| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously GHC would always assume that section types began with `@` while
producing assembly, which is not true. For instance, in ARM assembly syntax
section types begin with `%`. This abstracts out section type pretty-printing
and adjusts it to correctly account for the target architectures assembly
flavor.
Reviewers: austin, hvr, Phyx
Reviewed By: Phyx
Subscribers: Phyx, rwbarton, thomie, erikd
GHC Trac Issues: #13937
Differential Revision: https://phabricator.haskell.org/D3712
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The binutils documentation states that .short is a synonym for .word,
which I assumed to mean "machine word", leading me to believe that we
needed to use .hword to render half-machine-words. However, Darwin's
toolchain doesn't understand .hword, so there we instead used .short.
However, as it turns out the binutils documentation confusingly uses
"word" to refer to a 16-bit word, so .short should work fine. Moreover,
LLVM's internal assembler also doesn't understand .hword, so using
.short consistently simplies things remarkably.
Test Plan: Validate using binutils and LLVM internal assembler,
validate on Darwin
Reviewers: niteria, austin
Reviewed By: niteria
Subscribers: rwbarton, thomie
GHC Trac Issues: #13866
Differential Revision: https://phabricator.haskell.org/D3667
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
While investigating #12545, I discovered several places in the code
that performed length-checks like so:
```
length ts == 4
```
This is not ideal, since the length of `ts` could be much longer than 4,
and we'd be doing way more work than necessary! There are already a slew
of helper functions in `Util` such as `lengthIs` that are designed to do
this efficiently, so I found every place where they ought to be used and
did just that. I also defined a couple more utility functions for list
length that were common patterns (e.g., `ltLength`).
Test Plan: ./validate
Reviewers: austin, hvr, goldfire, bgamari, simonmar
Reviewed By: bgamari, simonmar
Subscribers: goldfire, rwbarton, thomie
Differential Revision: https://phabricator.haskell.org/D3622
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This is a bit smaller than the alternative, DW_CFA_val_expression. Also
a bit of refactoring.
Test Plan: Validate
Reviewers: scpmw, simonmar, austin
Reviewed By: simonmar
Subscribers: thomie
Differential Revision: https://phabricator.haskell.org/D2745
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
And use to mark `stg_stack_underflow_frame`, which we are unable to
determine a caller from.
To simplify parsing at the moment we steal the `return` keyword to
indicate an undefined unwind value. Perhaps this should be revisited.
Reviewers: scpmw, simonmar, austin, erikd
Subscribers: dfeuer, thomie
Differential Revision: https://phabricator.haskell.org/D2738
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
As discussed in D1532, Trac Trac #11337, and Trac Trac #11338, the stack
unwinding information produced by GHC is currently quite approximate.
Essentially we assume that register values do not change at all within a
basic block. While this is somewhat true in normal Haskell code, blocks
containing foreign calls often break this assumption. This results in
unreliable call stacks, especially in the code containing foreign calls.
This is worse than it sounds as unreliable unwinding information can at
times result in segmentation faults.
This patch set attempts to improve this situation by tracking unwinding
information with finer granularity. By dispensing with the assumption of
one unwinding table per block, we allow the compiler to accurately
represent the areas surrounding foreign calls.
Towards this end we generalize the representation of unwind information
in the backend in three ways,
* Multiple CmmUnwind nodes can occur per block
* CmmUnwind nodes can now carry unwind information for multiple
registers (while not strictly necessary; this makes emitting
unwinding information a bit more convenient in the compiler)
* The NCG backend is given an opportunity to modify the unwinding
records since it may need to make adjustments due to, for instance,
native calling convention requirements for foreign calls (see
#11353).
This sets the stage for resolving #11337 and #11338.
Test Plan: Validate
Reviewers: scpmw, simonmar, austin, erikd
Subscribers: qnikst, thomie
Differential Revision: https://phabricator.haskell.org/D2741
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Test Plan: Look at DWARF output.
Reviewers: scpmw, austin
Reviewed By: austin
Subscribers: thomie
Differential Revision: https://phabricator.haskell.org/D1734
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Overhaul the Overhauled Pattern Match Checker
* Changed the representation of Value Set Abstractions. Instead of
using a prefix tree, we now use a list of Value Vector Abstractions.
The set of constraints Delta for every Value Vector Abstraction is the
oracle state so that we solve everything only once.
* Instead of doing everything lazily, we prune at once (and in general
everything is much stricter). Hence, an example written with pattern
guards is checked in almost the same time as the equivalent with
pattern matching.
* Do not store the covered and the divergent sets at all. Since what we
only need is a yes/no (does this clause cover anything? Does it force
any thunk?) We just keep a boolean for each.
* Removed flags `-Wtoo-many-guards` and `-ffull-guard-reasoning`.
Replaced with `fmax-pmcheck-iterations=n`. Still debatable what should
the default `n` be.
* When a guard is for sure not going to contribute anything, we treat
it as such: The oracle is not called and cases `CGuard`, `UGuard` and
`DGuard` from the paper are not happening at all (the generation of a
fresh variable, the unfolding of the pattern list etc.). his combined
with the above seems to be enough to drop the memory increase for test
T783 down to 18.7%.
* Do not export function `dsPmWarn` (it is now called directly from
within `checkSingle` and `checkMatches`).
* Make `PmExprVar` hold a `Name` instead of an `Id`. The term oracle
does not handle type information so using `Id` was a waste of
time/space.
* Added testcases T11195, T11303b (data families) and T11374
The patch addresses at least the following:
Trac #11195, #11276, #11303, #11374, #11162
Test Plan: validate
Reviewers: goldfire, bgamari, hvr, austin
Subscribers: simonpj, thomie
Differential Revision: https://phabricator.haskell.org/D1795
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Summary:
In the past the canonical way for constructing an SDoc string literal was the
composition `ptext . sLit`. But for some time now we have function `text` that
does the same. Plus it has some rules that optimize its runtime behaviour.
This patch takes all uses of `ptext . sLit` in the compiler and replaces them
with calls to `text`. The main benefits of this patch are clener (shorter) code
and less dependencies between module, because many modules now do not need to
import `FastString`. I don't expect any performance benefits - we mostly use
SDocs to report errors and it seems there is little to be gained here.
Test Plan: ./validate
Reviewers: bgamari, austin, goldfire, hvr, alanz
Subscribers: goldfire, thomie, mpickering
Differential Revision: https://phabricator.haskell.org/D1784
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Introduction of two new flags, for more precise control over the new
pattern match checker's behaviour when reasoning about guards. This is
supposed to address #11195 (and maybe more performance bugs related to
the NP-Hardness of coverage checking).
Expected behaviour:
* When `-ffull-guard-reasoning` is on, run the new pattern match
checker in its full power
* When `-ffull-guard-reasoning` is off (the default), for every
match, check a metric to see whether pattern match checking for it
has high probability of being non performant (at the the moment we
check whether the number of guards is over 20 but I would like to
use a more precise measure in the future). If the probability is
high:
- Oversimplify the guards (less expressive but more performant)
and run the checker, and
- Issue a warning about the simplification that happened.
A new flag `-Wtoo-many-guards/-Wno-too-many-guards` suppresses the
warning about the simplification (useful when combined with -Werror).
Test Plan: validate
Reviewers: goldfire, austin, hvr, bgamari
Reviewed By: bgamari
Subscribers: mpickering, thomie
Differential Revision: https://phabricator.haskell.org/D1676
GHC Trac Issues: #11195
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Apparently gnu as uses `.short` as a synonym for `.word`. To emit a
16-bit value one would use `.hword`. However, Darwin doesn't support
`.hword`, instead taking `.short` to mean a 16-bit value. The
insanity is nearly unbearable!
OS X reference:
https://developer.apple.com/library/mac/documentation/DeveloperTools/Ref
erence/Assembler/040-Assembler_Directives/asm_directives.html#//apple_re
f/doc/uid/TP30000823-TPXREF101
gnu as reference:
https://sourceware.org/binutils/docs/as/hword.html#hword
Test Plan: Validate
Reviewers: austin
Reviewed By: austin
Subscribers: thomie
Differential Revision: https://phabricator.haskell.org/D1602
GHC Trac Issues: #11202
|
| |
|
|
|
|
|
|
| |
In order to accomplish this we need to ensure that emit DIEs for all
DebugBlocks, even those that have been optimized out, lest we end up
with undefined symbols of parents at link time.
Differential Revision: https://phabricator.haskell.org/D1279
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This adds a flag -split-sections that does similar things to
-split-objs, but using sections in single object files instead of
relying on the Satanic Splitter and other abominations. This is very
similar to the GCC flags -ffunction-sections and -fdata-sections.
The --gc-sections linker flag, which allows unused sections to actually
be removed, is added to all link commands (if the linker supports it) so
that space savings from having base compiled with sections can be
realized.
Supported both in LLVM and the native code-gen, in theory for all
architectures, but really tested on x86 only.
In the GHC build, a new SplitSections variable enables -split-sections
for relevant parts of the build.
Test Plan: validate with both settings of SplitSections
Reviewers: dterei, Phyx, austin, simonmar, thomie, bgamari
Reviewed By: simonmar, thomie, bgamari
Subscribers: hsyl20, erikd, kgardas, thomie
Differential Revision: https://phabricator.haskell.org/D1242
GHC Trac Issues: #8405
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
We will need to use these to setup proper unwinding information for the
stg_stop_thread closure. This pokes a hole in the STG abstraction,
exposing the machine's stack pointer register so that we can accomplish
this. We also expose a dummy return address register, which corresponds
to the register used to hold the DWARF return address.
Differential Revision: https://phabricator.haskell.org/D1225
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Here we add a `same_value $sp` instruction to our default unwinding
rules to ensure that the implicit `$sp = CFA` rule (which `libdw`
appears to exhibit on x86_64) doesn't overwrite it (necessary since we
don't use $sp to track our call stack).
See Phab Diff D1189 for details on how we arrived at this resolution.
Reviewers: scpmw, austin, simonmar
Reviewed By: austin, simonmar
Subscribers: thomie, simonmar
Differential Revision: https://phabricator.haskell.org/D1224
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This is supposed to be encoded with ULEB128 which the previous
implementation would only guarranty with short lengths. This likely
holds in nearly all cases, but I'd really rather not take changes.
I fix this using the `.uleb128` directive. I'm not certain that this is
portable across assemblers but it makes this quite straightforward and
at the moment I value correctness over portability.
Test Plan: Compare implementation to DWARF spec
Reviewers: scpmw, austin
Reviewed By: austin
Subscribers: thomie
Differential Revision: https://phabricator.haskell.org/D1220
|
| |
|
|
|
|
|
|
|
|
| |
Reviewers: austin
Reviewed By: austin
Subscribers: thomie
Differential Revision: https://phabricator.haskell.org/D1221
|
| |
|
|
|
|
|
|
|
|
| |
Test Plan: Check with `readelf --debug-dump=ranges`
Reviewers: scpmw, austin
Subscribers: thomie
Differential Revision: https://phabricator.haskell.org/D1174
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Some libraries (e.g. elfutils) need these otherwise they ignore our
DWARF annotations.
Test Plan: Test with elfutils' `readelf --debug-dump=cu_index`
Reviewers: scpmw, austin
Subscribers: thomie
Differential Revision: https://phabricator.haskell.org/D1173
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Previously this was given in the body but not in the abbreviation table.
Who knows what sort of havoc this was wrecking.
Test Plan: Verify against DWARF4 specification
Reviewers: scpmw, austin
Subscribers: Tarrasch, thomie
Differential Revision: https://phabricator.haskell.org/D1172
|
| |
|
|
|
|
|
| |
Fortunately this is relatively straightforward - all we need to do is
switch to a non-ELF-specific way of specifying object file sections and
make sure that section-relative addresses work correctly. This is enough
to make "gdb" work on MinGW builds.
|
| |
|
|
|
|
|
|
|
| |
- Don't bracket HsTick expression uneccessarily
- Generate debug information in UTF8
- Reduce amount of information generated - we do not currently need
block information, for example.
Special thanks to slyfox for the reports!
|
| |
|
|
|
|
|
|
| |
- Make abbrev offset absolute on Non-Mac systems
- Add another termination byte at the end of the abbrev section
(readelf complains)
- Scope combination was wrong for the simpler cases
- Shouldn't have a "global/" in front of all scopes
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This tells debuggers such as GDB how to "unwind" a program state,
which allows them to walk the stack up.
Notes:
* The code is quite general, perhaps unnecessarily so. Unless we get
more unwind information, only the first case of pprSetUnwind will
get used - and pprUnwindExpr and pprUndefUnwind will never be
called. It just so happens that this is a point where we can get a
lot of features cheaply, even if we don't use them.
* When determining what location to show for a return address, most
debuggers check the map for "rip-1", assuming that's where the
"call" instruction is. For tables-next-to-code, that happens to
always be the end of an info table. We therefore cheat a bit here by
shifting .debug_frame information so it covers the end of the info
table, as well as generating a .loc directive for the info table
data.
Debuggers will still show the wrong label for the return address,
though. Haven't found a way around that one yet.
(From Phabricator D396)
|
|
|
This is where we actually make GHC emit DWARF code. The info section
contains all the general meta information bits as well as an entry for
every block of native code.
Notes:
* We need quite a few new labels in order to properly address starts
and ends of blocks.
* Thanks to Nathan Howell for taking the iniative to get our own Haskell
language ID for DWARF!
(From Phabricator D396)
|
