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

3 files changed, 37 insertions, 5 deletions

diff --git a/compiler/stranal/WwLib.hs b/compiler/stranal/WwLib.hs
index 0057f6fb78..1a096052db 100644
--- a/compiler/stranal/WwLib.hs
+++ b/compiler/stranal/WwLib.hs
@@ -143,7 +143,7 @@ mkWwBodies dflags fam_envs fun_ty demands res_info
               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 is_small_enough work_args && (useful1 && not only_one_void_argument || useful2)
           then return (Just (worker_args_dmds, wrapper_body, worker_body))
           else return Nothing
         }
@@ -163,6 +163,10 @@ mkWwBodies dflags fam_envs fun_ty demands res_info
       = True
       | otherwise
       = False
+    is_small_enough args = count isId args <= maxWorkerArgs dflags
+          -- See Note [Limit w/w arity]
+          -- We count only Free variables (isId) to skip Type, Kind
+          -- variables which have no runtime representation.
 
 {-
 Note [Always do CPR w/w]
@@ -177,6 +181,30 @@ 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.
 
 ************************************************************************
 *                                                                      *
diff --git a/testsuite/tests/perf/compiler/all.T b/testsuite/tests/perf/compiler/all.T
index 3c8cbdabf9..130ba44691 100644
--- a/testsuite/tests/perf/compiler/all.T
+++ b/testsuite/tests/perf/compiler/all.T
@@ -428,8 +428,9 @@ test('T5631',
 test('parsing001',
      [compiler_stats_num_field('bytes allocated',
           [(wordsize(32), 274000576, 10),
-           (wordsize(64), 587079016, 5)]),
+           (wordsize(64), 581551384, 5)]),
         # expected value: 587079016 (amd64/Linux)
+        # 2016-09-01:     581551384 (amd64/Linux) Restore w/w limit (#11565)
        only_ways(['normal']),
       ],
      compile_fail, [''])
@@ -767,7 +768,7 @@ test('T9872d',
 test('T9961',
      [ only_ways(['normal']),
        compiler_stats_num_field('bytes allocated',
-          [(wordsize(64), 568526784, 5),
+          [(wordsize(64), 537297968, 5),
           # 2015-01-12    807117816   Initally created
           # 2015-spring   772510192   Got better
           # 2015-05-22    663978160   Fix for #10370 improves it more
@@ -775,6 +776,7 @@ test('T9961',
           # 2015-12-17    745044392   x86_64/Darwin  Creep upwards
           # 2016-03-20    519436672   x64_64/Linux   Don't use build desugaring for large lists (#11707)
           # 2016-03-24    568526784   x64_64/Linux   Add eqInt* variants (#11688)
+          # 2016-09-01    537297968   x64_64/Linux   Restore w/w limit (#11565)
            (wordsize(32), 275264188, 5)
           # was           375647160
           # 2016-04-06    275264188   x86/Linux
diff --git a/testsuite/tests/perf/space_leaks/all.T b/testsuite/tests/perf/space_leaks/all.T
index c6b1d925a6..301029cf58 100644
--- a/testsuite/tests/perf/space_leaks/all.T
+++ b/testsuite/tests/perf/space_leaks/all.T
@@ -58,17 +58,19 @@ test('T4018',
 
 test('T4029',
      [stats_num_field('peak_megabytes_allocated',
-          [(wordsize(64), 82, 10)]),
+          [(wordsize(64), 71, 10)]),
             # 2016-02-26: 66 (amd64/Linux)           INITIAL
             # 2016-05-23: 82 (amd64/Linux)           Use -G1
             # 2016-07-13: 92 (amd64/Linux)           Changes to tidyType
+            # 2016-09-01: 71 (amd64/Linux)           Restore w/w limit (#11565)
       stats_num_field('max_bytes_used',
-          [(wordsize(64), 22920616, 5)]),
+          [(wordsize(64), 21648488, 5)]),
             # 2016-02-26: 24071720 (amd64/Linux)     INITIAL
             # 2016-04-21: 25542832 (amd64/Linux)
             # 2016-05-23: 25247216 (amd64/Linux)     Use -G1
             # 2016-07-13: 27575416 (amd64/Linux)     Changes to tidyType
             # 2016-07-20: 22920616 (amd64/Linux)     Fix laziness of instance matching
+            # 2016-09-01: 21648488 (amd64/Linux)     Restore w/w limit (#11565)
       extra_hc_opts('+RTS -G1 -RTS' ),
       ],
      ghci_script,