When using `-mprefer-vector-width=128` with `-march=sandybridge` copying
3 cache lines in one go (192B) gets converted into `rep;movsb` which
translate into a 60% hit in performance.
Consecutive calls to `__builtin_memcpy_inline` (implementation behind
`builtin::Memcpy::block_offset`) are not coalesced by the compiler and
so calling it three times in a row generates the desired assembly. It
only differs in the interleaving of the loads and stores and does not
affect performance.
This is needed to reland
https://github.com/llvm/llvm-project/pull/108939.
So far, isValidAddrSpaceCast only allows casts to the flat address
space and between the constant(32) address spaces. It does not allow
casting between the global and constant address spaces, even though they
alias. That affects, e.g., the lowering of memmoves from the constant to
the global address space in LowerMemIntrinsics, since that requires
aliasing address spaces to be castable.
This patch relaxes isValidAddrSpaceCast and allows such casts. It also
includes a memmove test that would crash with the previous
implementation because the memmove IR lowering would not be
applicable for the move from constant AS to global AS.
Fix missing extern templates for llvm::Registry use in other projects of
llvm
Windows doesn't implicitly import and merge exported symbols across
shared libraries
like Linux does so we need to explicitly export/import each
instantiation of llvm::Registry.
Updated LLVM_INSTANTIATE_REGISTRY to just be a full explicit template
instantiation.
This is part of the work to enable LLVM_BUILD_LLVM_DYLIB and LLVM
plugins on window.
**Change relanded after feedback on failures and improvements to the
check of the addend. Original PR #111970**
Changes from original patch:
- The value that is being checked has changed, it is now correctly
checking any Addend for the instruction, rather than the Value. The
addend is kept within the Target data structure from my investigation.
- Removed changes to the following tests due to the original behaviour
being correct, and my original patch causing unexpected errors
- llvm/test/MC/ARM/Windows/mov32t-range.s
- llvm/test/MC/MachO/ARM/thumb2-movw-fixup.s
As per the ARM ABI, the MOVT and MOVW instructions should have addends
that fall within a 16bit signed range. LLVM does not check this so it is
possible to use addends that are beyond the accepted range. These
addends are silently truncated.
A new check is added to ensure the addend falls within the expected
range, rejecting an addend that falls outside with an error.
Information relating to the ABI requirements can be found here:
https://github.com/ARM-software/abi-aa/blob/main/aaelf32/aaelf32.rst#addends-and-pc-bias-compensation
This PR fixes multiple bugs in `DuplicateFunctionElimination`.
- Prevents elimination of function declarations.
- Updates all symbol uses to reference unique function representatives.
Fixes#93483.
This patch is inspired by @Snowy1803 excellent work in swift and the
patch: https://github.com/swiftlang/swift/pull/73334/files
Add an instrumentation pass to llvm to collect dropped debug information
variable statistics for every Function-level and Module-level IR pass.
This patch creates adds the class DroppedVariableStats which iterates
over every DbgRecord in a function or module before and after an
optimization pass and counts the number of variables who's debug
information has been dropped due to that pass, then prints that output
to stdout in a csv format.
I ran this patch on optdriver.cpp can see:
Pass Name, Dropped Variables
'InstCombinePass', 1
'SimplifyCFGPass', 6
'JumpThreadingPass', 25
Add the llvm-canon tool. Description from the [original
PR](https://reviews.llvm.org/D66029#change-wZv3yOpDdxIu):
> Added a new llvm-canon tool which aims to transform LLVM Modules into
a canonical form by reordering and renaming instructions while
preserving the same semantics. This tool makes it easier to spot
semantic differences while diffing two modules which have undergone
different transformation passes.
The current version of this tool can:
- Reorder instructions within a function.
- Rename instructions based on the operands.
- Sort commutative operands.
This code was originally written by @michalpaszkowski and [submitted to
mainline
LLVM](14d358537f).
However, it was quickly
[reverted](335de55fa3)
to do BuildBot errors.
Michal presented his version of the tool in [LLVM-Canon: Shooting for
Clear Diffs](https://www.youtube.com/watch?v=c9WMijSOEUg).
@AidanGoldfarb and I ported the code to the new pass manager, added more
tests, and fixed some bugs related to PHI nodes that may have been the
root cause of the BuildBot errors that caused the patch to be reverted.
Additionally, we rewrote the implementation of instruction reordering to
fix cases where the original algorithm would break use-def chains.
Note that this is @AidanGoldfarb and I's first time submitting to LLVM.
Please liberally critique the PR!
CC @plotfi for initial review.
---------
Co-authored-by: Aidan <aidan.goldfarb@mail.mcgill.ca>
For Hand-In-Hand we need float to string to support a wider set of
architectures, specifically the ones that libc++ supports. This includes
powerpc, which apparently uses "double double" as its long double type.
Since Hand-In-Hand isn't currently using long double, this just opts
them out.
With this change and appropriate linker changes
(https://r.android.com/3236256)
AOSP boots with memtag-global throughout the platform.
Without this change, we would sometimes generate PC-relative references
to tagged globals, which then do not have the proper tag.
Fix a number of dependencies issue to build mlir-linalg-ods-yaml-gen
host binary which make a cross-build using the Make generator fail.
Namely:
- do not use binary path for the custom target created when
LLVM_USE_HOST_TOOLS is true;
- use target name instead of name of variable holding the target name
for add_custom_target and set_target_properties in setup_host_tool();
- remove dependency on target defined in different directory in
add_linalg_ods_yaml_gen() since add_custom_target DEPENDS can only be
used on "files and outputs of custom commands created with
add_custom_command() command calls in the same directory";
- remove unneeded dependency on ${MLIR_LINALG_ODS_YAML_GEN_EXE}, the
target dependency will ensure the binary will be built.
Note that we keep using ${MLIR_LINALG_ODS_YAML_GEN_EXE} in the COMMAND
rather than use ${MLIR_LINALG_ODS_YAML_GEN_TARGET} because when
LLVM_NATIVE_TOOL_DIR is used the latter is an empty string.
Testing-wise, all three codepaths in get_host_tool_path() were tested
with both GNU Make and Ninja generators:
- cross-compiling with LLVM_NATIVE_TOOL_DIR checks the if path;
- cross-compiling without LLVM_NATIVE_TOOL_DIR checks the elseif path;
- native build without LLVM_NATIVE_TOOL_DIR checks the else path.
Implements std::from_chars for float and double.
The implementation uses LLVM-libc to do the real parsing. Since this is
the first time libc++
uses LLVM-libc there is a bit of additional infrastructure code. The
patch is based on the
[RFC] Project Hand In Hand (LLVM-libc/libc++ code sharing)
https://discourse.llvm.org/t/rfc-project-hand-in-hand-llvm-libc-libc-code-sharing/77701
Currently all the headings marked as `#` show up
as a top-level entry in the `Developing LLDB`
toctree. This patch marks these as `##` so only
`Adding Programming Language Support` is displayed
in the table of contents.
This improves performance of doing a `bazel test @llvm-project//...` a
lot because previously every lit test would have some symlink tree
configured for it.
When an operation has no properties, no property struct is emitted. To avoid a compilation error, we should also skip emitting `setPropertiesFromParsedAttr`, `parseProperties` and `printProperties` in such cases.
Compilation error:
```
error: ‘Properties’ has not been declared
static ::llvm::LogicalResult setPropertiesFromParsedAttr(Properties &prop, ::mlir::Attribute attr, ::llvm::function_ref<::mlir::InFlightDiagnostic()> emitError);
```
If we split these features in the compiler (see relevant pull request
https://github.com/llvm/llvm-project/pull/109299), we would only be able
to hand-write a 'memtag2' version using inline assembly since the
compiler cannot generate the instructions that become available with
FEAT_MTE2. However these instructions only work at Exception Level 1, so
they would be unusable since FMV is a user space facility. I am
therefore unifying them.
Approved in ACLE as https://github.com/ARM-software/acle/pull/351
When adding new SEH pseudo instructions in #110024 I noticed that some
of the tests were changing their output since these new instructions
were counting towards thresholds for branching versus folding decisions.
These instructions do not result in real machine instructions being
emitted, so they should be marked as meta instructions.
This is a re-do of #110889 as we hit an issue where some of the SEH
pseudo instructions in the prolog were being duplicated, which resulted
errors being raised as the CodeView generator was seeing prolog
directives after an end-prolog directive:
<https://github.com/llvm/llvm-project/pull/110889#issuecomment-2393405613>.
The fix for this is to mark the prolog related SEH pseudo instructions
as being non-duplicatable.
With bazel 8.x these are strongly encouraged, and this disambiguates
which version of these rules we get for older versions. Specifically the
native.py_test was using the wrong version of py_test.
Enables initial non-power-of-2 support (but still requires number of
elements, forming whole registers) for reductions.
Enables extra vectorization for
MultiSource/Benchmarks/7zip/7zip-benchmark, CINT2006/464.h264ref and
CFP2017rate/526.blender_r (checked for SSE2)
Reviewers: RKSimon
Reviewed By: RKSimon
Pull Request: https://github.com/llvm/llvm-project/pull/112361
Change the spill weight calculations for `optsize` functions to remove
the block frequency multiplier. For those functions, we do not want to
consider the runtime cost of spilling, only the codesize cost.
I built a large app with the basic and greedy (default) register
allocator enabled.
| Regalloc Type | Uncompressed Size Delta | Compressed Size Delta |
| - | - | - |
| Basic | -303.8 KiB (-0.23%) | -232.0 KiB (-0.39%) |
| Greedy | 159.1 KiB (0.12%) | 130.1 KiB (0.22%) |
Since I only saw a size win with the basic register allocator, I decided
to only change the behavior for that type.
