Summary:
This patch teaches ConstantFolding to constant fold
both scalar and vector variants of llvm.smul.fix and
llvm.smul.fix.sat.
As described in the LangRef rounding is unspecified for
these instrinsics. If the result cannot be represented
exactly the default behavior in ConstantFolding is to
round down towards negative infinity. If a target has a
preferred rounding that is different some kind of target
hook would be needed (same strategy as used by the
SelectionDAG legalizer).
Reviewers: nikic, leonardchan, RKSimon
Reviewed By: leonardchan
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D63385
llvm-svn: 363811
This patch splits ConstantFoldScalarCall into several
functions.
Benefits:
- Reduces indentation levels and avoids long if-statements.
- Makes it easier to add support for > 3 operands.
llvm-svn: 363810
Summary:
This patch strengthens the tests introduced in D63009 by:
- adding new test for default device ID.
- modifying existing tests to pass device ID local variable to the task allocation function.
Reviewers: ABataev, Hahnfeld, caomhin, jdoerfert
Reviewed By: ABataev
Subscribers: guansong, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D63454
llvm-svn: 363809
Summary:
The problems with __syncthreads() were fixed in clang >= 9.0 and the
original __syncthreads() can be used instead of the ptx instruction.
Reviewers: grokos
Subscribers: guansong, jdoerfert, openmp-commits, kkwli0, caomhin
Tags: #openmp
Differential Revision: https://reviews.llvm.org/D63515
llvm-svn: 363807
Vector load/store instructions support an optional alignment field
that the compiler can use to provide known alignment info to the
hardware. If the field is used (and the information is correct),
the hardware may be able (on some models) to perform faster memory
accesses than otherwise.
This patch adds support for alignment hints in the assembler and
disassembler, and fills in known alignment during codegen.
llvm-svn: 363806
Summary:
The first use of this is a code tweak to expand macro calls.
Will later be used to build syntax trees.
The memory overhead is small as we only store tokens of the main file.
Reviewers: sammccall
Reviewed By: sammccall
Subscribers: mgorny, MaskRay, jkorous, arphaman, kadircet, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D62956
llvm-svn: 363803
The test only checks the existence of the `Types CU list` line.
Unfortunately I can't make a better test because
{gcc,clang} -fuse-ld={lld,gold} --gdb-index do not give me a non-empty types CU list.
Reviewed By: ikudrin
Differential Revision: https://reviews.llvm.org/D63537
llvm-svn: 363800
Using the -fdeclare-opencl-builtins option will require a way to
predefine types and macros such as `int4`, `CLK_GLOBAL_MEM_FENCE`,
etc. Move these out of opencl-c.h into opencl-c-base.h such that the
latter can be shared by -fdeclare-opencl-builtins and
-finclude-default-header.
This changes the behaviour of -finclude-default-header when
-fdeclare-opencl-builtins is specified: instead of including the full
header, it will include the header with only the base definitions.
Differential revision: https://reviews.llvm.org/D63256
llvm-svn: 363794
Use matchBinaryPredicate instead of isConstOrConstSplat to let us handle non-uniform shift cases.
This requires us to tweak matchBinaryPredicate to allow it to (optionally) handle constants with different type widths.
llvm-svn: 363792
This introduced MMX instructions in code that wasn't previously using
them, breaking programs using 64-bit vectors and x87 floating-point in
the same application. See discussion on the code review for more
details.
> According to System V i386 ABI: the __m64 type paramater and return
> value are passed by MMX registers. But current implementation treats
> __m64 as i64 which results in parameter passing by stack and returning
> by EDX and EAX.
>
> This patch fixes the bug (https://bugs.llvm.org/show_bug.cgi?id=41029)
> for Linux and NetBSD.
>
> Patch by Wei Xiao (wxiao3)
>
> Differential Revision: https://reviews.llvm.org/D59744
llvm-svn: 363790
As discussed in the review for D62952, this patch pre-normalizes the
reference expected output sarif files by removing lines containing
fields for which we expect differences that should be ignored.
llvm-svn: 363788
Summary:
Bug: https://bugs.llvm.org/show_bug.cgi?id=39024
The bug reports that a vectorized loop is stepped through 4 times and each step through the loop seemed to show a different path. I found two problems here:
A) An incorrect line number on a preheader block (for.body.preheader) instruction causes a step into the loop before it begins.
B) Instructions in the middle block have different line numbers which give the impression of another iteration.
In this patch I give all of the middle block instructions the line number of the scalar loop latch terminator branch. This seems to provide the smoothest debugging experience because the vectorized loops will always end on this line before dropping into the scalar loop. To solve problem A I have altered llvm::SplitBlockPredecessors to accommodate loop header blocks.
I have set up a separate review D61933 for a fix which is required for this patch.
Reviewers: samsonov, vsk, aprantl, probinson, anemet, hfinkel, jmorse
Reviewed By: hfinkel, jmorse
Subscribers: jmorse, javed.absar, eraman, kcc, bjope, jmellorcrummey, hfinkel, gbedwell, hiraditya, zzheng, llvm-commits
Tags: #llvm, #debug-info
Differential Revision: https://reviews.llvm.org/D60831
> llvm-svn: 363046
llvm-svn: 363786
Summary:
The test case does an (out of bounds) load from a global constant with
type <3 x float>. InstSimplify tried to turn this into an integer load
of the whole alloc size of the vector, which is 128 bits due to
alignment padding, and then bitcast this to <3 x vector> which failed
an assertion due to the type size mismatch.
The fix is to do an integer load of the normal size of the vector, with
no alignment padding.
Reviewers: tpr, arsenm, majnemer, dstuttard
Reviewed By: arsenm
Subscribers: hfinkel, wdng, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D63375
llvm-svn: 363784
This patch allows immediates (and CSR alias immediates) which start with
a tilde token or an exclaim (!) token to be parsed as intended.
Differential Revision: https://reviews.llvm.org/D57320
llvm-svn: 363783
Since the parser attempts to parse an operand as a register with
parentheses before parsing it as an immediate, immediates in
parentheses should not be parsed by parseRegister. However in the case
where the immediate does not start with an identifier, the LParen is not
unlexed and so the RParen causes an unexpected token error.
This patch adds the missing UnLex, and modifies the existing UnLex to
not use a buffered token, as it should always be unlexing an LParen.
Differential Revision: https://reviews.llvm.org/D57319
llvm-svn: 363782
This caused Chromium's clang package to stop building, see comment on
https://reviews.llvm.org/D61242 for details.
> Summary:
> The main problem here is that `-*-version_min=` was not being passed to
> the compiler when building test cases. This can cause problems when
> testing on devices running older OSs because Clang would previously
> assume the minimum deployment target is the the latest OS in the SDK
> which could be much newer than what the device is running.
>
> Previously the generated value looked like this:
>
> `-arch arm64 -isysroot
> <path_to_xcode>/Contents/Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS12.1.sdk`
>
> With this change it now looks like:
>
> `-arch arm64 -stdlib=libc++ -miphoneos-version-min=8.0 -isysroot
> <path_to_xcode>/Contents/Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS12.1.sdk`
>
> This mirrors the setting of `config.target_cflags` on macOS.
>
> This change is made for ASan, LibFuzzer, TSan, and UBSan.
>
> To implement this a new `get_test_cflags_for_apple_platform()` function
> has been added that when given an Apple platform name and architecture
> returns a string containing the C compiler flags to use when building
> tests. This also calls a new helper function `is_valid_apple_platform()`
> that validates Apple platform names.
>
> This is the second attempt at landing the patch. The first attempt (r359305)
> had to be reverted (r359327) due to a buildbot failure. The problem was
> that calling `get_test_cflags_for_apple_platform()` can trigger a CMake
> error if the provided architecture is not supported by the current
> CMake configuration. Previously, this could be triggered by passing
> `-DCOMPILER_RT_ENABLE_IOS=OFF` to CMake. The root cause is that we were
> generating test configurations for a list of architectures without
> checking if the relevant Sanitizer actually supported that architecture.
> We now intersect the list of architectures for an Apple platform
> with `<SANITIZER>_SUPPORTED_ARCH` (where `<SANITIZER>` is a Sanitizer
> name) to iterate through the correct list of architectures.
>
> rdar://problem/50124489
>
> Reviewers: kubamracek, yln, vsk, juliehockett, phosek
>
> Subscribers: mgorny, javed.absar, kristof.beyls, #sanitizers, llvm-commits
>
> Tags: #llvm, #sanitizers
>
> Differential Revision: https://reviews.llvm.org/D61242
llvm-svn: 363779
A recent build of Solaris 11.5 Beta (st_047) gained madvise(MADV_DONTDUMP)
support for Linux compatibility. This broke the compiler-rt build:
/vol/llvm/src/llvm/dist/projects/compiler-rt/lib/sanitizer_comm/sanitizer_posix_libcdep.cc: In function ‘bool __sanitizer::DontDumpShadowMemory(__sanitizer::uptr, __sanitizer::uptr)’:
/vol/llvm/src/llvm/dist/projects/compiler-rt/lib/sanitizer_common/sanitizer_posix_libcdep.cc:81:18: error: invalid conversion from ‘void*’ to ‘caddr_t’ {aka ‘char*’} [-fpermissive]
81 | return madvise((void *)addr, length, MADV_DONTDUMP) == 0;
| ^~~~~~~~~~~~
| |
| void*
In file included from
/vol/llvm/src/llvm/dist/projects/compiler-rt/lib/sanitizer_common/sanitizer_posix_libcdep.cc:32:
/usr/include/sys/mman.h:231:20: note: initializing argument 1 of ‘int
madvise(caddr_t, std::size_t, int)’
231 | extern int madvise(caddr_t, size_t, int);
| ^~~~~~~
The obvious fix is to use the same solution that has already been used a
couple of lines earlier:
// In the default Solaris compilation environment, madvise() is declared
// to take a caddr_t arg; casting it to void * results in an invalid
// conversion error, so use char * instead.
This allowed the compiler-rt build to finish and was tested successfully on
i386-pc-solaris2.11 and x86_64-pc-linux-gnu.
Differential Revision: https://reviews.llvm.org/D62892
llvm-svn: 363778
Summary:
llvm.x86.sse.stmxcsr only writes to memory.
llvm.x86.sse.ldmxcsr only reads from memory, and might generate an FPE.
Reviewers: craig.topper, RKSimon
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62896
llvm-svn: 363773
on some systems this test fails because the two methods it uses to
cross-reference the data don't match in the case of the vdso module. The
"read from /proc/%pid/maps" method returns "[vdso]", while the method
which reads it from the linker rendezvous structures returns
"linux-vdso.so.1". Neither of the two names match any actual file.
This restricts the test to only consider the libraries that we ourselves
have added to the test, minimizing the impact of system dependencies
that we cannot control.
llvm-svn: 363772
This patch adds lowering for global TLS addresses for the TLS models of
InitialExec, GlobalDynamic, LocalExec and LocalDynamic.
LocalExec support required using a 4-operand add instruction, which uses
the fourth operand to express a relocation on the symbol. The necessary
fixup is emitted when the instruction is emitted.
Differential Revision: https://reviews.llvm.org/D55305
llvm-svn: 363771
Turns out this worked on my machine because I still had VS2017 installed, but
it didn't actually work in general.
Since the extension is unmaintained and MS is doing their own LLVM toolset
integration for VS2019, let's just revert.
llvm-svn: 363768
Summary:
This patch makes the DIERef class always valid by default constructor
and operator bool. This allows one to express the validity of a DIERef
in the type system. Places which are working with potentially-invalid
DIERefs have been updated to use Optional<DIERef> instead.
The constructor taking a DWARFFormValue was not needed, as all places
which were constructing a DIERef this way were immediately converting it
into a DWARFDIE or a user_id. This can be done without constructing an
intermediate DIERef.
Reviewers: JDevlieghere, clayborg, aprantl
Subscribers: arphaman, lldb-commits
Differential Revision: https://reviews.llvm.org/D63399
llvm-svn: 363767
Simon Pilgrim