[Git][ghc/ghc][wip/marge_bot_batch_merge_job] 11 commits: X86 CodeGen: fix assign_eax_sse_regs
Marge Bot pushed to branch wip/marge_bot_batch_merge_job at Glasgow Haskell Compiler / GHC
Commits:
ca5c902f by sheaf at 2025-12-17T08:46:38-05:00
X86 CodeGen: fix assign_eax_sse_regs
We must set %al to the number of SSE2 registers that contain arguments
(in case we are dealing with a varargs function). The logic for counting
how many arguments reside in SSE2 registers was incorrect, as it used
'isFloatFormat', which incorrectly ignores vector registers.
We now instead do case analysis on the register class:
is_sse_reg r =
case targetClassOfReg platform r of
RcFloatOrVector -> True
RcInteger -> False
This change is necessary to prevent segfaults in T20030_test1j, because
subsequent commits change the format calculations, resulting in vector
formats more often.
- - - - -
ba601f9b by sheaf at 2025-12-17T08:46:38-05:00
X86 regUsageOfInstr: fix format for IMUL
When used with 8-bit operands, the IMUL instruction returns the result
in the lower 16 bits of %rax (also known as %ax). This is different
than for the other sizes, where an input at 16, 32 or 64 bits will
result in 16, 32 or 64 bits of output in both %rax and %rdx.
This doesn't affect the behaviour of the compiler, because we don't
allow partial writes at sub-word sizes. The rationale is explained
in Wrinkle [Don't allow scalar partial writes] in Note [Register formats in liveness analysis],
in GHC.CmmToAsm.Reg.Liveness.
- - - - -
90114349 by sheaf at 2025-12-17T08:46:38-05:00
Liveness analysis: consider register formats
This commit updates the register allocator to be a bit more careful in
situations in which a single register is used at multiple different
formats, e.g. when xmm1 is used both to store a Double# and a DoubleX2#.
This is done by introducing the 'Regs' newtype around 'UniqSet RegWithFormat',
for which the combining operations take the larger of the two formats
instead of overriding the format.
Operations on 'Regs' are defined in 'GHC.CmmToAsm.Reg.Regs'. There is
a modest compile-time cost for the additional overhead for tracking
register formats, which causes the metric increases of this commit.
The subtle aspects of the implementation are outlined in
Note [Register formats in liveness analysis] in GHC.CmmToAsm.Reg.Liveness.
Fixes #26411 #26611
-------------------------
Metric Increase:
T12707
T26425
T3294
-------------------------
- - - - -
bf43168f by sheaf at 2025-12-17T08:46:38-05:00
Register allocator: reload at same format as spill
This commit ensures that if we spill a register onto the stack at a
given format, we then always reload the register at this same format.
This ensures we don't end up in a situation where we spill F64x2 but end
up only reloading the lower F64. This first reload would make us believe
the whole data is in a register, thus silently losing the upper 64 bits
of the spilled register's contents.
Fixes #26526
- - - - -
4237e52a by sheaf at 2025-12-17T08:46:38-05:00
Register allocation: writes redefine format
As explained in Note [Allocated register formats] in GHC.CmmToAsm.Reg.Linear,
we consider all writes to redefine the format of the register.
This ensures that in a situation such as
movsd .Ln6m(%rip),%v1
shufpd $0,%v1,%v1
we properly consider the broadcast operation to change the format of %v1
from F64 to F64x2.
This completes the fix to #26411 (test in T26411b).
- - - - -
df5e42fc by Vladislav Zavialov at 2025-12-17T08:46:39-05:00
Parser: improve mkModuleImpExp, remove checkImportSpec
1. The `mkModuleImpExp` helper now knows whether it is processing an import or
export list item, and uses this information to produce a more accurate error
message for `import M (T(..,x))` with PatternSynonyms disabled.
The old message incorrectly referred to this case as an export form.
2. The `checkImportSpec` helper is removed in favor of more comprehensive error
checking in `mkModuleImpExp`.
3. Additionaly, the invariants of `ImpExpList` and `ImpExpAllWith` have been
made more explicit in the comments and assertions (calls to 'panic').
Test case: import-syntax-no-ext
- - - - -
92493eb2 by Vladislav Zavialov at 2025-12-17T08:46:40-05:00
Subordinate namespace-specified wildcards (#25901)
Add support for subordinate namespace-specified wildcards
`X(type ..)` and `X(data ..)` to import and export lists.
Examples:
import M (Cls(type ..)) -- imports Cls and all its associated types
import M (Cls(data ..)) -- imports Cls and all its methods
module M (R(data ..), C(type ..)) where
-- exports R and all its data constructors and record fields;
-- exports C and all its associated types, but not its methods
The scope of this change is limited to the case where the wildcard is the only
subordinate import/export item, whereas the more complex forms `X(type .., f)`
or `X(type .., data ..)` are unsupported and raise the newly introduced
PsErrUnsupportedExplicitNamespace error. This restriction may be lifted later.
Summary of the changes:
1. Refactor IEThingAll to store its extension field XIEThingAll as a record
IEThingAllExt instead of a tuple.
2. Extend the AST by adding a NamespaceSpecifier field to IEThingAllExt,
representing an optional namespace specifier `type` or `data` in front
of a subordinate wildcard `X(..)`.
3. Extend the grammar in Parser.y with productions for `type ..` and `data ..`
in subordinate import/export items.
4. Introduce `filterByNamespaceGREs` to filter [GlobalRdrElt] by a
NamespaceSpecifier; use it in `filterImports` and `exports_from_avail`
to account for the namespace specifier in IEThingAll.
5. Improve diagnostics by storing more information in DodgyImportsEmptyParent
and DodgyExportsEmptyParent.
Test cases:
T25901_sub_e T25901_sub_f T25901_sub_g T25901_sub_a
T25901_sub_b T25901_sub_c T25901_sub_d T25901_sub_w
DodgyImports02 DodgyImports03 DodgyImports04
- - - - -
83d927d3 by Cheng Shao at 2025-12-17T08:46:41-05:00
testsuite: improve coverage of foundation test
This patch refactors the `foundation` test a bit to improve coverage:
- Instead of using a hard-coded seed, a random seed is now taken from
the command line, and printed upon test failure. This improves test
coverage over many future CI runs, and shall a failure occur, the
seed is available in the CI log for local reproduction.
- The iterations count is bumped to 10000 instead of 100, similar to
the bump in `test-primops`. Runtime timeout is bumped 2x just to be
safe, though it's 2m39s on my machine and well within default
timeout 300s.
- Improve `newLCGGen` by using non-atomic loads/stores on a
`MutableByteArray#` for storing mutable `Word64`, this test doesn't
use parallelism in the first place
- Fixed a few compiler warnings and removed redundant pragmas and
imports
Co-authored-by: Codex
participants (1)
-
Marge Bot (＠marge-bot)