Currently we assert that only non target specific opcodes can have
missing RegisterClass constraints in the MCDesc. The backend can have
instructions with register operands but don't have RegisterClass
constraints (say using unknown_class) in which case the instruction
defining the register will constrain it.
Change the assert to only fire if a def has no regclass.
https://reviews.llvm.org/D43409
llvm-svn: 326142
When writing an inline test, there is no way to make sure that any of
the inline commands are actually executed, so this patch adds a sanity
check that at least one breakpoint was hit. This avoids a test with no
breakpoints being hit passing.
Differential Revision: https://reviews.llvm.org/D43694
llvm-svn: 326140
This fixes pr36475.
I think this code can be simplified a bit, but I would like to check
in the more direct fix if we are in agreement on the direction and
then refactor.
This is not something that bfd does. The issue is not noticed in bfd
because it keeps fewer sections from the linkerscript in the output.
The reasons why it seems reasonable to do this:
- As George noticed, we would still keep the flags if the output
section had both an empty synthetic section and a regular section
- We need an heuristic to find the flags of output sections. Using the
flags of a synthetic section that would have been there seems a
reasonable heuristic.
llvm-svn: 326137
See D42775 for discussion. Turns out, just exploring nodes which
weren't explored first is not quite enough, as e.g. the first quick
traversal resulting in a report can mark everything as "visited", and
then subsequent traversals of the same region will get all the pitfalls
of DFS.
Priority queue-based approach in comparison shows much greater
increase in coverage and even performance, without sacrificing memory.
Differential Revision: https://reviews.llvm.org/D43354
llvm-svn: 326136
Bison/YACC generated files result in a very large number of (presumably)
false positives from the analyzer.
These false positives are "true" in a sense of the information analyzer
sees: assuming that the lexer can return any token at any point a number
of uninitialized reads does occur.
(naturally, the analyzer can not capture a complex invariant that
certain tokens can only occur under certain conditions).
Current fix simply stops analysis on those files.
I have examined a very large number of such auto-generated files, and
they do all start with such a comment.
Conversely, user code is very unlikely to contain such a comment.
rdar://33608161
Differential Revision: https://reviews.llvm.org/D43421
llvm-svn: 326135
Further investigation required; tests will be enabled on the buildbot
worker soon. Marking failing tests for now in order to start with a
green buildbot while investigation takes place.
llvm.org/pr36527
llvm-svn: 326134
Agner's tables indicate that for SSE42+ targets (Core2 and later) we can reduce the FADD/FSUB/FMUL costs down to 1, which should fix the Himeno benchmark.
Note: the AVX512 FDIV costs look rather dodgy, but this isn't part of this patch.
Differential Revision: https://reviews.llvm.org/D43733
llvm-svn: 326133
Summary:
Fixes Bug 32434
See https://bugs.llvm.org/show_bug.cgi?id=32434
Short summary:
std::rethrow_exception does not use __cxa_throw to rethrow the exception, so if
it is called from uninstrumented code, it will leave the stack poisoned. This
can lead to false positives.
Long description:
For functions which don't return normally (e.g. via exceptions), asan needs to
unpoison the entire stack. It is not known before a call to such a function
where execution will continue, some function which don't contain cleanup code
like destructors might be skipped. After stack unwinding, execution might
continue in uninstrumented code.
If the stack has been poisoned before such a function is called, but the stack
is unwound during the unconventional return, then zombie redzones (entries) for
no longer existing stack variables can remain in the shadow memory. Normally,
this is avoided by asan generating a call to asan_handle_no_return before all
functions marked as [[noreturn]]. This asan_handle_no_return unpoisons the
entire stack. Since these [[noreturn]] functions can be called from
uninstrumented code, asan also introduces interceptor functions which call
asan_handle_no_return before running the original [[noreturn]] function;
for example, cxa_throw is intercepted.
If a [[noreturn]] function is called from uninstrumented code (so the stack is
left poisoned) and additionally, execution continues in uninstrumented code, new
stack variables might be introduced and overlap with the stack variables
which have been removed during stack unwinding. Since the redzones are not
cleared nor overwritten by uninstrumented code, they remain but now contain
invalid data.
Now, if the redzones are checked against the new stack variables, false
positive reports can occur. This can happen for example by the uninstrumented
code calling an intercepted function such as memcpy, or an instrumented
function.
Intercepting std::rethrow_exception directly is not easily possible since it
depends on the C++ standard library implementation (e.g. libcxx vs libstdc++)
and the mangled name it produces for this function. As a rather simple
workaround, we're intercepting _Unwind_RaiseException for libstdc++. For
libcxxabi, we can intercept the ABI function __cxa_rethrow_primary_exception.
Patch by Robert Schneider.
Reviewers: kcc, eugenis, alekseyshl, vitalybuka
Reviewed By: vitalybuka
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D42644
llvm-svn: 326132
Without this fix, the test ERRORs because the link of the inferior fails. This
patch adds the LLDB_TEST_API macro where needed and uses the new -2 magic
value for num_expected_locations to account for lazy-loading of module symbols
on Windows.
With this fix, the test itself still fails: conflicting_symbol isn't in the
debug info nor the export table, and Windows binaries don't have an equivalent
of the ELF .symtab. We need to understand why the test works to keep the
symbol out of the debug info. In the mean time, having the test fail at this
point is a better indication of the remaining problem than a build error.
Differential Revision: https://reviews.llvm.org/D43688
llvm-svn: 326130
There's still some shortcoming in our ability to combine binops of constants with different sizes separated by an extend. I'll try to look at that next.
llvm-svn: 326128
Addresses https://bugs.llvm.org/show_bug.cgi?id=36206
rdar://37159026
A proper fix would be much harder, and would involve changing the
appropriate code in ExprEngine to be aware of the size limitations of
the type used for addressing.
Differential Revision: https://reviews.llvm.org/D43218
llvm-svn: 326122
These sections will be retired. Also, explicitly list llvm-objdump
as a dependency. This should've been done in the previous commit,
but I failed to squash the two changes together.
Thanks to Adrian for pointing the first problem out in a comment.
llvm-svn: 326121
test/std/numerics/numeric.ops/exclusive.scan/exclusive_scan.pass.cpp
test/std/numerics/numeric.ops/exclusive.scan/exclusive_scan_init_op.pass.cpp
test/std/numerics/numeric.ops/inclusive.scan/inclusive_scan.pass.cpp
test/std/numerics/numeric.ops/inclusive.scan/inclusive_scan_op.pass.cpp
test/std/numerics/numeric.ops/inclusive.scan/inclusive_scan_op_init.pass.cpp
test/std/numerics/numeric.ops/transform.exclusive.scan/transform_exclusive_scan_init_bop_uop.pass.cpp
test/std/numerics/numeric.ops/transform.inclusive.scan/transform_inclusive_scan_bop_uop.pass.cpp
test/std/numerics/numeric.ops/transform.inclusive.scan/transform_inclusive_scan_bop_uop_init.pass.cpp
Fix MSVC x64 truncation warnings.
warning C4267: conversion from 'size_t' to 'int', possible loss of data
test/std/strings/basic.string/string.modifiers/string_append/push_back.pass.cpp
Fix MSVC uninitialized memory warning.
warning C6001: Using uninitialized memory 'vl'.
test/std/utilities/tuple/tuple.tuple/tuple.cnstr/PR20855_tuple_ref_binding_diagnostics.pass.cpp
Include <cassert> for the assert() macro.
Fixes D43273.
llvm-svn: 326120
Summary:
We have an early DAG combine to turn these patterns into MOVMSK, but that combine doesn't work if the vXi1 type has more elements than the widest legal vXi8 type. Type legalization will eventually split it down to v16i1 or v32i1 and then the bitcast gets legalized to a truncstore and a scalar load. The truncstore will get lowered to a series of extracts and bit math.
This patch adds a custom legalization to use a sign extend and MOVMSK instead. This prevents the eventual scalarization.
Reviewers: spatel, RKSimon, zvi
Reviewed By: RKSimon
Subscribers: mgorny, llvm-commits
Differential Revision: https://reviews.llvm.org/D43593
llvm-svn: 326119
This reverts commit e17911006548518634fad66bb8648bcad49a1d64.
This is failing on ASAN bots because asan expects column info,
and it's also failing on some linux bots for unknown reasons which
i need to investigate.
llvm-svn: 326116
Windows debuggers don't work properly when column info is emitted
with lines. We handled this by checking if the driver mode was
cl, but it's possible to cause the gcc driver to emit codeview as
well, and in that path we were emitting column info with codeview.
Differential Revision: https://reviews.llvm.org/D43700
llvm-svn: 326113
Summary:
The command takes two input arguments: a module to use as a debug target
and a file containing a list of commands. The command will execute each
of the breakpoint commands in the file and dump the breakpoint state
after each one.
The commands are expected to be breakpoint set/remove/etc. commands, but
I explicitly allow any lldb command here, so you can do things like
change setting which impact breakpoint resolution, etc. There is also a
"-persistent" flag, which causes lldb-test to *not* automatically clear
the breakpoint list after each command. Right now I don't use it, but
the idea behind it was that it could be used to test more complex
combinations of breakpoint commands (set+modify, set+disable, etc.).
Right now the command prints out only the basic breakpoint state, but
more information can be easily added there. To enable easy matching of
the "at least one breakpoint location found" state, the command
explicitly prints out the string "At least one breakpoint location.".
To enable testing of breakpoints set with an absolute paths, I add the
ability to perform rudimentary substitutions on the commands: right now
the string %p is replaced by the directory which contains the command
file (so, under normal circumstances, this will perform the same
substitution as lit would do for %p).
I use this command to rewrite the TestBreakpointCaseSensitivity test --
the test was checking about a dozen breakpoint commands, but it was
launching a new process for each one, so it took about 90 seconds to
run. The new test takes about 0.3 seconds for me, which is approximately
a 300x speedup.
Reviewers: davide, zturner, jingham
Subscribers: luporl, lldb-commits
Differential Revision: https://reviews.llvm.org/D43686
llvm-svn: 326112
This change improves incremental rebuild performance on dual Xeon 8168
machines by 54%. This change also improves run time code gen by not
forcing the case values to be lvalues.
llvm-svn: 326109
This wires up -pass-remarks-hotness-threshold to LTO and ThinLTO.
Next is to change the clang driver to pass this
with -fdiagnostics-hotness-threshold.
Differential Revision: https://reviews.llvm.org/D41465
llvm-svn: 326107
FindAvailableMemoryRange can currently overwrite existing memory (by restricting the VM below addresses that are already used). This patch adds a check to make sure we don't restrict the VM space too much. We are also now more explicit about why the lookup failed and print out verbose values.
Differential Revision: https://reviews.llvm.org/D43318
llvm-svn: 326106
Summary:
Print current stack on CHECK violation to aid debugging and
match other sanitizers functionality.
Reviewers: eugenis
Subscribers: delcypher, llvm-commits, #sanitizers
Differential Revision: https://reviews.llvm.org/D43692
llvm-svn: 326105
In DWARF v5 the Line Number Program Header is extensible, allowing values with
new content types. This vendor extension to DWARF v5 allows source text to be
embedded directly in the line tables of the debug line section.
Add new flag (-g[no-]embed-source) to Driver and CC1 which indicates
that source should be passed through to LLVM during CodeGen.
Differential Revision: https://reviews.llvm.org/D42766
llvm-svn: 326102
Summary:
Start making the Scudo tests less Linux-y:
- `malloc_usable_size` doesn't exist everywhere, so replace them with
`__sanitizer_get_allocated_size` which we provide;
- move all the `memalign` related tests into `memalign.c` since it's also not
available everywhere.
I also noticed that the `memalign.c` was missing a line in one of the loops.
Reviewers: alekseyshl
Reviewed By: alekseyshl
Subscribers: delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D43393
llvm-svn: 326100
Note: gcc appears to allow this fold with -freciprocal-math alone,
but clang/llvm require more than that with this patch. The wording
in the definitions seems fuzzy enough that it could go either way,
but we'll err on the conservative side of FMF interpretation.
This patch also changes the newly created fmul to have FMF propagated
by the last fdiv rather than intersecting the FMF of the fdivs. This
matches the behavior of other folds near here. The new fmul is only
used to produce an intermediate op for the final fdiv result, so it
shouldn't be any stricter than that result. The previous behavior
could result in dropping FMF via other folds in instcombine or CSE.
Differential Revision: https://reviews.llvm.org/D43398
llvm-svn: 326098
In r322867, we introduced IsStandalone when printing MIR in -debug
output. The default behaviour for that was:
1) If any of MBB, MI, or MO are -debug-printed separately, don't omit any
redundant information.
2) When -debug-printing a MF entirely, don't print any redundant
information.
3) When printing MIR, don't print any redundant information.
I'd like to change 2) to:
2) When -debug-printing a MF entirely, don't omit any redundant information.
Differential Revision: https://reviews.llvm.org/D43337
llvm-svn: 326094
Aditya Nandakumar