Refactored the function `target` to make preparation for fixing the
issue of ahead-of-time device memory deallocation.
Reviewed By: ye-luo
Differential Revision: https://reviews.llvm.org/D84816
Fix for the issue raised in https://github.com/rust-lang/rust/issues/74632.
The current heuristic for inserting LFENCEs uses a quadratic-time algorithm. This can apparently cause substantial compilation slowdowns for building Rust projects, where functions > 5000 LoC are apparently common.
The updated heuristic in this patch implements a linear-time algorithm. On a set of benchmarks, the slowdown factor for the generated code was comparable (2.55x geo mean for the quadratic-time heuristic, vs. 2.58x for the linear-time heuristic). Both heuristics offer the same security properties, namely, mitigating LVI.
This patch also includes some formatting fixes.
Differential Revision: https://reviews.llvm.org/D84471
After the recent change to the tuning settings for pentium4 to improve our default 32-bit behavior, I've decided to see about implementing -mtune support. This way we could have a default architecture CPU of "pentium4" or "x86-64" and a default tuning cpu of "generic". And we could change our "pentium4" tuning settings back to what they were before.
As a step to supporting this, this patch separates all of the features lists for the CPUs into 2 lists. I'm using the Proc class and a new ProcModel class to concat the 2 lists before passing to the target independent ProcessorModel. Future work to truly support mtune would change ProcessorModel to take 2 lists separately. I've diffed the X86GenSubtargetInfo.inc file before and after this patch to ensure that the final feature list for the CPUs isn't changed.
Differential Revision: https://reviews.llvm.org/D84879
This patch addes time trace functionality to have a better understanding
of the analysis times.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D84980
This includes basic support for computeKnownBits on abs. I've left FIXMEs for more complicated things we could do.
Differential Revision: https://reviews.llvm.org/D84963
InstrProfilingBuffer.c.o is generic code that must support compilation
into freestanding projects. This gets rid of its dependence on the
_getpagesize symbol from libc, shifting it to InstrProfilingFile.c.o.
This fixes a build failure seen in a firmware project.
rdar://66249701
Just the obvious implementation that rewrites the result type. Also fix
warning from EXTRACT_SUBVECTOR legalization that triggers on the test.
Differential Revision: https://reviews.llvm.org/D84706
This tool will be used to generate C wrappers for the C++ LLVM libc
implementations. This change does not hook this tool up to anything yet.
However, it can be useful for cases where one does not want to run the
objcopy step (to insert the C symbol in the object file) but can make use
of LTO to eliminate the cost of the additional wrapper call. This can be
relevant for certain downstream platforms. If this tool can benefit other
libc platforms in general, then it can be integrated into the build system
with options to use or not use the wrappers. An example of such a
platform is CUDA.
Reviewed By: abrachet
Differential Revision: https://reviews.llvm.org/D84848
fptosi/fptoui have similar, but not identical, semantics. In
particular, the behavior on overflow is different.
Fixes https://bugs.llvm.org/show_bug.cgi?id=46844 for 64-bit. (The
corresponding patch for 32-bit is more involved because the equivalent
intrinsics don't exist, as far as I can tell.)
Differential Revision: https://reviews.llvm.org/D84703
The -harness option enables new testing use-cases for llvm-jitlink. It takes a
list of objects to treat as a test harness for any regular objects passed to
llvm-jitlink.
If any files are passed using the -harness option then the following
transformations are applied to all other files:
(1) Symbols definitions that are referenced by the harness files are promoted
to default scope. (This enables access to statics from test harness).
(2) Symbols definitions that clash with definitions in the harness files are
deleted. (This enables interposition by test harness).
(3) All other definitions in regular files are demoted to local scope.
(This causes untested code to be dead stripped, reducing memory cost and
eliminating spurious unresolved symbol errors from untested code).
These transformations allow the harness files to reference and interpose
symbols in the regular object files, which can be used to support execution
tests (including fuzz tests) of functions in relocatable objects produced by a
build.
This allows clients to detect invalid transformations applied by JITLink passes
(e.g. inserting or removing symbols in unexpected ways) and terminate linking
with an error.
This change is used to simplify the error propagation logic in
ObjectLinkingLayer.
Avoid recursively calling copyPhysReg for AGPR handling. This was
dropping the necessary super register implicit defs to avoid liveness
verifier errors.
We previously used a non-aggregate RValue to represent the passed value,
which violated the assumptions of call arg lowering in some cases, in
particular on 32-bit Windows, where we'd end up producing an FCA store
with TBAA metadata, that the IR verifier would reject.
codesign (or more specifically libstuff) checks that each section in
__LINKEDIT ends where the next one starts -- no gaps are permitted. This
diff achieves it by aligning every section's start and end points to
WordSize.
Remarks: ld64 appears to satisfy the constraint by adding padding bytes
when generating the __LINKEDIT data, e.g. by emitting BIND_OPCODE_DONE
(which is a 0x0 byte) repeatedly. I think the approach this diff takes
is a bit more elegant, but I'm not sure if it's too restrictive. In
particular, it assumes padding always uses the zero byte. But we can
revisit this later.
Reviewed By: #lld-macho, compnerd
Differential Revision: https://reviews.llvm.org/D84718
Tools like `install_name_tool` and `codesign` may modify the Mach-O
header and increase its size. The linker has to provide padding to make this
possible. This diff does that, plus sets its default value to 32 bytes (which
is what ld64 does).
Unlike ld64, however, we lay out our sections *exactly* `-headerpad` bytes from
the header, whereas ld64 just treats the padding requirement as a lower bound.
ld64 actually starts laying out the non-header sections in the __TEXT segment
from the end of the (page-aligned) segment rather than the front, so its
binaries typically have more than `-headerpad` bytes of actual padding.
We should consider implementing the same alignment behavior.
Reviewed By: #lld-macho, compnerd
Differential Revision: https://reviews.llvm.org/D84714
The C++ ABI requires dylibs to pass a pointer to __cxa_atexit which does
e.g. cleanup of static global variables. The C++ spec says that the pointer
can point to any address in one of the dylib's segments, but in practice
ld64 seems to set it to point to the header, so that's what's implemented
here.
Reviewed By: #lld-macho, smeenai
Differential Revision: https://reviews.llvm.org/D83603
Pass the abs poison flag to the underlying ConstantRange
implementation, allowing CVP to simplify based on it.
Importantly, this recognizes that abs with poison flag is actually
non-negative...
This fixes an assertion failure that was being triggered in
SelectionDAG::getZeroExtendInReg(), where it was trying to extend the <2xi32>
to i64 (which should have been <2xi64>).
Fixes: rdar://66016901
Differential Revision: https://reviews.llvm.org/D84884
When declaring the same variable twice with an initialization, we were failing
an internal check. I fixed this by checking to see if the associated symbol
already had an error.
I added tests for pointer and non-pointer initialization of duplicate names.
Differential Revision: https://reviews.llvm.org/D84969
Clang and GCC have a feature (-MD flag) to create a dependency file
in a format that build systems such as Make or Ninja can read, which
specifies all the additional inputs such .h files.
This change introduces the same functionality to lld bringing it to
feature parity with ld and gold which gained this feature recently.
See https://sourceware.org/bugzilla/show_bug.cgi?id=22843 for more
details and discussion.
The implementation corresponds to -MD -MP compiler flag where the
generated dependency file also includes phony targets which works
around the errors where the dependency is removed. This matches the
format used by ld and gold.
Fixes PR42806
Differential Revision: https://reviews.llvm.org/D82437
Determine whether switch edges are feasible based on range information,
and remove non-feasible edges lateron.
This does not try to determine whether the default edge is dead,
as we'd have to determine that the range is fully covered by the
cases for that.
Another limitation here is that we don't remove dead cases that
have the same successor as a live case. I'm not handling this
because I wanted to keep the edge removal based on feasible edges
only, rather than inspecting ranges again there -- this does not
seem like a particularly useful case to handle.
Differential Revision: https://reviews.llvm.org/D84270
Currently we skip alias sets with only reads or a single write and no
reads, but still add the pointers to the list of pointers in RtCheck.
This can lead to cases where we try to access a pointer that does not
exist when grouping checks. In most cases, the way we access
PositionMap masked that, as the value would default to index 0.
But in the example in PR46854 it causes a crash.
This patch updates the logic to avoid adding pointers for alias sets
that do not need any checks. It makes things slightly more verbose, by
first checking the numbers of reads/writes and bailing out early if we don't
need checks for the alias set.
I think this makes the logic a bit simpler to follow.
Reviewed By: anemet
Differential Revision: https://reviews.llvm.org/D84608
ThinLTO is run using a single thread on Linux on Power. The
compute_thread_count() routine calls getHostNumPhysicalCores which
returns -1 by default, and so `MaxThreadCount is set to 1.
unsigned llvm::ThreadPoolStrategy::compute_thread_count() const {
int MaxThreadCount = UseHyperThreads
? computeHostNumHardwareThreads()
: sys::getHostNumPhysicalCores();
if (MaxThreadCount <= 0)
MaxThreadCount = 1;
…
}
Fix: provide custom implementation of getHostNumPhysicalCores for
Linux on Power and Linux on Z.
Reviewed By: Kai, uweigand
Differential Revision: https://reviews.llvm.org/D84764
LLVM selection dag assumes "switch" indices are pointer sized, which causes problems for our 32-bit br_table. The new function ensures 32-bit operands don't get unnecessarily extended, and 64-bit operands get truncated.
Note that the changes to the existing test test exactly that: the addition of -NEXT in 2 places ensures no extension is inserted (which the test previously ignored) and that the wrap is present (previously omitted in wasm64 mode).
Differential Revision: https://reviews.llvm.org/D84705
We are using undef on the indirect move source subreg and then
using implicit super-reg. This creates a problem in RA when
Greedy decides to split the register. It reassigns the implicit
super-reg but does not bother to change undef source because
it is really does not matter. The fix is to stop lying to RA and
drop undef flag.
This has also hit a problem in SIFoldOperands as it can fold
immediate into an indirect move since there is no undef flag
anymore. That results in multiple test failures, so added the
check for this case.
Differential Revision: https://reviews.llvm.org/D84899
This test was added in D74217 (and the `.categories` file later added in ccf1c30cde6e1e763e7c9cdd48a609a805166699) around the same time I moved the test tree from `lldb/packages/Python/lldbsuite/test` to `lldb/test/API` (D71151). Since this got lost in the move, it isn't running. (I introduced an intentional syntax error, and `ninja check-lldb` passes).
I moved it to the correct location, and now it runs and passes -- locally, at least -- as `ninja check-lldb-api-tools-lldb-server-registers-target-xml-reading`.
