Accord the discussion in D120953, we should firstly exclude all scalable vector
extending loads and then selectively enable those which we directly support.
This patch is intend to refactor for above (truncating stores is not touched),and
more scalable vector types will try to reduce the number of masked loads in favour
of more unpklo/hi instructions.
Reviewed By: paulwalker-arm
Differential Revision: https://reviews.llvm.org/D122281
We wish to support emitting a pre-processed output for an importable
header unit, that can be consumed to produce the same header units as
the original source.
This means that ee need to find the original filename used to produce
the re-preprocessed output, so that it can be assigned as the module
name. This is peeked from the first line of the pre-processed source
when the action sets up the files.
Differential Revision: https://reviews.llvm.org/D121098
This is used for f16 emulation. We emulate f16 for SSE2 targets and
above. Refactoring makes the future code to be more clean.
Reviewed By: LuoYuanke
Differential Revision: https://reviews.llvm.org/D122475
In DAGISel, the parameter alignment only have 4 bits to hold the value.
The encode(alignment) would plus the value by 1, so the max aligment that
ISel can support is 2^14. This patch verify align attribute for parameter.
Differential Revision: https://reviews.llvm.org/D122130
In DAGISel, the parameter alignment only have 4 bits to hold the value.
The encode(alignment) would plus the shift value by 1, so the max aligment
ISel can support is 2^14. This patch verify the parameter and return
value for alignment.
Differential Revision: https://reviews.llvm.org/D121898
The default expansion for buildvectors is to extract each element and
insert them into a new vector. That involves a lot of copying to/from
the GPR registers. TLB3 and TLB4 can be relatively slow instructions
with the mask needing to be loaded from a constant pool, but they should
always be better than all the moves to/from GPRs.
Differential Revision: https://reviews.llvm.org/D121137
Some signal handlers were set up within an !_MSC_VER condition,
i.e. omitted in MSVC builds but included in mingw builds. Previously
sigtstp_handler was defined in all builds, but since
4bcadd6686 / D120320 it's only
defined non platforms other than Windows.
Change the condition to !_WIN32 for consistency between the MSVC
and mingw builds, fixing the build for mingw.
Differential Revision: https://reviews.llvm.org/D122486
Now the decoded thread has Append methods that provide more flexibility
in terms of the underlying data structure that represents the
instructions. In this case, we are able to represent the sporadic errors
as map and thus reduce the size of each instruction.
Differential Revision: https://reviews.llvm.org/D122293
For header units we build the top level module directly from the header
that it represents and macros defined in this TU need to be emitted (when
such a definition is live at the end of the TU).
Differential Revision: https://reviews.llvm.org/D121097
Instead of using a temporary `string` in `__vformat_to_wrapped` use a new
generic iterator. This aids to reduce the number of template instantions
and avoids using a `string` to buffer the entire formatted output.
This changes the type of `format_context` and `wformat_context`, this can
still be done since the code isn't ABI stable yet.
Several approaches have been evaluated:
- Using a __output_buffer base class with:
- a put function to store the buffer in its internal buffer
- a virtual flush function to copy the internal buffer to the output
- Using a `function` to forward the output operation to the output buffer,
much like the next method.
- Using a type erased function point to store the data in the buffer.
The last version resulted in the best performance. For some cases there's
still a loss of speed over the original method. This loss many becomes
apparent when large strings are copied to a pointer like iterator, before
the compiler optimized this using `memcpy`.
Reviewed By: ldionne, vitaut, #libc
Differential Revision: https://reviews.llvm.org/D110495
1. Add comments to explain why we set `isAsmParserOnly` for XACQUIRE and XRELEASE
2. Check `X86Inst` in the constructor of `RecognizableInstrBase` so that
we can avoid the case where one of it's field is not initialized but
accessed by user. (e.g. in X86EVEX2VEXTablesEmitter.cpp)
3. Move `Rec` from `RecognizableInstrBase` to `RecognizableInstr` to reduce
size of `RecognizableInstrBase`
4. Remove out-of-date comments for shouldBeEmitted() (filter() was removed)
5. Add a basic field `IsAsmParserOnly` and remove the field
`ShouldBeEmitted` b/c we can deduce it w/ little overhead
In C, assignment expressions result in an rvalue whose type is the type
of the lhs of the assignment after it undergoes lvalue to rvalue
conversion. lvalue to rvalue conversion in C strips all qualifiers
including _Atomic.
We used getUnqualifiedType() which does not strip the _Atomic qualifier
when we should have used getAtomicUnqualifiedType(). This corrects the
usage and adds some comments to getUnqualifiedType() to make it more
clear that it does not strip _Atomic and that's on purpose (see C11
6.2.5p27).
This addresses Issue 48742.
D117296 removed wording for __builtin_assume, D120205 restored the
wording, but the last sentence was only partly restored. This restores
the rest of the last sentence.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D122423
Now, perform last active true vector combine only where
we're extracting from a flag-setting operation. But in
fact, the last active extracting will output LASTB + WHILELS,
and the WHILELS itself is a flag-setting operation, so
precommit this case to test the potentially further optimization.
Reviewed By: paulwalker-arm
Differential Revision: https://reviews.llvm.org/D122453
In fact, an instruction can not be emitted to disassemble table when
`isAsmParserOnly` is true, so `isAsmParserOnly=true` implies
`ShouldBeEmitted=false`.
We check `isAsmParserOnly` in X86FoldTablesEmitter.cpp at a early stage
b/c none of them is foldable.
This is support for the user-facing options to create importable header units
from headers in the user or system search paths (or to be given an absolute path).
This means that an incomplete header path will be passed by the driver and the
lookup carried out using the search paths present when the front end is run.
To support this, we introduce file fypes for c++-{user,system,header-unit}-header.
These terms are the same as the ones used by GCC, to minimise the differences for
tooling (and users).
The preprocessor checks for headers before issuing a warning for
"#pragma once" in a header build. We ensure that the importable header units
are recognised as headers in order to avoid such warnings.
Differential Revision: https://reviews.llvm.org/D121096
As we mentioned in the code comments for function `ResourcePoolTy::release`,
at some point there could be two identical resources on the two sides of `Next`
mark. It is usually not an issue, unless the following case:
1. Some resources are not returned.
2. We need to iterate the pool and free the element.
That will cause double free, which is the case for event pool. Since we don't release
events hold by the data map, it can happen that the `Next` mark is not reset, and
we have two identical items in the pool. When the pool is destroyed, we will call
`cuEventDestroy` twice on the same event. In the best case, we can only observe
CUDA errors. In the worst case, it can cause internal failures in CUDART and further
crash.
This patch fixes the issue by tracking all resources that have been given using
an `unordered_set`. We don't remove it when a resource is returned. When the pool
is destroyed, we merge the pool (a `vector`) and the set. In this way, we can make
sure that the set contains all resources allocated from the device. We just need
to iterate the set and free the resource accordingly.
For now, only event pool is set to use it. Stream pool is not because we can make
sure all streams are returned when the plugin is destroyed.
Someone might be wondering, why don't we release all events hold in the data map.
That is because, plugins are determined to be destroyed *before* `libomptarget`.
If we can somehow make the plugin outlast `libomptarget`, life will be much
easier.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D122014
This patch adds the necessary AMDGPU calling convention to the ctor /
dtor kernels. These are fundamentally device kenels called by the host
on image load. Without this calling convention information the AMDGPU
plugin is unable to identify them.
Depends on D122504
Fixes#54091
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D122515
The default construction of constructor functions by LLVM tends to make
them have internal linkage. When we call a ctor / dtor function in the
target region we are actually creating a kernel that is called at
registration. Because the ctor is a kernel we need to make sure it's
externally visible so we can actually call it. This prevented AMDGPU
from correctly using constructors while NVPTX could use them simply
because it ignored internal visibility.
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D122504
All LLVM backends use MCDisassembler as a base class for their
instruction decoders. Use "const MCDisassembler *" for the decoder
instead of "const void *". Remove unnecessary static casts.
Reviewed By: skan
Differential Revision: https://reviews.llvm.org/D122245
STATUS='NEW' and 'REPLACE' require FILE= to be present.
STATUS='SCRATCH' may not appear with FILE=.
These errors are caught at compilation time when constant character
strings are used in an OPEN statement, but the runtime needs
to enforce them as well to catch errors in OPEN statements
with character variables and expressions.
Differential Revision: https://reviews.llvm.org/D122509
With the shared cache getting split into multiple files, the current
way we created ObjectFileMachO objects for shared cache dylib images
will break.
This patch conditionally adopts new SPIs which will do the right
thing in the new world of multi-file caches.
Some functions can end up non-externally visible despite not being
declared "static" or in an unnamed namespace in C++ - such as by having
parameters that are of non-external types.
Such functions aren't mistakenly intended to be defining some function
that needs a declaration. They could be maybe more legible (except for
the operator new example) with an explicit static, but that's a
stylistic thing outside what should be addressed by a warning.
This reapplies 275c56226d - once we figure
out what to do about the change in behavior for -Wnon-c-typedef-for-linkage
(this reverts the revert commit 85ee1d3ca1)
Differential Revision: https://reviews.llvm.org/D121328
We could only do this in limited ways (since we emit the TUs first, we
can't use ref_addr (& we can't use that in Split DWARF either) - so we
had to synthesize declarations into the TUs) and they were ambiguous in
some cases (if the CU type had internal linkage, parsing the TU would
require knowing which CU was referencing the TU to know which type the
declaration was for, which seems not-ideal). So to avoid all that, let's
just not reference types defined in the CU from TUs - instead moving the
TU type into the CU (recursively).
This does increase debug info size (by pulling more things out of type
units, into the compile unit) - about 2% of uncompressed dwp file size
for clang -O0 -g -gsplit-dwarf. (5% .debug_info.dwo section size
increase in the .dwp)
chenglin.bi