- Mar 19, 2022
-
-
- Mar 18, 2022
-
-
-
Previously while constructing the jump table index we would zero-extend the discriminant before subtracting the start of the jump-table. This goes subtly wrong in the case of a sub-word, signed discriminant, as described in the included Note. Fix this in both the PPC and X86 NCGs. Fixes #21186.
-
Rewrite the critical notes and fix outdated ones, use `HsQuote GhcRn` (in `HsBracketTc`) for desugaring regardless of the bracket being typed or untyped, remove unused `EpAnn` from `Hs*Bracket GhcRn`, zonkExpr factor out common brackets code, ppr_expr factor out common brackets code, and fix tests, to finish MR !4782. ------------------------- Metric Decrease: hard_hole_fits -------------------------
-
An untyped bracket `HsQuote p` can never be constructed with `p ~ GhcTc`. This is because we don't typecheck `HsQuote` at all. That's OK, because we also never use `HsQuote GhcTc`. To enforce this at the type level we make `HsQuote GhcTc` isomorphic to `NoExtField` and impossible to construct otherwise, by using TTG field extensions to make all constructors, except for `XQuote` (which takes `NoExtField`), unconstructable, with `DataConCantHappen` This is explained more in detail in Note [The life cycle of a TH quotation] Related discussion: !4782
-
When desugaring a bracket we want to desugar /renamed/ rather than /typechecked/ code; So in (HsExpr GhcTc) tree, we must have a (HsExpr GhcRn) for the quotation itself. This commit reworks the TTG refactor on typed and untyped brackets by storing the /renamed/ code in the bracket field extension rather than in the constructor extension in `HsQuote` (previously called `HsUntypedBracket`) See Note [The life cycle of a TH quotation] and !4782
-
Split HsBracket into HsTypedBracket and HsUntypedBracket. Unfortunately, we still cannot get rid of instance XXTypedBracket GhcTc = HsTypedBracket GhcRn despite no longer requiring it for typechecking, but rather because the TH desugarer works on GhcRn rather than GhcTc (See GHC.HsToCore.Quote) -
-
-
- Mar 17, 2022
-
-
- Mar 16, 2022
-
-
-
Previously, we let `Unboxed` win in `lubBoxity`, which is unsoundly optimistic in terms ob Boxity analysis. "Unsoundly" in the sense that we sometimes unbox parameters that we better shouldn't unbox. Examples are #18907 and T19871.absent. Until now, we thought that this hack pulled its weight becuase it worked around some shortcomings of the phase separation between Boxity analysis and CPR analysis. But it is a gross hack which caused regressions itself that needed all kinds of fixes and workarounds. See for example #20767. It became impossible to work with in !7599, so I want to remove it. For example, at the moment, `lubDmd B dmd` will not unbox `dmd`, but `lubDmd A dmd` will. Given that `B` is supposed to be the bottom element of the lattice, it's hardly justifiable to get a better demand when `lub`bing with `A`. The consequence of letting `Boxed` win in `lubBoxity` is that we *would* regress #2387, #16040 and parts of #5075 and T19871.sumIO, until Boxity and CPR are able to communicate better. Fortunately, that is not the case since I could tweak the other source of optimism in Boxity analysis that is described in `Note [Unboxed demand on function bodies returning small products]` so that we *recursively* assume unboxed demands on function bodies returning small products. See the updated Note. `Note [Boxity for bottoming functions]` describes why we need bottoming functions to have signatures that say that they deeply unbox their arguments. In so doing, I had to tweak `finaliseArgBoxities` so that it will never unbox recursive data constructors. This is in line with our handling of them in CPR. I updated `Note [Which types are unboxed?]` to reflect that. In turn we fix #21119, #20767, #18907, T19871.absent and get a much simpler implementation (at least to think about). We can also drop the very ad-hoc definition of `deferAfterPreciseException` and its Note in favor of the simple, intuitive definition we used to have. Metric Decrease: T16875 T18223 T18698a T18698b hard_hole_fits Metric Increase: LargeRecord MultiComponentModulesRecomp T15703 T8095 T9872d Out of all the regresions, only the one in T9872d doesn't vanish in a perf build, where the compiler is bootstrapped with -O2 and thus SpecConstr. Reason for regressions: * T9872d is due to `ty_co_subst` taking its `LiftingContext` boxed. That is because the context is passed to a function argument, for example in `liftCoSubstTyVarBndrUsing`. * In T15703, LargeRecord and T8095, we get a bit more allocations in `expand_syn` and `piResultTys`, because a `TCvSubst` isn't unboxed. In both cases that guards against reboxing in some code paths. * The same is true for MultiComponentModulesRecomp, where we get less unboxing in `GHC.Unit.Finder.$wfindInstalledHomeModule`. In a perf build, allocations actually *improve* by over 4%! Results on NoFib: -------------------------------------------------------------------------------- Program Allocs Instrs -------------------------------------------------------------------------------- awards -0.4% +0.3% cacheprof -0.3% +2.4% fft -1.5% -5.1% fibheaps +1.2% +0.8% fluid -0.3% -0.1% ida +0.4% +0.9% k-nucleotide +0.4% -0.1% last-piece +10.5% +13.9% lift -4.4% +3.5% mandel2 -99.7% -99.8% mate -0.4% +3.6% parser -1.0% +0.1% puzzle -11.6% +6.5% reverse-complem -3.0% +2.0% scs -0.5% +0.1% sphere -0.4% -0.2% wave4main -8.2% -0.3% -------------------------------------------------------------------------------- Summary excludes mandel2 because of excessive bias Min -11.6% -5.1% Max +10.5% +13.9% Geometric Mean -0.2% +0.3% -------------------------------------------------------------------------------- Not bad for a bug fix. The regression in `last-piece` could become a win if SpecConstr would work on non-recursive functions. The regression in `fibheaps` is due to `Note [Reboxed crud for bottoming calls]`, e.g., #21128.
-
-
- Use extension suggestion hints instead of suggesting extensions in the error message body for several FFI errors. - Adds a test case for `TcRnForeignImportPrimExtNotSet`
-
Converts all uses of 'TcRnUnknownMessage' to proper diagnostics.
-
- Mar 15, 2022
-
-
Vladislav Zavialov authored* Users can define their own (~) type operator * Haddock can display documentation for the built-in (~) * New transitional warnings implemented: -Wtype-equality-out-of-scope -Wtype-equality-requires-operators Updates the haddock submodule.
-
- Mar 14, 2022
-
-
As `Note [Demand analysis for recursive data constructors]` describes, we now refrain from unboxing recursive data type arguments, for two reasons: 1. Relating to run/alloc perf: Similar to `Note [CPR for recursive data constructors]`, it seldomly improves run/alloc performance if we just unbox a finite number of layers of a potentially huge data structure. 2. Relating to ghc/alloc perf: Inductive definitions on single-product recursive data types like the one in T11545 will (diverge, and) have very deep demand signatures before any other abortion mechanism in Demand analysis is triggered. That leads to great and unnecessary churn on Demand analysis when ultimately we will never make use of any nested strictness information anyway. Conclusion: Discard nested demand and boxity information on such recursive types with the help of `Note [Detecting recursive data constructors]`. I also implemented `GHC.Types.Unique.MemoFun.memoiseUniqueFun` in order to avoid the overhead of repeated calls to `GHC.Core.Opt.WorkWrap.Utils.isRecDataCon`. It's nice and simple and guards against some smaller regressions in T9233 and T16577. ghc/alloc performance-wise, this patch is a very clear win: Test Metric value New value Change --------------------------------------------------------------------------------------- LargeRecord(normal) ghc/alloc 6,141,071,720 6,099,871,216 -0.7% MultiLayerModulesTH_OneShot(normal) ghc/alloc 2,740,973,040 2,705,146,640 -1.3% T11545(normal) ghc/alloc 945,475,492 85,768,928 -90.9% GOOD T13056(optasm) ghc/alloc 370,245,880 326,980,632 -11.7% GOOD T18304(normal) ghc/alloc 90,933,944 76,998,064 -15.3% GOOD T9872a(normal) ghc/alloc 1,800,576,840 1,792,348,760 -0.5% T9872b(normal) ghc/alloc 2,086,492,432 2,073,991,848 -0.6% T9872c(normal) ghc/alloc 1,750,491,240 1,737,797,832 -0.7% TcPlugin_RewritePerf(normal) ghc/alloc 2,286,813,400 2,270,957,896 -0.7% geo. mean -2.9% No noteworthy change in run/alloc either. NoFib results show slight wins, too: -------------------------------------------------------------------------------- Program Allocs Instrs -------------------------------------------------------------------------------- constraints -1.9% -1.4% fasta -3.6% -2.7% reverse-complem -0.3% -0.9% treejoin -0.0% -0.3% -------------------------------------------------------------------------------- Min -3.6% -2.7% Max +0.1% +0.1% Geometric Mean -0.1% -0.1% Metric Decrease: T11545 T13056 T18304 -
As #20837 pointed out, `isLiftedType_maybe` returned `Just False` in many situations where it should return `Nothing`, because it didn't take into account type families or type variables. In this patch, we fix this issue. We rename `isLiftedType_maybe` to `typeLevity_maybe`, which now returns a `Levity` instead of a boolean. We now return `Nothing` for types with kinds of the form `TYPE (F a1 ... an)` for a type family `F`, as well as `TYPE (BoxedRep l)` where `l` is a type variable. This fix caused several other problems, as other parts of the compiler were relying on `isLiftedType_maybe` returning a `Just` value, and were now panicking after the above fix. There were two main situations in which panics occurred: 1. Issues involving the let/app invariant. To uphold that invariant, we need to know whether something is lifted or not. If we get an answer of `Nothing` from `isLiftedType_maybe`, then we don't know what to do. As this invariant isn't particularly invariant, we can change the affected functions to not panic, e.g. by behaving the same in the `Just False` case and in the `Nothing` case (meaning: no observable change in behaviour compared to before). 2. Typechecking of data (/newtype) constructor patterns. Some programs involving patterns with unknown representations were accepted, such as T20363. Now that we are stricter, this caused further issues, culminating in Core Lint errors. However, the behaviour was incorrect the whole time; the incorrectness only being revealed by this change, not triggered by it. This patch fixes this by overhauling where the representation polymorphism involving pattern matching are done. Instead of doing it in `tcMatches`, we instead ensure that the `matchExpected` functions such as `matchExpectedFunTys`, `matchActualFunTySigma`, `matchActualFunTysRho` allow return argument pattern types which have a fixed RuntimeRep (as defined in Note [Fixed RuntimeRep]). This ensures that the pattern matching code only ever handles types with a known runtime representation. One exception was that patterns with an unknown representation type could sneak in via `tcConPat`, which points to a missing representation-polymorphism check, which this patch now adds. This means that we now reject the program in #20363, at least until we implement PHASE 2 of FixedRuntimeRep (allowing type families in RuntimeRep positions). The aforementioned refactoring, in which checks have been moved to `matchExpected` functions, is a first step in implementing PHASE 2 for patterns. Fixes #20837
-
-
AbsBinds and ABExport both depended on the typechecker, and were thus removed from the main AST Expr. CollectPass now has a new function `collectXXHsBindsLR` used for the new HsBinds extension point Bumped haddock submodule to work with AST changes. The removed Notes from Language.Haskell.Syntax.Binds were duplicated (and not referenced) and the copies in GHC.Hs.Binds are kept (and referenced there). (See #19252)
-
Move GHC-specific comments from Language.Haskell.Syntax.Binds to GHC.Hs.Binds It looks like the Note was deleted but there were actually two copies of it. L.H.S.B no longer references it, and GHC.Hs.Binds keeps an updated copy. (See #19252) There are other duplicated notes -- they will be fixed in the next commit
-
- Mar 13, 2022
-
-
Issue #21150 shows that worker/wrapper allocated a worker function for a function with multiple calls that said "called at most once" when the first argument was absent. That's bad! This patch makes it so that WW preserves at least one non-one-shot value lambda (see `Note [Preserving float barriers]`) by passing around `void#` in place of absent arguments. Fixes #21150. Since the fix is pretty similar to `Note [Protecting the last value argument]`, I put the logic in `mkWorkerArgs`. There I realised (#21204) that `-ffun-to-thunk` is basically useless with `-ffull-laziness`, so I deprecated the flag, simplified and split into `needsVoidWorkerArg`/`addVoidWorkerArg`. SpecConstr is another client of that API. Fixes #21204. Metric Decrease: T14683
-
Note [Weak loop breakers] explains why we need to track variables free in RHS of rules. But we need to do this for /inactive/ rules as well as active ones, unlike the rhs_fv_env stuff. So we now have two fields in node Details, one for free vars of active rules, and one for free vars of all rules. This was shown up by #20820, which is now fixed.
-
- Mar 12, 2022
-
-
-
-
Added for no reason in 7ce1b694
-
-
Cf ticky report before the change: Entries Alloc Alloc'd Non-void Arguments STG Name -------------------------------------------------------------------------------- 696987 29044128 0 1 L main:Target.trackArgument_go5{v r24kY} (fun) -
-
This enables GHC to report more parse errors in a single pass.
-
The result of ORing two BigNats is always greater or equal to the larger of the two. Therefore it is safe to skip the magnitude checks of integerFromBigNat#.
-
We were failing to pass the style file to `makeindex`, as is done by the mklatex configuration generated by Sphinx. Fixes #20913.
-
When ASSIGN_BaseReg is a no-op, we shouldn't generate any C code, otherwise C compiler complains a bunch of -Wunused-value warnings when doing unregisterised codegen.
-
Use castFloatToWord32/castDoubleToWord64 in base to perform the reinterpret cast.
-
Otherwise the C compiler may complain "warning: non-void function does not return a value in all control paths [-Wreturn-type]".
-
Our memcpyish primop's type signatures doesn't match the C type signatures. It's not a problem for typical archs, since their C ABI permits dropping the result, but it doesn't work for wasm. The previous logic would cast the memcpyish function pointer to an incorrect type and perform an indirect call, which results in a runtime trap on wasm. The most straightforward fix is: don't emit EFF_ for memcpyish functions. Since we don't want to include extra headers in .hc to bring in their prototypes, we can just use the __builtin versions.
-
