Summary:
Support for OpenCL 2.0 pipe type.
This is a bug-fix version for bader's patch reviews.llvm.org/D14441
Reviewers: pekka.jaaskelainen, Anastasia
Subscribers: bader, Anastasia, cfe-commits
Differential Revision: http://reviews.llvm.org/D15603
llvm-svn: 257254
This patch attempts to fix the regressions identified when the patch was committed initially.
Thanks to Michael Liao for identifying the fix in the offloading metadata generation
related with side effects in evaluation of function arguments.
llvm-svn: 256933
Summary:
In order to offloading work properly two things need to be in place:
- a descriptor with all the offloading information (device entry functions, and global variable) has to be created by the host and registered in the OpenMP offloading runtime library.
- all the device functions need to be emitted for the device and a convention has to be in place so that the runtime library can easily map the host ID of an entry point with the actual function in the device.
This patch adds support for these two things. However, only entry functions are being registered given that 'declare target' directive is not yet implemented.
About offloading descriptor:
The details of the descriptor are explained with more detail in http://goo.gl/L1rnKJ. Basically the descriptor will have fields that specify the number of devices, the pointers to where the device images begin and end (that will be defined by the linker), and also pointers to a the begin and end of table whose entries contain information about a specific entry point. Each entry has the type:
```
struct __tgt_offload_entry{
void *addr;
char *name;
int64_t size;
};
```
and will be implemented in a pre determined (ELF) section `.omp_offloading.entries` with 1-byte alignment, so that when all the objects are linked, the table is in that section with no padding in between entries (will be like a C array). The code generation ensures that all `__tgt_offload_entry` entries are emitted in the same order for both host and device so that the runtime can have the corresponding entries in both host and device in same index of the table, and efficiently implement the mapping.
The resulting descriptor is registered/unregistered with the runtime library using the calls `__tgt_register_lib` and `__tgt_unregister_lib`. The registration is implemented in a high priority global initializer so that the registration happens always before any initializer (that can potentially include target regions) is run.
The driver flag -omptargets= was created to specify a comma separated list of devices the user wants to support so that the new functionality can be exercised. Each device is specified with its triple.
About target codegen:
The target codegen is pretty much straightforward as it reuses completely the logic of the host version for the same target region. The tricky part is to identify the meaningful target regions in the device side. Unlike other programming models, like CUDA, there are no already outlined functions with attributes that mark what should be emitted or not. So, the information on what to emit is passed in the form of metadata in host bc file. This requires a new option to pass the host bc to the device frontend. Then everything is similar to what happens in CUDA: the global declarations emission is intercepted to check to see if it is an "interesting" declaration. The difference is that instead of checking an attribute, the metadata information in checked. Right now, there is only a form of metadata to pass information about the device entry points (target regions). A class `OffloadEntriesInfoManagerTy` was created to manage all the information and queries related with the metadata. The metadata looks like this:
```
!omp_offload.info = !{!0, !1, !2, !3, !4, !5, !6}
!0 = !{i32 0, i32 52, i32 77426347, !"_ZN2S12r1Ei", i32 479, i32 13, i32 4}
!1 = !{i32 0, i32 52, i32 77426347, !"_ZL7fstatici", i32 461, i32 11, i32 5}
!2 = !{i32 0, i32 52, i32 77426347, !"_Z9ftemplateIiET_i", i32 444, i32 11, i32 6}
!3 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 99, i32 11, i32 0}
!4 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 272, i32 11, i32 3}
!5 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 127, i32 11, i32 1}
!6 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 159, i32 11, i32 2}
```
The fields in each metadata entry are (in sequence):
Entry 1) an ID of the type of metadata - right now only zero is used meaning "OpenMP target region".
Entry 2) a unique ID of the device where the input source file that contain the target region lives.
Entry 3) a unique ID of the file where the input source file that contain the target region lives.
Entry 4) a mangled name of the function that encloses the target region.
Entries 5) and 6) line and column number where the target region was found.
Entry 7) is the order the entry was emitted.
Entry 2) and 3) are required to distinguish files that have the same function name.
Entry 4) is required to distinguish different instances of the same declaration (usually templated ones)
Entries 5) and 6) are required to distinguish the particular target region in body of the function (it is possible that a given target region is not an entry point - if clause can evaluate always to zero - and therefore we need to identify the "interesting" target regions. )
This patch replaces http://reviews.llvm.org/D12306.
Reviewers: ABataev, hfinkel, tra, rjmccall, sfantao
Subscribers: FBrygidyn, piotr.rak, Hahnfeld, cfe-commits
Differential Revision: http://reviews.llvm.org/D12614
llvm-svn: 256842
than reusing the "overridden buffer" mechanism. This will allow us to make
embedded files and overridden files behave differently in future.
llvm-svn: 254121
https://gcc.gnu.org/onlinedocs/gcc/Typeof.html
Differences from the GCC extension:
* __auto_type is also permitted in C++ (but only in places where
it could appear in C), allowing its use in headers that might
be shared across C and C++, or used from C++98
* __auto_type can be combined with a declarator, as with C++ auto
(for instance, "__auto_type *p")
* multiple variables can be declared in a single __auto_type
declaration, with the C++ semantics (the deduced type must be
the same in each case)
This patch also adds a missing restriction on applying typeof to
a bit-field, which GCC has historically rejected in C (due to
lack of clarity as to whether the operand should be promoted).
The same restriction also applies to __auto_type in C (in both
GCC and Clang).
This also fixes PR25449.
Patch by Nicholas Allegra!
llvm-svn: 252690
we can't load that file due to a configuration mismatch, and implicit module
building is disabled, and the user turns off the error-by-default warning for
that situation, then fall back to textual inclusion for the module rather than
giving an error if any of its headers are included.
llvm-svn: 252114
This new builtin template allows for incredibly fast instantiations of
templates like std::integer_sequence.
Performance numbers follow:
My work station has 64 GB of ram + 20 Xeon Cores at 2.8 GHz.
__make_integer_seq<std::integer_sequence, int, 90000> takes 0.25
seconds.
std::make_integer_sequence<int, 90000> takes unbound time, it is still
running. Clang is consuming gigabytes of memory.
Differential Revision: http://reviews.llvm.org/D13786
llvm-svn: 252036
Introduce the notion of a module file extension, which introduces
additional information into a module file at the time it is built that
can then be queried when the module file is read. Module file
extensions are identified by a block name (which must be unique to the
extension) and can write any bitstream records into their own
extension block within the module file. When a module file is loaded,
any extension blocks are matched up with module file extension
readers, that are per-module-file and are given access to the input
bitstream.
Note that module file extensions can only be introduced by
programmatic clients that have access to the CompilerInvocation. There
is only one such extension at the moment, which is used for testing
the module file extension harness. As a future direction, one could
imagine allowing the plugin mechanism to introduce new module file
extensions.
llvm-svn: 251955
Use the *current* state of "is-moduleness" rather than the state at
serialization time so that if we read a builtin identifier from a module
that wasn't "interesting" to that module, we will still write it out to
a PCH that imports that module.
Otherwise, we would get mysterious "unknown builtin" errors when using
PCH+modules.
rdar://problem/23287656
llvm-svn: 251565
Summary: It breaks the build for the ASTMatchers
Subscribers: klimek, cfe-commits
Differential Revision: http://reviews.llvm.org/D13893
llvm-svn: 250827
Summary:
This change adds support for `__builtin_ms_va_list`, a GCC extension for
variadic `ms_abi` functions. The existing `__builtin_va_list` support is
inadequate for this because `va_list` is defined differently in the Win64
ABI vs. the System V/AMD64 ABI.
Depends on D1622.
Reviewers: rsmith, rnk, rjmccall
CC: cfe-commits
Differential Revision: http://reviews.llvm.org/D1623
llvm-svn: 247941
them directly to the control block. These are fairly large, and in a build with
lots of modules / chained PCH, we don't need to read most of them. No
functionality change intended.
llvm-svn: 247055
r246546, with a workaround for an MSVC 2013 miscompile and an MSVC 2015
rejects-valid.
Original commit message:
[modules] Rework serialized DeclContext lookup table management. Instead of
walking the loaded ModuleFiles looking for lookup tables for the context, store
them all in one place, and merge them together if we find we have too many
(currently, more than 4). If we do merge, include the merged form in our
serialized lookup table, so that downstream readers never need to look at our
imports' tables.
This gives a huge performance improvement to builds with very large numbers of
modules (in some cases, more than a 2x speedup was observed).
llvm-svn: 246582
predictable diagnostic experience. The hash-of-DeclID order we were using
before gave different results on Win32 due to a different predefined
declaration of __builtin_va_list.
llvm-svn: 246521
walking the loaded ModuleFiles looking for lookup tables for the context, store
them all in one place, and merge them together if we find we have too many
(currently, more than 4). If we do merge, include the merged form in our
serialized lookup table, so that downstream readers never need to look at our
imports' tables.
This gives a huge performance improvement to builds with very large numbers of
modules (in some cases, more than a 2x speedup was observed).
llvm-svn: 246497
DeclarationName (because all ctor names are considered the same, and so on).
Reflect this in the type used as the lookup table key. As a side-effect, remove
one copy of the duplicated code used to compute the hash of the key.
llvm-svn: 246124
Adds parsing/sema analysis/serialization/deserialization for array sections in OpenMP constructs (introduced in OpenMP 4.0).
Currently it is allowed to use array sections only in OpenMP clauses that accepts list of expressions.
Differential Revision: http://reviews.llvm.org/D10732
llvm-svn: 245937
Instead of eagerly deserializing a list of DeclIDs when we load a module file
and doing a binary search to find the redeclarations of a decl, store a list of
redeclarations of each chain before the first declaration and load it directly.
llvm-svn: 245789
all modules and reduce the number of declarations we load when loading a
redeclaration chain.
The new approach is:
* when loading the first declaration of an entity within a module file, we
first load all declarations of the entity that were imported into that
module file, and then load all the other declarations of that entity from
that module file and build a suitable decl chain from them
* when loading any other declaration of an entity, we first load the first
declaration from the same module file
As before, we complete redecl chains through name lookup where necessary.
To make this work, I also had to change the way that template specializations
are stored -- it no longer suffices to track only canonical specializations; we
now emit all "first local" declarations when emitting a list of specializations
for a template.
On one testcase with several thousand imported module files, this reduces the
total runtime by 72%.
llvm-svn: 245779
file in the .pcm files. This allows a smaller set of files to be sent to a
remote build worker when building with explicit modules (for instance, module
map files need not be sent along with the corresponding precompiled modules).
This doesn't actually make the embedded files visible to header search, so
it's not useful as a packaging format for public header files.
llvm-svn: 245028
files: include the .pcm file itself in the .d output, rather than including its
own input files. Other forms of module file continue to be transparent for .d
output.
Arguably, the input files for the .pcm file are still inputs to the
compilation, but that's unnecessary for make-like build systems (where the
mtime of the .pcm file is sufficient) and harmful for smarter build systems
that know about module files and want to track only the local dependencies.
llvm-svn: 244923
emit lexical contents for a declaration for another module. Track which module
those contents came from, and ensure that we only grab the lexical contents
from a single such instantiation.
llvm-svn: 244682
arguments because the reloaded form might have become non-canonical across the
serialization/deserialization step (this particularly happens when the
canonical form of the type involves an expression).
llvm-svn: 244409
Xiuli Pan