The x86 tests are now broken (in paticular add-scalar.ll now hits the
DAG fallback) due to not handling G_UADDO. The DAG x86 backend has a
custom lowering for this, so that will need to be implemented.
llvm-svn: 369673
Local symbols in the indirect symbol table contain the value
`INDIRECT_SYMBOL_LOCAL` and the corresponding __pointers entry must
contain the address of the target.
In r349060, I added support for local symbols in the indirect symbol
table, which was checking if the symbol `isDefined` && `!isExternal` to
determine if the symbol is local or not.
It turns out that `isDefined` will return false if the user of the
symbol comes before its definition, and we'll again generate .long 0
which will be the symbol at the adress 0x0.
Instead of doing that, use GlobalValue::hasLocalLinkage() to check if
the symbol is local.
Differential Revision: https://reviews.llvm.org/D66563
llvm-svn: 369671
It appears the FIXME here was handled at some point. r159728 from 2012 seems to be at least aportion of fixing it.
Differential Revision: https://reviews.llvm.org/D66570
llvm-svn: 369665
Patch https://reviews.llvm.org/D43256 introduced more aggressive loop layout optimization which depends on profile information. If profile information is not available, the statically estimated profile information(generated by BranchProbabilityInfo.cpp) is used. If user program doesn't behave as BranchProbabilityInfo.cpp expected, the layout may be worse.
To be conservative this patch restores the original layout algorithm in plain mode. But user can still try the aggressive layout optimization with -force-precise-rotation-cost=true.
Differential Revision: https://reviews.llvm.org/D65673
llvm-svn: 369664
Summary: These nodes end up being processed regardless due to DAGCombiner ensuring arguments are processed. This changes the order in which nodes are processed, which fixes an issue on PowerPC.
Reviewers: craig.topper, efriedma, RKSimon, lebedev.ri, mcberg2017, stefanp, hfinkel
Subscribers: nemanjai, MaskRay, jsji, steven.zhang, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66548
llvm-svn: 369662
A lot of places in the code combine checks for both ABI (SVR4/Darwin/AIX) and
addressing mode (64-bit vs 32-bit). In an attempt to make some of the code more
readable I've added a couple functions that combine checking for the ELF abi and
64-bit/32-bit code at once. As we add more AIX support I intend to add similar
functions for the AIX ABI.
Differential Revision: https://reviews.llvm.org/D65814
llvm-svn: 369658
Previously we would get the csect a symbol was contained in through its
fragment. This works only if we are writing an object file, and only for
defined symbols. To fix this we set the contating csect explicitly on the
MCSymbolXCOFF object.
Differential Revision: https://reviews.llvm.org/D66032
llvm-svn: 369657
Summary:
This would make client know which codeActionKinds that clangd may
return.
VSCode will add a new entry "Refactor..." (which shows all
refactoring-kind code actions) in the right-click menu.
Reviewers: ilya-biryukov
Subscribers: MaskRay, jkorous, arphaman, kadircet, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D66592
llvm-svn: 369656
Since D66174 I see failures of TestDataFormatterStdList in about 50% of runs on
Fedora 30 x86_64 libstdc++. I have found out that LLDB internally expects these
RegularExpressions to be matched in their alphabetical order:
^std::(__cxx11::)?list<.+>(( )?&)?$
^std::__[[:alnum:]]+::list<.+>(( )?&)?$
But since D66174 they are sometimes matched in reverse order. In fact it was
only some luck it worked before as there is internally
std::map<lldb::RegularExpressionSP, FormatterImpl> (FormattersContainer).
Differential Revision: https://reviews.llvm.org/D66398
llvm-svn: 369655
Summary: In D65402, I want to get DerefState from AADereferenceable but it was not allowed. This patch moves DerefState definition into Attributor.h and makes AADerefenceable inherit StateWrapper.
Reviewers: jdoerfert, sstefan1
Reviewed By: jdoerfert
Subscribers: hiraditya, jfb, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66585
llvm-svn: 369653
Summary:
When we print the IR with --print-after/before-*,
SlotIndexes will be printed whenever available (We haven't freed it).
This introduces some noises when we try to compare the IR
among different optimizations.
eg:
-print-before=machine-cp will print SlotIndexes for 1st machine-cp
pass, but NOT for 2nd machine-cp;
-print-after=machine-cp will NOT print SlotIndexes for both
machine-cp passes.
So SlotIndexes in 1st pass introduce noises when differing these IRs.
This patch introduces an option to hide indexes.
Reviewers: stoklund, thegameg, qcolombet
Reviewed By: thegameg
Subscribers: hiraditya, arphaman, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66500
llvm-svn: 369650
This broke compiling some ASan tests with never versions of MSVC/the Win
SDK, see https://crbug.com/996675
> MSVC 2017 update 3 (_MSC_VER 1911) enables /Zc:twoPhase by default, and
> so should clang-cl:
> https://docs.microsoft.com/en-us/cpp/build/reference/zc-twophase
>
> clang-cl takes the MSVC version it emulates from the -fmsc-version flag,
> or if that's not passed it tries to check what the installed version of
> MSVC is and uses that, and failing that it uses a default version that's
> currently 1911. So this changes the default if no -fmsc-version flag is
> passed and no installed MSVC is detected. (It also changes the default
> if -fmsc-version is passed or MSVC is detected, and either indicates
> _MSC_VER >= 1911.)
>
> As mentioned in the MSDN article, the Windows SDK header files in
> version 10.0.15063.0 (Creators Update or Redstone 2) and earlier
> versions do not work correctly with /Zc:twoPhase. If you need to use
> these old SDKs with a new clang-cl, explicitly pass /Zc:twoPhase- to get
> the old behavior.
>
> Fixes PR43032.
>
> Differential Revision: https://reviews.llvm.org/D66394
llvm-svn: 369647
This fixes https://bugs.llvm.org/show_bug.cgi?id=40337.
Previously, it was always assumed that relocations referenced symbols in the static symbol table.
Now, if the Link field references a section called ".dynsym" it will look up these symbols
in the dynamic symbol table.
This patch is heavily based on D59097 by James Henderson
Differential revision: https://reviews.llvm.org/D66532
llvm-svn: 369645
It was causing some errors like:
Encoding error:
'ascii' codec can't decode byte 0xe2 in position 341: ordinal not in range(128)
The full traceback has been saved in /tmp/sphinx-err-y2fq4dtb.log, if you want to report the issue to the developers.
llvm-svn: 369644
Summary:
We've been running doxygen with the autobrief option for a couple of
years now. This makes the \brief markers into our comments
redundant. Since they are a visual distraction and we don't want to
encourage more \brief markers in new code either, this patch removes
them all.
Patch produced by
for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done
[This is analogous to LLVM r331272 and CFE r331834]
Subscribers: srhines, nemanjai, javed.absar, kbarton, MaskRay, jkorous, arphaman, jfb, kadircet, jsji, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D66578
llvm-svn: 369643
On Jaguar, XCHG has a latency of 1cy and decodes to 2 macro-opcodes. Maximum
throughput for XCHG is 1 IPC. The byte exchange has worse latency and decodes to
1 extra uOP; maximum observed throughput is 0.5 IPC.
```
xchgb %cl, %dl # Latency: 2cy - uOPs: 3 - 2 ALU
xchgw %cx, %dx # Latency: 1cy - uOPs: 2 - 2 ALU
xchgl %ecx, %edx # Latency: 1cy - uOPs: 2 - 2 ALU
xchgq %rcx, %rdx # Latency: 1cy - uOPs: 2 - 2 ALU
```
The reg-mem forms of XCHG are atomic operations with an observed latency of
16cy. The resource usage is similar to the XCHGrr variants. The biggest
difference is obviously the bus-locking, which prevents the LS to issue other
memory uOPs in parallel until the unlocking store uOP is executed.
```
xchgb %cl, (%rsp) # Latency: 16cy - uOPs: 3 - ECX latency: 11cy
xchgw %cx, (%rsp) # Latency: 16cy - uOPs: 3 - ECX latency: 11cy
xchgl %ecx, (%rsp) # Latency: 16cy - uOPs: 3 - ECX latency: 11cy
xchgq %rcx, (%rsp) # Latency: 16cy - uOPs: 3 - ECX latency: 11cy
```
The exchanged in/out register operand becomes available after 11cy from the
start of execution. Added test xchg.s to verify that we correctly see that
register write committed in 11cy (and not 16cy).
Reg-reg XADD instructions have the same latency/throughput than the byte
exchange (register-register variant).
```
xaddb %cl, %dl # latency: 2cy - uOPs: 3 - 3 ALU
xaddw %cx, %dx # latency: 2cy - uOPs: 3 - 3 ALU
xaddl %ecx, %edx # latency: 2cy - uOPs: 3 - 3 ALU
xaddq %rcx, %rdx # latency: 2cy - uOPs: 3 - 3 ALU
```
The non-atomic RM variants have a latency of 11cy, and decode to 4
macro-opcodes. They still consume 2 ALU pipes, and the exchange in/out register
operand becomes available in 3cy (it matches the 'load-to-use latency').
```
xaddb %cl, (%rsp) # latency: 11cy - uOPs: 4 - 3 ALU
xaddw %cx, (%rsp) # latency: 11cy - uOPs: 4 - 3 ALU
xaddl %ecx, (%rsp) # latency: 11cy - uOPs: 4 - 3 ALU
xaddq %rcx, (%rsp) # latency: 11cy - uOPs: 4 - 3 ALU
```
The atomic XADD variants execute in 16cy. The in/out register operand is
available after 11cy from the start of execution.
```
lock xaddb %cl, (%rsp) # latency: 16cy - uOPs: 4 - 3 ALU -- ECX latency: 11cy
lock xaddw %cx, (%rsp) # latency: 16cy - uOPs: 4 - 3 ALU -- ECX latency: 11cy
lock xaddl %ecx, (%rsp) # latency: 16cy - uOPs: 4 - 3 ALU -- ECX latency: 11cy
lock xaddq %rcx, (%rsp) # latency: 16cy - uOPs: 4 - 3 ALU -- ECX latency: 11cy
```
Added test xadd.s to verify those latencies as well as read-advance values.
Differential Revision: https://reviews.llvm.org/D66535
llvm-svn: 369642
The CodeGen/Thumb2/mve-vaddv.ll test needed to be amended to reflect the
changes from the above patch.
This reverts commit cd53ff6, reapplying 7c6b229.
llvm-svn: 369638
It broke the bots, see e.g. http://lab.llvm.org:8011/builders/clang-cuda-build/builds/36275/
> This patch fixes shifts by a 128/256 bit shift amount. It also fixes
> codegen for shifts of 32 by delegating to LLVM's default optimisation
> instead of emitting a long shift.
>
> Tests that used to generate long shifts of 32 are updated to check for the
> more optimised codegen.
>
> Differential revision: https://reviews.llvm.org/D66519
>
> llvm-svn: 369626
llvm-svn: 369636
It causes the build to fail with
"inlinable function call in a function with debug info must have a !dbg location"
in Chromium. See llvm-commits thread for more info.
(This also reverts the follow-up in r369474.)
> Fixes PR43012
>
> Differential Revision: https://reviews.llvm.org/D66328
llvm-svn: 369633
We do not need it, std::error_code is used mostly for COFF and
this patch rewrites the calls to use a different overload.
Having reportError(std::error_code EC, ... is excessive by itself,
because API that use error codes actually needs refactoring to
use Error/Expected<> instead.
DIfferential revision: https://reviews.llvm.org/D66521
llvm-svn: 369630
I don't really understand the costs we're using for fp_to_sint,
but prior to widening legalization we used 20 as the cost for this
via the v2i64->v2f64 entry. That number seems better than the 40
we got with widening legalization. So now we need either a
v2i32->v2f64 entry or a v4i32->v2f64 entry depending on whether
AVX is enabled or not since we skip the first SSE2 table look up
under AVX.
llvm-svn: 369628
Summary:
There was a subtle, but pretty important difference between the Slice
and regular versions of this function. The Slice function was
zero-initializing the rest of the buffer when the read syscall returned
less bytes than expected, while the regular function did not.
This patch removes the inconsistency by making both functions *not*
zero-initialize the buffer. The zeroing code is moved to the
MemoryBuffer class, which is currently the only user of this code. This
makes the API more consistent, and the code shorter.
While in there, I also refactor the functions to return the number of
bytes through the regular return value (via Expected<size_t>) instead of
a separate by-ref argument.
Reviewers: aganea, rnk
Subscribers: kristina, Bigcheese, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66471
llvm-svn: 369627
This patch fixes shifts by a 128/256 bit shift amount. It also fixes
codegen for shifts of 32 by delegating to LLVM's default optimisation
instead of emitting a long shift.
Tests that used to generate long shifts of 32 are updated to check for the
more optimised codegen.
Differential revision: https://reviews.llvm.org/D66519
llvm-svn: 369626
Summary:
We currently have a bunch of code that is supposed to handle invalid command options, but
all this code is unreachable because invalid options are already handled in `Options::Parse`.
The only way we can reach this code is when we declare but then not implement an option
(which will be made impossible with D65386, which is also when we can completely remove
the `default` cases).
This patch replaces all this code with `llvm_unreachable` to make clear this is dead code
that can't be reached.
Reviewers: JDevlieghere
Reviewed By: JDevlieghere
Subscribers: lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D66522
llvm-svn: 369625
Summary:
We still have some leftovers of the old completion API in the internals of
LLDB that haven't been replaced by the new CompletionRequest. These leftovers
are:
* The return values (int/size_t) in all completion functions.
* Our result array that starts indexing at 1.
* `WordComplete` mode.
I didn't replace them back then because it's tricky to figure out what exactly they
are used for and the completion code is relatively untested. I finally got around
to writing more tests for the API and understanding the semantics, so I think it's
a good time to get rid of them.
A few words why those things should be removed/replaced:
* The return values are really cryptic, partly redundant and rarely documented.
They are also completely ignored by Xcode, so whatever information they contain will end up
breaking Xcode's completion mechanism. They are also partly impossible to even implement
as we assign negative values special meaning and our completion API sometimes returns size_t.
Completion functions are supposed to return -2 to rewrite the current line. We seem to use this
in some untested code path to expand the history repeat character to the full command, but
I haven't figured out why that doesn't work at the moment.
Completion functions return -1 to 'insert the completion character', but that isn't implemented
(even though we seem to activate this feature in LLDB sometimes).
All positive values have to match the number of results. This is obviously just redundant information
as the user can just look at the result list to get that information (which is what Xcode does).
* The result array that starts indexing at 1 is obviously unexpected. The first element of the array is
reserved for the common prefix of all completions (e.g. "foobar" and "footar" -> "foo"). The idea is
that we calculate this to make the life of the API caller easier, but obviously forcing people to have
1-based indices is not helpful (or even worse, forces them to manually copy the results to make it
0-based like Xcode has to do).
* The `WordComplete` mode indicates that LLDB should enter a space behind the completion. The
idea is that we let the top-level API know that we just provided a full completion. Interestingly we
`WordComplete` is just a single bool that somehow represents all N completions. And we always
provide full completions in LLDB, so in theory it should always be true.
The only use it currently serves is providing redundant information about whether we have a single
definitive completion or not (which we already know from the number of results we get).
This patch essentially removes `WordComplete` mode and makes the result array indexed from 0.
It also removes all return values from all internal completion functions. The only non-redundant information
they contain is about rewriting the current line (which is broken), so that functionality was moved
to the CompletionRequest API. So you can now do `addCompletion("blub", "description", CompletionMode::RewriteLine)`
to do the same.
For the SB API we emulate the old behaviour by making the array indexed from 1 again with the common
prefix at index 0. I didn't keep the special negative return codes as we either never sent them before (e.g. -2) or we
didn't even implement them in the Editline handler (e.g. -1).
I tried to keep this patch minimal and I'm aware we can probably now even further simplify a bunch of related code,
but I would prefer doing this in follow-up NFC commits
Reviewers: JDevlieghere
Reviewed By: JDevlieghere
Subscribers: arphaman, abidh, lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D66536
llvm-svn: 369624
The patch introduces MakeLibCallOptions struct as suggested by @efriedma on D65497.
The struct contain argument flags which will pass to makeLibCall function.
The patch should not has any functionality changes.
Differential Revision: https://reviews.llvm.org/D65795
llvm-svn: 369622
Without this patch, failing to provide a subcommand to lit's internal
`env` results in either a python `IndexError` or an attempt to execute
the final `env` argument, such as `FOO=1`, as a command. This patch
diagnoses those cases with a more helpful message.
Reviewed By: stella.stamenova
Differential Revision: https://reviews.llvm.org/D66482
llvm-svn: 369620
Benjamin Kramer