Recent changes in #174746 to use GISel for optnone functions broke this.
Now at O3 -aarch64-enable-global-isel-at-O=-1 is having the opposite
affect of actually enabling GISel instead of SDAG and at O0 FastISel is
no longer used. I've added a check for if this is disabled.
Currently, when SDAG is run on AArch64 and an `optnone` function is
encountered, the selector is chosen as FastISel. AArch64 makes use of
GlobalISel at O0 and this patch aims to align `optnone` with this
functionality.
A flag is exposed to enable this functionality for a given backend but,
as AArch64 is currently the only backend I could find using GlobalISel
at O0 this is the only one with it implemented. This flag is set when
the target supports GlobalISel & GlobalISel hasn't been forced by the
user, the target machine or by being at an optimisation level lower than
`EnableGlobalISelAtO`.
If this happens, the GlobalISel passes are included as shown in
`llvm/test/CodeGen/AArch64/O3-pipeline.ll` and skipped by IRTranslator
for functions not marked as `optnone`.
In updating the tests based on this functionality, I found some unused
check lines or run lines that mixed SDAG with GlobalISel pass names
which have been fixed.
---------
Co-authored-by: Matt Arsenault <arsenm2@gmail.com>
When SubRegister Liveness Tracking (SRLT) is enabled, this pass adds
extra implicit-def's to instructions that define the low N bits of a
GPR/FPR register to represent that the top bits are written, because all
AArch64 instructions that write the low bits of a GPR/FPR also
implicitly zero the top bits.
These semantics are originally represented in the MIR using
`SUBREG_TO_REG`, but during register coalescing this information is lost
and when rewriting virtual -> physical registers the implicit-defs are
not added to represent the the top bits are written.
There have been several attempts to fix this in the coalescer (#168353),
but each iteration has exposed new bugs and the patch had to be
reverted. Additionally, the concept of adding 'implicit-def' of a
virtual register during the register allocation process is particularly
fragile and many places don't expect it (for example in
`X86::commuteInstructionImpl` the code only looks at specific operands
and does not consider implicit-defs. Similar in
`SplitEditor::addDeadDef` where it traverses operand 'defs' rather than
'all_defs').
We want a temporary solution that doesn't impact other targets and is
simpler and less intrusive than the patch proposed for the register
coalescer so that we can enable SRLT to make better use of SVE/SME
multi-vector instructions while we work on a more permanent solution
that requires rewriting a large part of the AArch64 instructions (32-bit
and NEON).
The previous SelectionDAG lowering is still available via
`-aarch64-new-sme-abi=false` (this will stay around until at least LLVM
23).
In tests that contained `CHECK-NEWLOWERING` the checks have been updated
so:
* `CHECK-NEWLOWERING` -> `CHECK` (the new default)
* `CHECK` -> `CHECK-SDAG` (the old SelectionDAG lowering)
But otherwise, the check lines have not changed.
Tests that were not explicitly checking the SME lowering have been
updated to match the new default lowering.
Those tests are:
* llvm/test/CodeGen/AArch64/O0-pipeline.ll
* llvm/test/CodeGen/AArch64/O3-pipeline.ll
* llvm/test/CodeGen/AArch64/sme-disable-gisel-fisel.ll
* llvm/test/CodeGen/AArch64/sme-framelower-use-bp.ll
* llvm/test/CodeGen/AArch64/stack-hazard.ll
In certain situations, especially with zero phi operands propagated after tail
duplications, we can end up with CBZ/CBNZ/TBZ/TBNZ with a zero register. It
can can be introduced late in the pipeline.
This patch adds a basic pass to fold them away to either a direct branch or
removing the instruction entirely. It runs quite late n the pipeline, so doesnt
fit into any of the existing passes. It only needs to look at the terminators
to each BB, so the new pass should have a limited in compile-time impact.
This patch adds a step to the MachineSMEABIPass that propagates desired
ZA states.
This aims to pick better ZA states for edge bundles, as when many (or
all) blocks in a bundle do not have a preferred ZA state, the ZA state
assigned to a bundle can be less than ideal.
An important case is nested loops, where only the inner loop has a
preferred ZA state. Here we'd like to propagate the ZA state from the
inner loop to the outer loops (to avoid saves/restores in any loop).
GlobalMerge has been enabled for minsize for a while, this patch enables
it more generally. In my testing it did not affect performance very
much, especially with the linker relaxations we already perform, but
should help reduce code size a little.
Clang and other frontends generally need the LLVM data layout string in
order to generate LLVM IR modules for LLVM. MLIR clients often need it
as well, since MLIR users often lower to LLVM IR.
Before this change, the LLVM datalayout string was computed in the
LLVM${TGT}CodeGen library in the relevant TargetMachine subclass.
However, none of the logic for computing the data layout string requires
any details of code generation. Clients who want to avoid duplicating
this information were forced to link in LLVMCodeGen and all registered
targets, leading to bloated binaries. This happened in PR #145899,
which measurably increased binary size for some of our users.
By moving this information to the TargetParser library, we
can delete the duplicate datalayout strings in Clang, and retain the
ability to generate IR for unregistered targets.
This is intended to be a very mechanical LLVM-only change, but there is
an immediately obvious follow-up to clang, which will be prepared
separately.
The vast majority of data layouts are computable with two inputs: the
triple and the "ABI name". There is only one exception, NVPTX, which has
a cl::opt to enable short device pointers. I invented a "shortptr" ABI
name to pass this option through the target independent interface.
Everything else fits. Mips is a bit awkward because it uses a special
MipsABIInfo abstraction, which includes members with codegen-like
concepts like ABI physical registers that can't live in TargetParser. I
think the string logic of looking for "n32" "n64" etc is reasonable to
duplicate. We have plenty of other minor duplication to preserve
layering.
---------
Co-authored-by: Matt Arsenault <arsenm2@gmail.com>
Co-authored-by: Sergei Barannikov <barannikov88@gmail.com>
## Short Summary
This patch adds a new pass `aarch64-machine-sme-abi` to handle the ABI
for ZA state (e.g., lazy saves and agnostic ZA functions). This is
currently not enabled by default (but aims to be by LLVM 22). The goal
is for this new pass to more optimally place ZA saves/restores and to
work with exception handling.
## Long Description
This patch reimplements management of ZA state for functions with
private and shared ZA state. Agnostic ZA functions will be handled in a
later patch. For now, this is under the flag `-aarch64-new-sme-abi`,
however, we intend for this to replace the current SelectionDAG
implementation once complete.
The approach taken here is to mark instructions as needing ZA to be in a
specific ("ACTIVE" or "LOCAL_SAVED"). Machine instructions implicitly
defining or using ZA registers (such as $zt0 or $zab0) require the
"ACTIVE" state. Function calls may need the "LOCAL_SAVED" or "ACTIVE"
state depending on the callee (having shared or private ZA).
We already add ZA register uses/definitions to machine instructions, so
no extra work is needed to mark these.
Calls need to be marked by glueing Arch64ISD::INOUT_ZA_USE or
Arch64ISD::REQUIRES_ZA_SAVE to the CALLSEQ_START.
These markers are then used by the MachineSMEABIPass to find
instructions where there is a transition between required ZA states.
These are the points we need to insert code to set up or restore a ZA
save (or initialize ZA).
To handle control flow between blocks (which may have different ZA state
requirements), we bundle the incoming and outgoing edges of blocks.
Bundles are formed by assigning each block an incoming and outgoing
bundle (initially, all blocks have their own two bundles). Bundles are
then combined by joining the outgoing bundle of a block with the
incoming bundle of all successors.
These bundles are then assigned a ZA state based on the blocks that
participate in the bundle. Blocks whose incoming edges are in a bundle
"vote" for a ZA state that matches the state required at the first
instruction in the block, and likewise, blocks whose outgoing edges are
in a bundle vote for the ZA state that matches the last instruction in
the block. The ZA state with the most votes is used, which aims to
minimize the number of state transitions.
Remove Linker Optimization Hints (LOHs) from outlining candidates
instead of simply preventing outlining if LOH labels are found in the
candidate. This will improve the effectiveness of the machine outliner
when LOHs are enabled (which is the default).
In
https://discourse.llvm.org/t/loh-conflicting-with-machineoutliner/83279/1
it was observed that the machine outliner is much more effective when
LOHs are disabled. Rather than completely disabling LOH, this PR aims to
keep LOH in most places and removing them from outlined functions where
it could be illegal. Note that we are conservatively removing all LOHs
from outlined functions for simplicity, but I believe we could retain
LOHs that are in the intersection of all candidates.
It should be ok to remove these LOHs since these blocks are being
outlined anyway, which will harm performance much more than the gain
from keeping the LOHs.
## Purpose
This patch is one in a series of code-mods that annotate LLVM’s public
interface for export. This patch annotates the `llvm/Target` library.
These annotations currently have no meaningful impact on the LLVM build;
however, they are a prerequisite to support an LLVM Windows DLL (shared
library) build.
## Background
This effort is tracked in #109483. Additional context is provided in
[this
discourse](https://discourse.llvm.org/t/psa-annotating-llvm-public-interface/85307),
and documentation for `LLVM_ABI` and related annotations is found in the
LLVM repo
[here](https://github.com/llvm/llvm-project/blob/main/llvm/docs/InterfaceExportAnnotations.rst).
A sub-set of these changes were generated automatically using the
[Interface Definition Scanner (IDS)](https://github.com/compnerd/ids)
tool, followed formatting with `git clang-format`.
The bulk of this change is manual additions of `LLVM_ABI` to
`LLVMInitializeX` functions defined in .cpp files under llvm/lib/Target.
Adding `LLVM_ABI` to the function implementation is required here
because they do not `#include "llvm/Support/TargetSelect.h"`, which
contains the declarations for this functions and was already updated
with `LLVM_ABI` in a previous patch. I considered patching these files
with `#include "llvm/Support/TargetSelect.h"` instead, but since
TargetSelect.h is a large file with a bunch of preprocessor x-macro
stuff in it I was concerned it would unnecessarily impact compile times.
In addition, a number of unit tests under llvm/unittests/Target required
additional dependencies to make them build correctly against the LLVM
DLL on Windows using MSVC.
## Validation
Local builds and tests to validate cross-platform compatibility. This
included llvm, clang, and lldb on the following configurations:
- Windows with MSVC
- Windows with Clang
- Linux with GCC
- Linux with Clang
- Darwin with Clang
We rename `createGenericSchedLive` and `createGenericSchedPostRA`
to `createSchedLive` and `createSchedPostRA`, and add a template
parameter `Strategy` which is the generic implementation by default.
This can simplify some code for targets that have custom scheduler
strategy.
Register assembly printer passes in the pass registry.
This makes it possible to use `llc -start-before=<target>-asm-printer ...` in tests.
Adds a `char &ID` parameter to the AssemblyPrinter constructor to allow
targets to use the `INITIALIZE_PASS` macros and register the pass in the
pass registry. This currently has a default parameter so it won't break
any targets that have not been updated.
Replace "concept based polymorphism" with simpler PImpl idiom.
This pursues two goals:
* Enforce static type checking. Previously, target implementations hid
base class methods and type checking was impossible. Now that they
override the methods, the compiler will complain on mismatched
signatures.
* Make the code easier to navigate. Previously, if you asked your
favorite LSP server to show a method (e.g. `getInstructionCost()`), it
would show you methods from `TTI`, `TTI::Concept`, `TTI::Model`,
`TTIImplBase`, and target overrides. Now it is two less :)
There are three commits to hopefully simplify the review.
The first commit removes `TTI::Model`. This is done by deriving
`TargetTransformInfoImplBase` from `TTI::Concept`. This is possible
because they implement the same set of interfaces with identical
signatures.
The first commit makes `TargetTransformImplBase` polymorphic, which
means all derived classes should `override` its methods. This is done in
second commit to make the first one smaller. It appeared infeasible to
extract this into a separate PR because the first commit landed
separately would result in tons of `-Woverloaded-virtual` warnings (and
break `-Werror` builds).
The third commit eliminates `TTI::Concept` by merging it with the only
derived class `TargetTransformImplBase`. This commit could be extracted
into a separate PR, but it touches the same lines in
`TargetTransformInfoImpl.h` (removes `override` added by the second
commit and adds `virtual`), so I thought it may make sense to land these
two commits together.
Pull Request: https://github.com/llvm/llvm-project/pull/136674
The call to the initializeAArch64CondBrTuningPass function is missing in
the AArch64TargetMachine LLVMInitializeAArch64Target function.
This means that the pass is not in the pass registry and options such as
-run-pass=aarch64-cond-br-tuning and
-stop-after=aarch64-cond-br-tuning cannot be used. This patch fixes that
issue.
The createSIMachineScheduler & createPostMachineScheduler
target hooks are currently placed in the PassConfig interface.
Moving it out to TargetMachine so that both legacy and
the new pass manager can effectively use them.
This makes sure no optimizations are applied that assume the
bigger alignment or size, which could be incorrect if we link
together with non-instrumented code.
Following discussions in #110443, and the following earlier discussions
in https://lists.llvm.org/pipermail/llvm-dev/2017-October/117907.html,
https://reviews.llvm.org/D38482, https://reviews.llvm.org/D38489, this
PR attempts to overhaul the `TargetMachine` and `LLVMTargetMachine`
interface classes. More specifically:
1. Makes `TargetMachine` the only class implemented under
`TargetMachine.h` in the `Target` library.
2. `TargetMachine` contains target-specific interface functions that
relate to IR/CodeGen/MC constructs, whereas before (at least on paper)
it was supposed to have only IR/MC constructs. Any Target that doesn't
want to use the independent code generator simply does not implement
them, and returns either `false` or `nullptr`.
3. Renames `LLVMTargetMachine` to `CodeGenCommonTMImpl`. This renaming
aims to make the purpose of `LLVMTargetMachine` clearer. Its interface
was moved under the CodeGen library, to further emphasis its usage in
Targets that use CodeGen directly.
4. Makes `TargetMachine` the only interface used across LLVM and its
projects. With these changes, `CodeGenCommonTMImpl` is simply a set of
shared function implementations of `TargetMachine`, and CodeGen users
don't need to static cast to `LLVMTargetMachine` every time they need a
CodeGen-specific feature of the `TargetMachine`.
5. More importantly, does not change any requirements regarding library
linking.
cc @arsenm @aeubanks
On Windows there is no platform support for ifunc but we could lower
them to global function pointers.
This also enables FMV for Windows with Clang and Compiler-rt.
Depends on #111961
- [x] Add `clearSubtargetInfo` API to TargetMachine and each backend to
make it possible to release memory used in each backend's
`SubtargetInfo` map if needed. Keep this API as `protected` so that it
will be used with precautions.
MSVC has a set of qualifiers to allow using 32-bit signed/unsigned
pointers when building 64-bit targets. This is useful for WoW code
(i.e., the part of Windows that handles running 32-bit application on a
64-bit OS). Currently this is supported on x64 using the 270, 271 and
272 address spaces, but does not work for AArch64 at all.
This change adds the same 270, 271 and 272 address spaces to AArch64 and
adjusts the data layout string accordingly. Clang will generate the
correct address space casts, but these will currently be ignored until
the AArch64 backend is updated to handle them.
Partially fixes#62536
This is a resurrected version of <https://reviews.llvm.org/D158857>
(originally created by @a_vorobev) - I've cleaned it up a little, fixed
the rest of the tests and added to auto-upgrade for the data layout.
This pass removes back-to-back smstart/smstop instructions
to reduce the number of streaming mode changes in a function.
The implementation as proposed doesn't aim to solve all problems
yet and suggests a number of cases that can be optimized in the
future.
In preparation for implementing hardening of BLRA* instructions,
restrict thunk function generation to only the thunks being actually
called from any function. As described in the existing comments,
emitting all possible thunks for BLRAA and BLRAB instructions would mean
adding about 1800 functions in total, most of which are likely not to be
called.
This commit merges AArch64SLSHardening class into SLSBLRThunkInserter,
so thunks can be created as needed while rewriting a machine function.
The usages of TII, TRI and ST fields of AArch64SLSHardening class are
replaced with requesting them in-place, as ThunkInserter assumes
multiple "entry points" in contrast to the only runOnMachineFunction
method of AArch64SLSHardening.
The runOnMachineFunction method essentially replaces pre-existing
insertThunks implementation as there is no more need to insert all
possible thunks unconditionally. Instead, thunks are created on first
use from inside of insertThunks method.
On MSVC the `this` uses inside `decltype` require a lambda capture. On
clang they result in an unused capture warning instead. Add the capture
and suppress the warning with `(void)this`.
-----
Initializing this map is somewhat expensive (especially for O0), so we
currently only do it if certain flags are used. I would like to make use
of it for crash dumps (#96078), where we don't know in advance whether
it will be needed or not.
This patch changes the initialization to a lazy approach, where a
callback is registered that does the actual initialization. The
callbacks will be run the first time the pass name is requested.
This way there is no compile-time impact if the mapping is not used.
My attempt to fix the Windows build made things worse,
revert entirely for now.
This reverts commit e7137f2fed5cfee822ae3c4c6d39188adb59a16c.
This reverts commit 6eaf204dbb0a6a81cddfd02f625c130f7bb1aae5.
This reverts commit 957dc4366dd2ce9d5d2991c3ad76bbf438e9954e.
Initializing this map is somewhat expensive (especially for O0), so we
currently only do it if certain flags are used. I would like to make use
of it for crash dumps (#96078), where we don't know in advance whether
it will be needed or not.
This patch changes the initialization to a lazy approach, where a
callback is registered that does the actual initialization. The
callbacks will be run the first time the pass name is requested.
This way there is no compile-time impact if the mapping is not used.
