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
|
commit 5efbf0d243984444cf352ad6f0d147e226c64498
Author: Sergei Trofimovich <slyfox@gentoo.org>
Date: Thu Sep 1 17:34:58 2016 +0100
restore -fmax-worker-args handling (Trac #11565)
maxWorkerArgs handling was accidentally lost 3 years ago
in a major update of demand analysis
commit 0831a12ea2fc73c33652eeec1adc79fa19700578
Old regression is noticeable as:
- code bloat (requires stack reshuffling)
- compilation slowdown (more code to optimise/generate)
- and increased heap usage (DynFlags unboxing/reboxing?)
On a simple compile benchmark this change causes heap
allocation drop from 70G don to 67G (ghc perf build).
Signed-off-by: Sergei Trofimovich <siarheit@google.com>
Reviewers: simonpj, ezyang, goldfire, austin, bgamari
Reviewed By: simonpj, ezyang
Subscribers: thomie
Differential Revision: https://phabricator.haskell.org/D2503
GHC Trac Issues: #11565
diff --git a/compiler/specialise/SpecConstr.hs b/compiler/specialise/SpecConstr.hs
index 10d5614..7166f57 100644
--- a/compiler/specialise/SpecConstr.hs
+++ b/compiler/specialise/SpecConstr.hs
@@ -29,7 +29,7 @@ import CoreFVs ( exprsFreeVarsList )
import CoreMonad
import Literal ( litIsLifted )
import HscTypes ( ModGuts(..) )
-import WwLib ( mkWorkerArgs )
+import WwLib ( isWorkerSmallEnough, mkWorkerArgs )
import DataCon
import Coercion hiding( substCo )
import Rules
@@ -1533,10 +1533,14 @@ specialise env bind_calls (RI { ri_fn = fn, ri_lam_bndrs = arg_bndrs
| Just all_calls <- lookupVarEnv bind_calls fn
= -- pprTrace "specialise entry {" (ppr fn <+> ppr (length all_calls)) $
- do { (boring_call, pats) <- callsToPats env specs arg_occs all_calls
-
+ do { (boring_call, all_pats) <- callsToPats env specs arg_occs all_calls
-- Bale out if too many specialisations
- ; let n_pats = length pats
+ ; let pats = filter (is_small_enough . fst) all_pats
+ is_small_enough vars = isWorkerSmallEnough (sc_dflags env) vars
+ -- We are about to construct w/w pair in 'spec_one'.
+ -- Omit specialisation leading to high arity workers.
+ -- See Note [Limit w/w arity]
+ n_pats = length pats
spec_count' = n_pats + spec_count
; case sc_count env of
Just max | not (sc_force env) && spec_count' > max
diff --git a/compiler/stranal/WwLib.hs b/compiler/stranal/WwLib.hs
index 09bc204..d9460d9 100644
--- a/compiler/stranal/WwLib.hs
+++ b/compiler/stranal/WwLib.hs
@@ -8,6 +8,7 @@
module WwLib ( mkWwBodies, mkWWstr, mkWorkerArgs
, deepSplitProductType_maybe, findTypeShape
+ , isWorkerSmallEnough
) where
#include "HsVersions.h"
@@ -144,7 +145,8 @@ mkWwBodies dflags fam_envs fun_ty demands res_info one_shots
wrapper_body = wrap_fn_args . wrap_fn_cpr . wrap_fn_str . applyToVars work_call_args . Var
worker_body = mkLams work_lam_args. work_fn_str . work_fn_cpr . work_fn_args
- ; if useful1 && not (only_one_void_argument) || useful2
+ ; if isWorkerSmallEnough dflags work_args
+ && (useful1 && not only_one_void_argument || useful2)
then return (Just (worker_args_dmds, wrapper_body, worker_body))
else return Nothing
}
@@ -165,6 +167,12 @@ mkWwBodies dflags fam_envs fun_ty demands res_info one_shots
| otherwise
= False
+-- See Note [Limit w/w arity]
+isWorkerSmallEnough :: DynFlags -> [Var] -> Bool
+isWorkerSmallEnough dflags vars = count isId vars <= maxWorkerArgs dflags
+ -- We count only Free variables (isId) to skip Type, Kind
+ -- variables which have no runtime representation.
+
{-
Note [Always do CPR w/w]
~~~~~~~~~~~~~~~~~~~~~~~~
@@ -178,6 +186,30 @@ a disaster, because then the enclosing function might say it has the CPR
property, but now doesn't and there a cascade of disaster. A good example
is Trac #5920.
+Note [Limit w/w arity]
+~~~~~~~~~~~~~~~~~~~~~~~~
+Guard against high worker arity as it generates a lot of stack traffic.
+A simplified example is Trac #11565#comment:6
+
+Current strategy is very simple: don't perform w/w transformation at all
+if the result produces a wrapper with arity higher than -fmax-worker-args=.
+
+It is a bit all or nothing, consider
+
+ f (x,y) (a,b,c,d,e ... , z) = rhs
+
+Currently we will remove all w/w ness entirely. But actually we could
+w/w on the (x,y) pair... it's the huge product that is the problem.
+
+Could we instead refrain from w/w on an arg-by-arg basis? Yes, that'd
+solve f. But we can get a lot of args from deeply-nested products:
+
+ g (a, (b, (c, (d, ...)))) = rhs
+
+This is harder to spot on an arg-by-arg basis. Previously mkWwStr was
+given some "fuel" saying how many arguments it could add; when we ran
+out of fuel it would stop w/wing.
+Still not very clever because it had a left-right bias.
************************************************************************
* *
|
GitWeb