- Jul 22, 2023
- Jul 18, 2023
-
-
- Jul 12, 2023
-
-
- Jul 11, 2023
-
-
Here we introduce a new code generation option, `-forig-thunk-info`, which ensures that an `stg_orig_thunk_info` frame is pushed before every update frame. This can be invaluable when debugging thunk cycles and similar. See Note [Original thunk info table frames] for details. Closes #23255.
-
- Jul 05, 2023
-
-
In #22010 we established that Int was not always sufficient to store all the uniques we generate during compilation on 32-bit platforms. This commit addresses that problem by using Word64 instead of Int for uniques. The core of the change is in GHC.Core.Types.Unique and GHC.Core.Types.Unique.Supply. However, the representation of uniques is used in many other places, so those needed changes too. Additionally, the RTS has been extended with an atomic_inc64 operation. One major change from this commit is the introduction of the Word64Set and Word64Map data types. These are adapted versions of IntSet and IntMap from the containers package. These are planned to be upstreamed in the future. As a natural consequence of these changes, the compiler will be a bit slower and take more space on 32-bit platforms. Our CI tests indicate around a 5% residency increase. Metric Increase: CoOpt_Read CoOpt_Singletons LargeRecord ManyAlternatives ManyConstructors MultiComponentModules MultiComponentModulesRecomp MultiLayerModulesTH_OneShot RecordUpdPerf T10421 T10547 T12150 T12227 T12234 T12425 T12707 T13035 T13056 T13253 T13253-spj T13379 T13386 T13719 T14683 T14697 T14766 T15164 T15703 T16577 T16875 T17516 T18140 T18223 T18282 T18304 T18698a T18698b T18923 T1969 T19695 T20049 T21839c T3064 T3294 T4801 T5030 T5321FD T5321Fun T5631 T5642 T5837 T6048 T783 T8095 T9020 T9198 T9233 T9630 T9675 T9872a T9872b T9872b_defer T9872c T9872d T9961 TcPlugin_RewritePerf UniqLoop WWRec hard_hole_fits
-
As noted in #23577, CentOS's ancient toolchain throws spurious missing-field-initializer warnings.
-
- Jul 03, 2023
-
-
-
-
Previously, it was possible for pinned, aligned allocation requests to allocate beyond the end of the pinned accumulator block. Specifically, we failed to account for the padding needed to achieve the requested alignment in the "large object" check. With large alignment requests, this can result in the allocator using the capability's pinned object accumulator block to service a request which is larger than `PINNED_EMPTY_SIZE`. To fix this we reorganize `allocatePinned` to consistently account for the alignment padding in all large object checks. This is a bit subtle as we must handle the case of a small allocation request filling the accumulator block, as well as large requests. Fixes #23400.
-
- Jun 29, 2023
-
-
-
As debugTrace is a macro we must take care to ensure that the fact is clear to the compiler lest we see warnings.
-
The libffi shipped with Apple's XCode toolchain does not contain a definition of the FFI_GO_CLOSURES macro, despite containing references to said macro. Work around this by defining the macro, following the model of a similar workaround in OpenJDK [1]. [1] https://github.com/openjdk/jdk17u-dev/pull/741/files
-
-
x86-64/Darwin's toolchain inexplicably warns that collectFreshWeakPtrs needs to be a prototype.
-
-
This is no longer used.
-
-
-
-
This is no longer used.
-
rts: Drop write_barrier
-
-
- cache last elements of `relTable`, `relaTable` and `symbolTables` in `ocInit_ELF` - cache shndx table in ObjectCode - run `checkProddableBlock` only with debug rts
-
- Jun 26, 2023
-
-
Torsten Schmits authored
Tracking ticket: #23394 MR: !10448 * Add constructor `IfaceBreakpoint` to `IfaceTickish` * Store breakpoint data in interface files * Store `BreakArray` for the breakpoint's module, not the current module, in BCOs * Store module name in BCOs instead of `Unique`, since the `Unique` from an `Iface` doesn't match the modules in GHCi's state * Allocate module name in `ModBreaks`, like `BreakArray` * Lookup breakpoint by module name in GHCi * Skip creating breakpoint instructions when no `ModBreaks` are available, rather than injecting `ModBreaks` in the linker when breakpoints are enabled, and panicking when `ModBreaks` is missing
-
- thread knowledge about levity into PrimRep instead of panicking - JS: remove assumption that unlifted heap objects are rts objects (TVar#, etc.) Doing this also fixes #22291 (test added). There is a small performance hit (~1% more allocations). Metric Increase: T18698a T18698b
-
- Jun 24, 2023
-
-
Luite Stegeman authored
Bytecode instructions like PUSH_L (push a local variable) contain an operand that refers to the stack slot. Before this patch, the operand type was SmallOp (Word16), limiting the maximum stack offset to 65535 words. This could cause compiler panics in some cases (See #22888). This patch changes the operand type for stack offsets from SmallOp to Op, removing the stack offset limit. Fixes #22888
-
- Jun 23, 2023
-
-
Darwin's toolchain inexpliciably claims that `write_barrier` and friends have declarations without prototypes, despite the fact that (a) they are definitions, and (b) the prototypes appear only a few lines above. Work around this by making the definitions proper prototypes.
-
- Jun 20, 2023
-
-
-
-
Otherwise the build fails due to warnings. See #23530.
-
Previously the `+werror` transformer would only pass `-Werror` to GHC, which does not ensure that the same is passed to the C compiler when building the RTS. Arguably this is itself a bug but for now we will just work around this by passing `-optc-Werror` to GHC. I tried to enable `-Werror` in all C compilations but the boot libraries are something of a portability nightmare.
-
-
-
According to Alpine's warnings and poll(2), <poll.h> should be preferred.
-
-
-
-
-
Previously `shutdown_handler` would call `stg_exit` if the scheduler was Oalready found to be in `SCHED_INTERRUPTING` state (or higher). However, `stg_exit` is not signal-safe as it calls `exit` (which calls `atexit` handlers). The only safe thing to do in this situation is to call `_exit`, which terminates with minimal cleanup. Fixes #23417.
-
- Jun 19, 2023
-
-
Finley McIlwaine authored
IPE data resulting from the `-finfo-table-map` flag may now be compressed by configuring the GHC build with the `--enable-ipe-data-compression` flag. This results in about a 20% reduction in the size of IPE-enabled build results. The compression library, zstd, may optionally be statically linked by configuring with the `--enabled-static-libzstd` flag (on non-darwin platforms) libzstd version 1.4.0 or greater is required.
-
