muls 0 divs 0 sqrts 18.7% density
< 164 lines 8 adds 17 subs 19 muls 4 divs 0 sqrts 29.2% density
> 113 lines 10 adds 16 subs 19 muls 4 divs 0 sqrts 43.3% density
< 913 lines 66 adds 47 subs 87 muls 0 divs 0 sqrts 21.9% density
> 689 lines 58 adds 38 subs 79 muls 0 divs 0 sqrts 25.3% density
< 516 lines 51 adds 28 subs 75 muls 25 divs 4 sqrts 35.4% density
> 510 lines 49 adds 28 subs 75 muls 25 divs 4 sqrts 35.4% density
< 3130 lines 18 adds 0 subs 18 muls 0 divs 0 sqrts 1.1% density
> 3562 lines 8 adds 0 subs 4 muls 0 divs 0 sqrts .3% density
< 12485 lines 37 adds 17 subs 48 muls 13 divs 0 sqrts .9% density
> 6933 lines 19 adds 21 subs 32 muls 9 divs 0 sqrts 1.1% density
< 913 lines 66 adds 47 subs 87 muls 0 divs 0 sqrts 21.9% density
> 689 lines 58 adds 38 subs 79 muls 0 divs 0 sqrts 25.3% density
< 477 lines 53 adds 20 subs 73 muls 17 divs 4 sqrts 35.0% density
> 467 lines 51 adds 20 subs 73 muls 17 divs 4 sqrts 35.3% density
< 218 lines 4 adds 0 subs 4 muls 0 divs 0 sqrts 3.6% density
> 731 lines 15 adds 0 subs 15 muls 0 divs 0 sqrts 4.1% density
< 698 lines 21 adds 14 subs 24 muls 6 divs 0 sqrts 9.3% density
> 1066 lines 15 adds 13 subs 19 muls 6 divs 0 sqrts 4.9% density
< 1563 lines 61 adds 28 subs 63 muls 1 divs 2 sqrts 9.9% density
> 1196 lines 73 adds 28 subs 59 muls 1 divs 0 sqrts 13.4% density
< 1469 lines 121 adds 30 subs 121 muls 62 divs 12 sqrts 23.5% density
> 1239 lines 121 adds 29 subs 122 muls 60 divs 12 sqrts 27.7% density
< 6930 lines 7 adds 0 subs 15 muls 0 divs 0 sqrts .3% density
> 9450 lines 9 adds 0 subs 16 muls 0 divs 0 sqrts .2% density
should be no diff in result but there is for nr-sqrt-RF
< -0.0883295263511017642
> -0.0883295263511017503
c appears unable to compile nr-sqrt-RR