- ghc-commits - Haskell.org

[Git][ghc/ghc][wip/jeltsch/improve-closure-property-check] Improve the definition of the closure property
by Wolfgang Jeltsch (＠jeltsch) 15 Sep '26

15 Sep '26

Wolfgang Jeltsch pushed to branch wip/jeltsch/improve-closure-property-check at Glasgow Haskell Compiler / GHC Commits: bbeff40d by Wolfgang Jeltsch at 2026-09-15T20:44:07+03:00 Improve the definition of the closure property - - - - - 1 changed file: - compiler/GHC/Unit/Env.hs Changes: ===================================== compiler/GHC/Unit/Env.hs ===================================== @@ -443,11 +443,10 @@ The flow: Closure Property ---------------- -A unit environment must have the closure property, which means that, whenever -some units /h/₁ and /h/₂ have been loaded as home units, /h/₁ does not directly -or indirectly depend on an external unit that directly or indirectly depends -on /h/₂. 'GHC.Driver.Downsweep.checkHomeUnitsClosed' checks whether a given unit -environment indeed has this property. +A unit environment must have the closure property, which means that a home unit +must not directly or indirectly depend on an external unit that directly or +indirectly depends on a home unit. 'GHC.Driver.Downsweep.checkHomeUnitsClosed' +checks whether a given unit environment indeed has this property. Offsetting Paths ---------------- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/bbeff40da6d01f69a2741e9997e236e… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/bbeff40da6d01f69a2741e9997e236e… You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications: https://gitlab.haskell.org/-/profile/notifications | Help: https://gitlab.haskell.org/help

1 0

[Git][ghc/ghc][wip/apk/expose-overloaded-unfoldings] Enable -fexpose-overloaded-unfoldings by default when optimizations are on.
by Andreas Klebinger (＠AndreasK) 15 Sep '26

15 Sep '26

Andreas Klebinger pushed to branch wip/apk/expose-overloaded-unfoldings at Glasgow Haskell Compiler / GHC Commits: b1eb8241 by Andreas Klebinger at 2026-09-15T14:53:55+02:00 Enable -fexpose-overloaded-unfoldings by default when optimizations are on. - - - - - 4 changed files: - + changelog.d/T13090-expose-overloaded-unfoldings - compiler/GHC/Driver/DynFlags.hs - compiler/GHC/Iface/Tidy.hs - docs/users_guide/using-optimisation.rst Changes: ===================================== changelog.d/T13090-expose-overloaded-unfoldings ===================================== @@ -0,0 +1,6 @@ +section: compiler +issues: #13090 +mrs: !16684 +synopsis: + Enable :ghc-flag:`-fexpose-overloaded-unfoldings` by default with :ghc-flag:`-O` or higher. + ===================================== compiler/GHC/Driver/DynFlags.hs ===================================== @@ -1337,6 +1337,7 @@ optLevelFlags -- see Note [Documenting optimisation flags] , ([1,2], Opt_Specialise) , ([1,2], Opt_PolymorphicSpecialisation) -- Now on by default (#23559) , ([1,2], Opt_CrossModuleSpecialise) + , ([1,2], Opt_ExposeOverloadedUnfoldings) -- See Note [Exposing overloaded functions] in GHC.Iface.Tidy , ([1,2], Opt_InlineGenerics) , ([1,2], Opt_Strictness) , ([1,2], Opt_UnboxSmallStrictFields) ===================================== compiler/GHC/Iface/Tidy.hs ===================================== @@ -806,8 +806,19 @@ anyway we allow for some false positives and simply expose all functions which have a constraint. This means we might expose more unhelpful unfoldings. But it seems like the better choice. -Currently this option is off by default and has to be enabled manually. But -we might change this in the future. +We enable the flag by default with -O/O2 to allow users to opt into specialization +at call sites. We do not enable -fspecialize-aggressively. See also #13090. + +In terms of performance impact the cost is very modest: +On GHC and it's boot libraries overall enabling this flag increases overall disk +space used by ~1%. With a highly variant distribution across various .hi files. +In terms of compile times the overhead is low enough to be hard to measure, but +seemed to be less than 0.2% if anything at all. + +It seems tempting to do the same for -fspecialize-aggressively, but that comes in +at a hefty 20% compile time overhead. So we only provide the unfoldings and leave +the choice of how much specialization to enable up to the user. + -} {- ===================================== docs/users_guide/using-optimisation.rst ===================================== @@ -624,10 +624,10 @@ as such you shouldn't need to set any of them explicitly. A flag :reverse: -fno-expose-overloaded-unfoldings :category: - :default: off + :default: -O and above :since: 9.12.1 - This experimental flag is a slightly less heavy weight alternative + This flag is a slightly less heavy weight alternative to :ghc-flag:`-fexpose-all-unfoldings`. Instead of exposing all functions it exposes only those functions which View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/b1eb8241e455377551bba1c3c889ae4… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/b1eb8241e455377551bba1c3c889ae4… You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications: https://gitlab.haskell.org/-/profile/notifications | Help: https://gitlab.haskell.org/help

1 0

[Git][ghc/ghc][wip/T27724] Redesign the noinline magic
by Simon Peyton Jones (＠simonpj) 15 Sep '26

15 Sep '26

Simon Peyton Jones pushed to branch wip/T27724 at Glasgow Haskell Compiler / GHC Commits: dc55d4ce by Simon Peyton Jones at 2026-09-15T17:34:57+01:00 Redesign the noinline magic Main idea: * When simplifying `noinline (f e1 e2)`, switch off inlining when simplifying `f` * Mechanism: see `simplOutId`, the `noinlineIdKey` case * If we see `noinline (f e1) e2`, the simplifier moves the `e2` inside; see the `splitContArgs` call. This is important so you can write `noinliine f e1 e2` and have it mean the same as `noinline (f e1 e2)`. * Now `f` sees all its arguments, henc: * Fixes #27724 * Fixes #23911 * Also fixes #27803 where we were not respecting `noinline` if the RHS was trivial See Note [noinlineId magic] for lots of specifics. It's surprisingly tricky! For some reason a couple of perf tests allocate 1.2% less at compile time,n namely T29872 and T8095. Geom mean decease of 0.1% Metric Decrease: T9872d - - - - - 32 changed files: - + changelog.d/T27724 - compiler/GHC/Builtin/WiredIn/Ids.hs - compiler/GHC/Builtin/WiredIn/Prim.hs - compiler/GHC/Core/Opt/SetLevels.hs - compiler/GHC/Core/Opt/Simplify/Env.hs - compiler/GHC/Core/Opt/Simplify/Iteration.hs - compiler/GHC/Core/Opt/Simplify/Utils.hs - compiler/GHC/Core/TyCo/Rep.hs - compiler/GHC/Core/Type.hs - compiler/GHC/CoreToIface.hs - compiler/GHC/CoreToStg/Prep.hs - compiler/GHC/Driver/Config/Core/Opt/Simplify.hs - compiler/GHC/HsToCore.hs - compiler/GHC/HsToCore/Expr.hs - compiler/GHC/Tc/Instance/Class.hs - compiler/GHC/Tc/Types/Origin.hs - libraries/ghc-internal/src/GHC/Internal/Magic.hs - testsuite/tests/corelint/LintEtaExpand.hs - + testsuite/tests/dmdanal/should_compile/T23911.hs - + testsuite/tests/dmdanal/should_compile/T23911.stderr - testsuite/tests/dmdanal/should_compile/all.T - testsuite/tests/interface-stability/base-exports.stdout - testsuite/tests/interface-stability/ghc-experimental-exports.stdout - testsuite/tests/interface-stability/ghc-prim-exports.stdout - testsuite/tests/simplCore/should_compile/T18995.stderr - + testsuite/tests/simplCore/should_compile/T27724a.hs - + testsuite/tests/simplCore/should_compile/T27724a.stderr - + testsuite/tests/simplCore/should_compile/T27803.hs - + testsuite/tests/simplCore/should_compile/T27803.stderr - testsuite/tests/simplCore/should_compile/all.T - testsuite/tests/typecheck/should_fail/T17940.stderr - utils/genprimopcode/Main.hs The diff was not included because it is too large. View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/dc55d4ce900445241c228693fe3bc1a… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/dc55d4ce900445241c228693fe3bc1a… You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications: https://gitlab.haskell.org/-/profile/notifications | Help: https://gitlab.haskell.org/help

1 0

[Git][ghc/ghc][wip/wip/27786] hadrian: pass transitive dependencies to haddock
by Zubin (＠wz1000) 15 Sep '26

15 Sep '26

Zubin pushed to branch wip/wip/27786 at Glasgow Haskell Compiler / GHC Commits: 8a368ffb by Zubin Duggal at 2026-09-15T17:18:26+05:30 hadrian: pass transitive dependencies to haddock Haddock looks up the docs of a reexported name in the interface of the package defining it, which may not be a direct dependency. Since base reexports ghc-internal, packages reexporting from base (e.g. stm) couldn't find docs from transitive dependencies. Fixes #25850 Fixes #27786 - - - - - 2 changed files: - + changelog.d/hadrian-haddock-transitive-deps-27786 - hadrian/src/Rules/Documentation.hs Changes: ===================================== changelog.d/hadrian-haddock-transitive-deps-27786 ===================================== @@ -0,0 +1,6 @@ +section: packaging +synopsis: Hadrian now passes the interfaces of all transitive dependencies to + Haddock, so documentation of entities reexported from ``ghc-internal`` via + ``base`` (e.g. in ``stm``) is no longer missing. +issues: #27786 #25850 +mrs: !16691 ===================================== hadrian/src/Rules/Documentation.hs ===================================== @@ -422,9 +422,19 @@ buildManPage = do checkSphinxWarnings dir copyFileUntracked (dir -/- "ghc.1") file --- | Find the Haddock files for the dependencies of the current library. +-- | Find the Haddock files for the transitive dependencies of the current +-- library. Transitive because the package defining a reexported name may not be +-- a direct dependency (#27786). haddockDependencies :: Context -> Action [(Package, FilePath)] haddockDependencies context = do - depNames <- interpretInContext context (getContextData depNames) - sequence [ (,) <$> pure depPkg <*> (pkgHaddockFile $ vanillaContext (stage context) depPkg) - | Just depPkg <- map findPackageByName depNames, (pkgName depPkg) `notElem` haddockExclude ] + depPkgs <- go [] [Context.package context] + sequence [ (,) depPkg <$> pkgHaddockFile (vanillaContext (stage context) depPkg) + | depPkg <- depPkgs ] + where + go seen [] = return seen + go seen (pkg:pkgs) = do + names <- interpretInContext (context { Context.package = pkg }) (getContextData depNames) + let new = [ depPkg | Just depPkg <- map findPackageByName names + , pkgName depPkg `notElem` haddockExclude + , depPkg `notElem` seen ] + go (seen ++ new) (pkgs ++ new) View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/8a368ffb00003902fc5d33291921431… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/8a368ffb00003902fc5d33291921431… You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications: https://gitlab.haskell.org/-/profile/notifications | Help: https://gitlab.haskell.org/help

1 0

[Git][ghc/ghc][wip/wip/27786] hadrian: pass transitive dependencies to haddock
by Zubin (＠wz1000) 15 Sep '26

15 Sep '26

Zubin pushed to branch wip/wip/27786 at Glasgow Haskell Compiler / GHC Commits: d49468d9 by Zubin Duggal at 2026-09-15T17:18:03+05:30 hadrian: pass transitive dependencies to haddock Haddock looks up the docs of a reexported name in the interface of the package defining it, which may not be a direct dependency. Since base reexports ghc-internal, packages reexporting from base (e.g. stm) couldn't find docs from transitive dependencies. Fixes #27786 - - - - - 2 changed files: - + changelog.d/hadrian-haddock-transitive-deps-27786 - hadrian/src/Rules/Documentation.hs Changes: ===================================== changelog.d/hadrian-haddock-transitive-deps-27786 ===================================== @@ -0,0 +1,6 @@ +section: packaging +synopsis: Hadrian now passes the interfaces of all transitive dependencies to + Haddock, so documentation of entities reexported from ``ghc-internal`` via + ``base`` (e.g. in ``stm``) is no longer missing. +issues: #27786 #25850 +mrs: !16691 ===================================== hadrian/src/Rules/Documentation.hs ===================================== @@ -422,9 +422,19 @@ buildManPage = do checkSphinxWarnings dir copyFileUntracked (dir -/- "ghc.1") file --- | Find the Haddock files for the dependencies of the current library. +-- | Find the Haddock files for the transitive dependencies of the current +-- library. Transitive because the package defining a reexported name may not be +-- a direct dependency (#27786). haddockDependencies :: Context -> Action [(Package, FilePath)] haddockDependencies context = do - depNames <- interpretInContext context (getContextData depNames) - sequence [ (,) <$> pure depPkg <*> (pkgHaddockFile $ vanillaContext (stage context) depPkg) - | Just depPkg <- map findPackageByName depNames, (pkgName depPkg) `notElem` haddockExclude ] + depPkgs <- go [] [Context.package context] + sequence [ (,) depPkg <$> pkgHaddockFile (vanillaContext (stage context) depPkg) + | depPkg <- depPkgs ] + where + go seen [] = return seen + go seen (pkg:pkgs) = do + names <- interpretInContext (context { Context.package = pkg }) (getContextData depNames) + let new = [ depPkg | Just depPkg <- map findPackageByName names + , pkgName depPkg `notElem` haddockExclude + , depPkg `notElem` seen ] + go (seen ++ new) (pkgs ++ new) View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/d49468d97e7c3516c7497edf77bca2f… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/d49468d97e7c3516c7497edf77bca2f… You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications: https://gitlab.haskell.org/-/profile/notifications | Help: https://gitlab.haskell.org/help

1 0

[Git][ghc/ghc][wip/wip/27786] hadrian: pass transitive dependencies to haddock
by Zubin (＠wz1000) 15 Sep '26

15 Sep '26

Zubin pushed to branch wip/wip/27786 at Glasgow Haskell Compiler / GHC Commits: 37c4aee1 by Zubin Duggal at 2026-09-15T17:13:55+05:30 hadrian: pass transitive dependencies to haddock Haddock looks up the docs of a reexported name in the interface of the package defining it, which may not be a direct dependency. Since base reexports ghc-internal, packages reexporting from base (e.g. stm) couldn't find docs from transitive dependencies. Fixes #27786 - - - - - 2 changed files: - + changelog.d/hadrian-haddock-transitive-deps-27786 - hadrian/src/Rules/Documentation.hs Changes: ===================================== changelog.d/hadrian-haddock-transitive-deps-27786 ===================================== @@ -0,0 +1,6 @@ +section: packaging +synopsis: Hadrian now passes the interfaces of all transitive dependencies to + Haddock, so documentation of entities reexported from ``ghc-internal`` via + ``base`` (e.g. in ``stm``) is no longer missing. +issues: #27786 +mrs: !16691 ===================================== hadrian/src/Rules/Documentation.hs ===================================== @@ -422,9 +422,19 @@ buildManPage = do checkSphinxWarnings dir copyFileUntracked (dir -/- "ghc.1") file --- | Find the Haddock files for the dependencies of the current library. +-- | Find the Haddock files for the transitive dependencies of the current +-- library. Transitive because the package defining a reexported name may not be +-- a direct dependency (#27786). haddockDependencies :: Context -> Action [(Package, FilePath)] haddockDependencies context = do - depNames <- interpretInContext context (getContextData depNames) - sequence [ (,) <$> pure depPkg <*> (pkgHaddockFile $ vanillaContext (stage context) depPkg) - | Just depPkg <- map findPackageByName depNames, (pkgName depPkg) `notElem` haddockExclude ] + depPkgs <- go [] [Context.package context] + sequence [ (,) depPkg <$> pkgHaddockFile (vanillaContext (stage context) depPkg) + | depPkg <- depPkgs ] + where + go seen [] = return seen + go seen (pkg:pkgs) = do + names <- interpretInContext (context { Context.package = pkg }) (getContextData depNames) + let new = [ depPkg | Just depPkg <- map findPackageByName names + , pkgName depPkg `notElem` haddockExclude + , depPkg `notElem` seen ] + go (seen ++ new) (pkgs ++ new) View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/37c4aee1c36919778f805dc1f46f7e0… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/37c4aee1c36919778f805dc1f46f7e0… You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications: https://gitlab.haskell.org/-/profile/notifications | Help: https://gitlab.haskell.org/help

1 0

[Git][ghc/ghc][wip/T27731] More simplification
by Simon Peyton Jones (＠simonpj) 15 Sep '26

15 Sep '26

Simon Peyton Jones pushed to branch wip/T27731 at Glasgow Haskell Compiler / GHC Commits: 71669d7e by Simon Peyton Jones at 2026-09-15T12:42:39+01:00 More simplification - - - - - 1 changed file: - compiler/GHC/Tc/Errors.hs Changes: ===================================== compiler/GHC/Tc/Errors.hs ===================================== @@ -481,7 +481,8 @@ reportImplic ctxt implic@(Implic { ic_skols = tvs ctxt' = ctxt1 { cec_tidy = env1 , cec_encl = implic' : cec_encl ctxt - , cec_suppress = insoluble || cec_suppress ctxt +-- NO LONGER NEEDED +-- , cec_suppress = cec_suppress ctxt -- Suppress inessential errors if there -- are insolubles anywhere in the -- tree rooted here, or we've come across @@ -572,8 +573,9 @@ mkErrorItem ct , ei_suppress = suppressCtError ev }} -- | Actually report this 'ErrorItem'. -unsuppressErrorItem :: ErrorItem -> ErrorItem -unsuppressErrorItem ei = ei { ei_suppress = False } +-- NOT NEEDED ANY MORE +--unsuppressErrorItem :: ErrorItem -> ErrorItem +-- unsuppressErrorItem ei = ei { ei_suppress = False } -- | Should we completely ignore this constraint in error reporting? -- It *must* be the case that any constraint for which this returns True @@ -647,19 +649,23 @@ reportWanteds ctxt tc_lvl wc@(WC { wc_simple = simples, wc_impl = implics | isEmptyWC wc = traceTc "reportWanteds empty WC" empty | otherwise - = do { tidy_items1 <- mapMaybeM mkErrorItem tidy_cts + = do { let tidy_cts = bagToList (mapBag (tidyCt env) simples) + tidy_errs = bagToList (mapBag (tidyDelayedError env) errs) + + ; tidy_items1 <- mapMaybeM mkErrorItem tidy_cts -- Catch an awkward (and probably rare) case in which /all/ errors are -- suppressed: see (SCE3) in Note [ei_suppress: suppressing confusing errors] - ; errs_already <- ifErrsM (return True) (return False) +-- ; errs_already <- ifErrsM (return True) (return False) ; let tidy_items - | not errs_already -- Have not already reported an error (perhaps - -- from an outer implication); see #21405 - , not (any ignoreConstraint simples) -- No error is ignorable (is reported elsewhere) - , all ei_suppress tidy_items1 -- All errors are suppressed - = map unsuppressErrorItem tidy_items1 - - | otherwise +-- NO LONGER NEEDED +-- | not errs_already -- Have not already reported an error (perhaps +-- -- from an outer implication); see #21405 +-- , not (any ignoreConstraint simples) -- No error is ignorable (is reported elsewhere) +-- , all ei_suppress tidy_items1 -- All errors are suppressed +-- = map unsuppressErrorItem tidy_items1 +-- +-- | otherwise = tidy_items1 ; traceTc "reportWanteds 1" (vcat [ text "Simples =" <+> ppr simples @@ -668,23 +674,20 @@ reportWanteds ctxt tc_lvl wc@(WC { wc_simple = simples, wc_impl = implics , text "tidy_items1 =" <+> ppr tidy_items1 , text "tidy_errs =" <+> ppr tidy_errs ]) - -- First, deal with any out-of-scope errors: - ; let (out_of_scope, other_holes, not_conc_errs, mult_co_errs) = partition_errors tidy_errs + ; let (out_of_scope, other_holes, not_conc_errs, mult_co_errs) + = partition_errors tidy_errs + -- Don't suppress out-of-scope errors -- See (SLIE0) in Note [cec_suppress: suppressing less-important error messages] - ctxt_for_scope_errs = ctxt { cec_suppress = False } - ; (_, no_out_of_scope) <- askNoErrs $ - reportHoles tidy_items ctxt_for_scope_errs out_of_scope +-- ctxt_for_scope_errs = ctxt { cec_suppress = False } - -- Next, deal with things that are utterly wrong - -- These should not be suppressed by the incoming context - -- (but should be suppressed by out-of-scope errors) - -- See (SLIE0) in Note [cec_suppress: suppressing less-important error messages] - ; let ctxt_for_insols = ctxt { cec_suppress = not no_out_of_scope } + -- First, deal with any out-of-scope errors: + ; (_, no_out_of_scope) <- askNoErrs $ + reportHoles ctxt tidy_items out_of_scope - -- Don't suppress holes or concreteness errors unless we have scope errors - ; reportHoles tidy_items ctxt_for_insols other_holes - ; reportNotConcreteErrs ctxt_for_insols not_conc_errs + -- Don't suppress holes or concreteness errors, ever + ; reportHoles ctxt tidy_items other_holes + ; reportNotConcreteErrs ctxt not_conc_errs -- We only want to report multiplicity coercion errors for multiplicity -- constraints which are /solved/ with a non-reflexivity coercion. We @@ -692,29 +695,45 @@ reportWanteds ctxt tc_lvl wc@(WC { wc_simple = simples, wc_impl = implics -- when /all/ other constraints are solved. -- See wrinkle (DME1) in Note [Coercion errors in tcSubMult] in GHC.Tc.Utils.Unify. ; when (null simples) $ - reportMultiplicityCoercionErrs ctxt_for_insols mult_co_errs + reportMultiplicityCoercionErrs ctxt mult_co_errs -- Now the constraints that give serious errors -- Like Int ~ Bool (incl nullary TyCons) -- or Int ~ t a (AppTy on one side) - ; (items1, no_serious_errs) + -- + -- First visible ones, then invisible ones; but all serious + ; let suppress_insols = not no_out_of_scope + -- Serious errors are not suppressed by cec_suppress, + -- hence not including cec_suppress in suppress_insols + -- See (SLIE0) in Note [cec_suppress: suppressing less-important + -- error messages] + (vis_items, invis_items) + = partition (isVisibleOrigin . errorItemOrigin) tidy_items + + ; (vis_items1, no_vis_serious_errs) <- askNoErrs $ - tryReporters ctxt_for_insols serious_reporters tidy_items + tryReporters ctxt suppress_insols serious_reporters vis_items + + ; let suppress_invis = suppress_insols || not no_vis_serious_errs + ; (invis_items1, no_invis_serious_errs) + <- askNoErrs $ + tryReporters ctxt suppress_invis serious_reporters invis_items -- Now take account of the inherited cec_suppress, and augment it -- to suppress further errors if we have found any serious errors locally - ; let suppress_for_rest = cec_suppress ctxt - || not no_out_of_scope - || not no_serious_errs - ctxt_for_rest = ctxt { cec_suppress = suppress_for_rest } + ; let suppress_rest = cec_suppress ctxt + || suppress_invis + || not no_invis_serious_errs -- Next, deal with the less-serious constraints - ; leftovers <- tryReporters ctxt_for_rest less_serious_reporters items1 + ; leftovers <- tryReporters ctxt suppress_rest less_serious_reporters $ + (vis_items1 ++ invis_items1) ; massertPpr (null leftovers) (text "The following unsolved Wanted constraints \ \have not been reported to the user:" $$ ppr leftovers) + ; let ctxt_for_rest = ctxt { cec_suppress = suppress_rest } ; mapBagM_ (reportImplic ctxt_for_rest) implics -- NB ctxt2: don't suppress inner insolubles if there's only a -- wanted insoluble here; but do suppress inner insolubles @@ -722,8 +741,6 @@ reportWanteds ctxt tc_lvl wc@(WC { wc_simple = simples, wc_impl = implics } where env = cec_tidy ctxt - tidy_cts = bagToList (mapBag (tidyCt env) simples) - tidy_errs = bagToList (mapBag (tidyDelayedError env) errs) partition_errors :: [DelayedError] -> ([Hole], [Hole], [NotConcreteError], [(TcCoercion, CtLoc)]) partition_errors [] @@ -747,14 +764,14 @@ reportWanteds ctxt tc_lvl wc@(WC { wc_simple = simples, wc_impl = implics = [ -- We put implicit lifting errors first, because are solid errors -- See "Implicit lifting" in GHC.Tc.Gen.Splice -- Note [Lifecycle of an untyped splice, and PendingRnSplice] - ("implicit lifting", is_implicit_lifting, True, mkImplicitLiftingReporter) + ("implicit lifting", is_implicit_lifting, mkImplicitLiftingReporter) -- Next, solid equality errors - , given_eq_spec - , ("insoluble2", utterly_wrong, True, mkGroupReporter mkEqErr) - , ("skolem eq1", very_wrong, True, mkSkolReporter) - , ("FixedRuntimeRep", is_FRR, True, mkGroupReporter mkFRRErr) - , ("skolem eq2", skolem_eq, True, mkSkolReporter) + , ("given_eq", is_given_eq, mkGivenErrorReporter) + , ("insoluble2", utterly_wrong, mkGroupReporter mkEqErr) + , ("skolem eq1", very_wrong, mkSkolReporter) + , ("FixedRuntimeRep", is_FRR, mkGroupReporter mkFRRErr) + , ("skolem eq2", skolem_eq, mkSkolReporter) -- Next, custom type errors -- See Note [Custom type errors in constraints] in GHC.Tc.Types.Constraint @@ -770,29 +787,29 @@ reportWanteds ctxt tc_lvl wc@(WC { wc_simple = simples, wc_impl = implics -- we want to report it as a custom error, /not/ as a mis-match -- between TypeError and ()! Also see the Assert example -- in Note [Custom type errors in constraints] - , ("custom_error", is_user_type_error, True, mkUserTypeErrorReporter) + , ("custom_error", is_user_type_error, mkUserTypeErrorReporter) -- (Handles TypeError and Unsatisfiable) -- "non-tv-eq": equalities (ty1 ~ ty2) where ty1 is not a tyvar - , ("non-tv eq", non_tv_eq, True, mkSkolReporter) + , ("non-tv eq", non_tv_eq, mkSkolReporter) -- The only remaining equalities are alpha ~ ty, -- where alpha is untouchable; and representational equalities -- Prefer homogeneous equalities over hetero, because the -- former might be holding up the latter. -- See Note [Equalities with heterogeneous kinds] in GHC.Tc.Solver.Equality - , ("Homo eqs", is_homo_equality, True, mkGroupReporter mkEqErr) - , ("Other eqs", is_equality, True, mkGroupReporter mkEqErr) + , ("Homo eqs", is_homo_equality, mkGroupReporter mkEqErr) + , ("Other eqs", is_equality, mkGroupReporter mkEqErr) - , ("Insoluble fundeps", is_insoluble, True, mkGroupReporter mkDictErr) + , ("Insoluble fundeps", is_insoluble, mkGroupReporter mkDictErr) ] -- less_serious_reporters: we suppress these if have already -- reported a serious error, higher in the tree less_serious_reporters - = [ ("Irreds", is_irred, False, mkGroupReporter mkIrredErr) - , ("Dicts", is_dict, False, mkGroupReporter mkDictErr) - , ("Quantified", is_qc, False, mkGroupReporter mkQCErr) ] + = [ ("Irreds", is_irred, mkGroupReporter mkIrredErr) + , ("Dicts", is_dict, mkGroupReporter mkDictErr) + , ("Quantified", is_qc, mkGroupReporter mkQCErr) ] -- rigid_nom_eq, rigid_nom_tv_eq, is_dict, is_equality, is_FRR, is_irred :: ErrorItem -> Pred -> Bool @@ -857,30 +874,6 @@ reportWanteds ctxt tc_lvl wc@(WC { wc_simple = simples, wc_impl = implics is_qc _ (ForAllPred {}) = True is_qc _ _ = False - given_eq_spec -- See Note [Given errors] - | has_gadt_match_here - = ("insoluble1a", is_given_eq, True, mkGivenErrorReporter) - | otherwise - = ("insoluble1b", is_given_eq, False, ignoreErrorReporter) - -- False means don't suppress subsequent errors - -- Reason: we don't report all given errors - -- (see mkGivenErrorReporter), and we should only suppress - -- subsequent errors if we actually report this one! - -- #13446 is an example - - -- See Note [Given errors] - has_gadt_match_here = has_gadt_match (cec_encl ctxt) - has_gadt_match [] = False - has_gadt_match (implic : implics) - | PatSkol {} <- ic_info implic - , ic_given_eqs implic /= NoGivenEqs - , ic_warn_inaccessible implic - -- Don't bother doing this if -Winaccessible-code isn't enabled. - -- See Note [Avoid -Winaccessible-code when deriving] in GHC.Tc.TyCl.Instance. - = True - | otherwise - = has_gadt_match implics - --------------- isSkolemTy :: TcLevel -> Type -> Bool -- The type is a skolem tyvar @@ -1039,7 +1032,6 @@ type Reporter type ReporterSpec = ( String -- Name , ErrorItem -> Pred -> Bool -- Pick these ones - , Bool -- True <=> suppress subsequent reporters , Reporter) -- The reporter itself mkSkolReporter :: Reporter @@ -1058,9 +1050,12 @@ mkSkolReporter ctxt items | eq_lhs_type item1 item2 = True | otherwise = False -reportHoles :: [ErrorItem] -- other (tidied) constraints - -> SolverReportErrCtxt -> [Hole] -> TcM () -reportHoles tidy_items ctxt holes +reportHoles :: SolverReportErrCtxt + -> [ErrorItem] -- Other (tidied) constraints + -> [Hole] -- Holes to report + -> TcM () +-- Ignores cec_suppress +reportHoles ctxt tidy_items holes = do diag_opts <- initDiagOpts <$> getDynFlags let severity = diagReasonSeverity diag_opts (cec_type_holes ctxt) @@ -1115,6 +1110,7 @@ zonkTidyTcLclEnvs tidy_env lcls = foldM go (tidy_env, emptyNameEnv) (concatMap c return (tidy_env', extendNameEnv name_env name tidy_ty) reportNotConcreteErrs :: SolverReportErrCtxt -> [NotConcreteError] -> TcM () +-- Ignores cec_suppress reportNotConcreteErrs _ [] = return () reportNotConcreteErrs ctxt errs@(err0:_) = do { msg <- mkErrorReport (ctLocEnv (nce_loc err0)) diag (Just ctxt) [] [] @@ -1218,15 +1214,10 @@ mkImplicitLiftingReporter ctxt mkGivenErrorReporter :: Reporter -- See Note [Given errors] -mkGivenErrorReporter ctxt (item:|_) +mkGivenErrorReporter ctxt items@(item:|_) + | Just implic <- has_gadt_match (cec_encl ctxt) = do { (ctxt, relevant_binds, item) <- relevantBindings True ctxt item - ; let implic = - case cec_encl ctxt of - -- cec_encl is always non-empty when mkGivenErrorReporter is called - outer_implic:_ -> outer_implic - _ -> pprPanic "mkGivenErrorReporter" (ppr item) - - loc' = setCtLocEnv (ei_loc item) (ic_env implic) + ; let loc' = setCtLocEnv (ei_loc item) (ic_env implic) item' = item { ei_loc = loc' } -- For given constraints we overwrite the env (and hence src-loc) -- with one from the immediately-enclosing implication. @@ -1234,18 +1225,29 @@ mkGivenErrorReporter ctxt (item:|_) ; eq_err_msg <- mkEqErr_help ctxt item' ty1 ty2 ; let msg = TcRnInaccessibleCode implic (SolverReportWithCtxt ctxt eq_err_msg) - ; msg <- mkErrorReport (ctLocEnv loc') msg (Just ctxt) [SupplementaryBindings relevant_binds] [] + ; msg <- mkErrorReport (ctLocEnv loc') msg (Just ctxt) + [SupplementaryBindings relevant_binds] [] ; reportDiagnostic msg } - where - (ty1, ty2) = getEqPredTys (errorItemPred item) -ignoreErrorReporter :: Reporter --- Discard Given errors that don't come from --- a pattern match; maybe we should warn instead? -ignoreErrorReporter ctxt items + | otherwise -- Discard Given errors that don't come from + -- a pattern match; maybe we should warn instead? = do { traceTc "mkGivenErrorReporter no" (ppr items $$ ppr (cec_encl ctxt)) ; return () } + where + (ty1, ty2) = getEqPredTys (errorItemPred item) + + has_gadt_match :: [Implication] -> Maybe Implication + has_gadt_match [] = Nothing + has_gadt_match (implic : implics) + | PatSkol {} <- ic_info implic + , ic_given_eqs implic /= NoGivenEqs + , ic_warn_inaccessible implic + -- Don't bother doing this if -Winaccessible-code isn't enabled. + -- See Note [Avoid -Winaccessible-code when deriving] in GHC.Tc.TyCl.Instance. + = Just implic + | otherwise + = has_gadt_match implics {- Note [Given errors] ~~~~~~~~~~~~~~~~~~~~~~ @@ -1427,31 +1429,29 @@ mkErrorTerm ct_loc ty ctxt msg supp hints ; return $ evDelayedError ty err_str } -tryReporters :: SolverReportErrCtxt -> [ReporterSpec] -> [ErrorItem] +tryReporters :: SolverReportErrCtxt + -> Bool -- Suppress errors if this is True + -- (over-riding cec_suppress) + -> [ReporterSpec] + -> [ErrorItem] -> TcM [ErrorItem] -- Use the first reporter in the list whose predicate says True -tryReporters ctxt reporters items - = do { let (vis_items, invis_items) = partition (isVisibleOrigin . errorItemOrigin) items - ; traceTc "tryReporters {" (ppr vis_items $$ ppr invis_items) - ; items' <- go reporters vis_items invis_items +tryReporters ctxt suppress reporters items + = do { traceTc "tryReporters {" (ppr items) + ; items' <- go reporters items ; traceTc "tryReporters }" (ppr items') ; return items' } where - go [] vis_items invis_items - = return (vis_items ++ invis_items) - - go (r : rs) vis_items invis_items - -- Always look at *visible* Origins before invisible ones - -- this is the whole point of isVisibleOrigin - = do { vis_items' <- tryReporter ctxt r vis_items - ; invis_items' <- tryReporter ctxt r invis_items - ; go rs vis_items' invis_items' } - -- Carry on with the rest, because we must make - -- deferred bindings for them if we have -fdefer-type-errors - -- But suppress their error messages + -- Use `suppress` to set the `cec_supress` flag for these items + ctxt_w_suppress = ctxt { cec_suppress = suppress } + go [] items = return items + go (r:rs) items = do { items' <- tryReporter ctxt_w_suppress r items + ; go rs items' } + -- Carry on with the rest, because we must make + -- deferred bindings for them if we have -fdefer-type-errors tryReporter :: SolverReportErrCtxt -> ReporterSpec -> [ErrorItem] -> TcM [ErrorItem] -tryReporter ctxt (str, keep_me, _, reporter) items = case nonEmpty yeses of +tryReporter ctxt (str, keep_me, reporter) items = case nonEmpty yeses of Nothing -> return items Just yeses -> do { traceTc "tryReporter{ " (text str <+> ppr yeses) View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/71669d7e82a45e2a2f2f3eb1b45dbc4… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/71669d7e82a45e2a2f2f3eb1b45dbc4… You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications: https://gitlab.haskell.org/-/profile/notifications | Help: https://gitlab.haskell.org/help

1 0

[Git][ghc/ghc][wip/apk/unfold-dicts] 3 commits: Tweak inlining discout for unary class ops.
by Andreas Klebinger (＠AndreasK) 15 Sep '26

15 Sep '26

Andreas Klebinger pushed to branch wip/apk/unfold-dicts at Glasgow Haskell Compiler / GHC Commits: 001da929 by Andreas Klebinger at 2026-09-11T20:01:15+02:00 Tweak inlining discout for unary class ops. Consider the number of arguments applied to a unary class op, not just it'S unary-ness when computing class op application size. Fixes #27802 - - - - - 658c5cc9 by Andreas Klebinger at 2026-09-15T06:13:00+00:00 Try discounting args applied to a result of a class op - - - - - 697dddd3 by Andreas Klebinger at 2026-09-15T11:31:11+00:00 Generalize to the unknown call case - - - - - 5 changed files: - compiler/GHC/Core/Unfold.hs - compiler/GHC/Core/Unfold.hs-boot - compiler/GHC/Core/Utils.hs - compiler/GHC/Driver/Session.hs - docs/users_guide/using-optimisation.rst Changes: ===================================== compiler/GHC/Core/Unfold.hs ===================================== @@ -30,6 +30,7 @@ module GHC.Core.Unfold ( updateFunAppDiscount, updateDictDiscount, updateVeryAggressive, updateCaseScaling, updateCaseThreshold, updateReportPrefix, + updateUnknownCallArg, inlineBoringOk, calcUnfoldingGuidance, uncondInlineJoin @@ -87,6 +88,9 @@ data UnfoldingOpts = UnfoldingOpts , unfoldingCaseScaling :: !Int -- ^ Penalize depth with 1/x + , unfoldingUnknownCallArg :: !Int + -- ^ Penalize depth with 1/x + , unfoldingReportPrefix :: !(Maybe String) -- ^ Only report inlining decisions for names with this prefix } @@ -121,6 +125,11 @@ defaultUnfoldingOpts = UnfoldingOpts -- Penalize depth with (size*depth)/scaling , unfoldingCaseScaling = 30 + -- Makes functions more likely to inline if they apply + -- a function argument to interesting arguments. + -- See Note [Discounting for known-at-callsite function calls] + , unfoldingUnknownCallArg = 10 + -- Don't filter inlining decision reports , unfoldingReportPrefix = Nothing } @@ -149,6 +158,11 @@ updateCaseThreshold n opts = opts { unfoldingCaseThreshold = n } updateCaseScaling :: Int -> UnfoldingOpts -> UnfoldingOpts updateCaseScaling n opts = opts { unfoldingCaseScaling = n } +updateUnknownCallArg :: Int -> UnfoldingOpts -> UnfoldingOpts +updateUnknownCallArg n opts = opts { unfoldingUnknownCallArg = n } + + + updateReportPrefix :: Maybe String -> UnfoldingOpts -> UnfoldingOpts updateReportPrefix n opts = opts { unfoldingReportPrefix = n } @@ -335,6 +349,8 @@ isValFun :: CoreExpr -> Bool -- one top-level value lambda isValFun (Lam b e) | isRuntimeVar b = True | otherwise = isValFun e +isValFun (Cast e _c) = isValFun e +isValFun (Tick _t e) = isValFun e isValFun _ = False calcUnfoldingGuidance @@ -709,10 +725,10 @@ sizeExpr opts !bOMB_OUT_SIZE top_args expr FCallId _ -> sizeN (callSize (length val_args) voids) DataConWorkId dc -> conSize dc (length val_args) PrimOpId op _ -> primOpSize op (length val_args) - ClassOpId cls _ -> classOpSize opts cls top_args val_args + ClassOpId cls _ -> classOpSize opts cls top_args val_args voids _ | fun `hasKey` buildIdKey -> buildSize | fun `hasKey` augmentIdKey -> augmentSize - | otherwise -> funSize opts top_args fun (length val_args) voids + | otherwise -> funSize opts top_args fun val_args voids ------------ size_up_alt (Alt _con _bndrs rhs) = size_up rhs `addSizeN` 10 @@ -777,21 +793,29 @@ litSize _other = 0 -- Must match size of nullary constructors -- Key point: if x |-> 4, then x must inline unconditionally -- (eg via case binding) -classOpSize :: UnfoldingOpts -> Class -> [Id] -> [CoreExpr] -> ExprSize +classOpSize :: UnfoldingOpts -> Class -> [Id] -> [CoreExpr] -> Int -> ExprSize -- See (IA1) in Note [Interesting arguments] in GHC.Core.Opt.Simplify.Utils -classOpSize _opts _cls _top_args [] +classOpSize _opts _cls _top_args [] _voids = sizeZero -- A non-applied classop -classOpSize opts cls top_args (dict_arg:other_val_args) - = SizeIs size (arg_discount dict_arg) 0 +classOpSize opts cls top_args (dict_arg:other_val_args) voids + = SizeIs size dict_arg_discount 0 where - size | isUnaryClass cls = 0 -- See (UCM4) in Note [Unary class magic] in GHC.Core.TyCon - | otherwise = 20 + (10 * length other_val_args) + -- See (UCM4) in Note [Unary class magic] in GHC.Core.TyCon + op_app_size = if isUnaryClass cls then 0 else 20 + + -- Size penalty for applying the extracted class method to it's + -- arguments. + method_app_size = callSize (length other_val_args) voids + + size = op_app_size + method_app_size -- If the class op is scrutinising a lambda bound dictionary then -- give it a discount, to encourage the inlining of this function - arg_discount (Cast arg _co) = arg_discount arg - arg_discount (Var dict) | dict `elem` top_args = unitBag (dict, dict_discount) - arg_discount _ = emptyBag + dict_arg_discount = case getIdFromTrivialExpr_maybe dict_arg of + Nothing -> emptyBag + Just dict + | dict `elem` top_args -> unitBag (dict, dict_discount) + | otherwise -> emptyBag -- If we have (class-op d arg1 .. argn) then it's super-good to inline -- to expose `d`; not only can we do the dictionary selection @@ -800,7 +824,9 @@ classOpSize opts cls top_args (dict_arg:other_val_args) -- See the discussion on #26831, esp "Delicate inlining". dict_discount | null other_val_args = unfoldingDictDiscount opts - | otherwise = unfoldingDictDiscount opts + unfoldingFunAppDiscount opts + | otherwise = unfoldingDictDiscount opts + unfoldingFunAppDiscount opts + + -- See Note [Discounting for known-at-callsite function calls] + unknownFunArgDiscount opts top_args other_val_args -- | The size of a function call callSize @@ -826,12 +852,13 @@ jumpSize _n_val_args _voids = 0 -- Jumps are small, and we don't want penalise -- spectral/puzzle. TODO Perhaps adjusting the default threshold would be a -- better solution? -funSize :: UnfoldingOpts -> [Id] -> Id -> Int -> Int -> ExprSize +funSize :: UnfoldingOpts -> [Id] -> Id -> [CoreExpr] -> Int -> ExprSize -- Size for function calls where the function is not a constructor or primops -- Note [Function applications] -funSize opts top_args fun n_val_args voids +funSize opts top_args fun val_args voids | otherwise = SizeIs size arg_discount res_discount where + n_val_args = length val_args some_val_args = n_val_args > 0 is_join = isJoinId fun @@ -839,10 +866,14 @@ funSize opts top_args fun n_val_args voids | not some_val_args = 0 | otherwise = callSize n_val_args voids + fun_discount + | fun `elem` top_args = unfoldingFunAppDiscount opts + unknownFunArgDiscount opts top_args val_args + | otherwise = unfoldingFunAppDiscount opts + -- DISCOUNTS -- See Note [Function and non-function discounts] arg_discount | some_val_args && fun `elem` top_args - = unitBag (fun, unfoldingFunAppDiscount opts) + = unitBag (fun, fun_discount) | otherwise = emptyBag -- If the function is an argument and is applied -- to some values, give it an arg-discount @@ -864,7 +895,64 @@ conSize dc n_val_args -- See Note [Constructor size and result discount] | otherwise = SizeIs 10 emptyBag 10 -{- Note [Constructor size and result discount] +-- See Note [Discounting for known-at-callsite function calls] +unknownFunArgDiscount :: UnfoldingOpts -> [Id] -> [CoreExpr] -> Int +unknownFunArgDiscount opts top_args args = sum (map arg_discount args) + where + arg_discount arg + | interestingArg arg = unfoldingUnknownCallArg opts + | otherwise = 0 + + -- Little brother to the simplifier's interestingArg. + -- Use trivial_expr_fold to look through casts, ticks and type applications, + -- in a principled manner. + interestingVar v = v `elem` top_args || exprIsConLike (Var v) + interestingArg = trivial_expr_fold interestingVar (const True) False True + +{- +Note [Discounting for known-at-callsite function calls] +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +A regular unknown function call `f x y` inside some function `foo` provides no +optimization opportunities to improve `foo` no matter what shape x and y have. + +However we might have a pattern where a function argument is applied to +interesting arguments like: + + foo f g x = ... f x ... g g ... + bar = foo snd snd (1,2) + +Here we really want to inline `foo`, as it will expose `snd` to the tuple, causing +it to inline and eliminating the tuple allocation completely. + +Note that this doesn't always require the function argument itself to be inlined. +Inlining `foo` will also expose the strictness and W/W properties of foos function +arguments and allows specConstr to fire on them. There is simple a large number of +optimizations that can happen for known functions that are impossible for unknown +calls. + +So here is the plan: Whenever the argument to `f` could be useful for optimizing +`foo` if `f` is a known call we give `f` a discount. + +This includes: + +* The argument is a top_arg itself. (Think `f id 1` with `f g = g`) +* The argument is con like: This can help rules, W/W, SpecConstr, further inlining. +* The argument is a literal: Rules/Inlining +* The argument is non-trivial: + + If it's a thunk strictness might allow eager evaluationg + + If it's a pap/lambda we might be able to eta-expand it, removing an intermediate pap + or even removing the PAP fully. + +However we have to be careful. It's *not* a given that we actually +get a benefit from inlining such a call into it's context. It fully +depends on the specific arguments. So we simply make those interesting +arguments "free" rather charging the usual cost of 10 per applied arg. + +Similarly we *only* should give this discount to a unknown call. If the function +`f` being called inside `foo` is a known function all those optimizations can +happen inside foo without inlining it into it's call sites. + +Note [Constructor size and result discount] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Treat a constructors application as size 10, regardless of how many arguments it has; we are keen to expose them (and we charge separately ===================================== compiler/GHC/Core/Unfold.hs-boot ===================================== @@ -13,3 +13,4 @@ updateDictDiscount :: Int -> UnfoldingOpts -> UnfoldingOpts updateVeryAggressive :: Bool -> UnfoldingOpts -> UnfoldingOpts updateCaseThreshold :: Int -> UnfoldingOpts -> UnfoldingOpts updateCaseScaling :: Int -> UnfoldingOpts -> UnfoldingOpts +updateUnknownCallArg :: Int -> UnfoldingOpts -> UnfoldingOpts ===================================== compiler/GHC/Core/Utils.hs ===================================== @@ -1654,7 +1654,9 @@ it off at source. {-# INLINE trivial_expr_fold #-} trivial_expr_fold :: (Id -> r) -> (Literal -> r) -> r -> r -> CoreExpr -> r --- ^ The worker function for Note [exprIsTrivial] and Note [getIdFromTrivialExpr] +-- ^ k_id k_lit k_triv k_not_triv +-- +-- The worker function for Note [exprIsTrivial] and Note [getIdFromTrivialExpr] -- This is meant to have the code of both functions in one place and make it -- easy to derive custom predicates. -- ===================================== compiler/GHC/Driver/Session.hs ===================================== @@ -1881,6 +1881,8 @@ dynamic_flags_deps = [ (intSuffix (\n d -> d { unfoldingOpts = updateCaseThreshold n (unfoldingOpts d)})) , make_ord_flag defFlag "funfolding-case-scaling" (intSuffix (\n d -> d { unfoldingOpts = updateCaseScaling n (unfoldingOpts d)})) + , make_ord_flag defFlag "funfolding-unknown-call-arg-discount" + (intSuffix (\n d -> d { unfoldingOpts = updateUnknownCallArg n (unfoldingOpts d)})) , make_dep_flag defFlag "funfolding-keeness-factor" (floatSuffix (\_ d -> d)) ===================================== docs/users_guide/using-optimisation.rst ===================================== @@ -1902,6 +1902,26 @@ as such you shouldn't need to set any of them explicitly. A flag recommended. Values in the range 10 <= n <= 20 allow some inlining to take place while still allowing GHC to compile modules containing such inlining loops. +.. ghc-flag:: -funfolding-unknown-call-arg-discount=⟨n⟩ + :shortdesc: *default: 10.* Discount for each interesting arg applied to a unknown function. + :type: dynamic + :category: + + :default: 10 + + .. index:: + single: inlining, controlling + single: unfolding, controlling + + Typically a function like ``foo f x`` is not much more likely to inline if + ``f`` is a known argument. + + With this discount being given we make GHC more eager to inline ``foo`` + if ``f`` is applied to interesting arguments inside the RHS of ``foo``. + + This optimizes cases where other optimizations can fire on the call to ``f`` + after it has been made a known call by inlining ``foo`` into it's call site. + .. ghc-flag:: -fworker-wrapper :shortdesc: Enable the worker/wrapper transformation. Implied by :ghc-flag:`-O` View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/3520030fa377abd59eaca4a61d16f3… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/3520030fa377abd59eaca4a61d16f3… You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications: https://gitlab.haskell.org/-/profile/notifications | Help: https://gitlab.haskell.org/help

1 0

[Git][ghc/ghc] Pushed new branch wip/wip/27786
by Zubin (＠wz1000) 15 Sep '26

15 Sep '26

Zubin pushed new branch wip/wip/27786 at Glasgow Haskell Compiler / GHC -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/tree/wip/wip/27786 You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications: https://gitlab.haskell.org/-/profile/notifications | Help: https://gitlab.haskell.org/help

1 0

[Git][ghc/ghc][wip/10.0.1-backports] 4 commits: haddock: Bump version to 2.34.0
by Zubin (＠wz1000) 15 Sep '26

15 Sep '26

Zubin pushed to branch wip/10.0.1-backports at Glasgow Haskell Compiler / GHC Commits: 5a9e2598 by Zubin Duggal at 2026-09-15T15:46:47+05:30 haddock: Bump version to 2.34.0 - - - - - a92da7b2 by Zubin Duggal at 2026-09-15T15:50:06+05:30 Bump libffi-clib submodule to 3.6.0 - - - - - a31e70d0 by Zubin Duggal at 2026-09-15T16:01:04+05:30 hadrian: update bootstrap plans and add a plan for 9.12.4 - - - - - d99cbb9f by Zubin Duggal at 2026-09-15T16:50:19+05:30 Prepare 10.0.1-alpha1 - - - - - 91 changed files: - − changelog.d/27532 - − changelog.d/27626 - − changelog.d/27627 - − changelog.d/27703 - − changelog.d/27704 - − changelog.d/27717 - − changelog.d/T25636 - − changelog.d/T26423 - − changelog.d/T26716 - − changelog.d/T26978 - − changelog.d/T26979 - − changelog.d/T27046 - − changelog.d/T27047 - − changelog.d/T27121.md - − changelog.d/T27123.md - − changelog.d/T27124.md - − changelog.d/T27131 - − changelog.d/T27182.md - − changelog.d/T27202 - − changelog.d/T27225 - − changelog.d/T27308 - − changelog.d/T27314.md - − changelog.d/T27359 - − changelog.d/T27360 - − changelog.d/T27386 - − changelog.d/T27440 - − changelog.d/T27455 - − changelog.d/T27456 - − changelog.d/T27583 - − changelog.d/T27586 - − changelog.d/T27639 - − changelog.d/T27705 - − changelog.d/T27749 - − changelog.d/T27764 - − changelog.d/arm_ncg_fixes_T27430 - − changelog.d/bump-process - − changelog.d/dynamic-trace-flags - − changelog.d/fix-absent-dict-projection - − changelog.d/fix-blackhole-handling - − changelog.d/fix-cmm-atomic-load-store - − changelog.d/fix-compacting-gc-ap-27434 - − changelog.d/fix-exponential-case-desugar-27383 - − changelog.d/fix-finalizers-27072 - − changelog.d/fix-layout-stack-fcall - − changelog.d/fix-parallel-gc-selector-race-27477 - − changelog.d/fix-peekitbl-no-tntc - − changelog.d/fix-use-std-ap-thunk - − changelog.d/ghc-pkg-faster-closure - − changelog.d/hadrian-stale-package-confs-26661 - − changelog.d/ipe-event-class - − changelog.d/jobserver-leak-fix - − changelog.d/link-deterministic-order - − changelog.d/llvm-22 - − changelog.d/llvm-23 - − changelog.d/more-efficient-home-unit-imports-finding - − changelog.d/reexported-module-errors - − changelog.d/semaphore-v2 - − changelog.d/tag-inference-27005 - − changelog.d/tcplugin_init.md - − changelog.d/tcplugins-pmc.md - − changelog.d/typecheckModule-API.md - − changelog.d/unit-index - − changelog.d/unused-type - − changelog.d/wasm-fix-serviceworker - − changelog.d/windows-rethrow-overlapped-exception - − changelog.d/withTcPlugins.md - docs/users_guide/10.0.1-notes.rst - docs/users_guide/exts/static_pointers.rst - docs/users_guide/runtime_control.rst - hadrian/bootstrap/plan-9_10_1.json - hadrian/bootstrap/plan-9_10_2.json - hadrian/bootstrap/plan-9_10_3.json - hadrian/bootstrap/plan-9_12_1.json - hadrian/bootstrap/plan-9_12_2.json - + hadrian/bootstrap/plan-9_12_4.json - hadrian/bootstrap/plan-9_14_1.json - hadrian/bootstrap/plan-bootstrap-9_10_1.json - hadrian/bootstrap/plan-bootstrap-9_10_2.json - hadrian/bootstrap/plan-bootstrap-9_10_3.json - + hadrian/bootstrap/plan-bootstrap-9_12_4.json - libraries/base/changelog.md - libraries/ghc-experimental/CHANGELOG.md - libraries/ghc-internal/CHANGELOG.md - libraries/libffi-clib - testsuite/tests/ghci/T27532/Makefile - − testsuite/tests/ghci/T27532/T27532.stdout - − testsuite/tests/ghci/T27532/a.script - testsuite/tests/ghci/T27532/all.T - − testsuite/tests/ghci/T27532/b.script - utils/haddock/haddock-api/haddock-api.cabal - utils/haddock/haddock.cabal The diff was not included because it is too large. View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/d4cfcd13291182e55fa13a140aef89… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/d4cfcd13291182e55fa13a140aef89… You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications: https://gitlab.haskell.org/-/profile/notifications | Help: https://gitlab.haskell.org/help

1 0