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
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
|
%
% (c) The GRASP/AQUA Project, Glasgow University, 1993-1998
%
\section[Specialise]{Stamping out overloading, and (optionally) polymorphism}
\begin{code}
module Specialise ( specProgram ) where
#include "HsVersions.h"
import DynFlags ( DynFlags, DynFlag(..) )
import Id ( Id, idName, idType, mkUserLocal )
import TcType ( Type, mkTyVarTy, tcSplitSigmaTy,
tyVarsOfTypes, tyVarsOfTheta, isClassPred,
tcCmpType, isUnLiftedType
)
import CoreSubst ( Subst, mkEmptySubst, extendTvSubstList, lookupIdSubst,
substBndr, substBndrs, substTy, substInScope,
cloneIdBndr, cloneIdBndrs, cloneRecIdBndrs
)
import VarSet
import VarEnv
import CoreSyn
import CoreUtils ( applyTypeToArgs, mkPiTypes )
import CoreFVs ( exprFreeVars, exprsFreeVars, idRuleVars )
import CoreTidy ( tidyRules )
import CoreLint ( showPass, endPass )
import Rules ( addIdSpecialisations, mkLocalRule, lookupRule, emptyRuleBase, rulesOfBinds )
import PprCore ( pprRules )
import UniqSupply ( UniqSupply,
UniqSM, initUs_, thenUs, returnUs, getUniqueUs,
getUs, mapUs
)
import Name ( nameOccName, mkSpecOcc, getSrcLoc )
import MkId ( voidArgId, realWorldPrimId )
import FiniteMap
import Maybes ( catMaybes, maybeToBool )
import ErrUtils ( dumpIfSet_dyn )
import BasicTypes ( Activation( AlwaysActive ) )
import Bag
import List ( partition )
import Util ( zipEqual, zipWithEqual, cmpList, lengthIs,
equalLength, lengthAtLeast, notNull )
import Outputable
import FastString
infixr 9 `thenSM`
\end{code}
%************************************************************************
%* *
\subsection[notes-Specialise]{Implementation notes [SLPJ, Aug 18 1993]}
%* *
%************************************************************************
These notes describe how we implement specialisation to eliminate
overloading.
The specialisation pass works on Core
syntax, complete with all the explicit dictionary application,
abstraction and construction as added by the type checker. The
existing type checker remains largely as it is.
One important thought: the {\em types} passed to an overloaded
function, and the {\em dictionaries} passed are mutually redundant.
If the same function is applied to the same type(s) then it is sure to
be applied to the same dictionary(s)---or rather to the same {\em
values}. (The arguments might look different but they will evaluate
to the same value.)
Second important thought: we know that we can make progress by
treating dictionary arguments as static and worth specialising on. So
we can do without binding-time analysis, and instead specialise on
dictionary arguments and no others.
The basic idea
~~~~~~~~~~~~~~
Suppose we have
let f = <f_rhs>
in <body>
and suppose f is overloaded.
STEP 1: CALL-INSTANCE COLLECTION
We traverse <body>, accumulating all applications of f to types and
dictionaries.
(Might there be partial applications, to just some of its types and
dictionaries? In principle yes, but in practice the type checker only
builds applications of f to all its types and dictionaries, so partial
applications could only arise as a result of transformation, and even
then I think it's unlikely. In any case, we simply don't accumulate such
partial applications.)
STEP 2: EQUIVALENCES
So now we have a collection of calls to f:
f t1 t2 d1 d2
f t3 t4 d3 d4
...
Notice that f may take several type arguments. To avoid ambiguity, we
say that f is called at type t1/t2 and t3/t4.
We take equivalence classes using equality of the *types* (ignoring
the dictionary args, which as mentioned previously are redundant).
STEP 3: SPECIALISATION
For each equivalence class, choose a representative (f t1 t2 d1 d2),
and create a local instance of f, defined thus:
f@t1/t2 = <f_rhs> t1 t2 d1 d2
f_rhs presumably has some big lambdas and dictionary lambdas, so lots
of simplification will now result. However we don't actually *do* that
simplification. Rather, we leave it for the simplifier to do. If we
*did* do it, though, we'd get more call instances from the specialised
RHS. We can work out what they are by instantiating the call-instance
set from f's RHS with the types t1, t2.
Add this new id to f's IdInfo, to record that f has a specialised version.
Before doing any of this, check that f's IdInfo doesn't already
tell us about an existing instance of f at the required type/s.
(This might happen if specialisation was applied more than once, or
it might arise from user SPECIALIZE pragmas.)
Recursion
~~~~~~~~~
Wait a minute! What if f is recursive? Then we can't just plug in
its right-hand side, can we?
But it's ok. The type checker *always* creates non-recursive definitions
for overloaded recursive functions. For example:
f x = f (x+x) -- Yes I know its silly
becomes
f a (d::Num a) = let p = +.sel a d
in
letrec fl (y::a) = fl (p y y)
in
fl
We still have recusion for non-overloaded functions which we
speciailise, but the recursive call should get specialised to the
same recursive version.
Polymorphism 1
~~~~~~~~~~~~~~
All this is crystal clear when the function is applied to *constant
types*; that is, types which have no type variables inside. But what if
it is applied to non-constant types? Suppose we find a call of f at type
t1/t2. There are two possibilities:
(a) The free type variables of t1, t2 are in scope at the definition point
of f. In this case there's no problem, we proceed just as before. A common
example is as follows. Here's the Haskell:
g y = let f x = x+x
in f y + f y
After typechecking we have
g a (d::Num a) (y::a) = let f b (d'::Num b) (x::b) = +.sel b d' x x
in +.sel a d (f a d y) (f a d y)
Notice that the call to f is at type type "a"; a non-constant type.
Both calls to f are at the same type, so we can specialise to give:
g a (d::Num a) (y::a) = let f@a (x::a) = +.sel a d x x
in +.sel a d (f@a y) (f@a y)
(b) The other case is when the type variables in the instance types
are *not* in scope at the definition point of f. The example we are
working with above is a good case. There are two instances of (+.sel a d),
but "a" is not in scope at the definition of +.sel. Can we do anything?
Yes, we can "common them up", a sort of limited common sub-expression deal.
This would give:
g a (d::Num a) (y::a) = let +.sel@a = +.sel a d
f@a (x::a) = +.sel@a x x
in +.sel@a (f@a y) (f@a y)
This can save work, and can't be spotted by the type checker, because
the two instances of +.sel weren't originally at the same type.
Further notes on (b)
* There are quite a few variations here. For example, the defn of
+.sel could be floated ouside the \y, to attempt to gain laziness.
It certainly mustn't be floated outside the \d because the d has to
be in scope too.
* We don't want to inline f_rhs in this case, because
that will duplicate code. Just commoning up the call is the point.
* Nothing gets added to +.sel's IdInfo.
* Don't bother unless the equivalence class has more than one item!
Not clear whether this is all worth it. It is of course OK to
simply discard call-instances when passing a big lambda.
Polymorphism 2 -- Overloading
~~~~~~~~~~~~~~
Consider a function whose most general type is
f :: forall a b. Ord a => [a] -> b -> b
There is really no point in making a version of g at Int/Int and another
at Int/Bool, because it's only instancing the type variable "a" which
buys us any efficiency. Since g is completely polymorphic in b there
ain't much point in making separate versions of g for the different
b types.
That suggests that we should identify which of g's type variables
are constrained (like "a") and which are unconstrained (like "b").
Then when taking equivalence classes in STEP 2, we ignore the type args
corresponding to unconstrained type variable. In STEP 3 we make
polymorphic versions. Thus:
f@t1/ = /\b -> <f_rhs> t1 b d1 d2
We do this.
Dictionary floating
~~~~~~~~~~~~~~~~~~~
Consider this
f a (d::Num a) = let g = ...
in
...(let d1::Ord a = Num.Ord.sel a d in g a d1)...
Here, g is only called at one type, but the dictionary isn't in scope at the
definition point for g. Usually the type checker would build a
definition for d1 which enclosed g, but the transformation system
might have moved d1's defn inward. Solution: float dictionary bindings
outwards along with call instances.
Consider
f x = let g p q = p==q
h r s = (r+s, g r s)
in
h x x
Before specialisation, leaving out type abstractions we have
f df x = let g :: Eq a => a -> a -> Bool
g dg p q = == dg p q
h :: Num a => a -> a -> (a, Bool)
h dh r s = let deq = eqFromNum dh
in (+ dh r s, g deq r s)
in
h df x x
After specialising h we get a specialised version of h, like this:
h' r s = let deq = eqFromNum df
in (+ df r s, g deq r s)
But we can't naively make an instance for g from this, because deq is not in scope
at the defn of g. Instead, we have to float out the (new) defn of deq
to widen its scope. Notice that this floating can't be done in advance -- it only
shows up when specialisation is done.
User SPECIALIZE pragmas
~~~~~~~~~~~~~~~~~~~~~~~
Specialisation pragmas can be digested by the type checker, and implemented
by adding extra definitions along with that of f, in the same way as before
f@t1/t2 = <f_rhs> t1 t2 d1 d2
Indeed the pragmas *have* to be dealt with by the type checker, because
only it knows how to build the dictionaries d1 and d2! For example
g :: Ord a => [a] -> [a]
{-# SPECIALIZE f :: [Tree Int] -> [Tree Int] #-}
Here, the specialised version of g is an application of g's rhs to the
Ord dictionary for (Tree Int), which only the type checker can conjure
up. There might not even *be* one, if (Tree Int) is not an instance of
Ord! (All the other specialision has suitable dictionaries to hand
from actual calls.)
Problem. The type checker doesn't have to hand a convenient <f_rhs>, because
it is buried in a complex (as-yet-un-desugared) binding group.
Maybe we should say
f@t1/t2 = f* t1 t2 d1 d2
where f* is the Id f with an IdInfo which says "inline me regardless!".
Indeed all the specialisation could be done in this way.
That in turn means that the simplifier has to be prepared to inline absolutely
any in-scope let-bound thing.
Again, the pragma should permit polymorphism in unconstrained variables:
h :: Ord a => [a] -> b -> b
{-# SPECIALIZE h :: [Int] -> b -> b #-}
We *insist* that all overloaded type variables are specialised to ground types,
(and hence there can be no context inside a SPECIALIZE pragma).
We *permit* unconstrained type variables to be specialised to
- a ground type
- or left as a polymorphic type variable
but nothing in between. So
{-# SPECIALIZE h :: [Int] -> [c] -> [c] #-}
is *illegal*. (It can be handled, but it adds complication, and gains the
programmer nothing.)
SPECIALISING INSTANCE DECLARATIONS
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Consider
instance Foo a => Foo [a] where
...
{-# SPECIALIZE instance Foo [Int] #-}
The original instance decl creates a dictionary-function
definition:
dfun.Foo.List :: forall a. Foo a -> Foo [a]
The SPECIALIZE pragma just makes a specialised copy, just as for
ordinary function definitions:
dfun.Foo.List@Int :: Foo [Int]
dfun.Foo.List@Int = dfun.Foo.List Int dFooInt
The information about what instance of the dfun exist gets added to
the dfun's IdInfo in the same way as a user-defined function too.
Automatic instance decl specialisation?
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Can instance decls be specialised automatically? It's tricky.
We could collect call-instance information for each dfun, but
then when we specialised their bodies we'd get new call-instances
for ordinary functions; and when we specialised their bodies, we might get
new call-instances of the dfuns, and so on. This all arises because of
the unrestricted mutual recursion between instance decls and value decls.
Still, there's no actual problem; it just means that we may not do all
the specialisation we could theoretically do.
Furthermore, instance decls are usually exported and used non-locally,
so we'll want to compile enough to get those specialisations done.
Lastly, there's no such thing as a local instance decl, so we can
survive solely by spitting out *usage* information, and then reading that
back in as a pragma when next compiling the file. So for now,
we only specialise instance decls in response to pragmas.
SPITTING OUT USAGE INFORMATION
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
To spit out usage information we need to traverse the code collecting
call-instance information for all imported (non-prelude?) functions
and data types. Then we equivalence-class it and spit it out.
This is done at the top-level when all the call instances which escape
must be for imported functions and data types.
*** Not currently done ***
Partial specialisation by pragmas
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
What about partial specialisation:
k :: (Ord a, Eq b) => [a] -> b -> b -> [a]
{-# SPECIALIZE k :: Eq b => [Int] -> b -> b -> [a] #-}
or even
{-# SPECIALIZE k :: Eq b => [Int] -> [b] -> [b] -> [a] #-}
Seems quite reasonable. Similar things could be done with instance decls:
instance (Foo a, Foo b) => Foo (a,b) where
...
{-# SPECIALIZE instance Foo a => Foo (a,Int) #-}
{-# SPECIALIZE instance Foo b => Foo (Int,b) #-}
Ho hum. Things are complex enough without this. I pass.
Requirements for the simplifer
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The simplifier has to be able to take advantage of the specialisation.
* When the simplifier finds an application of a polymorphic f, it looks in
f's IdInfo in case there is a suitable instance to call instead. This converts
f t1 t2 d1 d2 ===> f_t1_t2
Note that the dictionaries get eaten up too!
* Dictionary selection operations on constant dictionaries must be
short-circuited:
+.sel Int d ===> +Int
The obvious way to do this is in the same way as other specialised
calls: +.sel has inside it some IdInfo which tells that if it's applied
to the type Int then it should eat a dictionary and transform to +Int.
In short, dictionary selectors need IdInfo inside them for constant
methods.
* Exactly the same applies if a superclass dictionary is being
extracted:
Eq.sel Int d ===> dEqInt
* Something similar applies to dictionary construction too. Suppose
dfun.Eq.List is the function taking a dictionary for (Eq a) to
one for (Eq [a]). Then we want
dfun.Eq.List Int d ===> dEq.List_Int
Where does the Eq [Int] dictionary come from? It is built in
response to a SPECIALIZE pragma on the Eq [a] instance decl.
In short, dfun Ids need IdInfo with a specialisation for each
constant instance of their instance declaration.
All this uses a single mechanism: the SpecEnv inside an Id
What does the specialisation IdInfo look like?
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The SpecEnv of an Id maps a list of types (the template) to an expression
[Type] |-> Expr
For example, if f has this SpecInfo:
[Int, a] -> \d:Ord Int. f' a
it means that we can replace the call
f Int t ===> (\d. f' t)
This chucks one dictionary away and proceeds with the
specialised version of f, namely f'.
What can't be done this way?
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
There is no way, post-typechecker, to get a dictionary for (say)
Eq a from a dictionary for Eq [a]. So if we find
==.sel [t] d
we can't transform to
eqList (==.sel t d')
where
eqList :: (a->a->Bool) -> [a] -> [a] -> Bool
Of course, we currently have no way to automatically derive
eqList, nor to connect it to the Eq [a] instance decl, but you
can imagine that it might somehow be possible. Taking advantage
of this is permanently ruled out.
Still, this is no great hardship, because we intend to eliminate
overloading altogether anyway!
A note about non-tyvar dictionaries
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Some Ids have types like
forall a,b,c. Eq a -> Ord [a] -> tau
This seems curious at first, because we usually only have dictionary
args whose types are of the form (C a) where a is a type variable.
But this doesn't hold for the functions arising from instance decls,
which sometimes get arguements with types of form (C (T a)) for some
type constructor T.
Should we specialise wrt this compound-type dictionary? We used to say
"no", saying:
"This is a heuristic judgement, as indeed is the fact that we
specialise wrt only dictionaries. We choose *not* to specialise
wrt compound dictionaries because at the moment the only place
they show up is in instance decls, where they are simply plugged
into a returned dictionary. So nothing is gained by specialising
wrt them."
But it is simpler and more uniform to specialise wrt these dicts too;
and in future GHC is likely to support full fledged type signatures
like
f ;: Eq [(a,b)] => ...
%************************************************************************
%* *
\subsubsection{The new specialiser}
%* *
%************************************************************************
Our basic game plan is this. For let(rec) bound function
f :: (C a, D c) => (a,b,c,d) -> Bool
* Find any specialised calls of f, (f ts ds), where
ts are the type arguments t1 .. t4, and
ds are the dictionary arguments d1 .. d2.
* Add a new definition for f1 (say):
f1 = /\ b d -> (..body of f..) t1 b t3 d d1 d2
Note that we abstract over the unconstrained type arguments.
* Add the mapping
[t1,b,t3,d] |-> \d1 d2 -> f1 b d
to the specialisations of f. This will be used by the
simplifier to replace calls
(f t1 t2 t3 t4) da db
by
(\d1 d1 -> f1 t2 t4) da db
All the stuff about how many dictionaries to discard, and what types
to apply the specialised function to, are handled by the fact that the
SpecEnv contains a template for the result of the specialisation.
We don't build *partial* specialisations for f. For example:
f :: Eq a => a -> a -> Bool
{-# SPECIALISE f :: (Eq b, Eq c) => (b,c) -> (b,c) -> Bool #-}
Here, little is gained by making a specialised copy of f.
There's a distinct danger that the specialised version would
first build a dictionary for (Eq b, Eq c), and then select the (==)
method from it! Even if it didn't, not a great deal is saved.
We do, however, generate polymorphic, but not overloaded, specialisations:
f :: Eq a => [a] -> b -> b -> b
{#- SPECIALISE f :: [Int] -> b -> b -> b #-}
Hence, the invariant is this:
*** no specialised version is overloaded ***
%************************************************************************
%* *
\subsubsection{The exported function}
%* *
%************************************************************************
\begin{code}
specProgram :: DynFlags -> UniqSupply -> [CoreBind] -> IO [CoreBind]
specProgram dflags us binds
= do
showPass dflags "Specialise"
let binds' = initSM us (go binds `thenSM` \ (binds', uds') ->
returnSM (dumpAllDictBinds uds' binds'))
endPass dflags "Specialise" Opt_D_dump_spec binds'
dumpIfSet_dyn dflags Opt_D_dump_rules "Top-level specialisations"
(pprRules (tidyRules emptyTidyEnv (rulesOfBinds binds')))
return binds'
where
-- We need to start with a Subst that knows all the things
-- that are in scope, so that the substitution engine doesn't
-- accidentally re-use a unique that's already in use
-- Easiest thing is to do it all at once, as if all the top-level
-- decls were mutually recursive
top_subst = mkEmptySubst (mkInScopeSet (mkVarSet (bindersOfBinds binds)))
go [] = returnSM ([], emptyUDs)
go (bind:binds) = go binds `thenSM` \ (binds', uds) ->
specBind top_subst bind uds `thenSM` \ (bind', uds') ->
returnSM (bind' ++ binds', uds')
\end{code}
%************************************************************************
%* *
\subsubsection{@specExpr@: the main function}
%* *
%************************************************************************
\begin{code}
specVar :: Subst -> Id -> CoreExpr
specVar subst v = lookupIdSubst subst v
specExpr :: Subst -> CoreExpr -> SpecM (CoreExpr, UsageDetails)
-- We carry a substitution down:
-- a) we must clone any binding that might flaot outwards,
-- to avoid name clashes
-- b) we carry a type substitution to use when analysing
-- the RHS of specialised bindings (no type-let!)
---------------- First the easy cases --------------------
specExpr subst (Type ty) = returnSM (Type (substTy subst ty), emptyUDs)
specExpr subst (Var v) = returnSM (specVar subst v, emptyUDs)
specExpr subst (Lit lit) = returnSM (Lit lit, emptyUDs)
specExpr subst (Note note body)
= specExpr subst body `thenSM` \ (body', uds) ->
returnSM (Note (specNote subst note) body', uds)
---------------- Applications might generate a call instance --------------------
specExpr subst expr@(App fun arg)
= go expr []
where
go (App fun arg) args = specExpr subst arg `thenSM` \ (arg', uds_arg) ->
go fun (arg':args) `thenSM` \ (fun', uds_app) ->
returnSM (App fun' arg', uds_arg `plusUDs` uds_app)
go (Var f) args = case specVar subst f of
Var f' -> returnSM (Var f', mkCallUDs subst f' args)
e' -> returnSM (e', emptyUDs) -- I don't expect this!
go other args = specExpr subst other
---------------- Lambda/case require dumping of usage details --------------------
specExpr subst e@(Lam _ _)
= specExpr subst' body `thenSM` \ (body', uds) ->
let
(filtered_uds, body'') = dumpUDs bndrs' uds body'
in
returnSM (mkLams bndrs' body'', filtered_uds)
where
(bndrs, body) = collectBinders e
(subst', bndrs') = substBndrs subst bndrs
-- More efficient to collect a group of binders together all at once
-- and we don't want to split a lambda group with dumped bindings
specExpr subst (Case scrut case_bndr ty alts)
= specExpr subst scrut `thenSM` \ (scrut', uds_scrut) ->
mapAndCombineSM spec_alt alts `thenSM` \ (alts', uds_alts) ->
returnSM (Case scrut' case_bndr' (substTy subst ty) alts', uds_scrut `plusUDs` uds_alts)
where
(subst_alt, case_bndr') = substBndr subst case_bndr
-- No need to clone case binder; it can't float like a let(rec)
spec_alt (con, args, rhs)
= specExpr subst_rhs rhs `thenSM` \ (rhs', uds) ->
let
(uds', rhs'') = dumpUDs args uds rhs'
in
returnSM ((con, args', rhs''), uds')
where
(subst_rhs, args') = substBndrs subst_alt args
---------------- Finally, let is the interesting case --------------------
specExpr subst (Let bind body)
= -- Clone binders
cloneBindSM subst bind `thenSM` \ (rhs_subst, body_subst, bind') ->
-- Deal with the body
specExpr body_subst body `thenSM` \ (body', body_uds) ->
-- Deal with the bindings
specBind rhs_subst bind' body_uds `thenSM` \ (binds', uds) ->
-- All done
returnSM (foldr Let body' binds', uds)
-- Must apply the type substitution to coerceions
specNote subst (Coerce t1 t2) = Coerce (substTy subst t1) (substTy subst t2)
specNote subst note = note
\end{code}
%************************************************************************
%* *
\subsubsection{Dealing with a binding}
%* *
%************************************************************************
\begin{code}
specBind :: Subst -- Use this for RHSs
-> CoreBind
-> UsageDetails -- Info on how the scope of the binding
-> SpecM ([CoreBind], -- New bindings
UsageDetails) -- And info to pass upstream
specBind rhs_subst bind body_uds
= specBindItself rhs_subst bind (calls body_uds) `thenSM` \ (bind', bind_uds) ->
let
bndrs = bindersOf bind
all_uds = zapCalls bndrs (body_uds `plusUDs` bind_uds)
-- It's important that the `plusUDs` is this way round,
-- because body_uds may bind dictionaries that are
-- used in the calls passed to specDefn. So the
-- dictionary bindings in bind_uds may mention
-- dictionaries bound in body_uds.
in
case splitUDs bndrs all_uds of
(_, ([],[])) -- This binding doesn't bind anything needed
-- in the UDs, so put the binding here
-- This is the case for most non-dict bindings, except
-- for the few that are mentioned in a dict binding
-- that is floating upwards in body_uds
-> returnSM ([bind'], all_uds)
(float_uds, (dict_binds, calls)) -- This binding is needed in the UDs, so float it out
-> returnSM ([], float_uds `plusUDs` mkBigUD bind' dict_binds calls)
-- A truly gruesome function
mkBigUD bind@(NonRec _ _) dbs calls
= -- Common case: non-recursive and no specialisations
-- (if there were any specialistions it would have been made recursive)
MkUD { dict_binds = listToBag (mkDB bind : dbs),
calls = listToCallDetails calls }
mkBigUD bind dbs calls
= -- General case
MkUD { dict_binds = unitBag (mkDB (Rec (bind_prs bind ++ dbsToPairs dbs))),
-- Make a huge Rec
calls = listToCallDetails calls }
where
bind_prs (NonRec b r) = [(b,r)]
bind_prs (Rec prs) = prs
dbsToPairs [] = []
dbsToPairs ((bind,_):dbs) = bind_prs bind ++ dbsToPairs dbs
-- specBindItself deals with the RHS, specialising it according
-- to the calls found in the body (if any)
specBindItself rhs_subst (NonRec bndr rhs) call_info
= specDefn rhs_subst call_info (bndr,rhs) `thenSM` \ ((bndr',rhs'), spec_defns, spec_uds) ->
let
new_bind | null spec_defns = NonRec bndr' rhs'
| otherwise = Rec ((bndr',rhs'):spec_defns)
-- bndr' mentions the spec_defns in its SpecEnv
-- Not sure why we couln't just put the spec_defns first
in
returnSM (new_bind, spec_uds)
specBindItself rhs_subst (Rec pairs) call_info
= mapSM (specDefn rhs_subst call_info) pairs `thenSM` \ stuff ->
let
(pairs', spec_defns_s, spec_uds_s) = unzip3 stuff
spec_defns = concat spec_defns_s
spec_uds = plusUDList spec_uds_s
new_bind = Rec (spec_defns ++ pairs')
in
returnSM (new_bind, spec_uds)
specDefn :: Subst -- Subst to use for RHS
-> CallDetails -- Info on how it is used in its scope
-> (Id, CoreExpr) -- The thing being bound and its un-processed RHS
-> SpecM ((Id, CoreExpr), -- The thing and its processed RHS
-- the Id may now have specialisations attached
[(Id,CoreExpr)], -- Extra, specialised bindings
UsageDetails -- Stuff to fling upwards from the RHS and its
) -- specialised versions
specDefn subst calls (fn, rhs)
-- The first case is the interesting one
| rhs_tyvars `lengthIs` n_tyvars -- Rhs of fn's defn has right number of big lambdas
&& rhs_bndrs `lengthAtLeast` n_dicts -- and enough dict args
&& notNull calls_for_me -- And there are some calls to specialise
-- At one time I tried not specialising small functions
-- but sometimes there are big functions marked INLINE
-- that we'd like to specialise. In particular, dictionary
-- functions, which Marcin is keen to inline
-- && not (certainlyWillInline fn) -- And it's not small
-- If it's small, it's better just to inline
-- it than to construct lots of specialisations
= -- Specialise the body of the function
specExpr subst rhs `thenSM` \ (rhs', rhs_uds) ->
-- Make a specialised version for each call in calls_for_me
mapSM spec_call calls_for_me `thenSM` \ stuff ->
let
(spec_defns, spec_uds, spec_rules) = unzip3 stuff
fn' = addIdSpecialisations fn spec_rules
in
returnSM ((fn',rhs'),
spec_defns,
rhs_uds `plusUDs` plusUDList spec_uds)
| otherwise -- No calls or RHS doesn't fit our preconceptions
= specExpr subst rhs `thenSM` \ (rhs', rhs_uds) ->
returnSM ((fn, rhs'), [], rhs_uds)
where
fn_type = idType fn
(tyvars, theta, _) = tcSplitSigmaTy fn_type
n_tyvars = length tyvars
n_dicts = length theta
(rhs_tyvars, rhs_ids, rhs_body)
= collectTyAndValBinders (dropInline rhs)
-- It's important that we "see past" any INLINE pragma
-- else we'll fail to specialise an INLINE thing
rhs_dicts = take n_dicts rhs_ids
rhs_bndrs = rhs_tyvars ++ rhs_dicts
body = mkLams (drop n_dicts rhs_ids) rhs_body
-- Glue back on the non-dict lambdas
calls_for_me = case lookupFM calls fn of
Nothing -> []
Just cs -> fmToList cs
----------------------------------------------------------
-- Specialise to one particular call pattern
spec_call :: (CallKey, ([DictExpr], VarSet)) -- Call instance
-> SpecM ((Id,CoreExpr), -- Specialised definition
UsageDetails, -- Usage details from specialised body
CoreRule) -- Info for the Id's SpecEnv
spec_call (CallKey call_ts, (call_ds, call_fvs))
= ASSERT( call_ts `lengthIs` n_tyvars && call_ds `lengthIs` n_dicts )
-- Calls are only recorded for properly-saturated applications
-- Suppose f's defn is f = /\ a b c d -> \ d1 d2 -> rhs
-- Supppose the call is for f [Just t1, Nothing, Just t3, Nothing] [dx1, dx2]
-- Construct the new binding
-- f1 = SUBST[a->t1,c->t3, d1->d1', d2->d2'] (/\ b d -> rhs)
-- PLUS the usage-details
-- { d1' = dx1; d2' = dx2 }
-- where d1', d2' are cloned versions of d1,d2, with the type substitution applied.
--
-- Note that the substitution is applied to the whole thing.
-- This is convenient, but just slightly fragile. Notably:
-- * There had better be no name clashes in a/b/c/d
--
let
-- poly_tyvars = [b,d] in the example above
-- spec_tyvars = [a,c]
-- ty_args = [t1,b,t3,d]
poly_tyvars = [tv | (tv, Nothing) <- rhs_tyvars `zip` call_ts]
spec_tyvars = [tv | (tv, Just _) <- rhs_tyvars `zip` call_ts]
ty_args = zipWithEqual "spec_call" mk_ty_arg rhs_tyvars call_ts
where
mk_ty_arg rhs_tyvar Nothing = Type (mkTyVarTy rhs_tyvar)
mk_ty_arg rhs_tyvar (Just ty) = Type ty
rhs_subst = extendTvSubstList subst (spec_tyvars `zip` [ty | Just ty <- call_ts])
in
cloneBinders rhs_subst rhs_dicts `thenSM` \ (rhs_subst', rhs_dicts') ->
let
inst_args = ty_args ++ map Var rhs_dicts'
-- Figure out the type of the specialised function
body_ty = applyTypeToArgs rhs fn_type inst_args
(lam_args, app_args) -- Add a dummy argument if body_ty is unlifted
| isUnLiftedType body_ty -- C.f. WwLib.mkWorkerArgs
= (poly_tyvars ++ [voidArgId], poly_tyvars ++ [realWorldPrimId])
| otherwise = (poly_tyvars, poly_tyvars)
spec_id_ty = mkPiTypes lam_args body_ty
in
newIdSM fn spec_id_ty `thenSM` \ spec_f ->
specExpr rhs_subst' (mkLams lam_args body) `thenSM` \ (spec_rhs, rhs_uds) ->
let
-- The rule to put in the function's specialisation is:
-- forall b,d, d1',d2'. f t1 b t3 d d1' d2' = f1 b d
spec_env_rule = mkLocalRule (mkFastString ("SPEC " ++ showSDoc (ppr fn)))
AlwaysActive (idName fn)
(poly_tyvars ++ rhs_dicts')
inst_args
(mkVarApps (Var spec_f) app_args)
-- Add the { d1' = dx1; d2' = dx2 } usage stuff
final_uds = foldr addDictBind rhs_uds (my_zipEqual "spec_call" rhs_dicts' call_ds)
-- NOTE: we don't add back in any INLINE pragma on the RHS, so even if
-- the original function said INLINE, the specialised copies won't.
-- The idea is that the point of inlining was precisely to specialise
-- the function at its call site, and that's not so important for the
-- specialised copies. But it still smells like an ad hoc decision.
in
returnSM ((spec_f, spec_rhs),
final_uds,
spec_env_rule)
where
my_zipEqual doc xs ys
| not (equalLength xs ys) = pprPanic "my_zipEqual" (ppr xs $$ ppr ys $$ (ppr fn <+> ppr call_ts) $$ ppr rhs)
| otherwise = zipEqual doc xs ys
dropInline :: CoreExpr -> CoreExpr
dropInline (Note InlineMe rhs) = rhs
dropInline rhs = rhs
\end{code}
%************************************************************************
%* *
\subsubsection{UsageDetails and suchlike}
%* *
%************************************************************************
\begin{code}
data UsageDetails
= MkUD {
dict_binds :: !(Bag DictBind),
-- Floated dictionary bindings
-- The order is important;
-- in ds1 `union` ds2, bindings in ds2 can depend on those in ds1
-- (Remember, Bags preserve order in GHC.)
calls :: !CallDetails
}
type DictBind = (CoreBind, VarSet)
-- The set is the free vars of the binding
-- both tyvars and dicts
type DictExpr = CoreExpr
emptyUDs = MkUD { dict_binds = emptyBag, calls = emptyFM }
type ProtoUsageDetails = ([DictBind],
[(Id, CallKey, ([DictExpr], VarSet))]
)
------------------------------------------------------------
type CallDetails = FiniteMap Id CallInfo
newtype CallKey = CallKey [Maybe Type] -- Nothing => unconstrained type argument
type CallInfo = FiniteMap CallKey
([DictExpr], VarSet) -- Dict args and the vars of the whole
-- call (including tyvars)
-- [*not* include the main id itself, of course]
-- The finite maps eliminate duplicates
-- The list of types and dictionaries is guaranteed to
-- match the type of f
-- Type isn't an instance of Ord, so that we can control which
-- instance we use. That's tiresome here. Oh well
instance Eq CallKey where
k1 == k2 = case k1 `compare` k2 of { EQ -> True; other -> False }
instance Ord CallKey where
compare (CallKey k1) (CallKey k2) = cmpList cmp k1 k2
where
cmp Nothing Nothing = EQ
cmp Nothing (Just t2) = LT
cmp (Just t1) Nothing = GT
cmp (Just t1) (Just t2) = tcCmpType t1 t2
unionCalls :: CallDetails -> CallDetails -> CallDetails
unionCalls c1 c2 = plusFM_C plusFM c1 c2
singleCall :: Id -> [Maybe Type] -> [DictExpr] -> CallDetails
singleCall id tys dicts
= unitFM id (unitFM (CallKey tys) (dicts, call_fvs))
where
call_fvs = exprsFreeVars dicts `unionVarSet` tys_fvs
tys_fvs = tyVarsOfTypes (catMaybes tys)
-- The type args (tys) are guaranteed to be part of the dictionary
-- types, because they are just the constrained types,
-- and the dictionary is therefore sure to be bound
-- inside the binding for any type variables free in the type;
-- hence it's safe to neglect tyvars free in tys when making
-- the free-var set for this call
-- BUT I don't trust this reasoning; play safe and include tys_fvs
--
-- We don't include the 'id' itself.
listToCallDetails calls
= foldr (unionCalls . mk_call) emptyFM calls
where
mk_call (id, tys, dicts_w_fvs) = unitFM id (unitFM tys dicts_w_fvs)
-- NB: the free vars of the call are provided
callDetailsToList calls = [ (id,tys,dicts)
| (id,fm) <- fmToList calls,
(tys, dicts) <- fmToList fm
]
mkCallUDs subst f args
| null theta
|| not (all isClassPred theta)
-- Only specialise if all overloading is on class params.
-- In ptic, with implicit params, the type args
-- *don't* say what the value of the implicit param is!
|| not (spec_tys `lengthIs` n_tyvars)
|| not ( dicts `lengthIs` n_dicts)
|| maybeToBool (lookupRule (\act -> True) (substInScope subst) emptyRuleBase f args)
-- There's already a rule covering this call. A typical case
-- is where there's an explicit user-provided rule. Then
-- we don't want to create a specialised version
-- of the function that overlaps.
= emptyUDs -- Not overloaded, or no specialisation wanted
| otherwise
= MkUD {dict_binds = emptyBag,
calls = singleCall f spec_tys dicts
}
where
(tyvars, theta, _) = tcSplitSigmaTy (idType f)
constrained_tyvars = tyVarsOfTheta theta
n_tyvars = length tyvars
n_dicts = length theta
spec_tys = [mk_spec_ty tv ty | (tv, Type ty) <- tyvars `zip` args]
dicts = [dict_expr | (_, dict_expr) <- theta `zip` (drop n_tyvars args)]
mk_spec_ty tyvar ty
| tyvar `elemVarSet` constrained_tyvars = Just ty
| otherwise = Nothing
------------------------------------------------------------
plusUDs :: UsageDetails -> UsageDetails -> UsageDetails
plusUDs (MkUD {dict_binds = db1, calls = calls1})
(MkUD {dict_binds = db2, calls = calls2})
= MkUD {dict_binds = d, calls = c}
where
d = db1 `unionBags` db2
c = calls1 `unionCalls` calls2
plusUDList = foldr plusUDs emptyUDs
-- zapCalls deletes calls to ids from uds
zapCalls ids uds = uds {calls = delListFromFM (calls uds) ids}
mkDB bind = (bind, bind_fvs bind)
bind_fvs (NonRec bndr rhs) = pair_fvs (bndr,rhs)
bind_fvs (Rec prs) = foldl delVarSet rhs_fvs bndrs
where
bndrs = map fst prs
rhs_fvs = unionVarSets (map pair_fvs prs)
pair_fvs (bndr, rhs) = exprFreeVars rhs `unionVarSet` idRuleVars bndr
-- Don't forget variables mentioned in the
-- rules of the bndr. C.f. OccAnal.addRuleUsage
addDictBind (dict,rhs) uds = uds { dict_binds = mkDB (NonRec dict rhs) `consBag` dict_binds uds }
dumpAllDictBinds (MkUD {dict_binds = dbs}) binds
= foldrBag add binds dbs
where
add (bind,_) binds = bind : binds
dumpUDs :: [CoreBndr]
-> UsageDetails -> CoreExpr
-> (UsageDetails, CoreExpr)
dumpUDs bndrs uds body
= (free_uds, foldr add_let body dict_binds)
where
(free_uds, (dict_binds, _)) = splitUDs bndrs uds
add_let (bind,_) body = Let bind body
splitUDs :: [CoreBndr]
-> UsageDetails
-> (UsageDetails, -- These don't mention the binders
ProtoUsageDetails) -- These do
splitUDs bndrs uds@(MkUD {dict_binds = orig_dbs,
calls = orig_calls})
= if isEmptyBag dump_dbs && null dump_calls then
-- Common case: binder doesn't affect floats
(uds, ([],[]))
else
-- Binders bind some of the fvs of the floats
(MkUD {dict_binds = free_dbs,
calls = listToCallDetails free_calls},
(bagToList dump_dbs, dump_calls)
)
where
bndr_set = mkVarSet bndrs
(free_dbs, dump_dbs, dump_idset)
= foldlBag dump_db (emptyBag, emptyBag, bndr_set) orig_dbs
-- Important that it's foldl not foldr;
-- we're accumulating the set of dumped ids in dump_set
-- Filter out any calls that mention things that are being dumped
orig_call_list = callDetailsToList orig_calls
(dump_calls, free_calls) = partition captured orig_call_list
captured (id,tys,(dicts, fvs)) = fvs `intersectsVarSet` dump_idset
|| id `elemVarSet` dump_idset
dump_db (free_dbs, dump_dbs, dump_idset) db@(bind, fvs)
| dump_idset `intersectsVarSet` fvs -- Dump it
= (free_dbs, dump_dbs `snocBag` db,
extendVarSetList dump_idset (bindersOf bind))
| otherwise -- Don't dump it
= (free_dbs `snocBag` db, dump_dbs, dump_idset)
\end{code}
%************************************************************************
%* *
\subsubsection{Boring helper functions}
%* *
%************************************************************************
\begin{code}
type SpecM a = UniqSM a
thenSM = thenUs
returnSM = returnUs
getUniqSM = getUniqueUs
mapSM = mapUs
initSM = initUs_
mapAndCombineSM f [] = returnSM ([], emptyUDs)
mapAndCombineSM f (x:xs) = f x `thenSM` \ (y, uds1) ->
mapAndCombineSM f xs `thenSM` \ (ys, uds2) ->
returnSM (y:ys, uds1 `plusUDs` uds2)
cloneBindSM :: Subst -> CoreBind -> SpecM (Subst, Subst, CoreBind)
-- Clone the binders of the bind; return new bind with the cloned binders
-- Return the substitution to use for RHSs, and the one to use for the body
cloneBindSM subst (NonRec bndr rhs)
= getUs `thenUs` \ us ->
let
(subst', bndr') = cloneIdBndr subst us bndr
in
returnUs (subst, subst', NonRec bndr' rhs)
cloneBindSM subst (Rec pairs)
= getUs `thenUs` \ us ->
let
(subst', bndrs') = cloneRecIdBndrs subst us (map fst pairs)
in
returnUs (subst', subst', Rec (bndrs' `zip` map snd pairs))
cloneBinders subst bndrs
= getUs `thenUs` \ us ->
returnUs (cloneIdBndrs subst us bndrs)
newIdSM old_id new_ty
= getUniqSM `thenSM` \ uniq ->
let
-- Give the new Id a similar occurrence name to the old one
name = idName old_id
new_id = mkUserLocal (mkSpecOcc (nameOccName name)) uniq new_ty (getSrcLoc name)
in
returnSM new_id
\end{code}
Old (but interesting) stuff about unboxed bindings
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
What should we do when a value is specialised to a *strict* unboxed value?
map_*_* f (x:xs) = let h = f x
t = map f xs
in h:t
Could convert let to case:
map_*_Int# f (x:xs) = case f x of h# ->
let t = map f xs
in h#:t
This may be undesirable since it forces evaluation here, but the value
may not be used in all branches of the body. In the general case this
transformation is impossible since the mutual recursion in a letrec
cannot be expressed as a case.
There is also a problem with top-level unboxed values, since our
implementation cannot handle unboxed values at the top level.
Solution: Lift the binding of the unboxed value and extract it when it
is used:
map_*_Int# f (x:xs) = let h = case (f x) of h# -> _Lift h#
t = map f xs
in case h of
_Lift h# -> h#:t
Now give it to the simplifier and the _Lifting will be optimised away.
The benfit is that we have given the specialised "unboxed" values a
very simplep lifted semantics and then leave it up to the simplifier to
optimise it --- knowing that the overheads will be removed in nearly
all cases.
In particular, the value will only be evaluted in the branches of the
program which use it, rather than being forced at the point where the
value is bound. For example:
filtermap_*_* p f (x:xs)
= let h = f x
t = ...
in case p x of
True -> h:t
False -> t
==>
filtermap_*_Int# p f (x:xs)
= let h = case (f x) of h# -> _Lift h#
t = ...
in case p x of
True -> case h of _Lift h#
-> h#:t
False -> t
The binding for h can still be inlined in the one branch and the
_Lifting eliminated.
Question: When won't the _Lifting be eliminated?
Answer: When they at the top-level (where it is necessary) or when
inlining would duplicate work (or possibly code depending on
options). However, the _Lifting will still be eliminated if the
strictness analyser deems the lifted binding strict.
|
