This is an artefact of split-mode CUDA compilation that we need to
mimic. HD functions are sometimes allowed to call H or D functions. Due
to split compilation mode device-side compilation will not see host-only
function and thus they will not be considered at all. For clang both H
and D variants will become function overloads visible to
compiler. Normally target attribute is considered only if C++ rules can
not determine which function is better. However in this case we need to
ignore functions that would not be present during current compilation
phase before we apply normal overload resolution rules.
Changes:
* introduced another level of call preference to better describe
possible call combinations.
* removed WrongSide functions from consideration if the set contains
SameSide function.
* disabled H->D, D->H and G->H calls. These combinations are
not allowed by CUDA and we were reluctantly allowing them to work
around device-side calls to math functions in std namespace.
We no longer need it after r258880.
Differential Revision: http://reviews.llvm.org/D16870
llvm-svn: 260697
Rewrite the code to handle all pseudo-instructions in a single pass.
This temporarily reverts spill slot optimization that used general-
purpose registers to hold values of spilled predicate registers.
llvm-svn: 260696
For some cases, InstCombine replaces the sequence of xor/sub instruction
followed by cmp instruction into a single cmp instruction.
However, this replacement may result suboptimal result especially when
the xor/sub has more than one use, as discussed in
bug 26465 (https://llvm.org/bugs/show_bug.cgi?id=26465).
This patch make the replacement happen only when xor/sub has only one
use.
Differential Revision: http://reviews.llvm.org/D16915
Patch by Taewook Oh!
llvm-svn: 260695
Historically, AMD internal sp3 assembler has flat_store* addr, data
format. To match existing code and to enable reuse, change LLVM
definitions to match. Also update MC and CodeGen tests.
Differential Revision: http://reviews.llvm.org/D16927
Patch by: Nikolay Haustov
llvm-svn: 260694
Summary:
It is possible that the loop condition can be a boolean constant (infinite loop,
for example). So we sould handle constant condition in annotating a loop. This
patch adds this functionality to support annotating constant condition.
Reviewers: tstellarAMD, arsenm
Subscribers: llvm-commits, arsenm
Differential Revision: http://reviews.llvm.org/D15093
llvm-svn: 260692
We can generate the actual instructions from the intrinsics without the
need for pseudo-instructions. Also, since the intrinsics have a side-
effect in a form of a store, attempt to optimize away loads from the
store location.
llvm-svn: 260690
Summary:
Before this change, callee-save registers would be rounded up to even
pairs of GPRs and FPRs. This change eliminates these extra padding
load/stores, though it does keep the stack allocation the same size
unless both the GPR and FPR sets have an odd size, in which case one
full pair stack slot (16 bytes) is saved.
This optimization cannot currently be done for MachO targets since they
rely on a fast-path .debug_frame equivalent that can only encode
callee-save registers as pairs.
Reviewers: t.p.northover, rengolin, mcrosier, jmolloy
Subscribers: aemerson, rengolin, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D17000
llvm-svn: 260689
Expand the simplify boolean expression check to handle implicit conversion of integral types to bool and improve the handling of implicit conversion of member pointers to bool.
Implicit conversion of member pointers are replaced with explicit comparisons to nullptr.
Implicit conversions of integral types are replaced with explicit comparisons to 0.
Patch by Richard Thomson.
llvm-svn: 260681
The DataLayout can calculate alignment of vectors based on the alignment
of the element type and the number of elements. In fact, it is the product
of these two values. The problem is that for vectors of N x i1, this will
return the alignment of N bytes, since the alignment of i1 is 8 bits. The
vector types of vNi1 should be aligned to N bits instead. Provide explicit
alignment for HVX vectors to avoid such complications.
llvm-svn: 260680
The DataLayout can calculate alignment of vectors based on the alignment
of the element type and the number of elements. In fact, it is the product
of these two values. The problem is that for vectors of N x i1, this will
return the alignment of N bytes, since the alignment of i1 is 8 bits. The
vector types of vNi1 should be aligned to N bits instead. Provide explicit
alignment for HVX vectors to avoid such complications.
llvm-svn: 260678
This patch adds some TLS relocations and relaxations for AArch64.
Some Global-Dynamic relocation are handled by optimizing them to
Local-Exec (Initial-Exec is not yet supported). They are:
- R_AARCH64_TLSDESC_ADR_PAGE21
- R_AARCH64_TLSDESC_LD64_LO12_NC
- R_AARCH64_TLSDESC_ADD_LO12_NC
- R_AARCH64_TLSDESC_CALL
Also some Init-Exec is optimized to Local-Exec if possible. They are:
- R_AARCH64_TLSIE_ADR_GOTTPREL_PAGE21
- R_AARCH64_TLSIE_LD64_GOTTPREL_LO12_NC
Finally some Local-Exec relocation are handled in relocateOne:
- R_AARCH64_TLSLE_ADD_TPREL_HI12
- R_AARCH64_TLSLE_ADD_TPREL_LO12_NC
This work is mainly for compiler bootstrap, where static binaries is
showing good progress (although shared object still lacking support
from both TLS aarch64 backend and some other issues).
llvm-svn: 260677
Fix a crash while parsing this code:
struct X {
friend constexpr int foo(X*) { return 12; }
static constexpr int j = foo(static_cast<X*>(nullptr));
};
Differential Revision: http://reviews.llvm.org/D16973
llvm-svn: 260675
For pragma omp taskwait the runtime is called from the task context.
Therefore, the reentry frame information should be updated.
The information should be available for both taskwait event calls; therefore,
set before the first event and reset after the last event.
Patch by Joachim Protze
Differential Revision: http://reviews.llvm.org/D17145
llvm-svn: 260674
There's no obvious reason it should fail in this way but it's the only change
on the blamelist. I suspect stale lit*.cfg's from previous builds.
llvm-svn: 260672
The library functions defined in the C99 standard headers
are not available (OpenCL v1.2 s6.9.f).
This change stops treating OpenCL builtin functions as standard C lib
functions to eliminate warning messages about printf format string.
Patch by Liu Yaxun (Sam)!
Differential Revision: http://reviews.llvm.org/D16812
llvm-svn: 260671
Summary:
LLVMGetTargetMachineData has been removed, and LLVMGetDataLayout is
suggested to use. The LLVMGetDataLayout is exposed in go bindings.
So it's safe to remove the function.
Reviewers: bkramer
Subscribers: llvm-commits, axw
Differential Revision: http://reviews.llvm.org/D17193
llvm-svn: 260670
The lit test-suite containing the unit tests needs to be explicitly specified
as an argument to lit.py since it is no longer discovered when the other tests
are run (because they are one directory deeper).
dfsan, lsan, and sanitizer_common don't show the same problem.
llvm-svn: 260669
node set rather than walking the SCC directly.
This directly exposes the functions and has already had null entries
filtered out. We also don't need need to handle optnone as it has
already been handled in the caller -- we never try to remove convergent
when there are optnone functions in the SCC.
With this change, the code for removing convergent should work with the
new pass manager and a different SCC analysis.
llvm-svn: 260668
I added portability warnings when int results are casted to long. I forgot to handle uint, ulong and ulonglong.
Tested on x86 and powerpc targets, hope it works now on all buildbots.
llvm-svn: 260667
with the test for a non-convergent intrinsic call.
While it is possible to use the call records to search for function
calls, we're going to do an instruction scan anyways to find the
intrinsics, we can handle both cases while scanning instructions. This
will also make the logic more amenable to the new pass manager which
doesn't use the same call graph structure.
My next patch will remove use of CallGraphNode entirely and allow this
code to work with both the old and new pass manager. Fortunately, it
should also get strictly simpler without changing functionality.
llvm-svn: 260666
However, they also contain fallback logic that - in cases where LLDB can't recognize the specific subclass - actually does run code in order to inspect those objects.
The argument for this logic was that these data types are critical enough that the risk of getting it wrong is outweighed by the advantage of always providing accurate child information.
Practical experience however shows that "po" - a code running data-inspection command - is quite frequently used, and not considered burdensome by users.
As such, this makes the code-running fallback in the data formatters a risk that carries very little actual reward. Also, unlike the time this code was originally written, we now have accurate class information for Objective-C, and thus we are less likely to improperly identify classes.
This commit removes support for the code-running fallback, and aligns the data formatters for NSArray, NSDictionary and NSSet to the general no-code-running behavior of other data formatters.
While it is possible for us to add support for some subclasses that are now no longer covered by static inspection alone, this is beyond the scope of this commit.
llvm-svn: 260664
With this option one can optionally override the architecture dependent
default library to use if no -stdlib= is provided on compiler invocation.
Differential Revision: http://reviews.llvm.org/D15920
llvm-svn: 260662
This was hardcoded to the static private size, but this
would be missing the offset and additional size for someday
when we have dynamic sizing.
Also stops always initializing flat_scratch even when unused.
In the future we should stop emitting this unless flat instructions
are used to access private memory. For example this will initialize
it almost always on VI because flat is used for global access.
llvm-svn: 260658
Introduce a subtarget feature for this, and leave the default with
the current behavior which assumes up to 16-byte loads/stores can
be used. The field also seems to have the ability to be set to 2 bytes,
but I'm not sure what that would be used for.
llvm-svn: 260651
Mehdi Amini