Summary:
Negative arguments in kernel attributes are silently bitcast'ed to
unsigned, for example:
__attribute__((reqd_work_group_size(1, -1, 1)))
__kernel void k() {}
is a complete equivalent of:
__attribute__((reqd_work_group_size(1, 4294967294, 1)))
__kernel void k() {}
This is likely an error, so the patch forbids negative arguments in
several OpenCL attributes. Users who really want 4294967294 can still
use it as an unsigned representation.
Reviewers: Anastasia, yaxunl, bader
Reviewed By: Anastasia, yaxunl, bader
Subscribers: bader, cfe-commits
Differential Revision: https://reviews.llvm.org/D50259
llvm-svn: 341539
Summary:
Given the following kernel:
__kernel void foo() {
double d;
double4 dd;
}
and cl_khr_fp64 is disabled, the compilation would fail due to
the presence of 'double d', but when removed, it passes.
The expectation is that extended vector types of unsupported types
will also be unsupported.
The patch adds the check for this scenario.
Patch by: Ofir Cohen
Reviewers: bader, Anastasia, AlexeySotkin, yaxunl
Reviewed By: Anastasia
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D51296
llvm-svn: 341309
Recommit of r335084 after revert in r335516.
... instead of prepending it at the beginning (the original behavior
since implemented in r122535 2010-12-23). This builds up an
AttributeList in the the order in which the attributes appear in the
source.
The reverse order caused nodes for attributes in the AST (e.g. LoopHint)
to be in the reverse order, and therefore printed in the wrong order in
-ast-dump. Some TODO comments mention this. The order was explicitly
reversed for enable_if attribute overload resolution and name mangling,
which is not necessary anymore with this patch.
The change unfortunately has some secondary effect, especially on
diagnostic output. In the simplest cases, the CHECK lines or expected
diagnostic were changed to the the new output. If the kind of
error/warning changed, the attributes' order was changed instead.
This unfortunately causes some 'previous occurrence here' hints to be
textually after the main marker. This typically happens when attributes
are merged, but are incompatible to each other. Interchanging the role
of the the main and note SourceLocation will also cause the case where
two different declaration's attributes (in contrast to multiple
attributes of the same declaration) are merged to be reverse. There is
no easy fix because sometimes previous attributes are merged into a new
declaration's attribute list, sometimes new attributes are added to a
previous declaration's attribute list. Since 'previous occurrence here'
pointing to locations after the main marker is not rare, I left the
markers as-is; it is only relevant when the attributes are declared in
the same declaration anyway.
Differential Revision: https://reviews.llvm.org/D48100
llvm-svn: 338800
Summary:
C++11 onwards specs the non-member functions atomic_load and atomic_load_explicit as taking the atomic<T> by const (potentially volatile) pointer. C11, in its infinite wisdom, decided to drop the const, and C17 will fix this with DR459 (the current draft forgot to fix B.16, but that’s not the normative part).
clang’s lib/Headers/stdatomic.h implements these as #define to the __c11_* equivalent, which are builtins with custom typecheck. Fix the typecheck.
D47613 takes care of the libc++ side.
Discussion: http://lists.llvm.org/pipermail/cfe-dev/2018-May/058129.html
<rdar://problem/27426936>
Reviewers: rsmith
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D47618
llvm-svn: 338743
The way address space declarations for builtins currently work
is nearly useless. The code assumes the address spaces used for
builtins is a confusingly named "target address space" from user
code using __attribute__((address_space(N))) that matches
the builtin declaration. There's no way to use this to declare
a builtin that returns a language specific address space.
The terminology used is highly cofusing since it has nothing
to do with the the address space selected by the target to use
for a language address space.
This feature is essentially unused as-is. AMDGPU and NVPTX
are the only in-tree targets attempting to use this. The AMDGPU
builtins certainly do not behave as intended (i.e. all of the
builtins returning pointers can never compile because the numbered
address space never matches the expected named address space).
The NVPTX builtins are missing tests for some, and the others
seem to rely on an implicit addrspacecast.
Change the used address space for builtins based on a target
hook to allow using a language address space for a builtin.
This allows the same builtin declaration to be used for multiple
languages with similarly purposed address spaces (e.g. the same
AMDGPU builtin can be used in OpenCL and CUDA even though the
constant address spaces are arbitarily different).
This breaks the possibility of using arbitrary numbered
address spaces alongside the named address spaces for builtins.
If this is an issue we probably need to introduce another builtin
declaration character to distinguish language address spaces from
so-called "target address spaces".
llvm-svn: 338707
Summary:
Size_t, intptr_t, uintptr_t and ptrdiff_t cannot be used as kernel
arguments, according to OpenCL Specification s6.9k:
The size in bytes of these types are implementation-defined and in
addition can also be different for the OpenCL device and the host
processor making it difficult to allocate buffer objects to be passed
as arguments to a kernel declared as pointer to these types.
Patch by: Andrew Savonichev
Reviewers: Anastasia, yaxunl
Subscribers: yaxunl, Anastasia, cfe-commits
Differential Revision: https://reviews.llvm.org/D49725
llvm-svn: 338432
Summary:
OpenCL specification forbids use of several types as kernel arguments.
This patch improves existing diagnostic to look through arrays.
Patch by: Andrew Savonichev
Reviewers: Anastasia, yaxunl
Subscribers: yaxunl, Anastasia, cfe-commits
Differential Revision: https://reviews.llvm.org/D49723
llvm-svn: 338427
Summary:
As discussed in [[ https://bugs.llvm.org/show_bug.cgi?id=38166 | PR38166 ]], we need to be able to distinqush whether the cast
we are visiting is actually a cast, or part of an `ExplicitCast`.
There are at least four ways to get there:
1. Introduce a new `CastKind`, and use it instead of `IntegralCast` if we are in `ExplicitCast`.
Would work, but does not scale - what if we will need more of these cast kinds?
2. Introduce a flag in `CastExprBits`, whether this cast is part of `ExplicitCast` or not.
Would work, but it isn't immediately clear where it needs to be set.
2. Fix `ScalarExprEmitter::VisitCastExpr()` to visit these `NoOp` casts.
As pointed out by @rsmith, CodeGenFunction::EmitMaterializeTemporaryExpr calls
skipRValueSubobjectAdjustments, which steps over the CK_NoOp cast`,
which explains why we currently don't visit those.
This is probably impossible, as @efriedma points out, that is intentional as per `[class.temporary]` in the standard
3. And the simplest one, just record which NoOp casts we skip.
It just kinda works as-is afterwards.
But, the approach with a flag is the least intrusive one, and is probably the best one overall.
Reviewers: rsmith, rjmccall, majnemer, efriedma
Reviewed By: rsmith
Subscribers: cfe-commits, aaron.ballman, vsk, llvm-commits, rsmith
Differential Revision: https://reviews.llvm.org/D49508
llvm-svn: 337815
conversions are only applied to operands of class type, and the second
standard conversion sequence is not applied.
When diagnosing an invalid builtin binary operator, talk about the
original types rather than the converted types. If these differ by a
user-defined conversion, tell the user what happened.
llvm-svn: 335781
Summary:
The comment with the OpenCL clause about this clearly
says: "No type shall be qualified by qualifiers for
two or more different address spaces."
This must mean that two or more qualifiers for the
_same_ address space is allowed. However, it is
likely unintended by the programmer, so emit a
warning.
For dependent address space types, reject them like
before since we cannot know what the address space
will be.
Patch by Bevin Hansson (ebevhan).
Reviewers: Anastasia
Reviewed By: Anastasia
Subscribers: bader, cfe-commits
Differential Revision: https://reviews.llvm.org/D47630
llvm-svn: 335103
... instead of prepending it at the beginning (the original behavior
since implemented in r122535 2010-12-23). This builds up an
AttributeList in the the order in which the attributes appear in the
source.
The reverse order caused nodes for attributes in the AST (e.g. LoopHint)
to be in the reverse, and therefore printed in the wrong order by
-ast-dump. Some TODO comments mention this. The order was explicitly
reversed for enable_if attribute overload resolution and name mangling,
which is not necessary anymore with this patch.
The change unfortunately has some secondary effects, especially for
diagnostic output. In the simplest cases, the CHECK lines or expected
diagnostic were changed to the the new output. If the kind of
error/warning changed, the attribute's order was changed instead.
It also causes some 'previous occurrence here' hints to be textually
after the main marker. This typically happens when attributes are
merged, but are incompatible. Interchanging the role of the the main
and note SourceLocation will also cause the case where two different
declaration's attributes (in contrast to multiple attributes of the
same declaration) are merged to be reversed. There is no easy fix
because sometimes previous attributes are merged into a new
declaration's attribute list, sometimes new attributes are added to a
previous declaration's attribute list. Since 'previous occurrence here'
pointing to locations after the main marker is not rare, I left the
markers as-is; it is only relevant when the attributes are declared in
the same declaration anyway, which often is on the same line.
Differential Revision: https://reviews.llvm.org/D48100
llvm-svn: 335084
There are HIP applications e.g. Tensorflow 1.3 using amdgpu kernel attributes, however
currently they are only allowed on OpenCL kernel functions.
This patch will allow amdgpu kernel attributes to be applied to CUDA/HIP __global__
functions.
Differential Revision: https://reviews.llvm.org/D47958
llvm-svn: 334561
1. added restrictions to memory scope, order and volatile parameters
2. added custom processing for these builtins - currently is not used code,
needed to switch off GCCBuiltin link to the builtins (ongoing change to llvm
tree)
3. builtins renamed as requested
Differential Revision: https://reviews.llvm.org/D43281
llvm-svn: 332848
Added string literal helper function to obtain the type
attributed by a constant address space.
Also fixed predefind __func__ expr to use the helper
to constract the string literal correctly.
Differential Revision: https://reviews.llvm.org/D46049
llvm-svn: 331877
Restrict the following keywords in the OpenCL C++ language mode,
according to Sections 2.2 & 2.9 of the OpenCL C++ 1.0 Specification.
- dynamic_cast
- typeid
- register (already restricted in OpenCL C, update the diagnostic)
- thread_local
- exceptions (try/catch/throw)
- access qualifiers read_only, write_only, read_write
Support the `__global`, `__local`, `__constant`, `__private`, and
`__generic` keywords in OpenCL C++. Leave the unprefixed address
space qualifiers such as global available, i.e., do not mark them as
reserved keywords in OpenCL C++. libclcxx provides explicit address
space pointer classes such as `global_ptr` and `global<T>` that are
implemented using the `__`-prefixed qualifiers.
Differential Revision: https://reviews.llvm.org/D46022
llvm-svn: 331874
Summary:
For OpenCL 1.1 embedded profile 64 bit integers i.e. long,
ulong including the appropriate vector data types and operations
on 64-bit integers are optional. The "cles_khr_int64" extension
string will be reported if the embedded profile implementation
supports 64-bit integers.
Reviewers: Anastasia, bader
Reviewed By: Anastasia, bader
Subscribers: bader, yaxunl, Anastasia, cfe-commits
Differential Revision: https://reviews.llvm.org/D42532
llvm-svn: 323522
Summary:
Constant samplers are handled as static variables and clang's code generation
library, which leads to llvm::unreachable. We bypass emitting sampler variable
as static since it's translated to a function call later.
Reviewers: yaxunl, Anastasia
Reviewed By: yaxunl, Anastasia
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D34342
llvm-svn: 318290
Changes behavior introduced in r298369 to only error out on
vector component invalid length access on OpenCL mode.
Differential Revision: https://reviews.llvm.org/D38868
rdar://problem/33568748
llvm-svn: 316016
Currently Clang uses default address space (0) to represent private address space for OpenCL
in AST. There are two issues with this:
Multiple address spaces including private address space cannot be diagnosed.
There is no mangling for default address space. For example, if private int* is emitted as
i32 addrspace(5)* in IR. It is supposed to be mangled as PUAS5i but it is mangled as
Pi instead.
This patch attempts to represent OpenCL private address space explicitly in AST. It adds
a new enum LangAS::opencl_private and adds it to the variable types which are implicitly
private:
automatic variables without address space qualifier
function parameter
pointee type without address space qualifier (OpenCL 1.2 and below)
Differential Revision: https://reviews.llvm.org/D35082
llvm-svn: 315668
Treat 'f()' as 'f(void)' rather than a function w/o a prototype.
Reviewers: Anastasia, yaxunl
Reviewed By: Anastasia, yaxunl
Subscribers: cfe-commits, echuraev, chapuni
Differential Revision: https://reviews.llvm.org/D33681
Re-apply revision 306653.
llvm-svn: 315453
Currently clang allows the following code
int a;
int b = (const int) a;
However it does not the following code
int4 a;
int4 b = (const int4) a;
This is because Clang compares the qualified types instead of unqualified types for vector type casting, which causes the inconsistency.
This patch fixes that.
Differential Revision: https://reviews.llvm.org/D38463
llvm-svn: 314802
Block captures can have different physical locations
in memory segments depending on the use case (as a function
call or as a kernel enqueue) and in different vendor
implementations.
Therefore it's unclear how to add address space to capture
addresses uniformly. Currently it has been decided to disallow
taking addresses of captured variables until further
clarifications in the spec.
Differential Revision: https://reviews.llvm.org/D36410
llvm-svn: 312728
OpenCL 2.0 atomic builtin functions have a scope argument which is ideally
represented as synchronization scope argument in LLVM atomic instructions.
Clang supports translating Clang atomic builtin functions to LLVM atomic
instructions. However it currently does not support synchronization scope
of LLVM atomic instructions. Without this, users have to use LLVM assembly
code to implement OpenCL atomic builtin functions.
This patch adds OpenCL 2.0 atomic builtin functions as Clang builtin
functions, which supports generating LLVM atomic instructions with
synchronization scope operand.
Currently only constant memory scope argument is supported. Support of
non-constant memory scope argument will be added later.
Differential Revision: https://reviews.llvm.org/D28691
llvm-svn: 310082
Summary:
does it make sense to enable K&R function declaration style for OpenCL?
clang throws following error message for the declaration w/o arguments:
```
int my_func();
error: function with no prototype cannot use the spir_function calling convention
```
Current way to fix this issue is to specify that parameter list is empty by using 'void':
```
int my_func(void);
```
Let me know what do you think about this patch.
Reviewers: Anastasia, yaxunl
Reviewed By: Anastasia
Subscribers: cfe-commits, echuraev
Differential Revision: https://reviews.llvm.org/D33681
llvm-svn: 306653
Produce an error if variables qualified with a local or
a constant address space are not declared in the outermost
scope of a kernel.
Patch by Simon Perretta.
Differential Revision: https://reviews.llvm.org/D34024
llvm-svn: 305798
Summary:
This is the fix for patch https://reviews.llvm.org/D33353
@uweigand, could you please verify that everything will be good on SystemZ?
I added triple spir-unknown-unknown.
Thank you in advance!
Reviewers: uweigand
Reviewed By: uweigand
Subscribers: yaxunl, cfe-commits, bader, Anastasia, uweigand
Differential Revision: https://reviews.llvm.org/D33648
llvm-svn: 304191
Andrew Savonichev