[Git][ghc/ghc][wip/runRW] Further improvement to the runRW# rule
Simon Peyton Jones
gitlab at gitlab.haskell.org
Wed Apr 22 10:20:08 UTC 2020
Simon Peyton Jones pushed to branch wip/runRW at Glasgow Haskell Compiler / GHC
Commits:
fad536ea by Simon Peyton Jones at 2020-04-22T11:19:40+01:00
Further improvement to the runRW# rule
Now we avoid repeated simplification
- - - - -
1 changed file:
- compiler/GHC/Core/Opt/Simplify.hs
Changes:
=====================================
compiler/GHC/Core/Opt/Simplify.hs
=====================================
@@ -1880,14 +1880,36 @@ rebuildCall env info (CastIt co cont)
rebuildCall env info (ApplyToTy { sc_arg_ty = arg_ty, sc_cont = cont })
= rebuildCall env (addTyArgTo info arg_ty) cont
-rebuildCall env info@(ArgInfo { ai_encl = encl_rules, ai_type = fun_ty
+---------- The runRW# rule. Do this after absorbing all arguments ------
+-- runRW# :: forall (r :: RuntimeRep) (o :: TYPE r). (State# RealWorld -> o) -> o
+-- K[ runRW# rr ty (\s. body) ] --> runRW rr' ty' (\s. K[ body ])
+rebuildCall env (ArgInfo { ai_fun = fun, ai_args = rev_args })
+ (ApplyToVal { sc_arg = arg, sc_env = arg_se, sc_cont = cont })
+ | fun `hasKey` runRWKey
+ , not (contIsStop cont) -- Don't fiddle around if the continuation is boring
+ , [ TyArg {}, TyArg {} ] <- rev_args
+ = do { s <- newId (fsLit "s") realWorldStatePrimTy
+ ; let env' = (arg_se `setInScopeFromE` env) `addNewInScopeIds` [s]
+ cont' = ApplyToVal { sc_dup = Simplified, sc_arg = Var s
+ , sc_env = env', sc_cont = cont }
+ ; body' <- simplExprC env' arg cont'
+ ; let arg' = Lam s body'
+ ty' = contResultType cont
+ rr' = getRuntimeRep ty'
+ call' = mkApps (Var fun) [mkTyArg rr', mkTyArg ty', arg']
+ ; return (emptyFloats env, call') }
+
+rebuildCall env info@(ArgInfo { ai_type = fun_ty, ai_encl = encl_rules
, ai_strs = str:strs, ai_discs = disc:discs })
(ApplyToVal { sc_arg = arg, sc_env = arg_se
, sc_dup = dup_flag, sc_cont = cont })
+
+ -- Argument is already simplified
| isSimplified dup_flag -- See Note [Avoid redundant simplification]
= rebuildCall env (addValArgTo info' arg) cont
- | str -- Strict argument
+ -- Strict arguments
+ | str
, sm_case_case (getMode env)
= -- pprTrace "Strict Arg" (ppr arg $$ ppr (seIdSubst env) $$ ppr (seInScope env)) $
simplExprF (arg_se `setInScopeFromE` env) arg
@@ -1895,7 +1917,8 @@ rebuildCall env info@(ArgInfo { ai_encl = encl_rules, ai_type = fun_ty
, sc_dup = Simplified, sc_cont = cont })
-- Note [Shadowing]
- | otherwise -- Lazy argument
+ -- Lazy arguments
+ | otherwise
-- DO NOT float anything outside, hence simplExprC
-- There is no benefit (unlike in a let-binding), and we'd
-- have to be very careful about bogus strictness through
@@ -1923,25 +1946,6 @@ rebuildCall env info@(ArgInfo { ai_encl = encl_rules, ai_type = fun_ty
-- It's worth an 18% improvement in allocation for this
-- particular benchmark; 5% on 'mate' and 1.3% on 'multiplier'
----------- The runRW# rule. Do this after absorbing all arguments ------
--- runRW# :: forall (r :: RuntimeRep) (o :: TYPE r). (State# RealWorld -> o) -> o
--- K[ runRW# rr ty (\s. body) ] --> runRW rr' ty' (\s. K[ body ])
-rebuildCall env (ArgInfo { ai_fun = fun, ai_args = rev_args }) cont
- | fun `hasKey` runRWKey
- , not (contIsStop cont) -- Don't fiddle around if the continuation is boring
- , [ ValArg arg
- , TyArg {}, TyArg {} ] <- rev_args
- = do { s <- newId (fsLit "s") realWorldStatePrimTy
- ; let env' = zapSubstEnv env `addNewInScopeIds` [s]
- cont' = ApplyToVal { sc_dup = Simplified, sc_arg = Var s
- , sc_env = env', sc_cont = cont }
- ; body' <- simplExprC env' arg cont'
- ; let arg' = Lam s body'
- ty' = contResultType cont
- rr' = getRuntimeRep ty'
- call' = mkApps (Var fun) [mkTyArg rr', mkTyArg ty', arg']
- ; return (emptyFloats env, call') }
-
---------- No further useful info, revert to generic rebuild ------------
rebuildCall env (ArgInfo { ai_fun = fun, ai_args = rev_args }) cont
= rebuild env (argInfoExpr fun rev_args) cont
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/fad536ea7e2622fb1102df55bd642b4153603503
--
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/fad536ea7e2622fb1102df55bd642b4153603503
You're receiving this email because of your account on gitlab.haskell.org.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-commits/attachments/20200422/55f4739f/attachment-0001.html>
More information about the ghc-commits
mailing list