- ghc-devs - Haskell.org

simple pictures about GHC development flow
by Takenobu Tani 28 Jan '17

28 Jan '17

Hi devs, For myself and new contributors, I drew overview pictures about GHC development flow. GHC development flow http://takenobu-hs.github.io/downloads/ghc_development_flow.pdf https://github.com/takenobu-hs/ghc-development-flow If I misunderstood something, please teach me. I'll correct or remove them. Regards, Takenobu

4 15

Cannot build GHC using the Newcomers guide
by Marius Ghita 27 Jan '17

27 Jan '17

Following is a list of steps that I ran and their output linked: - clone repo https://gist.github.com/mhitza/f5d4516b6c8386fe8e064f95b5ad62 0b - build.mk configuration https://gist.github.com/mhitza/ 2d979c64a646bdd3e097f65fd650c675 - boot https://gist.github.com/mhitza/e23df8b9ed2aac5b1b8881c70165bf3f - configure https://gist.github.com/mhitza/88c09179be3bb82024192bf6181aef13 - make FAILS https://gist.github.com/mhitza/95738bf49c8c87ce46c9319b4c266a 2c I'm using a 'stack-ghc' executable, that's only a shell wrapper to run ghc from stack (since I don't have a globally installed ghc) (source https://gist.github.com/mhitza/38fe96fb440daab28e57a50de47863d5 ), and I also have 'ghc-pkg' wrapped in the same way with a stack-ghc-pkg script (source https://gist.github.com/mhitza/ 6c2b1978ef802707161041abe1d2699e ) -- Google+: https://plus.google.com/111881868112036203454

5 7

Problems with `arc` and `phabricator`
by Iavor Diatchki 26 Jan '17

26 Jan '17

Hello, I just fixed a small bug in GHC (#11406) and I'm trying to push the change to `phabricator` for review, but the command is failing. Here is my output, could you please advise on what to do? > arc diff Linting... LINT OKAY No lint problems. Running unit tests... No unit test engine is configured for this project. PUSH STAGING Pushing changes to staging area... Permission denied (publickey,keyboard-interactive). fatal: Could not read from remote repository. Please make sure you have the correct access rights and the repository exists. STAGING FAILED Unable to push changes to the staging area. Usage Exception: Failed to push changes to staging area. Correct the issue, or use --skip-staging to skip this step. Clearly, there is some sort of permission issue... -Iavor

3 3

RE: [commit: ghc] wip/discount-fv: Discount scrutinized free variables (fd9608e)
by Simon Peyton Jones 25 Jan '17

25 Jan '17

Long story short: learning and experimenting how GHC works and eventually contribute my findings (if any). OK great! Let us know if you need help. Simon From: Alex Biehl [mailto:alex.biehl@gmail.com] Sent: 25 January 2017 10:32 To: Simon Peyton Jones <simonpj(a)microsoft.com> Cc: ghc-devs(a)haskell.org Subject: Re: [commit: ghc] wip/discount-fv: Discount scrutinized free variables (fd9608e) I believe it was a false alarm. Unfortunately I could reproduce the reduced allocations even without my patch (I hadn't ran `validate` before, so I didn't know at that time). Ben was kind enough to push it to a branch so gipedia could pick up but it hadn't any effect either. What leaves me wondering though why are the allocations reduced drastically (by ~30% for haddock.cabal and haddock.base and even ~57% for T9203. c.f. https://ghc.haskell.org/trac/ghc/ticket/4960#comment:14) And not for others? I am using `./validate --testsuite-only --fast` (with a perf build GHC). The reason I did this was that I thought if I reduce `dupAppSize` in `CoreUtils` I could reduce code duplication in `case` expressions where GHC currently duplicates lots of alternatives (I only realized later that `dupAppSize` does not account for `case` expressions at all, so its probably some case-of-case stuff or something) in some of my code and I wanted to confirm if that is actually a good thing. I noticed the ticket and though before tackling that I will try myself on that discount stuff. Long story short: learning and experimenting how GHC works and eventually contribute my findings (if any). Simon Peyton Jones <simonpj(a)microsoft.com<mailto:simonpj@microsoft.com>> schrieb am Mi., 25. Jan. 2017 um 11:11 Uhr: Alex Interesting. Care to give us any background on what you are working on? I've often thought about discounting for free vars. Do you have some compelling examples? (Also fine if you just want to noodle privately for now.) Simon | -----Original Message----- | From: ghc-commits [mailto:ghc-commits-bounces@haskell.org<mailto:ghc-commits-bounces@haskell.org>] On Behalf Of | git(a)git.haskell.org<mailto:git@git.haskell.org> | Sent: 24 January 2017 17:20 | To: ghc-commits(a)haskell.org<mailto:ghc-commits@haskell.org> | Subject: [commit: ghc] wip/discount-fv: Discount scrutinized free | variables (fd9608e) | | Repository : ssh://git@git.haskell.org/ghc<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fgit%40git.h…> | | On branch : wip/discount-fv | Link : | https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fghc.haske | ll.org<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fll.org&data…>%2Ftrac%2Fghc%2Fchangeset%2Ffd9608ea93fc2389907b82c3fe540805d986c28 | e%2Fghc&data=02%7C01%7Csimonpj%40microsoft.com<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2F40microsoft…>%7C6b18dd9581bc459c203b08d4 | 447d482c%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636208752257772884& | sdata=3%2F1y5zQjDsa5j1%2FhTEjnKc4mg0qNtCD8WyqMaNUq5mA%3D&reserved=0 | | >--------------------------------------------------------------- | | commit fd9608ea93fc2389907b82c3fe540805d986c28e | Author: alexbiehl <alex.biehl(a)gmail.com<mailto:alex.biehl@gmail.com>> | Date: Mon Jan 23 20:34:20 2017 +0100 | | Discount scrutinized free variables | | | >--------------------------------------------------------------- | | fd9608ea93fc2389907b82c3fe540805d986c28e | compiler/coreSyn/CoreUnfold.hs | 95 +++++++++++++++++++++++++----------- | ------ | 1 file changed, 56 insertions(+), 39 deletions(-) | | diff --git a/compiler/coreSyn/CoreUnfold.hs | b/compiler/coreSyn/CoreUnfold.hs index 574d841..36ea382 100644 | --- a/compiler/coreSyn/CoreUnfold.hs | +++ b/compiler/coreSyn/CoreUnfold.hs | @@ -62,8 +62,11 @@ import Bag | import Util | import Outputable | import ForeignCall | +import VarEnv | | +import Control.Applicative ((<|>)) | import qualified Data.ByteString as BS | +import Debug.Trace | | {- | ************************************************************************ | @@ -501,43 +504,51 @@ sizeExpr :: DynFlags | -- Note [Computing the size of an expression] | | sizeExpr dflags bOMB_OUT_SIZE top_args expr | - = size_up expr | + = size_up emptyInScopeSet expr | where | - size_up (Cast e _) = size_up e | - size_up (Tick _ e) = size_up e | - size_up (Type _) = sizeZero -- Types cost nothing | - size_up (Coercion _) = sizeZero | - size_up (Lit lit) = sizeN (litSize lit) | - size_up (Var f) | isRealWorldId f = sizeZero | + size_up :: InScopeSet -> CoreExpr -> ExprSize | + size_up is (Cast e _) = size_up is e | + size_up is (Tick _ e) = size_up is e | + size_up _ (Type _) = sizeZero -- Types cost nothing | + size_up _ (Coercion _) = sizeZero | + size_up _ (Lit lit) = sizeN (litSize lit) | + size_up _ (Var f) | isRealWorldId f = sizeZero | -- Make sure we get constructor discounts even | -- on nullary constructors | - | otherwise = size_up_call f [] 0 | - | - size_up (App fun arg) | - | isTyCoArg arg = size_up fun | - | otherwise = size_up arg `addSizeNSD` | - size_up_app fun [arg] (if isRealWorldExpr arg | then 1 else 0) | - | - size_up (Lam b e) | - | isId b && not (isRealWorldId b) = lamScrutDiscount dflags | (size_up e `addSizeN` 10) | - | otherwise = size_up e | - | - size_up (Let (NonRec binder rhs) body) | - = size_up rhs `addSizeNSD` | - size_up body `addSizeN` | + | otherwise = size_up_call f [] 0 | + | + size_up is (App fun arg) | + | isTyCoArg arg = size_up is fun | + | otherwise = size_up is arg `addSizeNSD` | + size_up_app is fun [arg] (if isRealWorldExpr | + arg then 1 else 0) | + | + size_up is (Lam b e) | + | isId b && not (isRealWorldId b) = lamScrutDiscount dflags | (size_up is e `addSizeN` 10) | + | otherwise = size_up is e | + | + size_up is (Let (NonRec binder rhs) body) | + = let | + is' = extendInScopeSet is binder | + in | + size_up is rhs `addSizeNSD` | + size_up is' body `addSizeN` | (if isUnliftedType (idType binder) then 0 else 10) | -- For the allocation | -- If the binder has an unlifted type there is no | allocation | | - size_up (Let (Rec pairs) body) | - = foldr (addSizeNSD . size_up . snd) | - (size_up body `addSizeN` (10 * length pairs)) -- | (length pairs) for the allocation | + size_up is (Let (Rec pairs) body) | + = let | + is' = extendInScopeSetList is (map fst pairs) | + in | + foldr (addSizeNSD . size_up is' . snd) | + (size_up is' body | + `addSizeN` (10 * length pairs)) -- (length pairs) | for the allocation | pairs | | - size_up (Case e _ _ alts) | - | Just v <- is_top_arg e -- We are scrutinising an argument | variable | + size_up is (Case e _ _ alts) | + | Just v <- is_top_arg e <|> is_free_var e -- We are | + scrutinising an argument variable or a free variable | = let | - alt_sizes = map size_up_alt alts | + alt_sizes = map (size_up_alt is) alts | | -- alts_size tries to compute a good discount for | -- the case when we are scrutinising an argument | variable @@ -569,9 +580,12 @@ sizeExpr dflags bOMB_OUT_SIZE top_args expr | is_top_arg (Cast e _) = is_top_arg e | is_top_arg _ = Nothing | | + is_free_var (Var v) | not (v `elemInScopeSet` is) = Just v | + is_free_var (Cast e _) = is_free_var e | + is_free_var _ = Nothing | | - size_up (Case e _ _ alts) = size_up e `addSizeNSD` | - foldr (addAltSize . size_up_alt) | case_size alts | + size_up is (Case e _ _ alts) = size_up is e `addSizeNSD` | + foldr (addAltSize . size_up_alt is) | + case_size alts | where | case_size | | is_inline_scrut e, not (lengthExceeds alts 1) = sizeN (- | 10) @@ -608,15 +622,15 @@ sizeExpr dflags bOMB_OUT_SIZE top_args expr | | ------------ | -- size_up_app is used when there's ONE OR MORE value args | - size_up_app (App fun arg) args voids | - | isTyCoArg arg = size_up_app fun args voids | - | isRealWorldExpr arg = size_up_app fun (arg:args) | (voids + 1) | - | otherwise = size_up arg `addSizeNSD` | - size_up_app fun (arg:args) | voids | - size_up_app (Var fun) args voids = size_up_call fun args voids | - size_up_app (Tick _ expr) args voids = size_up_app expr args voids | - size_up_app (Cast expr _) args voids = size_up_app expr args voids | - size_up_app other args voids = size_up other `addSizeN` | + size_up_app is (App fun arg) args voids | + | isTyCoArg arg = size_up_app is fun args voids | + | isRealWorldExpr arg = size_up_app is fun (arg:args) | (voids + 1) | + | otherwise = size_up is arg `addSizeNSD` | + size_up_app is fun (arg:args) | voids | + size_up_app _ (Var fun) args voids = size_up_call fun args | voids | + size_up_app is (Tick _ expr) args voids = size_up_app is expr args | voids | + size_up_app is (Cast expr _) args voids = size_up_app is expr args | voids | + size_up_app is other args voids = size_up is other | `addSizeN` | callSize (length args) voids | -- if the lhs is not an App or a Var, or an invisible thing like | a | -- Tick or Cast, then we should charge for a complete call plus | the @@ -633,7 +647,10 @@ sizeExpr dflags bOMB_OUT_SIZE top_args expr | _ -> funSize dflags top_args fun (length | val_args) voids | | ------------ | - size_up_alt (_con, _bndrs, rhs) = size_up rhs `addSizeN` 10 | + size_up_alt :: InScopeSet -> Alt Var -> ExprSize | + size_up_alt is (_con, bndrs, rhs) = size_up is' rhs `addSizeN` 10 | + where is' = extendInScopeSetList is bndrs | + | -- Don't charge for args, so that wrappers look cheap | -- (See comments about wrappers with Case) | -- | | _______________________________________________ | ghc-commits mailing list | ghc-commits(a)haskell.org<mailto:ghc-commits@haskell.org> | https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fmail.hask | ell.org<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fell.org&dat…>%2Fcgi-bin%2Fmailman%2Flistinfo%2Fghc- | commits&data=02%7C01%7Csimonpj%40microsoft.com<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2F40microsoft…>%7C6b18dd9581bc459c203b08d4 | 447d482c%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636208752257772884& | sdata=rGeUVlgqjfwCl%2FEdTX3%2BX0mQGX5UcS7bY9qadLT%2FSE4%3D&reserved=0

1 0

RE: [commit: ghc] wip/discount-fv: Discount scrutinized free variables (fd9608e)
by Simon Peyton Jones 25 Jan '17

25 Jan '17

Alex Interesting. Care to give us any background on what you are working on? I've often thought about discounting for free vars. Do you have some compelling examples? (Also fine if you just want to noodle privately for now.) Simon | -----Original Message----- | From: ghc-commits [mailto:ghc-commits-bounces@haskell.org] On Behalf Of | git(a)git.haskell.org | Sent: 24 January 2017 17:20 | To: ghc-commits(a)haskell.org | Subject: [commit: ghc] wip/discount-fv: Discount scrutinized free | variables (fd9608e) | | Repository : ssh://git@git.haskell.org/ghc | | On branch : wip/discount-fv | Link : | https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fghc.haske | ll.org%2Ftrac%2Fghc%2Fchangeset%2Ffd9608ea93fc2389907b82c3fe540805d986c28 | e%2Fghc&data=02%7C01%7Csimonpj%40microsoft.com%7C6b18dd9581bc459c203b08d4 | 447d482c%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636208752257772884& | sdata=3%2F1y5zQjDsa5j1%2FhTEjnKc4mg0qNtCD8WyqMaNUq5mA%3D&reserved=0 | | >--------------------------------------------------------------- | | commit fd9608ea93fc2389907b82c3fe540805d986c28e | Author: alexbiehl <alex.biehl(a)gmail.com> | Date: Mon Jan 23 20:34:20 2017 +0100 | | Discount scrutinized free variables | | | >--------------------------------------------------------------- | | fd9608ea93fc2389907b82c3fe540805d986c28e | compiler/coreSyn/CoreUnfold.hs | 95 +++++++++++++++++++++++++----------- | ------ | 1 file changed, 56 insertions(+), 39 deletions(-) | | diff --git a/compiler/coreSyn/CoreUnfold.hs | b/compiler/coreSyn/CoreUnfold.hs index 574d841..36ea382 100644 | --- a/compiler/coreSyn/CoreUnfold.hs | +++ b/compiler/coreSyn/CoreUnfold.hs | @@ -62,8 +62,11 @@ import Bag | import Util | import Outputable | import ForeignCall | +import VarEnv | | +import Control.Applicative ((<|>)) | import qualified Data.ByteString as BS | +import Debug.Trace | | {- | ************************************************************************ | @@ -501,43 +504,51 @@ sizeExpr :: DynFlags | -- Note [Computing the size of an expression] | | sizeExpr dflags bOMB_OUT_SIZE top_args expr | - = size_up expr | + = size_up emptyInScopeSet expr | where | - size_up (Cast e _) = size_up e | - size_up (Tick _ e) = size_up e | - size_up (Type _) = sizeZero -- Types cost nothing | - size_up (Coercion _) = sizeZero | - size_up (Lit lit) = sizeN (litSize lit) | - size_up (Var f) | isRealWorldId f = sizeZero | + size_up :: InScopeSet -> CoreExpr -> ExprSize | + size_up is (Cast e _) = size_up is e | + size_up is (Tick _ e) = size_up is e | + size_up _ (Type _) = sizeZero -- Types cost nothing | + size_up _ (Coercion _) = sizeZero | + size_up _ (Lit lit) = sizeN (litSize lit) | + size_up _ (Var f) | isRealWorldId f = sizeZero | -- Make sure we get constructor discounts even | -- on nullary constructors | - | otherwise = size_up_call f [] 0 | - | - size_up (App fun arg) | - | isTyCoArg arg = size_up fun | - | otherwise = size_up arg `addSizeNSD` | - size_up_app fun [arg] (if isRealWorldExpr arg | then 1 else 0) | - | - size_up (Lam b e) | - | isId b && not (isRealWorldId b) = lamScrutDiscount dflags | (size_up e `addSizeN` 10) | - | otherwise = size_up e | - | - size_up (Let (NonRec binder rhs) body) | - = size_up rhs `addSizeNSD` | - size_up body `addSizeN` | + | otherwise = size_up_call f [] 0 | + | + size_up is (App fun arg) | + | isTyCoArg arg = size_up is fun | + | otherwise = size_up is arg `addSizeNSD` | + size_up_app is fun [arg] (if isRealWorldExpr | + arg then 1 else 0) | + | + size_up is (Lam b e) | + | isId b && not (isRealWorldId b) = lamScrutDiscount dflags | (size_up is e `addSizeN` 10) | + | otherwise = size_up is e | + | + size_up is (Let (NonRec binder rhs) body) | + = let | + is' = extendInScopeSet is binder | + in | + size_up is rhs `addSizeNSD` | + size_up is' body `addSizeN` | (if isUnliftedType (idType binder) then 0 else 10) | -- For the allocation | -- If the binder has an unlifted type there is no | allocation | | - size_up (Let (Rec pairs) body) | - = foldr (addSizeNSD . size_up . snd) | - (size_up body `addSizeN` (10 * length pairs)) -- | (length pairs) for the allocation | + size_up is (Let (Rec pairs) body) | + = let | + is' = extendInScopeSetList is (map fst pairs) | + in | + foldr (addSizeNSD . size_up is' . snd) | + (size_up is' body | + `addSizeN` (10 * length pairs)) -- (length pairs) | for the allocation | pairs | | - size_up (Case e _ _ alts) | - | Just v <- is_top_arg e -- We are scrutinising an argument | variable | + size_up is (Case e _ _ alts) | + | Just v <- is_top_arg e <|> is_free_var e -- We are | + scrutinising an argument variable or a free variable | = let | - alt_sizes = map size_up_alt alts | + alt_sizes = map (size_up_alt is) alts | | -- alts_size tries to compute a good discount for | -- the case when we are scrutinising an argument | variable @@ -569,9 +580,12 @@ sizeExpr dflags bOMB_OUT_SIZE top_args expr | is_top_arg (Cast e _) = is_top_arg e | is_top_arg _ = Nothing | | + is_free_var (Var v) | not (v `elemInScopeSet` is) = Just v | + is_free_var (Cast e _) = is_free_var e | + is_free_var _ = Nothing | | - size_up (Case e _ _ alts) = size_up e `addSizeNSD` | - foldr (addAltSize . size_up_alt) | case_size alts | + size_up is (Case e _ _ alts) = size_up is e `addSizeNSD` | + foldr (addAltSize . size_up_alt is) | + case_size alts | where | case_size | | is_inline_scrut e, not (lengthExceeds alts 1) = sizeN (- | 10) @@ -608,15 +622,15 @@ sizeExpr dflags bOMB_OUT_SIZE top_args expr | | ------------ | -- size_up_app is used when there's ONE OR MORE value args | - size_up_app (App fun arg) args voids | - | isTyCoArg arg = size_up_app fun args voids | - | isRealWorldExpr arg = size_up_app fun (arg:args) | (voids + 1) | - | otherwise = size_up arg `addSizeNSD` | - size_up_app fun (arg:args) | voids | - size_up_app (Var fun) args voids = size_up_call fun args voids | - size_up_app (Tick _ expr) args voids = size_up_app expr args voids | - size_up_app (Cast expr _) args voids = size_up_app expr args voids | - size_up_app other args voids = size_up other `addSizeN` | + size_up_app is (App fun arg) args voids | + | isTyCoArg arg = size_up_app is fun args voids | + | isRealWorldExpr arg = size_up_app is fun (arg:args) | (voids + 1) | + | otherwise = size_up is arg `addSizeNSD` | + size_up_app is fun (arg:args) | voids | + size_up_app _ (Var fun) args voids = size_up_call fun args | voids | + size_up_app is (Tick _ expr) args voids = size_up_app is expr args | voids | + size_up_app is (Cast expr _) args voids = size_up_app is expr args | voids | + size_up_app is other args voids = size_up is other | `addSizeN` | callSize (length args) voids | -- if the lhs is not an App or a Var, or an invisible thing like | a | -- Tick or Cast, then we should charge for a complete call plus | the @@ -633,7 +647,10 @@ sizeExpr dflags bOMB_OUT_SIZE top_args expr | _ -> funSize dflags top_args fun (length | val_args) voids | | ------------ | - size_up_alt (_con, _bndrs, rhs) = size_up rhs `addSizeN` 10 | + size_up_alt :: InScopeSet -> Alt Var -> ExprSize | + size_up_alt is (_con, bndrs, rhs) = size_up is' rhs `addSizeN` 10 | + where is' = extendInScopeSetList is bndrs | + | -- Don't charge for args, so that wrappers look cheap | -- (See comments about wrappers with Case) | -- | | _______________________________________________ | ghc-commits mailing list | ghc-commits(a)haskell.org | https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fmail.hask | ell.org%2Fcgi-bin%2Fmailman%2Flistinfo%2Fghc- | commits&data=02%7C01%7Csimonpj%40microsoft.com%7C6b18dd9581bc459c203b08d4 | 447d482c%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636208752257772884& | sdata=rGeUVlgqjfwCl%2FEdTX3%2BX0mQGX5UcS7bY9qadLT%2FSE4%3D&reserved=0

2 1

Intended meaning of nubBy use in GHC's compiler/cmm/Debug.hs
by David Feuer 24 Jan '17

24 Jan '17

The meaning of nubBy when applied to functions other than equivalence relations changed around the time you created compiler/cmm/Debug.hs. This makes it extra tricky to figure out exactly what that nubBy is expected to do. Do you think you could explain? Furthermore, it would be helpful to know what (if anything) is known about the relationships between the source spans in the the argument to the nubBy. This code is proving to be a bit of a hot spot in #11095 and the better we understand what's going on the better our chances of coming up with good solutions. Thanks, David Feuer

1 0

RE: Floating lazy primops
by David Feuer 24 Jan '17

24 Jan '17

I've opened #13182 to explore one possible approach to dataToTag# that strikes me as likely to be simpler and to have fewer potential gotchas. But there could be critical points I'm missing about why we do it as we do now. -------- Original message --------From: Simon Peyton Jones <simonpj(a)microsoft.com> Date: 1/23/17 4:47 PM (GMT-05:00) To: David Feuer <david(a)well-typed.com> Cc: ghc-devs(a)haskell.org Subject: RE: Floating lazy primops We should have this conversation on a ticket, perhaps #13027. | good at the time). Are there actually any primops with lifted arguments | where we *want* speculation? Perhaps the most important primop to | consider is seq#, which is I don't understand this question. comment:23 of #13027 specifically says to skip the ok-for-spec test for lifted args. So you as "are the any" whereas comment:23 says "all". | 2. If dataToTag# is marked can_fail (an aspect of https:// | phabricator.haskell.org/rGHC5a9a1738023a) is it still possible for it to | end up being applied to an unevaluated argument? If not, perhaps the | CorePrep specials can be removed altogether. That may be true, but it's not easy to GUARANTEE in the way that Lint guarantees types. So I'm happier leaving in the CorePrep stuff.. but please do add a comment there that points to the Note in primops.txt.pp and says that it seems unlikely this will ever occur. | One more question: do you think it's *better* to let dataToTag# float and | then fix it up later, or better to mark it can_fail? Unlike | reallyUnsafePtrEquality#, dataToTag# is used all over the place, so it is | important that it interacts as well as possible with the optimizer, | whatever that entails. I think better to do as now. That way the simplifier has the opportunity to common-up multiple evals into one. If we add them later that's harder. By all means add Notes to explain this. Thanks! Simon | -----Original Message----- | From: David Feuer [mailto:david@well-typed.com] | Sent: 18 January 2017 19:45 | To: Simon Peyton Jones <simonpj(a)microsoft.com> | Cc: ghc-devs(a)haskell.org | Subject: Floating lazy primops | | I opened up https://phabricator.haskell.org/D2987 to mark | reallyUnsafePtrEquality# can_fail, but in the process I realized a couple | things. | | 1. Part of https://phabricator.haskell.org/rGHC5a9a1738023a may actually | not have been such a hot idea after all (although it certainly sounded | good at the time). Are there actually any primops with lifted arguments | where we *want* speculation? Perhaps the most important primop to | consider is seq#, which is | (mysteriously?) marked neither can_fail nor has_side_effects, but another | to look at is unpackClosure#, which seems likely to give different | results before and after forcing. Most other primops with lifted | arguments are marked has_side_effects, and therefore won't be floated out | anyway. | | 2. If dataToTag# is marked can_fail (an aspect of https:// | phabricator.haskell.org/rGHC5a9a1738023a) is it still possible for it to | end up being applied to an unevaluated argument? If not, perhaps the | CorePrep specials can be removed altogether. | | David Feuer

1 0

StablePtr / StableName ?
by Johannes Waldmann 24 Jan '17

24 Jan '17

Dear ghc devs, would the StablePtr performance issue (slow hash table) https://ghc.haskell.org/trac/ghc/ticket/13165 also affect StableNames? Cf. https://github.com/ekmett/ersatz/issues/30 Could making 10^5 stable names, and accessing each just once, take noticeable time? Where would this show up in a profile? I guess there's no easy way to change the ersatz library (StableName this is the fundamental mechanism for detecting sharing) but if these issues are related, then ersatz provides a performance test case. Thanks, Johannes.

3 3

Floating lazy primops
by David Feuer 23 Jan '17

23 Jan '17

I opened up https://phabricator.haskell.org/D2987 to mark reallyUnsafePtrEquality# can_fail, but in the process I realized a couple things. 1. Part of https://phabricator.haskell.org/rGHC5a9a1738023a may actually not have been such a hot idea after all (although it certainly sounded good at the time). Are there actually any primops with lifted arguments where we *want* speculation? Perhaps the most important primop to consider is seq#, which is (mysteriously?) marked neither can_fail nor has_side_effects, but another to look at is unpackClosure#, which seems likely to give different results before and after forcing. Most other primops with lifted arguments are marked has_side_effects, and therefore won't be floated out anyway. 2. If dataToTag# is marked can_fail (an aspect of https:// phabricator.haskell.org/rGHC5a9a1738023a) is it still possible for it to end up being applied to an unevaluated argument? If not, perhaps the CorePrep specials can be removed altogether. David Feuer

2 2

RE: [Diffusion] [Build Failed] rGHCb78fa759bfb4: Simplify and improve CSE
by Simon Peyton Jones 20 Jan '17

20 Jan '17

I'm sorry about all these Phab failures. I'm struggling with arc, and so committing directly. It all validates fine on my machine. They seem to be all about T1969. It validates fine on my machine. But only just! On my machine I get peak_megabytes = 64; but on Harbormaster it seems to be 65 or 66. But I think 64 is *just* within the 20% margin from 55, and 65 is just outside. The increase might well be peak-memory sampling noise. Anyway it looks as if the 20% increase all occurred some time ago. Maybe we should look at when and why? But meanwhile, Ben/Reid/David, if you agree that the increase is some time ago, would you like to re-centre the number and (maybe) open a ticket to find where the increase actually happened. I have to go home now... apologies. Simon From: noreply(a)phabricator.haskell.org [mailto:noreply@phabricator.haskell.org] Sent: 20 January 2017 17:13 To: Simon Peyton Jones <simonpj(a)microsoft.com> Subject: [Diffusion] [Build Failed] rGHCb78fa759bfb4: Simplify and improve CSE Harbormaster failed to build B13285: rGHCb78fa759bfb4: Simplify and improve CSE! BRANCHES master USERS simonpj (Author) O7 (Auditor) O12 (Auditor) COMMIT https://phabricator.haskell.org/rGHCb78fa759bfb4 EMAIL PREFERENCES https://phabricator.haskell.org/settings/panel/emailpreferences/ To: simonpj, Harbormaster

1 1