This patch upstreams support for BFloat Matrix Multiplication Intrinsics
and Code Generation from __bf16 to AArch64. This includes IR intrinsics. Unittests are
provided as needed. AArch32 Intrinsics + CodeGen will come after this
patch.
This patch is part of a series implementing the Bfloat16 extension of
the
Armv8.6-a architecture, as detailed here:
https://community.arm.com/developer/ip-products/processors/b/processors-ip-blog/posts/arm-architecture-developments-armv8-6-a
The bfloat type, and its properties are specified in the Arm
Architecture
Reference Manual:
https://developer.arm.com/docs/ddi0487/latest/arm-architecture-reference-manual-armv8-for-armv8-a-architecture-profile
The following people contributed to this patch:
Luke Geeson
- Momchil Velikov
- Mikhail Maltsev
- Luke Cheeseman
Reviewers: SjoerdMeijer, t.p.northover, sdesmalen, labrinea, miyuki,
stuij
Reviewed By: miyuki, stuij
Subscribers: kristof.beyls, hiraditya, danielkiss, cfe-commits,
llvm-commits, miyuki, chill, pbarrio, stuij
Tags: #clang, #llvm
Differential Revision: https://reviews.llvm.org/D80752
Change-Id: I174f0fd0f600d04e3799b06a7da88973c6c0703f
Summary:
EOF macro token coming from a PCH file on macOS while marked as literal,
doesn't contain any literal data. This causes crash on every project
using PCHs.
This commit doesn't resolve the problem with PCH (maybe it was
designed like this for a purpose) or with `tryExpandAsInteger`, but
rather simply shoots off a crash itself.
Differential Revision: https://reviews.llvm.org/D81916
Previously we only printed a symbol value when it has a non-empty name
or non-zero value.
This patch changes the behavior. Now we only omit a symbols value when
a relocation does not reference a symbol (i.e. symbol index == 0).
Seems it is what GNU readelf does, looking on its output.
Differential revision: https://reviews.llvm.org/D81842
Currently, llvm-readelf crashes when there is a STT_SECTION symbol for the null section
and this symbol is used in a relocation.
Differential revision: https://reviews.llvm.org/D81840
An instruction like this will need to allocate some stack space for the
last parameter:
%x = call addrspace(1) i16 @bar(i64 undef, i64 undef, i16 undef, i16 0)
This worked fine when passing an actual value (in this case 0). However,
when passing undef, no value was pushed to the stack and therefore no
push instructions were created. This caused an unbalanced stack leading
to interesting results.
This commit fixes that by replacing the push logic with a regular stack
adjustment and stack-relative load/stores. This is less efficient but at
least it correctly compiles the code.
I can think of a few improvements in the future:
* The stack should have been adjusted in the function prologue when
there are no allocas in the function.
* Many (if not most) stack adjustments can be replaced by
pushing/popping the values directly. Exactly like the previous code
attempted but didn't do correctly.
* Small stack adjustments can be done more efficiently with a few
push/pop instructions (pushing/popping bogus values), both for code
size and for speed.
All in all, as long as there are no allocas in the function I think that
it is almost always more efficient to emit regular push/pop
instructions. This is however left for future optimizations.
Differential Revision: https://reviews.llvm.org/D78581
This patch fixes a bug in stack save/restore code. Because the frame
pointer was saved/restored manually (not by marking it as clobbered) the
StackSize variable was not updated accordingly. Most code still worked,
but code that tried to load a parameter passed on the stack did not.
This commit fixes this by marking the frame pointer as a
callee-clobbered register. This will let it be saved without any effort
in prolog/epilog code and will make sure the correct address is
calculated for loading parameters that are passed on the stack.
This approach is used by most other targets (such as X86, AArch64 and
RISC-V).
Differential Revision: https://reviews.llvm.org/D78579
These code patterns attempt to call isVMOVModifiedImm on a splat of i1
values, leading to an unreachable being hit. I've guarded the call on a
more specific set of sizes, as i1 vectors are legal under MVE.
Differential Revision: https://reviews.llvm.org/D81860
Summary:
this reduces significantly the number of assumes generated without aftecting too much
the information that is preserved. this improves the compile-time cost
of enable-knowledge-retention significantly.
Reviewers: jdoerfert, sstefan1
Reviewed By: jdoerfert
Subscribers: hiraditya, asbirlea, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D79650
Fix a crash in clangd caused by an (admittidly incorrect) Remark diagnositic being emitted from readability-else-after-return.
This crash doesn't occur in clang-tidy so there are no tests there for this.
Reviewed By: hokein
Differential Revision: https://reviews.llvm.org/D81785
GCC 7 was reporting "enumeral and non-enumeral type in conditional expression"
as a warning.
The code casts an instruction opcode enum to unsigned implicitly, in
line with intentions; so this commit silences the warning by making the
cast to unsigned explicit.
We are planning to add the bf16 value type in the HPR register class
and this will make the codegen patterns ambiguous.
Differential Revision: https://reviews.llvm.org/D81505
Summary:
It makes it much harder to use from other modules when one of the
parameters is an argparse Namespace. This commit makes it easier
to use CmpRuns programmatically.
Differential Revision: https://reviews.llvm.org/D81566
Summary:
JSON format is a bit more verbose and easier to reason about
and extend. For this reason, before extending SATestBuild
functionality it is better to refactor the part of how we
configure the whole system.
Differential Revision: https://reviews.llvm.org/D81563
Summary:
The field decl (in the testcase) was still valid, which results in a
valid RecordDecl, it led to crash when performing struct layout,
and computing struct size etc.
Reviewers: sammccall
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D81913
Summary:
Fix in getDescAsStringRef to not use a reference to a
temporary type and added a testcase.
Reviewers: arsenm, saiislam, scott.linder
Reviewed By: scott.linder
Subscribers: wdng, mgorny, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81653
Summary: Couple of tests to showcase what will be done and what to expect with ICV tracking.
Reviewers: jdoerfert, JonChesterfield
Subscribers: yaxunl, guansong, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81114
Summary:
Before unwinding the stack, `__asan_handle_no_return` is supposed to
unpoison the entire stack - that is, remove the entries in the shadow
memory corresponding to stack (e.g. redzone markers around variables).
This does not work correctly if `__asan_handle_no_return` is called from
the alternate stack used in signal handlers, because the stack top is
read from a cache, which yields the default stack top instead of the
signal alternate stack top.
It is also possible to jump between the default stack and the signal
alternate stack. Therefore, __asan_handle_no_return needs to unpoison
both.
Reviewers: vitalybuka, kubamracek, kcc, eugenis
Reviewed By: vitalybuka
Subscribers: phosek, #sanitizers
Tags: #sanitizers
Differential Revision: https://reviews.llvm.org/D76986
Summary: This adds a customization point to support unpoisoning of signal alternate stacks on POSIX.
Reviewers: vitalybuka
Reviewed By: vitalybuka
Subscribers: #sanitizers
Tags: #sanitizers
Differential Revision: https://reviews.llvm.org/D81577
Note that .eh_frame sections are generated in the 32-bit format even
when debug sections are 64-bit, for compatibility reasons. They use
relative references between entries, so they hardly benefit from the
64-bit format.
Differential Revision: https://reviews.llvm.org/D81149
DW_FORM_sec_offset was introduced in DWARFv4, so, for 64-bit DWARFv3,
DW_FORM_data8 should be used instead.
Differential Revision: https://reviews.llvm.org/D81148
The patch enables producing DWARF64 compilation units and fixes
generating references to .debug_abbrev and .debug_line sections.
A similar change for .debug_ranges/.debug_rnglists will be added
in a forthcoming patch.
Differential Revision: https://reviews.llvm.org/D81145
The patch adds an option `--dwarf64` to instruct a tool to generate
debug information in the 64-bit DWARF format. There is no real
implementation yet, only a few compatibility checks.
Differential Revision: https://reviews.llvm.org/D81143
Luke Geeson
Luke Geeson
Michael Liao
Valeriy Savchenko
Nathan James
Georgii Rymar
David Truby
Tyker
Ayke van Laethem
Alexander Belyaev
Ilya Leoshkevich
David Green
sstefan1
Simon Pilgrim
Kristof Beyls
Alexandros Lamprineas
LLVM GN Syncbot
Haojian Wu
Dineshkumar Bhaskaran
Jay Foad
Vitaly Buka
Andi-Bogdan Postelnicu
Kadir Cetinkaya
Kirill Bobyrev
Hans Wennborg
Robert Schneider
Igor Kudrin