Add in-source documentation on how CanonicalLoopInfo is intended to be used. In particular, clarify what parts of a CanonicalLoopInfo is considered part of the loop, that those parts must be side-effect free, and that InsertPoints to instructions outside those parts can be expected to be preserved after method calls implementing loop-associated directives.
CanonicalLoopInfo are now invalidated after it does not describe canonical loop anymore and asserts when trying to use it afterwards.
In addition, rename `createXYZWorkshareLoop` to `applyXYZWorkshareLoop` and remove the update location to avoid that the impression that they insert something from scratch at that location where in reality its InsertPoint is ignored. createStaticWorkshareLoop does not return a CanonicalLoopInfo anymore. First, it was not a canonical loop in the clarified sense (containing side-effects in form of calls to the OpenMP runtime). Second, it is ambiguous which of the two possible canonical loops it should actually return. It will not be needed before a feature expected to be introduced in OpenMP 6.0
Also see discussion in D105706.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D107540
- They need to be preserved even if there's no reference within the
device code as the host code may need to initialize them based on the
application logic.
Reviewed By: tra
Differential Revision: https://reviews.llvm.org/D107718
We were using an OpaqueValueExpr allocated on the stack to store
the size of a VLA. Because the VLASizeMap in CodegenFunction
uses the address of the expression to avoid recomputing VLAs,
we were accidentally reusing an earlier llvm::Value. This led to
invalid LLVM IR.
This is a temporary solution until VLASizeMap can be pushed and popped
based on the context.
Differential Revision: https://reviews.llvm.org/D107666
After D94315 we add the `NoInline` attribute to the outlined function to handle
data environments in the OpenMP if clause. This conflicted with the `AlwaysInline`
attribute added to the outlined function. for better performance in D106799.
The data environments should ideally not require NoInline, but for now this
fixes PR51349.
Reviewed By: mikerice
Differential Revision: https://reviews.llvm.org/D107649
This is recommit of the patch 16ff91ebcc,
reverted in 0c28a7c990 because it had
an error in call of getFastMathFlags (base type should be FPMathOperator
but not Instruction). The original commit message is duplicated below:
Clang has builtin function '__builtin_isnan', which implements C
library function 'isnan'. This function now is implemented entirely in
clang codegen, which expands the function into set of IR operations.
There are three mechanisms by which the expansion can be made.
* The most common mechanism is using an unordered comparison made by
instruction 'fcmp uno'. This simple solution is target-independent
and works well in most cases. It however is not suitable if floating
point exceptions are tracked. Corresponding IEEE 754 operation and C
function must never raise FP exception, even if the argument is a
signaling NaN. Compare instructions usually does not have such
property, they raise 'invalid' exception in such case. So this
mechanism is unsuitable when exception behavior is strict. In
particular it could result in unexpected trapping if argument is SNaN.
* Another solution was implemented in https://reviews.llvm.org/D95948.
It is used in the cases when raising FP exceptions by 'isnan' is not
allowed. This solution implements 'isnan' using integer operations.
It solves the problem of exceptions, but offers one solution for all
targets, however some can do the check in more efficient way.
* Solution implemented by https://reviews.llvm.org/D96568 introduced a
hook 'clang::TargetCodeGenInfo::testFPKind', which injects target
specific code into IR. Now only SystemZ implements this hook and it
generates a call to target specific intrinsic function.
Although these mechanisms allow to implement 'isnan' with enough
efficiency, expanding 'isnan' in clang has drawbacks:
* The operation 'isnan' is hidden behind generic integer operations or
target-specific intrinsics. It complicates analysis and can prevent
some optimizations.
* IR can be created by tools other than clang, in this case treatment
of 'isnan' has to be duplicated in that tool.
Another issue with the current implementation of 'isnan' comes from the
use of options '-ffast-math' or '-fno-honor-nans'. If such option is
specified, 'fcmp uno' may be optimized to 'false'. It is valid
optimization in general, but it results in 'isnan' always returning
'false'. For example, in some libc++ implementations the following code
returns 'false':
std::isnan(std::numeric_limits<float>::quiet_NaN())
The options '-ffast-math' and '-fno-honor-nans' imply that FP operation
operands are never NaNs. This assumption however should not be applied
to the functions that check FP number properties, including 'isnan'. If
such function returns expected result instead of actually making
checks, it becomes useless in many cases. The option '-ffast-math' is
often used for performance critical code, as it can speed up execution
by the expense of manual treatment of corner cases. If 'isnan' returns
assumed result, a user cannot use it in the manual treatment of NaNs
and has to invent replacements, like making the check using integer
operations. There is a discussion in https://reviews.llvm.org/D18513#387418,
which also expresses the opinion, that limitations imposed by
'-ffast-math' should be applied only to 'math' functions but not to
'tests'.
To overcome these drawbacks, this change introduces a new IR intrinsic
function 'llvm.isnan', which realizes the check as specified by IEEE-754
and C standards in target-agnostic way. During IR transformations it
does not undergo undesirable optimizations. It reaches instruction
selection, where is lowered in target-dependent way. The lowering can
vary depending on options like '-ffast-math' or '-ffp-model' so the
resulting code satisfies requested semantics.
Differential Revision: https://reviews.llvm.org/D104854
GCC supports multiple forms of -falign-loops=.
-falign-loops= is currently ignored in Clang.
This patch implements the simplest but the most useful form where N is a
power of 2.
The underlying implementation uses a `llvm::TargetOptions` option for now.
Bitcode generation ignores this option.
Differential Revision: https://reviews.llvm.org/D106701
Implement target builtins for gfx90a including fadd64, fadd32, add2h,
max and min on various global, flat and ds address spaces for which
intrinsics are implemented.
Differential Revision: https://reviews.llvm.org/D106909
This matches the behavior of GCC.
Patch does not change remapping logic itself, so adding one simple smoke test should be enough.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D107393
For fixed SVE types, predicates are represented using vectors of i8,
where as for scalable types they are represented using vectors of i1. We
can avoid going through memory for casts between these by bitcasting the
i1 scalable vectors to/from a scalable i8 vector of matching size, which
can then use the existing vector insert/extract logic.
Differential Revision: https://reviews.llvm.org/D106860
Clang has builtin function '__builtin_isnan', which implements C
library function 'isnan'. This function now is implemented entirely in
clang codegen, which expands the function into set of IR operations.
There are three mechanisms by which the expansion can be made.
* The most common mechanism is using an unordered comparison made by
instruction 'fcmp uno'. This simple solution is target-independent
and works well in most cases. It however is not suitable if floating
point exceptions are tracked. Corresponding IEEE 754 operation and C
function must never raise FP exception, even if the argument is a
signaling NaN. Compare instructions usually does not have such
property, they raise 'invalid' exception in such case. So this
mechanism is unsuitable when exception behavior is strict. In
particular it could result in unexpected trapping if argument is SNaN.
* Another solution was implemented in https://reviews.llvm.org/D95948.
It is used in the cases when raising FP exceptions by 'isnan' is not
allowed. This solution implements 'isnan' using integer operations.
It solves the problem of exceptions, but offers one solution for all
targets, however some can do the check in more efficient way.
* Solution implemented by https://reviews.llvm.org/D96568 introduced a
hook 'clang::TargetCodeGenInfo::testFPKind', which injects target
specific code into IR. Now only SystemZ implements this hook and it
generates a call to target specific intrinsic function.
Although these mechanisms allow to implement 'isnan' with enough
efficiency, expanding 'isnan' in clang has drawbacks:
* The operation 'isnan' is hidden behind generic integer operations or
target-specific intrinsics. It complicates analysis and can prevent
some optimizations.
* IR can be created by tools other than clang, in this case treatment
of 'isnan' has to be duplicated in that tool.
Another issue with the current implementation of 'isnan' comes from the
use of options '-ffast-math' or '-fno-honor-nans'. If such option is
specified, 'fcmp uno' may be optimized to 'false'. It is valid
optimization in general, but it results in 'isnan' always returning
'false'. For example, in some libc++ implementations the following code
returns 'false':
std::isnan(std::numeric_limits<float>::quiet_NaN())
The options '-ffast-math' and '-fno-honor-nans' imply that FP operation
operands are never NaNs. This assumption however should not be applied
to the functions that check FP number properties, including 'isnan'. If
such function returns expected result instead of actually making
checks, it becomes useless in many cases. The option '-ffast-math' is
often used for performance critical code, as it can speed up execution
by the expense of manual treatment of corner cases. If 'isnan' returns
assumed result, a user cannot use it in the manual treatment of NaNs
and has to invent replacements, like making the check using integer
operations. There is a discussion in https://reviews.llvm.org/D18513#387418,
which also expresses the opinion, that limitations imposed by
'-ffast-math' should be applied only to 'math' functions but not to
'tests'.
To overcome these drawbacks, this change introduces a new IR intrinsic
function 'llvm.isnan', which realizes the check as specified by IEEE-754
and C standards in target-agnostic way. During IR transformations it
does not undergo undesirable optimizations. It reaches instruction
selection, where is lowered in target-dependent way. The lowering can
vary depending on options like '-ffast-math' or '-ffp-model' so the
resulting code satisfies requested semantics.
Differential Revision: https://reviews.llvm.org/D104854
In LLVM IR terms the ACLE type 'data512_t' is essentially an aggregate
type { [8 x i64] }. When emitting code for inline assembly operands,
clang tries to scalarize aggregate types to an integer of the equivalent
length, otherwise it passes them by-reference. This patch adds a target
hook to tell whether a given inline assembly operand is scalarizable
so that clang can emit code to pass/return it by-value.
Differential Revision: https://reviews.llvm.org/D94098
Pulled out the OptimizationLevel class from PassBuilder in order to be able to access it from within the PassManager and avoid include conflicts.
Reviewed By: mtrofin
Differential Revision: https://reviews.llvm.org/D107025
@kpn pointed out that the global variable initialization functions didn't
have the "strictfp" metadata set correctly, and @rjmccall said that there
was buggy code in SetFPModel and StartFunction, this patch is to solve
those problems. When Sema creates a FunctionDecl, it sets the
FunctionDeclBits.UsesFPIntrin to "true" if the lexical FP settings
(i.e. a combination of command line options and #pragma float_control
settings) correspond to ConstrainedFP mode. That bit is used when CodeGen
starts codegen for a llvm function, and it translates into the
"strictfp" function attribute. See bugs.llvm.org/show_bug.cgi?id=44571
Reviewed By: Aaron Ballman
Differential Revision: https://reviews.llvm.org/D102343
On ELF, an SHT_INIT_ARRAY outside a section group is a GC root. The current
codegen abuses SHT_INIT_ARRAY in a section group to mean a GC root.
On PE/COFF, the dynamic initialization for `__declspec(selectany)` in a comdat
can be garbage collected by `-opt:ref`.
Call `addUsedGlobal` for the two cases to fix the abuse/bug.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D106925
The device runtime contains several calls to __kmpc_get_hardware_num_threads_in_block
and __kmpc_get_hardware_num_blocks. If the thread_limit and the num_teams are constant,
these calls can be folded to the constant value.
In commit D106033 we have the optimization phase. This commit adds the attributes to
the outlined function for the grid size. the two attributes are `omp_target_num_teams` and
`omp_target_thread_limit`. These values are added as long as they are constant.
Two functions are created `getNumThreadsExprForTargetDirective` and
`getNumTeamsExprForTargetDirective`. The original functions `emitNumTeamsForTargetDirective`
and `emitNumThreadsForTargetDirective` identify the expresion and emit the code.
However, for the Device version of the outlined function, we cannot emit anything.
Therefore, this is a first attempt to separate emision of code from deduction of the
values.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D106298
Replace the clang builtins and LLVM intrinsics for the SIMD extmul instructions
with normal codegen patterns.
Differential Revision: https://reviews.llvm.org/D106724
Allegedly the DWARF backend ignores this field of DIEnumerator, but we
set it nonetheless in case we decide to use it in the future.
Alternatively, we could remove it, but it is simpler to pass down the
signed bit as it is in the AST for now.
Implemented to address comments on D106585
This patch adds the always inline attribute to the outlined functions generated
by OpenMP regions. Because there is only a single instance of this function and
it always has internal linkage it is safe to inline in every instance it is
created. This could potentially lead to performance degredation due to
inflated register counts in the parallel region.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D106799
DIEnumerator stores an APInt as of April 2020, so now we don't need to
truncate the enumerator value to 64 bits. Fixes assertions during IRGen.
Split from D105320, thanks to Matheus Izvekov for the test case and
report.
Differential Revision: https://reviews.llvm.org/D106585
XL provides functions __vec_ldrmb/__vec_strmb for loading/storing a
sequence of 1 to 16 bytes in big endian order, right justified in the
vector register (regardless of target endianness).
This is equivalent to vec_xl_len_r/vec_xst_len_r which are only
available on Power9.
This patch simply uses the Power9 functions when compiled for Power9,
but provides a more general implementation for Power8.
Differential revision: https://reviews.llvm.org/D106757
In OpenMP 5.1:
> If the `write` or `update` clause is specifieded, the atomic operation is not an atomic conditional update for which the comparison fails, and the effective memory ordering is `release`, `acq_rel`, or `seq_cst`, the strong flush on entry to the atomic operation is also a release flush. If the `read` or `update` clause is specified and the effective memory ordering is `acquire`, `acq_rel`, or `seq_cst` then the strong flush on exit from the atomic operation is also an acquire flush.
In OpenMP 5.0:
> If the `write`, `update`, or **`capture`** clause is specified and the `release`, `acq_rel`, or `seq_cst` clause is specified then the strong flush on entry to the atomic operation is also a release flush. If the `read` or `capture` clause is specified and the `acquire`, `acq_rel`, or `seq_cst` clause is specified then the strong flush on exit from the atomic operation is also an acquire flush.
From my understanding, in OpenMP 5.1, `capture` is removed from the requirement for flush, therefore we don't have to enforce it.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D100768
Replace the clang builtins and LLVM intrinsics for {f32x4,f64x2}.{pmin,pmax}
with standard codegen patterns. Since wasm_simd128.h uses an integer vector as
the standard single vector type, the IR for the pmin and pmax intrinsic
functions contains bitcasts that would not be there otherwise. Add extra codegen
patterns that can still select the pmin and pmax instructions in the presence of
these bitcasts.
Differential Revision: https://reviews.llvm.org/D106612
Address sanitizer passes may generate call of ASAN bitcode library
functions after bitcode linking in lld, therefore lld cannot add
those symbols since it does not know they will be used later.
To solve this issue, clang emits a reference to a bicode library
function which calls all ASAN functions which need to be
preserved. This basically force all ASAN functions to be
linked in.
Reviewed by: Artem Belevich
Differential Revision: https://reviews.llvm.org/D106315
This is part of a patch series working towards the ability to make
SourceLocation into a 64-bit type to handle larger translation units.
!srcloc is generated in clang codegen, and pulled back out by llvm
functions like AsmPrinter::emitInlineAsm that need to report errors in
the inline asm. From there it goes to LLVMContext::emitError, is
stored in DiagnosticInfoInlineAsm, and ends up back in clang, at
BackendConsumer::InlineAsmDiagHandler(), which reconstitutes a true
clang::SourceLocation from the integer cookie.
Throughout this code path, it's now 64-bit rather than 32, which means
that if SourceLocation is expanded to a 64-bit type, this error report
won't lose half of the data.
The compiler will tolerate both of i32 and i64 !srcloc metadata in
input IR without faulting. Test added in llvm/MC. (The semantic
accuracy of the metadata is another matter, but I don't know of any
situation where that matters: if you're reading an IR file written by
a previous run of clang, you don't have the SourceManager that can
relate those source locations back to the original source files.)
Original version of the patch by Mikhail Maltsev.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D105491
This patch changes `__kmpc_free_shared` to take an additional argument
corresponding to the associated allocation's size. This makes it easier to
implement the allocator in the runtime.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D106496
These builtins were added to capture the fact that the underlying Wasm
instructions return i32s and implicitly sign or zero extend the extracted lanes
in the case of the i8x16 and i16x8 variants. But we do sufficient optimizations
during code gen that these low-level details do not need to be exposed to users.
This commit replaces the use of the builtins in wasm_simd128.h with normal
target-independent vector code. As a result, we can switch the relevant
intrinsics to use functions rather than macros and can use more user-friendly
return types rather than trying to precisely expose the underlying Wasm types.
Note, however, that the generated LLVM IR is no different after this change.
Differential Revision: https://reviews.llvm.org/D106500
Replace the experimental clang builtins and LLVM intrinsics for these
instructions with normal instruction selection patterns. The wasm_simd128.h
intrinsics header was already using portable code for the corresponding
intrinsics, so now it produces the correct instructions.
Differential Revision: https://reviews.llvm.org/D106400
This patch is in a series of patches to provide
builtins for compatibility with the XL compiler.
This patch adds builtins related to floating point
operations
Reviewed By: #powerpc, nemanjai, amyk, NeHuang
Differential Revision: https://reviews.llvm.org/D103986
This is part of a patch series working towards the ability to make
SourceLocation into a 64-bit type to handle larger translation units.
NFC: this patch introduces typedefs for the integer type used by
SourceLocation and makes all the boring changes to use the typedefs
everywhere, but for the moment, they are unconditionally defined to
uint32_t.
Patch originally by Mikhail Maltsev.
Reviewed By: tmatheson
Differential Revision: https://reviews.llvm.org/D105492
Implemented builtins for mtmsr, mfspr, mtspr on PowerPC;
the patch is intended for XL Compatibility.
Differential revision: https://reviews.llvm.org/D106130
This patch implements store, load, move from and to registers related
builtins, as well as the builtin for stfiw. The patch aims to provide
feature parady with xlC on AIX.
Differential revision: https://reviews.llvm.org/D105946
This patch is in a series of patches to provide builtins for compatibility
with the XL compiler. This patch add the builtin and emit target independent
code for __cmpb.
Reviewed By: nemanjai, #powerpc
Differential revision: https://reviews.llvm.org/D105194
This patch fixes `__builtin_ppc_recipdivf`, `__builtin_ppc_recipdivd`,
`__builtin_ppc_rsqrtf`, and `__builtin_ppc_rsqrtd`. FastMathFlags are
set to fast immediately before emitting these builtins. Now the flags
are restored to their previous values after the builtins are emitted.
Reviewed By: nemanjai, #powerpc
Differential Revision: https://reviews.llvm.org/D105984
Added a number of different builtins that exist in the XL compiler. Most of
these builtins already exist in clang under a different name.
Reviewed By: nemanjai, #powerpc
Differential Revision: https://reviews.llvm.org/D104386
This patch is in a series of patches to provide builtins for
compatibility with the XL compiler. This patch adds software divide
builtins with no checking. These builtins are each emitted as a fast
fdiv.
Reviewed By: #powerpc, nemanjai
Differential Revision: https://reviews.llvm.org/D106150
Summary:
The AIX linker will produce errors on unresolved weak symbols. Change the
generated code to not check for the initialization function but just call
it and ensure that it always exists. Also, the AIX atexit routine has a
different name (and signature) so call it correctly. Update the lit tests
to test on AIX appropriately.
Author: Jamie Schmeiser <schmeise@ca.ibm.com>
Reviewed By: hubert.reinterpretcast (Hubert Tong)
Differential Revision: https://reviews.llvm.org/D104420
It's noteworthy that GCC has the same bug here, which is a bit
surprising. Both Clang and GCC's bug is only for function template
arguments that are themselves templates with default template arguments
(f1<t1<int[, missing_default_here]>>). Probably because function name
matching isn't generally necessary - whereas type matching is necessary
for DWARF consumers to associate declarations and definitions across
translation units, so the bug's been addressed there already - but
continued to exist for function templates since it's fairly benign
there.
I came across this while working on a change that could reconstitute
these pretty printed names based on the rest of the DWARF, reducing the
size of the DWARF by not having to encode all the template parameters in
the name string. That reconstitution code can't tell the difference
between a defaulted argument or not, so couldn't create the current
buggy-ish output.
Making the names more consistent between direct and indirect references,
and between function and class templates seems all to the good.
(I fixed the function template version of this a few years back in
9fdd09a4cc - clearly I should've looked
more closely and generalized the code better so it only had to be fixed
once - well, doing that here now)
Remove uses of to-be-deprecated API. In cases where the correct
element type was not immediately obvious to me, fall back to
explicit getPointerElementType().
Remove uses of to-be-deprecated API.
Unfortunately this one mostly just makes the use of
getPointerElementType() explicit, as the correct type to use
wasn't immediately available (deriving it from QualType is left
as an excercise to the reader).
Remove uses of to-be-deprecated API. I've fallen back to calling
getPointerElementType() in some cases where the correct type wasn't
immediately obvious to me.
Parallel regions are outlined as functions with capture variables explicitly generated as distinct parameters in the function's argument list. That complicates the fork_call interface in the OpenMP runtime: (1) the fork_call is variadic since there is a variable number of arguments to forward to the outlined function, (2) wrapping/unwrapping arguments happens in the OpenMP runtime, which is sub-optimal, has been a source of ABI bugs, and has a hardcoded limit (16) in the number of arguments, (3) forwarded arguments must cast to pointer types, which complicates debugging. This patch avoids those issues by aggregating captured arguments in a struct to pass to the fork_call.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D102107
This provides intrinsics for emitting instructions that set the FPSCR (`mtfsf/mtfsfi`).
The patch also conservatively marks the rounding mode as an implicit def for both since they both may set the rounding mode depending on the operands.
Reviewed By: #powerpc, qiucf
Differential Revision: https://reviews.llvm.org/D105957
This patch is in a series of patches to provide builtins for compatibility
with the XL compiler. This patch adds the builtins and instrisics for population
count, reversed load and store related operations.
Reviewed By: nemanjai, #powerpc
Differential revision: https://reviews.llvm.org/D106021
This patch implements the `__popcntb` XL compatibility builtin for 32bit in the frontend and backend. This patch also updates tests for `__popcntb` and other XL Compat sync related builtins.
Reviewed By: #powerpc, nemanjai, amyk
Differential Revision: https://reviews.llvm.org/D105360
This patch is in a series of patches to provide builtins for compatibility
with the XL compiler. This patch adds the builtins and emit target independent
code for rotate related operations.
Reviewed By: nemanjai, #powerpc
Differential revision: https://reviews.llvm.org/D104744
This patch make coroutine passes run by default in LLVM pipeline. Now
the clang and opt could handle IR inputs containing coroutine intrinsics
without special options.
It should be fine. On the one hand, the coroutine passes seems to be stable
since there are already many projects using coroutine feature.
On the other hand, the coroutine passes should do nothing for IR who doesn't
contain coroutine intrinsic.
Test Plan: check-llvm
Reviewed by: lxfind, aeubanks
Differential Revision: https://reviews.llvm.org/D105877
Replace the experimental clang builtins and LLVM intrinsics for these
instructions with normal codegen patterns. Resolves PR50435.
Differential Revision: https://reviews.llvm.org/D106019
Replace the experimental clang builtin and LLVM intrinsics for these
instructions with normal codegen patterns. Resolves PR50433.
Differential Revision: https://reviews.llvm.org/D105950
LDARX and LWARX sometimes gets optimized out by the compiler
when it is critical to the correctness of the code. This inline asm generation
ensures that it preserved.
Differential Revision: https://reviews.llvm.org/D105754
Properties that were declared `@property(copy, nonatomic) id foo` make an
unnecessary call to objc_get_property(). This call can be replaced with a
direct access to the backing variable identical to how a `@property(nonatomic)
id foo` would do it.
This reduces codegen by 4 bytes (x86_64/arm64) and removes a cross linkage unit
function call per property declared as copy/nonatomic.
Differential Revision: https://reviews.llvm.org/D105311
Replace the clang builtin function and LLVM intrinsic for
f32x4.demote_zero_f64x2 with combines from normal SDNodes. Also add missing
combines for i32x4.trunc_sat_zero_f64x2_{s,u}, which share the same pattern.
Differential Revision: https://reviews.llvm.org/D105755
In the spirit of TRegions [0], this patch provides a simpler and uniform
interface for a kernel to set up the device runtime. The OMPIRBuilder is
used for reuse in Flang. A custom state machine will be generated in the
follow up patch.
The "surplus" threads of the "master warp" will not exit early anymore
so we need to use non-aligned barriers. The new runtime will not have an
extra warp but also require these non-aligned barriers.
[0] https://link.springer.com/chapter/10.1007/978-3-030-28596-8_11
This was in parts extracted from D59319.
Reviewed By: ABataev, JonChesterfield
Differential Revision: https://reviews.llvm.org/D101976
Broke check-clang, see https://reviews.llvm.org/D102307#2869065
Ran `git revert -n ebbe149a6f08535ede848a531a601ae6591cfbc5..269416d41908bb670f67af689155d5ab8eea689a`
In the spirit of TRegions [0], this patch provides a simpler and uniform
interface for a kernel to set up the device runtime. The OMPIRBuilder is
used for reuse in Flang. A custom state machine will be generated in the
follow up patch.
The "surplus" threads of the "master warp" will not exit early anymore
so we need to use non-aligned barriers. The new runtime will not have an
extra warp but also require these non-aligned barriers.
[0] https://link.springer.com/chapter/10.1007/978-3-030-28596-8_11
This was in parts extracted from D59319.
Reviewed By: ABataev, JonChesterfield
Differential Revision: https://reviews.llvm.org/D101976
Replace the clang builtin function and LLVM intrinsic previously used to select
the f64x2.promote_low_f32x4 instruction with custom combines from standard
SelectionDAG nodes. Implement the new combines to share code with the similar
combines for f64x2.convert_low_i32x4_{s,u}. Resolves PR50232.
Differential Revision: https://reviews.llvm.org/D105675
If the base is used in a map clause and later we have a memberexpr with
this base, and the member is a pointer, and this pointer is dereferenced
anyhow (subscript, array section, dereference, etc.), such components
should be considered as overlapped, otherwise it may lead to incorrect
size computations, since we try to map a pointee as a part of the whole
struct, which is not true for the pointer members.
Differential Revision: https://reviews.llvm.org/D105562
This change is intended as initial setup. The plan is to add
more semantic checks later. I plan to update the documentation
as more semantic checks are added (instead of documenting the
details up front). Most of the code closely mirrors that for
the Swift calling convention. Three places are marked as
[FIXME: swiftasynccc]; those will be addressed once the
corresponding convention is introduced in LLVM.
Reviewed By: rjmccall
Differential Revision: https://reviews.llvm.org/D95561
C++23 will make these conversions ambiguous - so fix them to make the
codebase forward-compatible with C++23 (& a follow-up change I've made
will make this ambiguous/invalid even in <C++23 so we don't regress
this & it generally improves the code anyway)
No need to emit private copyfor firstprivate constants in target
regions, we can use the original copy instead.
Differential Revision: https://reviews.llvm.org/D105647
In preparation for dropping support for it. I've replaced it with
a proper type where the correct type was obvious and left an
explicit getPointerElementType() where it wasn't.
This fixes a gap in the `overloadable` attribute support (K&R declared
functions would get mangled symbol names, but that name wouldn't be
represented in the debug info linkage name field for the function) and
in -funique-internal-linkage-names (this came up in review discussion on
D98799) where K&R static declarations would not get the uniqued linkage
names.
Same as other CreateLoad-style APIs, these need an explicit type
argument to support opaque pointers.
Differential Revision: https://reviews.llvm.org/D105395
This patch adds a new clang builtin, __arithmetic_fence. The purpose of the
builtin is to provide the user fine control, at the expression level, over
floating point optimization when -ffast-math (-ffp-model=fast) is enabled.
The builtin prevents the optimizer from rearranging floating point expression
evaluation. The new option fprotect-parens has the same effect on
parenthesized expressions, forcing the optimizer to respect the parentheses.
Reviewed By: aaron.ballman, kpn
Differential Revision: https://reviews.llvm.org/D100118
functions implicitly generated by the compiler
These fake functions would cause clang to crash if the changes proposed
in https://reviews.llvm.org/D98799 were made.
the call already has the operand bundle
This bug was causing the call to `replaceAllUsesWith` to crash because
the old call instruction and the new call instruction were the same.
rdar://74957948
Differential Revision: https://reviews.llvm.org/D97824
Fix suggested by Yuanfang Chen:
Non-distinct debuginfo is attached to the function due to the undecorated declaration. Later, when seeing the function definition and `nodebug` attribute, the non-distinct debuginfo should be cleared.
Reviewed By: dblaikie
Differential Revision: https://reviews.llvm.org/D104777
According to AVR ABI (https://gcc.gnu.org/wiki/avr-gcc), returned struct value
within size 1-8 bytes should be returned directly (via register r18-r25), while
larger ones should be returned via an implicit struct pointer argument.
Reviewed By: dylanmckay
Differential Revision: https://reviews.llvm.org/D99237
C++ constructors/destructors need to go through a different constructor to construct a GlobalDecl object in order to retrieve their linkage type. This causes an assert failure in the default constructor of GlobalDecl. I'm chaning it to using the exsiting GlobalDecl object.
Reviewed By: dblaikie
Differential Revision: https://reviews.llvm.org/D102356
This patch adds a new clang builtin, __arithmetic_fence. The purpose of the
builtin is to provide the user fine control, at the expression level, over
floating point optimization when -ffast-math (-ffp-model=fast) is enabled.
The builtin prevents the optimizer from rearranging floating point expression
evaluation. The new option fprotect-parens has the same effect on
parenthesized expressions, forcing the optimizer to respect the parentheses.
Reviewed By: aaron.ballman, kpn
Differential Revision: https://reviews.llvm.org/D100118
We need to mask the immediate to the width of a single vector
rather than 2 vectors. If we use the width of 2 vectors then
any shift larger than the length of 1 vector is going to overflow
the shuffle indices.
Fixes PR50895.
This patch adds a module level metadata flag indicating that the module
was compiled with the `-fopenmp` flag. This will make it easier for
passes like OpenMPOpt to determine if it should be run.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D102361
Note regarding C++ for OpenCL:
When compiling C++ for OpenCL, DW_LANG_C_plus_plus* is emitted.
There is no DWARF language code defined for C++ for OpenCL as of yet,
but DWARF issue 210514.1 has been raised to request one.
In the mean time, continuing to emit DW_LANG_C_plus_plus* for C++ for
OpenCL allows the potential to distinguish between C++ for OpenCL and
OpenCL C in !DICompileUnit nodes, whereas using DW_LANG_OpenCL for
C++ for OpenCL would prevent this.
This change therefore leaves C++ for OpenCL as-is.
Reviewed By: shchenz, Anastasia
Differential Revision: https://reviews.llvm.org/D104118
This is mostly a mechanical change, but a testcase that contains
parts of the StringRef class (clang/test/Analysis/llvm-conventions.cpp)
isn't touched.
copy/dispose helper functions
We found out that these fake functions would cause clang to crash if the
changes proposed in https://reviews.llvm.org/D98799 were made.
The original patch was reverted in f681fd927e
because debug locations were missing in the body of the block byref
helper functions. This patch fixes the bug by calling CreateArtificial
after the calls to StartFunction.
Differential Revision: https://reviews.llvm.org/D104082
Summary:
Memory globalization is required to maintain OpenMP standard semantics for data sharing between
worker and master threads. The GPU cannot share data between its threads so must allocate global or
shared memory to store the data in. Currently this is implemented fully in the frontend using the
`__kmpc_data_sharing_push_stack` and __kmpc_data_sharing_pop_stack` functions to emulate standard
CPU stack sharing. The front-end scans the target region for variables that escape the region and
must be shared between the threads. Each variable then has a field created for it in a global record
type.
This patch replaces this functinality with a single allocation command, effectively mimicing an
alloca instruction for the variables that must be shared between the threads. This will be much
slower than the current solution, but makes it much easier to optimize as we can analyze each
variable independently and determine if it is not captured. In the future, we can replace these
calls with an `alloca` and small allocations can be pushed to shared memory.
Reviewed By: tianshilei1992
Differential Revision: https://reviews.llvm.org/D97680
The codegen for simd constructs was affected by the presence (or
absence) of the 'monotonic' schedule modifier for worksharing
loops. The modifier is only intended to apply to the scheduling of
chunks for a thread, not iterations of a loop inside a chunk.
In addition, the monotonic modifier was applied to worksharing loops
by default if no schedule clause was present; the referenced part of
the OpenMP 4.5 spec in the code (section 2.7.1) only applies if the
user specified a schedule clause with a static kind but no modifier.
Without a user-specified schedule clause we should default to
nonmonotonic scheduling.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D103793
This can be seen as a follow up to commit 0ee439b705,
that changed the second argument of __powidf2, __powisf2 and
__powitf2 in compiler-rt from si_int to int. That was to align with
how those runtimes are defined in libgcc.
One thing that seem to have been missing in that patch was to make
sure that the rest of LLVM also handle that the argument now depends
on the size of int (not using the si_int machine mode for 32-bit).
When using __builtin_powi for a target with 16-bit int clang crashed.
And when emitting libcalls to those rtlib functions, typically when
lowering @llvm.powi), the backend would always prepare the exponent
argument as an i32 which caused miscompiles when the rtlib was
compiled with 16-bit int.
The solution used here is to use an overloaded type for the second
argument in @llvm.powi. This way clang can use the "correct" type
when lowering __builtin_powi, and then later when emitting the libcall
it is assumed that the type used in @llvm.powi matches the rtlib
function.
One thing that needed some extra attention was that when vectorizing
calls several passes did not support that several arguments could
be overloaded in the intrinsics. This patch allows overload of a
scalar operand by adding hasVectorInstrinsicOverloadedScalarOpd, with
an entry for powi.
Differential Revision: https://reviews.llvm.org/D99439
Also:
- add driver test (fsanitize-use-after-return.c)
- add basic IR test (asan-use-after-return.cpp)
- (NFC) cleaned up logic for generating table of __asan_stack_malloc
depending on flag.
for issue: https://github.com/google/sanitizers/issues/1394
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D104076
<string> is currently the highest impact header in a clang+llvm build:
https://commondatastorage.googleapis.com/chromium-browser-clang/llvm-include-analysis.html
One of the most common places this is being included is the APInt.h header, which needs it for an old toString() implementation that returns std::string - an inefficient method compared to the SmallString versions that it actually wraps.
This patch replaces these APInt/APSInt methods with a pair of llvm::toString() helpers inside StringExtras.h, adjusts users accordingly and removes the <string> from APInt.h - I was hoping that more of these users could be converted to use the SmallString methods, but it appears that most end up creating a std::string anyhow. I avoided trying to use the raw_ostream << operators as well as I didn't want to lose having the integer radix explicit in the code.
Differential Revision: https://reviews.llvm.org/D103888
-Wframe-larger-than= is an interesting warning; we can't know the frame
size until PrologueEpilogueInsertion (PEI); very late in the compilation
pipeline.
-Wframe-larger-than= was propagated through CC1 as an -mllvm flag, then
was a cl::opt in LLVM's PEI pass; this meant it was dropped during LTO
and needed to be re-specified via -plugin-opt.
Instead, make it part of the IR proper as a module level attribute,
similar to D103048. Introduce -fwarn-stack-size CC1 option.
Reviewed By: rsmith, qcolombet
Differential Revision: https://reviews.llvm.org/D103928
Implementation of the unroll directive introduced in OpenMP 5.1. Follows the approach from D76342 for the tile directive (i.e. AST-based, not using the OpenMPIRBuilder). Tries to use `llvm.loop.unroll.*` metadata where possible, but has to fall back to an AST representation of the outer loop if the partially unrolled generated loop is associated with another directive (because it needs to compute the number of iterations).
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D99459
Clang will create a global value put in constant memory if an aggregate value
is declared firstprivate in the target device. The symbol name only uses the
name of the firstprivate variable, so symbol name conflicts will occur if the
variable is allowed to have different types through templates. An example of
this behvaiour is shown in https://godbolt.org/z/EsMjYh47n. This patch adds the
mangled type name to the symbol to avoid such naming conflicts. This fixes
https://bugs.llvm.org/show_bug.cgi?id=50642.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D103995
This renames the expression value categories from rvalue to prvalue,
keeping nomenclature consistent with C++11 onwards.
C++ has the most complicated taxonomy here, and every other language
only uses a subset of it, so it's less confusing to use the C++ names
consistently, and mentally remap to the C names when working on that
context (prvalue -> rvalue, no xvalues, etc).
Renames:
* VK_RValue -> VK_PRValue
* Expr::isRValue -> Expr::isPRValue
* SK_QualificationConversionRValue -> SK_QualificationConversionPRValue
* JSON AST Dumper Expression nodes value category: "rvalue" -> "prvalue"
Signed-off-by: Matheus Izvekov <mizvekov@gmail.com>
Reviewed By: rsmith
Differential Revision: https://reviews.llvm.org/D103720
This implements the 'using enum maybe-qualified-enum-tag ;' part of
1099. It introduces a new 'UsingEnumDecl', subclassed from
'BaseUsingDecl'. Much of the diff is the boilerplate needed to get the
new class set up.
There is one case where we accept ill-formed, but I believe this is
merely an extended case of an existing bug, so consider it
orthogonal. AFAICT in class-scope the c++20 rule is that no 2 using
decls can bring in the same target decl ([namespace.udecl]/8). But we
already accept:
struct A { enum { a }; };
struct B : A { using A::a; };
struct C : B { using A::a;
using B::a; }; // same enumerator
this patch permits mixtures of 'using enum Bob;' and 'using Bob::member;' in the same way.
Differential Revision: https://reviews.llvm.org/D102241
vtbl itself is in default global address space. When clang emits
ctor, it gets a pointer to the vtbl field based on the this pointer,
then stores vtbl to the pointer.
Since this pointer can point to any address space (e.g. an object
created in stack), this pointer points to default address space, therefore
the pointer to vtbl field in this object should also be in default
address space.
Currently, clang incorrectly casts the pointer to vtbl field in this object
to global address space. This caused assertions in backend.
This patch fixes that by removing the incorrect addr space cast.
Reviewed by: Artem Belevich
Differential Revision: https://reviews.llvm.org/D103835
This fixes inconsistencies in the ms_abi.c testcase.
Also add a couple cases of missing double pointers in the windows part
of the testcase; the outcome of building that testcase on windows hasn't
changed, but the previous form of the test was imprecise (checking
for "%[[STRUCT_FOO]]*" when clang actually generates "%[[STRUCT_FOO]]**"),
which still used to match.
Ideally this would share code with the native Windows case, but
X86_64ABIInfo and WinX86_64ABIInfo aren't superclasses/subclasses of
each other so it's impractical, and the code to share currently only
consists of a couple lines.
Differential Revision: https://reviews.llvm.org/D103837
On x86_64 mingw, long doubles are always passed indirectly as
arguments (see an existing case in WinX86_64ABIInfo::classify);
generalize the existing code for reading varargs - any non-aggregate
type that is larger than 64 bits (which would be both long double
in mingw, and __int128) are passed indirectly too.
This makes reading varargs consistent with how they're passed,
fixing interop with both gcc and clang callers, for long double
and __int128.
Differential Revision: https://reviews.llvm.org/D103452
Refactored code of dependence processing and added new inoutset dependence type.
Compiler can set dependence flag to 0x8 when call __kmpc_omp_task_with_deps.
Size of type of the dependence flag changed from 1 to 4 bytes in clang.
All dependence flags library gets so far and corresponding dependence types:
1 - IN, 2 - OUT, 3 - INOUT, 4 - MUTEXINOUTSET, 8 - INOUTSET.
Differential Revision: https://reviews.llvm.org/D97085
If the memory object is scalable type, we do not know the exact size of
it at compile time. Set the size of lifetime marker to unknown if the
object is scalable one.
Differential Revision: https://reviews.llvm.org/D102822
This is a pre-patch for adding using-enum support. It breaks out
the shadow decl handling of UsingDecl to a new intermediate base
class, BaseUsingDecl, altering the decl hierarchy to
def BaseUsing : DeclNode<Named, "", 1>;
def Using : DeclNode<BaseUsing>;
def UsingPack : DeclNode<Named>;
def UsingShadow : DeclNode<Named>;
def ConstructorUsingShadow : DeclNode<UsingShadow>;
Differential Revision: https://reviews.llvm.org/D101777
Use llvm.experimental.vector.insert instead of storing into an alloca
when generating code for these intrinsics. This defers the codegen of
the generated vector to instruction selection, allowing existing
shufflevector style optimizations to apply.
Additionally, introduce a new target transform that can recognise fixed
predicate patterns in the svbool variants of these intrinsics.
Differential Revision: https://reviews.llvm.org/D103082
This fixes PR49198: Wrong usage of __dso_handle in user code leads to
a compiler crash.
When Init is an address of the global itself, we need to track it
across RAUW. Otherwise the initializer can be destroyed if the global
is replaced.
Differential Revision: https://reviews.llvm.org/D101156
This patch fixes a Windows -EHa crash induced by previous commit 797ad70152.
The crash was caused by "LifetimeMarker" scope (with option -O2) that should not be considered as SEH Scope.
This change also turns off -fasync-exceptions by default under -EHa option for now.
Differential Revision: https://reviews.llvm.org/D103664#2799944
Need to emit a call for __kmpc_cancel_barrier in the exit block for
__kmpc_cancel function call if cancellation of the parallel block is
requested.
Differential Revision: https://reviews.llvm.org/D103646
Michael Kruse