Force -fuse-ld=ld, as some other tests in the same file do.
Loosen the regex matching the linker tool name as well, as this
can end up being <triple>-ld in case such a named tool exists.
llvm-svn: 353946
It looks like I was too hasty to submit the previous patch. It does fix
some tests on python3, but it also breaks one tests with python2.
This happens because the gdb-remote protocol can sometimes (but not very
often) contain binary data, and attempting to parse this as utf8
characters fails.
This reverts commit r353944.
llvm-svn: 353945
FindLastStoreBRVisitor tries to find the first node in the exploded graph where
the current value was assigned to a region. This node is called the "store
site". It is identified by a pair of Pred and Succ nodes where Succ already has
the binding for the value while Pred does not have it. However the visitor
mistakenly identifies a node pair as the store site where the value is a
`LazyCompoundVal` and `Pred` does not have a store yet but `Succ` has it. In
this case the `LazyCompoundVal` is different in the `Pred` node because it also
contains the store which is different in the two nodes. This error may lead to
crashes (a declaration is cast to a parameter declaration without check) or
misleading bug path notes.
In this patch we fix this problem by checking for unequal `LazyCompoundVals`: if
their region is equal, and their store is the same as the store of their nodes
we consider them as equal when looking for the "store site". This is an
approximation because we do not check for differences of the subvalues
(structure members or array elements) in the stores.
Differential Revision: https://reviews.llvm.org/D58067
llvm-svn: 353943
This patch introduces a new instruction stage named 'IS_PENDING'.
An instruction transitions from the IS_DISPATCHED to the IS_PENDING stage if
input registers are not available, but their latency is known.
This patch also adds a new set of instructions named 'PendingSet' to class
Scheduler. The idea is that the PendingSet will only contain instructions that
have reached the IS_PENDING stage.
By construction, an instruction in the PendingSet is only dependent on
instructions that have already reached the execution stage. The plan is to use
this knowledge to identify bottlenecks caused by data dependencies (see
PR37494).
Differential Revision: https://reviews.llvm.org/D58066
llvm-svn: 353937
When instcombine sinks an instruction between two basic blocks, it sinks any
dbg.value users in the source block with it, to prevent debug use-before-free.
However we can do better by attempting to salvage the debug users, which would
avoid moving where the variable location changes. If we successfully salvage,
still sink a (cloned) dbg.value with the sunk instruction, as the sunk
instruction is more likely to be "live" later in the compilation process.
If we can't salvage dbg.value users of a sunk instruction, mark the dbg.values
in the original block as being undef. This terminates any earlier variable
location range, and represents the fact that we've optimized out the variable
location for a portion of the program.
Differential Revision: https://reviews.llvm.org/D56788
llvm-svn: 353936
This patch adds support of guards expressed in explicit form via
`widenable_condition` in Guard Widening pass.
Differential Revision: https://reviews.llvm.org/D56075
Reviewed By: reames
llvm-svn: 353932
Summary:
This is a follow-up to D57510. This patch stops DebugHandlerBase from
changing the starting label for the first non-overlapping,
register-described parameter DBG_VALUEs to the beginning of the
function. That code did not consider what defined the registers, which
could result in the ranges for the debug values starting before their
defining instructions. We currently do not emit debug values for
constant values directly at the start of the function, so this code is
still useful for such values, but my intention is to remove the code
from DebugHandlerBase completely when we get there. One reason for
removing it is that the code violates the history map's ranges, which I
think can make it quite confusing when troubleshooting.
In D57510, PrologEpilogInserter was amended so that parameter DBG_VALUEs
now are kept at the start of the entry block, even after emission of
prologue code. That was done to reduce the degradation of debug
completeness from this patch. PR40638 is another example, where the
lexical-scope trimming that LDV does, in combination with scheduling,
results in instructions after the prologue being left without locations.
There might be other cases where the DBG_VALUEs are pushed further down,
for which the DebugHandlerBase code may be helpful, but as it now quite
often result in incorrect locations, even after the prologue, it seems
better to remove that code, and try to work our way up with accurate
locations.
In the long run we should maybe not aim to provide accurate locations
inside the prologue. Some single location descriptions, at least those
referring to stack values, generate inaccurate values inside the
epilogue, so we maybe should not aim to achieve accuracy for location
lists. However, it seems that we now emit line number programs that can
result in GDB and LLDB stopping inside the prologue when doing line
number stepping into functions. See PR40188 for more information.
A summary of some of the changed test cases is available in PR40188#c2.
Reviewers: aprantl, dblaikie, rnk, jmorse
Reviewed By: aprantl
Subscribers: jdoerfert, jholewinski, jvesely, javed.absar, llvm-commits
Tags: #debug-info, #llvm
Differential Revision: https://reviews.llvm.org/D57511
llvm-svn: 353928
Restore the previous behavior of using install directories for
LLVM_MAIN_INCLUDE_DIR, LLVM_LIBRARY_DIR and LLVM_BINARY_DIR. The update
from llvm-config to CMake has changed the values of those values to use
LLVM_BUILD_* which is plain wrong and breaks stand-alone builds.
Instead, use the CMake counterparts of the values returned
by llvm-config.
Differential Revision: https://reviews.llvm.org/D57995
llvm-svn: 353925
Modify LLVMConfig to provide split variables for in-source and generated
include paths. Currently, it uses a single value for both
LLVM_INCLUDE_DIRS and LLVM_INCLUDE_DIR which works for install tree but
fails hard at build tree (where LLVM_INCLUDE_DIR incorrectly contains
multiple values).
Instead, put the generated directory in LLVM_INCLUDE_DIR, and the source
tree in LLVM_MAIN_INCLUDE_DIR which is consistent with in-LLVM builds.
For install tree, both variables will have the same value.
Differential Revision: https://reviews.llvm.org/D58109
llvm-svn: 353924
Try to use 64-bit SLP vectorization. In addition to horizontal instrs
this change triggers optimizations for partial vector operations (for instance,
using low halfs of 128-bit registers xmm0 and xmm1 to multiply <2 x float> by
<2 x float>).
Fixes llvm.org/PR32433
llvm-svn: 353923
In 64-bit mode prior to avx512 we use Expand, but with avx512 we need to make f32/f64 conversions Legal so we use Custom and then do our own expansion for f80. But this seems to produce codegen differences relative to avx2. This patch corrects this.
llvm-svn: 353921
-Pull the final stack load creation from the two callers into the helper.
-Return a single SDValue instead of a std::pair.
-Remove the Replace flag which isn't really needed.
llvm-svn: 353920
This function doesn't use anything MSVC specific but works fine
for any _WIN32 target.
Differential Revision: https://reviews.llvm.org/D58106
llvm-svn: 353918
Profiling still doesn't seem to work properly, but this at least
hooks up the library and eases completing whatever is missing.
Differential Revision: https://reviews.llvm.org/D58107
llvm-svn: 353917
Summary:
This is coming from the discussion in D55356 (the most interesting part
happened on the mailing list, so it isn't reflected on the review page).
In short the issue is that lldb assumes that all bytes of a module image
in memory will be backed by a "section". This isn't the case for PECOFF
files because the initial bytes of the module image will contain the
file header, which does not correspond to any normal section in the
file. In particular, this means it is not possible to implement
GetBaseAddress function for PECOFF files, because that's supposed point
to the first byte of that header.
If my (limited) understanding of how PECOFF files work is correct, then
the OS is expecded to load the entire module into one continuous chunk
of memory. The address of that chunk (+/- ASLR) is given by the "image
base" field in the COFF header, and it's size by "image size". All of
the COFF sections are then loaded into this range.
If that's true, then we can model this behavior in lldb by creating a
"container" section to represent the entire module image, and then place
other sections inside that. This would make be consistent with how MachO
and ELF files are modelled (except that those can have multiple
top-level containers as they can be loaded into multiple discontinuous
chunks of memory).
This change required a small number of fixups in the PDB plugins, which
assumed a certain order of sections within the object file (which
obivously changes now). I fix this by changing the lookup code to use
section IDs (which are unchanged) instead of indexes. This has the nice
benefit of removing spurious -1s in the plugins as the section IDs in
the pdbs match the 1-based section IDs in the COFF plugin.
Besides making the implementation of GetBaseAddress possible, this also
improves the lookup of addresses in the gaps between the object file
sections, which will now be correctly resolved as belonging to the
object file.
Reviewers: zturner, amccarth, stella.stamenova, clayborg, lemo
Reviewed By: clayborg, lemo
Subscribers: JDevlieghere, abidh, lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D56537
llvm-svn: 353916
If we're not in a standalone build, this variable should be already
set, so there's no need to set it again or to cache it.
Differential Revision: https://reviews.llvm.org/D57993
llvm-svn: 353915
Apparently I introduced a new top level directory called `b` in the root
of the monorepo with r353906. I don't understand how or why this
happened, but I presume something went wrong doing `git llvm push`.
I'm not even sure if I can delete the directory again, but giving it a
shot in this commit.
Thanks to Shoaib Meenai for noticing this.
llvm-svn: 353913
The `ap` suffix is a remnant of lldb's former use of auto pointers,
before they got deprecated. Although all their uses were replaced by
unique pointers, some variables still carried the suffix.
In r353795 I removed another auto_ptr remnant, namely redundant calls to
::get for unique_pointers. Jim justly noted that this is a good
opportunity to clean up the variable names as well.
I went over all the changes to ensure my find-and-replace didn't have
any undesired side-effects. I hope I didn't miss any, but if you end up
at this commit doing a git blame on a weirdly named variable, please
know that the change was unintentional.
llvm-svn: 353912
Known underlying bugs have been fixed, intensive fuzz testing did not
find any new problems. Re-enabling by default. Feel free to revert if
it causes any functional failures.
llvm-svn: 353911
We stil don't have a source location, which is pretty lame, but at least
we won't tell the user to file a clang bug report anymore.
Fixes PR40712
llvm-svn: 353907
This patch hooks up clang and lldb's reproducers functionality. It
ensures that when capturing a reproducer, headers and modules imported
through the expression parser are collected.
Differential revision: https://reviews.llvm.org/D58076
llvm-svn: 353906
Summary:
A link error was encountered when using the Red Hat Developer Toolset.
In the RHDTS, `libstdc++.so` is a linker script that may resolve symbols
to a static library. This patch places `-lstdc++` later in the ordering.
Reviewers: sfertile, nemanjai, tstellar, dberris
Reviewed By: dberris
Subscribers: dberris, mgorny, delcypher, jdoerfert, #sanitizers, llvm-commits
Tags: #llvm, #sanitizers
Differential Revision: https://reviews.llvm.org/D58144
llvm-svn: 353905
The DWARF standard says that an empty compile unit is not valid:
> Each such contribution consists of a compilation unit header (see
> Section 7.5.1.1 on page 200) followed by a single DW_TAG_compile_unit or
> DW_TAG_partial_unit debugging information entry, together with its
> children.
Therefore we shouldn't clone them in dsymutil.
Differential revision: https://reviews.llvm.org/D57979
llvm-svn: 353903
buildgo.sh puts most of sanitizer_common together in a single source
file. These single-letter macros end up affecting a lot of unrelated
code; #undef them as early as possible.
llvm-svn: 353902
Add plumbing to get MemorySSA in the remaining loop passes.
Also update unit test to add the dependency.
[EnableMSSALoopDependency remains disabled].
llvm-svn: 353901
Subtargets are a function level property, so ideally we would
eliminate everywhere that needs to check the global one. Rename the
function to try avoiding confusion.
llvm-svn: 353900
This was inhibiting inlining of library functions when clang was
invoking the inliner directly. This is covering a bit of a mess with
subtarget feature handling, and this shouldn't be a subtarget
feature. The behavior is different depending on whether you are using
a -mattr flag in clang, or llc, opt.
llvm-svn: 353899
Summary:
Unlimitted number of calls to getClobberingAccess can lead to high
compile times in pathological cases.
Switching EnableLicmCap flag from bool to int, and enabling to default 100.
(tested to be appropriate for current bechmarks)
We can revisit this value when enabling MemorySSA.
Reviewers: sanjoy, chandlerc, george.burgess.iv
Subscribers: jlebar, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D57968
llvm-svn: 353897
Martin Storsjo