ghc-commits

Download

ghc-commits@haskell.org

July 2025

1 participants
2 discussions

[Git][ghc/ghc][wip/andreask/overloaded_calls] Disable -fprof-late-overloaded-calls for join points.
by Andreas Klebinger (＠AndreasK) 01 Jul '25

01 Jul '25

Andreas Klebinger pushed to branch wip/andreask/overloaded_calls at Glasgow Haskell Compiler / GHC Commits: d8c22402 by Andreas Klebinger at 2025-07-01T11:41:50+02:00 Disable -fprof-late-overloaded-calls for join points. Currently GHC considers cost centres as destructive to join contexts. Or in other words this is not considered valid: join f x = ... in ... -> scc<tick> jmp This makes the functionality of `-fprof-late-overloaded-calls` not feasible for join points in general. We used to try to work around this by putting the ticks on the rhs of the join point rather than around the jump. However beyond the loss of accuracy this was broken for recursive join points as we ended up with something like: rec-join f x = scc<tick> ... jmp f x Which similarly is not valid as the tick once again destroys the tail call. One might think we could limit ourselves to non-recursive tail calls and do something clever like: join f x = scc<tick> ... in ... jmp f x And sometimes this works! But sometimes the full rhs would look something like: join g x = .... join f x = scc<tick> ... -> jmp g x Which, would again no longer be valid. I believe in the long run we can make cost centre ticks non-destructive to join points. Or we could keep track of where we are/are not allowed to insert a cost centre. But in the short term I will simply disable the annotation of join calls under this flag. - - - - - 2 changed files: - compiler/GHC/Core/LateCC/OverloadedCalls.hs - docs/users_guide/profiling.rst Changes: ===================================== compiler/GHC/Core/LateCC/OverloadedCalls.hs ===================================== @@ -20,7 +20,6 @@ import GHC.Core.Make import GHC.Core.Predicate import GHC.Core.Type import GHC.Core.Utils -import GHC.Tc.Utils.TcType import GHC.Types.Id import GHC.Types.Name import GHC.Types.SrcLoc @@ -29,6 +28,41 @@ import GHC.Types.Var type OverloadedCallsCCState = Strict.Maybe SrcSpan +{- Note [Overloaded Calls and join points] +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +Currently GHC considers cost centres as destructive to +join contexts. Or in other words this is not considered valid: + + join f x = ... + in + ... -> scc<tick> jmp + +This makes the functionality of `-fprof-late-overloaded-calls` not feasible +for join points in general. We used to try to work around this by putting the +ticks on the rhs of the join point rather than around the jump. However beyond +the loss of accuracy this was broken for recursive join points as we ended up +with something like: + + rec-join f x = scc<tick> ... jmp f x + +Which similarly is not valid as the tick once again destroys the tail call. +One might think we could limit ourselves to non-recursive tail calls and do +something clever like: + + join f x = scc<tick> ... + in ... jmp f x + +And sometimes this works! But sometimes the full rhs would look something like: + + join g x = .... + join f x = scc<tick> ... -> jmp g x + +Which, would again no longer be valid. I believe in the long run we can make +cost centre ticks non-destructive to join points. Or we could keep track of +where we are/are not allowed to insert a cost centre. But in the short term I will +simply disable the annotation of join calls under this flag. +-} + -- | Insert cost centres on function applications with dictionary arguments. The -- source locations attached to the cost centres is approximated based on the -- "closest" source note encountered in the traversal. @@ -52,21 +86,10 @@ overloadedCallsCC = CoreBndr -> LateCCM OverloadedCallsCCState CoreExpr -> LateCCM OverloadedCallsCCState CoreExpr - wrap_if_join b pexpr = do + wrap_if_join _b pexpr = do + -- See Note [Overloaded Calls and join points] expr <- pexpr - if isJoinId b && isOverloadedTy (exprType expr) then do - let - cc_name :: FastString - cc_name = fsLit "join-rhs-" `appendFS` getOccFS b - - cc_srcspan <- - fmap (Strict.fromMaybe (UnhelpfulSpan UnhelpfulNoLocationInfo)) $ - lift $ gets lateCCState_extra - - insertCC cc_name cc_srcspan expr - else - return expr - + return expr processExpr :: CoreExpr -> LateCCM OverloadedCallsCCState CoreExpr processExpr expr = @@ -99,6 +122,7 @@ overloadedCallsCC = -- Avoid instrumenting join points. -- (See comment in processBind above) + -- Also see Note [Overloaded Calls and join points] && not (isJoinVarExpr f) then do -- Extract a name and source location from the function being ===================================== docs/users_guide/profiling.rst ===================================== @@ -571,9 +571,7 @@ of your profiled program will be different to that of the unprofiled one. Some overloaded calls may not be annotated, specifically in cases where the optimizer turns an overloaded function into a join point. Calls to such functions will not be wrapped in ``SCC`` annotations, since it would make - them non-tail calls, which is a requirement for join points. Instead, - ``SCC`` annotations are added around the body of overloaded join variables - and given distinct names (``join-rhs-<var>``) to avoid confusion. + them non-tail calls, which is a requirement for join points. .. ghc-flag:: -fprof-cafs :shortdesc: Auto-add ``SCC``\\ s to all CAFs View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/d8c22402fe5bb986abb15250772e1b5… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/d8c22402fe5bb986abb15250772e1b5… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/romes/26039] Bump default language edition to GHC2024
by Rodrigo Mesquita (＠alt-romes) 01 Jul '25

01 Jul '25

Rodrigo Mesquita pushed to branch wip/romes/26039 at Glasgow Haskell Compiler / GHC Commits: dca20032 by Rodrigo Mesquita at 2025-07-01T09:43:35+01:00 Bump default language edition to GHC2024 As per the accepted ghc-proposal#632 Fixes #26039 - - - - - 42 changed files: - compiler/GHC/Driver/Flags.hs - testsuite/tests/ado/all.T - testsuite/tests/annotations/should_fail/all.T - testsuite/tests/array/should_run/all.T - testsuite/tests/core-to-stg/all.T - testsuite/tests/deSugar/should_fail/all.T - testsuite/tests/deriving/should_compile/all.T - testsuite/tests/deriving/should_fail/all.T - testsuite/tests/dmdanal/sigs/all.T - testsuite/tests/gadt/all.T - testsuite/tests/ghci.debugger/scripts/all.T - testsuite/tests/ghci/prog-mhu002/all.T - testsuite/tests/ghci/scripts/Makefile - testsuite/tests/ghci/should_run/all.T - testsuite/tests/indexed-types/should_compile/all.T - testsuite/tests/linear/should_fail/all.T - testsuite/tests/module/all.T - testsuite/tests/overloadedrecflds/should_fail/all.T - testsuite/tests/parser/should_fail/all.T - testsuite/tests/partial-sigs/should_compile/all.T - testsuite/tests/polykinds/all.T - testsuite/tests/programs/andy_cherry/test.T - testsuite/tests/rename/should_fail/all.T - testsuite/tests/roles/should_fail/all.T - testsuite/tests/showIface/DocsInHiFile1.stdout - testsuite/tests/showIface/DocsInHiFileTH.stdout - testsuite/tests/showIface/HaddockIssue849.stdout - testsuite/tests/showIface/HaddockOpts.stdout - testsuite/tests/showIface/HaddockSpanIssueT24378.stdout - testsuite/tests/showIface/MagicHashInHaddocks.stdout - testsuite/tests/showIface/Makefile - testsuite/tests/showIface/NoExportList.stdout - testsuite/tests/showIface/PragmaDocs.stdout - testsuite/tests/showIface/ReExports.stdout - testsuite/tests/simplCore/T9646/test.T - testsuite/tests/simplCore/should_compile/all.T - testsuite/tests/th/all.T - testsuite/tests/typecheck/should_compile/all.T - testsuite/tests/typecheck/should_fail/all.T - testsuite/tests/typecheck/should_run/all.T - testsuite/tests/vdq-rta/should_fail/all.T - testsuite/tests/warnings/should_fail/all.T The diff was not included because it is too large. View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/dca20032a13d456154532d97d1a59f1… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/dca20032a13d456154532d97d1a59f1… You're receiving this email because of your account on gitlab.haskell.org.

1 0