Summary:
Building MemorySSA gathers alias information for Defs/Uses.
Store and expose this information when optimizing uses (when building MemorySSA),
and when optimizing defs or updating uses (getClobberingMemoryAccess).
Current patch does not propagate alias information through MemoryPhis.
Reviewers: gbiv, dberlin
Subscribers: Prazek, sanjoy, llvm-commits
Differential Revision: https://reviews.llvm.org/D38569
llvm-svn: 327035
This fixes a false positive ODR violation that is reported by ASan when using LTO. In cases, where two constant globals have the same value, LTO will merge them, which breaks ASan's ODR detection.
Differential Revision: https://reviews.llvm.org/D43959
llvm-svn: 327029
This allows the customization of the performance report.
Users can specify their own custom sequence of views.
Each view contributes a portion of the performance report generated by the
BackendPrinter.
Internally, class BackendPrinter keeps a sequence of views; views are printed
out in sequence when method 'printReport()' is called.
This patch addresses one of the two review comments from Clement in D43951.
llvm-svn: 327018
Summary:
This patch adds basic .debug_names verification capabilities to the
DWARF verifier. Right now, it checks that the headers and abbreviation
tables of the individual name indexes can be parsed correctly, it
verifies the buckets table and the cross-checks the CU lists for
consistency. I intend to add further checks in follow-up patches.
Reviewers: JDevlieghere, aprantl, probinson, dblaikie
Subscribers: vleschuk, echristo, clayborg, llvm-commits
Differential Revision: https://reviews.llvm.org/D44211
llvm-svn: 327011
Append -Wl,-rpath-link conditionally to whether GNU ld.bfd is used
rather than the Linux+!gold conditionals. Also move it out of 'else'
branch of *BSD handling. This fixes build failures with ld.bfd
on Gentoo/FreeBSD, and should cause no harm on other systems using
ld.bfd.
This patch improves the original logic by reusing results of linker
detection introduced in r307852.
Differential Revision: https://reviews.llvm.org/D43751
llvm-svn: 327007
In an example like "clang -fxray-instrument .." the .. could be confused
with a literal .. (parent directory), which is used in commands like
"cmake -GNinja .."
llvm-svn: 327000
llvm-mca is an LLVM based performance analysis tool that can be used to
statically measure the performance of code, and to help triage potential
problems with target scheduling models.
llvm-mca uses information which is already available in LLVM (e.g. scheduling
models) to statically measure the performance of machine code in a specific cpu.
Performance is measured in terms of throughput as well as processor resource
consumption. The tool currently works for processors with an out-of-order
backend, for which there is a scheduling model available in LLVM.
The main goal of this tool is not just to predict the performance of the code
when run on the target, but also help with diagnosing potential performance
issues.
Given an assembly code sequence, llvm-mca estimates the IPC (instructions per
cycle), as well as hardware resources pressure. The analysis and reporting style
were mostly inspired by the IACA tool from Intel.
This patch is related to the RFC on llvm-dev visible at this link:
http://lists.llvm.org/pipermail/llvm-dev/2018-March/121490.html
Differential Revision: https://reviews.llvm.org/D43951
llvm-svn: 326998
Whilst working on improvements to the error handling of the debug line
parsing code, I noticed that if an invalid offset were to be specified
in a call to getOrParseLineTable(), an entry in the LineTableMap would
still be created, even if the offset was not within the section range.
The immediate parsing attempt afterwards would fail (it would end up
getting a version of 0), and thereafter, any subsequent calls to
getOrParseLineTable or getLineTable would return the default-
constructed, invalid line table. In reality, we shouldn't even attempt
to parse this table, and we should always return a nullptr from these
two functions for this situation.
I have tested this via a unit test, which required some new framework
for unit testing debug line. My plan is to add quite a few more unit
tests for the new error reporting mechanism that will follow shortly,
hence the reason why the supporting code for the tests are written the
way they are - I intend to extend the DwarfGenerator class to support
generating debug line. At that point, I'll make sure that there are a
few positive test cases for this and the parsing code too.
Differential Revision: https://reviews.llvm.org/D44200
Reviewers: JDevlieghere, aprantl
llvm-svn: 326995
With this patch, the tablegen 'SubtargetEmitter' always generates processor
resource names.
The impact of this patch on the code size of other llvm tools is small. I have
observed an average increase of 0.03% in code size when doing a release build of
LLVM (on windows, using MSVC) with all the default backends.
This change is done in preparation for the upcoming llvm-mca patch.
llvm-svn: 326993
This instruction can be thought of as reading either the even elements of a vXi32 input or the lower half of each element of a vXi64 input. We currently use the vXi32 interpretation, but vXi64 matches better with its broadcast behavior in EVEX.
I'm looking at moving MULDQ/MULUDQ creation to a DAG combine so we can do it when AVX512DQ is enabled without having to go through Custom lowering. But in some of the test cases we failed to use a broadcast load due to the size difference. This should help with that.
I'm also wondering if we can model these instructions in native IR and remove the intrinsics and I think using a vXi64 type will work better with that.
llvm-svn: 326991
- Improve description of XNACK ELF flag.
- Rename all uses of wave to wavefront to be consistent.
Differential Revision: https://reviews.llvm.org/D43983
llvm-svn: 326989
Summary:
Most of the time, compiler statistics can be obtained using a process that
performs a single compilation and terminates such as llc. However, this isn't
always the case. JITs for example, perform multiple compilations over their
lifetime and STATISTIC() will record cumulative values across all of them.
Provide tools like this with the facilities needed to measure individual
compilations by allowing them to reset the STATISTIC() values back to zero using
llvm::ResetStatistics(). It's still the tools responsibility to ensure that they
perform compilations in such a way that the results are meaningful to their
intended use.
Reviewers: qcolombet, rtereshin, bogner, aditya_nandakumar
Reviewed By: bogner
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D44181
llvm-svn: 326981
This reverts commit 1f3bd185c53beb6aa68446974b7e80837abd6ef0 (r326107)
because it fails
ThinLTO/X86/diagnostic-handler-remarks-with-hotness.ll.
llvm-svn: 326975
Summary:
Original change was D43313 (r326932) and reverted by r326953 because it
broke an LLD test and a windows build. The LLD test was already fixed in
lld commit r326944 (thanks maskray). This is the original change with
the windows build fixed.
llvm-svn: 326970
Summary:
Fixes an UB caught by sanitizer. The shift amount might be larger than 32 so the operand should be 1ULL.
In this patch, we replace the original expression with existing API with uint64_t type.
Reviewers: eli.friedman, rengolin
Reviewed By: rengolin
Subscribers: rengolin, javed.absar, llvm-commits, kristof.beyls
Differential Revision: https://reviews.llvm.org/D44234
llvm-svn: 326969
These patterns weren't checking the alignment of the load, but were using the aligned instructions. This will cause a GP fault if the data isn't aligned.
I believe these were introduced in r312450.
llvm-svn: 326967
Currently on Windows (_MSC_VER) LLVMSymbolizer supports only Microsoft mangling.
This fix just explicitly uses itaniumDemangle when mangled name starts with _Z.
Differential Revision: https://reviews.llvm.org/D44192
llvm-svn: 326959
Since there is no instruction for integer vector division, factor in the
cost of singling out each element to be used with the scalar division
instruction.
Differential revision: https://reviews.llvm.org/D43974
llvm-svn: 326955
The attached testcase started failing after the patch to define
isExtractSubvectorCheap with the following pattern mismatch:
ISEL: Starting pattern match
Initial Opcode index to 85068
Match failed at index 85076
LLVM ERROR: Cannot select: t47: v8i16 = insert_subvector undef:v8i16, t43, Constant:i64<0>
The code generated from llvm/lib/Target/AArch64/AArch64InstrInfo.td
def : Pat<(insert_subvector undef, (v4i16 FPR64:$src), (i32 0)),
(INSERT_SUBREG (v8i16 (IMPLICIT_DEF)), FPR64:$src, dsub)>;
is in ninja/lib/Target/AArch64/AArch64GenDAGISel.inc
At the location of the error it is:
/* 85076*/ OPC_CheckChild2Type, MVT::i32,
And it failed to match the type of operand 2.
Adding another def-pat for i64 fixes the failed def-pat error:
def : Pat<(insert_subvector undef, (v4i16 FPR64:$src), (i64 0)),
(INSERT_SUBREG (v8i16 (IMPLICIT_DEF)), FPR64:$src, dsub)>;
llvm-svn: 326949
Because of -ffunction-sections (and maybe other use cases I'm not aware of?) it
can occur that we need more than 0xfeff sections but ELF dosn't support that
many sections. To solve this problem SHN_XINDEX exists and with it come a whole
host of changes for section indexes everywhere. This change adds support for
those cases which should allow llvm-objcopy to copy binaries that have an
arbitrary number of sections.
Differential Revision: https://reviews.llvm.org/D42516
llvm-svn: 326940
