Summary:
For better vectorization result we should take into consideration the
cost of the user insertelement instructions when we try to
vectorize sequences that build the whole vector. I.e. if we have the
following scalar code:
```
<Scalar code>
insertelement <ScalarCode>, ...
```
we should consider the cost of the last `insertelement ` instructions as
the cost of the scalar code.
Reviewers: RKSimon, spatel, hfinkel, mkuper
Subscribers: javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D42657
llvm-svn: 324893
Armv8.1-A added an atomic load-add instruction, but not a load-subtract
instruction. Our current code-generation for atomic load-subtract always
inserts a NEG instruction to negate it's argument, even if it could be
folded into a constant or another instruction.
This adds lowering early in selection DAG to convert a load-subtract
operation into a subtract and a load-add, allowing the normal DAG
optimisations to work on it.
I've left the old tablegen patterns in because they are still needed for
global isel.
Some of the tests in this patch are copied from D35375 by Chad Rosier (which
was abandoned).
Differential revision: https://reviews.llvm.org/D42477
llvm-svn: 324892
Summary:
To aid debugging failures and crashes.
Only part of ignored diagnostics was logged before, now we log all of
them.
Reviewers: ioeric, hokein, sammccall
Reviewed By: hokein
Subscribers: klimek, jkorous-apple, cfe-commits
Differential Revision: https://reviews.llvm.org/D43123
llvm-svn: 324888
It asserts building Chromium; see PR36346.
(This also reverts the follow-up r324836.)
> If a load follows a store and reloads data that the store has written to memory, Intel microarchitectures can in many cases forward the data directly from the store to the load, This "store forwarding" saves cycles by enabling the load to directly obtain the data instead of accessing the data from cache or memory.
> A "store forward block" occurs in cases that a store cannot be forwarded to the load. The most typical case of store forward block on Intel Core microarchiticutre that a small store cannot be forwarded to a large load.
> The estimated penalty for a store forward block is ~13 cycles.
>
> This pass tries to recognize and handle cases where "store forward block" is created by the compiler when lowering memcpy calls to a sequence
> of a load and a store.
>
> The pass currently only handles cases where memcpy is lowered to XMM/YMM registers, it tries to break the memcpy into smaller copies.
> breaking the memcpy should be possible since there is no atomicity guarantee for loads and stores to XMM/YMM.
llvm-svn: 324887
Summary:
Before the fix, if clang-move decides to move the following macro statement, it only moves the first line `DEFINE(A,`.
```
DEFINE(A,
B);
```
Reviewers: ioeric
Reviewed By: ioeric
Subscribers: klimek, cfe-commits
Differential Revision: https://reviews.llvm.org/D43174
llvm-svn: 324886
In case of correct using of the 'l' constraint llvm now generates valid
code; otherwise it shows an error message. Initially these triggers an
assertion.
This commit is the same as r324869 with fixed the test's file name.
llvm-svn: 324885
Summary:
Right now clang is skipping array cookie poisoning for any operator
new[] which is not part of the set of replaceable global allocation
functions.
This commit adds a flag to tell clang to poison all operator new[]
cookies.
A previous review was poisoning all array cookies unconditionally, but
there is an edge case which would stop working under ASan (a custom
operator new[] saves whatever pointer it returned, and then accesses
it).
This newer revision adds a command line argument to toggle this feature.
Original revision: https://reviews.llvm.org/D41301
Compiler-rt test revision with an explanation of the edge case: https://reviews.llvm.org/D41664
Reviewers: rjmccall, kcc, rsmith
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D43013
llvm-svn: 324884
Summary:
It was deprecated and callback version and is used everywhere.
Only changes to the testing code were needed.
Reviewers: hokein, ioeric, sammccall
Reviewed By: sammccall
Subscribers: mgorny, klimek, jkorous-apple, cfe-commits
Differential Revision: https://reviews.llvm.org/D43068
llvm-svn: 324883
Add a common -trap-unreachable option, similar to the target
specific hexagon equivalent, which has been replaced. This
turns unreachable instructions into traps, which is useful for
debugging.
Differential Revision: https://reviews.llvm.org/D42965
llvm-svn: 324880
As a first step, pass '-c/--compile-only' to ptxas so that it
doesn't complain about references to external function. This
will successfully generate object files, but they won't work
at runtime because the registration routines need to adapted.
Differential Revision: https://reviews.llvm.org/D42921
llvm-svn: 324878
Summary:
These are functions like operator<<(raw_ostream&, Foo).
Previously these were only supported for messages. In the assertion
EXPECT_EQ(A, B) << C;
the local modifications would explicitly try to use raw_ostream printing for C.
However A and B would look for a std::ostream printing function, and often fall
back to gtest's default "168 byte object <00 01 FE 42 ...>".
This patch pulls out the raw_ostream support into a new header under `custom/`.
I changed the mechanism: instead of a convertible stream, we wrap the printed
value in a proxy object to allow it to be sent to a std::ostream.
I think the new way is clearer.
I also changed the policy: we prefer raw_ostream printers over std::ostream
ones. This is because the fallback printers are defined using std::ostream,
while all the raw_ostream printers should be "good".
Reviewers: ilya-biryukov, chandlerc
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D43091
llvm-svn: 324876
This test uses XML packets, but libxml is an optional dependency of
lldb, and this test fails if it is not present.
I'm leaving this enabled on mac, as thats the only platform that's
likely to have libxml always available, but ideally we should have a way
to skip this based on build configuration. I'll see if I can whip
something like that up soon, but for the time being, this unblocks the
buildbots.
llvm-svn: 324870
In case of correct using of the 'l' constraint llvm now generates valid
code; otherwise it shows an error message. Initially these triggers an
assertion.
llvm-svn: 324869
The current implementation of `getPostIncExpr` invokes `getAddExpr` for two recurrencies
and expects that it always returns it a recurrency. But this is not guaranteed to happen if we
have reached max recursion depth or refused to make SCEV simplification for other reasons.
This patch changes its implementation so that now it always returns SCEVAddRec without
relying on `getAddExpr`.
Differential Revision: https://reviews.llvm.org/D42953
llvm-svn: 324866
I don't believe we ever create an X86ISD::SUB with a 0 constant which is what the TEST handling needs. The ternary operator at the end of this code shows up as only going one way in the llvm-cov report from the bots.
llvm-svn: 324865
ISD::ADD implies individual vector element addition with no carries between elements. But for a vXi1 type that would be the same as XOR. And we already turn ISD::ADD into ISD::XOR for all vXi1 types during lowering. So the ISD::ADD pattern would never be able to match anyway.
KADD is different, it adds the elements but also propagates a carry between them. This just a way of doing an add in k-register without bitcasting to the scalar domain. There's still no way to match the pattern, but at least its not obviously wrong.
llvm-svn: 324861
Previously we just emitted this as a MOV8rm which would likely get folded during the peephole pass anyway. This just makes it explicit earlier.
The gpr-to-mask.ll test changed because the kaddb instruction has no memory form.
llvm-svn: 324860
Summary:
As noted in a discussion about testing the LLVM 6.0.0 release candidates
(with libc++) for FreeBSD, many tests turned out to fail with
"exception_ptr not yet implemented". This was because libc++ did not
choose the correct C++ ABI library, and therefore it fell back to the
`exception_fallback.ipp` header.
Since FreeBSD 10.x, we have been using libcxxrt as our C++ ABI library,
and its headers have always been installed in /usr/include/c++/v1,
together with the (system) libc++ headers. (Older versions of FreeBSD
used GNU libsupc++ by default, but these are now unsupported.)
Therefore, if we are building libc++ for FreeBSD, set:
* `LIBCXX_CXX_ABI_LIBNAME` to "libcxxrt"
* `LIBCXX_CXX_ABI_INCLUDE_PATHS` to "/usr/include/c++/v1"
by default.
Reviewers: emaste, EricWF, mclow.lists
Reviewed By: EricWF
Subscribers: mgorny, cfe-commits, krytarowski
Differential Revision: https://reviews.llvm.org/D43166
llvm-svn: 324855
Add GraphTraits definitions to the FunctionSummary and ModuleSummaryIndex classes. These GraphTraits will be used to construct find SCC's in ThinLTO analysis passes.
llvm-svn: 324854
Patch from charlieio@outlook.com
Reviewed as https://reviews.llvm.org/D42354
When the following command is used:
> clang-cl -std:c++17 -Iinclude\c++\v1 hello.cc c++.lib
An error occurred:
In file included from hello.cc:1:
In file included from include\c++\v1\iostream:38:
In file included from include\c++\v1\ios:216:
In file included from include\c++\v1\__locale:15:
In file included from include\c++\v1\string:477:
In file included from include\c++\v1\string_view:176:
In file included from include\c++\v1\__string:56:
In file included from include\c++\v1\algorithm:643:
In file included from include\c++\v1\memory:656:
include\c++\v1\new(165,29): error: redefinition of 'align_val_t'
enum class _LIBCPP_ENUM_VIS align_val_t : size_t { };
^
C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.12.25827\include\vcruntime_new.h(43,16): note:
previous definition is here
enum class align_val_t : size_t {};
^
1 error generated.
vcruntime_new.h has defined align_val_t, libcxx need hide align_val_t.
This patch fixes that error.
llvm-svn: 324853
Instead of reserving 0xF00 bytes for the fixed length portion of the CodeView
symbol name, calculate the actual length of the fixed length portion.
Differential Revision: https://reviews.llvm.org/D42125
llvm-svn: 324850
Currently NanoTime() on Darwin is unimplemented and always returns 0. Looks like there's quite a few things broken because of that (TSan periodic memory flush, ASan allocator releasing pages back to the OS). Let's fix that.
Differential Revision: https://reviews.llvm.org/D40665
llvm-svn: 324847
On Darwin, we currently use forkpty to communicate with the "atos" symbolizer. There are several problems that fork or forkpty has, e.g. that after fork, interceptors are still active and this sometimes causes crashes or hangs. This is especially problematic for TSan, which uses interceptors for OS-provided locks and mutexes, and even Libc functions use those.
This patch replaces forkpty with posix_spawn. Since posix_spawn doesn't fork (at least on Darwin), the interceptors are not a problem. Additionally, this also fixes a latent threading problem with ptsname (it's unsafe to use this function in multithreaded programs). Yet another benefit is that we'll handle post-fork failures (e.g. sandbox disallows "exec") gracefully now.
Differential Revision: https://reviews.llvm.org/D40032
llvm-svn: 324846
The related cases for (X * Y) / X were handled in rL124487.
https://rise4fun.com/Alive/6k9
The division in these tests is subsequently eliminated by existing instcombines
for 1/X.
llvm-svn: 324843
Summary:
Currently we only use min/max to help with ule/uge compares because it removes an invert of the result that would otherwise be needed. But we can also use it for ult/ugt compares if it will prevent the need for a sign bit flip needed to use pcmpgt at the cost of requiring an invert after the compare.
I also refactored the code so that the max/min code is self contained and does its own return instead of setting up a flag to manipulate the rest of the function's behavior.
Most of the test cases look ok with this. I did notice that we added instructions when one of the operands being sign flipped is a constant vector that we were able to constant fold the flip into.
I also noticed that sometimes the SSE min/max clobbers a register that is needed after the compare. This resulted in an extra move being inserted before the min/max to preserve the register. We could try to detect this and switch from min to max and change the compare operands to use the operand that gets reused in the compare.
Reviewers: spatel, RKSimon
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D42935
llvm-svn: 324842
This reverses instcombine's demanded bits' transform which always tries to clear bits in constants.
As noted in PR35792 and shown in the test diffs:
https://bugs.llvm.org/show_bug.cgi?id=35792
...we can do better in codegen by trying to form -1. The x86 sub test shows a missed opportunity.
I did investigate changing instcombine's behavior, but it would be more work to change
canonicalization in IR. Clearing bits / shrinking constants can allow killing instructions,
so we'd have to figure out how to not regress those cases.
Differential Revision: https://reviews.llvm.org/D42986
llvm-svn: 324839
Krzysztof Parzyszek