The global-isel feature has been required for a long time and was
removed in c9455d3c579292e7ae5b7559ad0302d459e69a95, so this was
causing all tests to be skipped.
This patch adjusts the following ARM/AArch64 LLVM IR intrinsics:
- neon_bfmmla
- neon_bfmlalb
- neon_bfmlalt
so that they take and return bf16 and float types. Previously these
intrinsics used <8 x i8> and <4 x i8> vectors (a rudiment from
implementation lacking bf16 IR type).
The neon_vbfdot[q] intrinsics are adjusted similarly. This change
required some additional selection patterns for vbfdot itself and
also for vector shuffles (in a previous patch) because of SelectionDAG
transformations kicking in and mangling the original code.
This patch makes the generated IR cleaner (less useless bitcasts are
produced), but it does not affect the final assembly.
Reviewed By: dmgreen
Differential Revision: https://reviews.llvm.org/D86146
Original Commit Message:
After the commit r368987 (rG643adb55769e) was landed, the frame record (FP and LR register)
may be placed in the middle of a stack frame if a function has both callee-saved
general-purpose registers and floating point registers. This will break the stack unwinders
that simply walk through the frame records (based on the guarantee from AAPCS64
"The Frame Pointer" section). This commit fixes the problem by adding the frame record offset.
Patch By: logan
Differential Revision: D70800
The NPM doesn't support call-site alwaysinline as described in the comments.
Also make NPM runs more similar to legacy PM runs.
Reviewed By: ychen, asbirlea
Differential Revision: https://reviews.llvm.org/D86663
https://reviews.llvm.org/D83833
Patch adds two new GICombinerRules for G_SELECT. The rules include:
combining selects with undef comparisons into their first selectee value,
and to combine away selects with constant comparisons. Patch additionally
adds a new combiner test for the AArch64 target to test these new G_SELECT
combiner rules and the existing select_same_val combiner rule.
Patch by mkitzan
Add a new flag that indicates which stage in the process we are in.
This flag is introduced for handling behavior of `getAAFor` according to the stage. (discussed in D86635)
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86678
https://reviews.llvm.org/D86676
Sometimes we can have the following code
x:gpr(s32) = G_OP
Say we build G_OP2 to the same x and then delete the previous instruction. Using something like
Register X = ...;
auto NewMIB = CSEBuilder.buildOp2(X, ... args);
Currently there's a mismatch in how NewMIB is profiled and inserted into the CSEMap (ie it doesn't consider register bank/register class along with type).Unify the profiling by refactoring and calling the common method.
This was found by turning on the CSEInfo::verify in at the end of each of our GISel passes which turns inconsistent state/non determinism in CSEing into crashes which likely usually indicates missing calls to Observer on mutations (the most common case). Here non determinism usually means not cseing sometimes, but almost never about producing incorrect code.
Also this patch adds this verification at the end of the combiners as well.
When joining the legal parts of vector arguments into its original value
during the lower of Formal Arguments in SelectionDAGBuilder, the Calling
Convention information was not being propagated for the handling of each
individual parts. The same did not happen when lowering calls, causing a
mismatch.
This patch fixes the issue by properly propagating the Calling
Convention details.
This fixes Bugzilla #47001.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D86715
See RFC for background:
http://lists.llvm.org/pipermail/llvm-dev/2020-June/142744.html
Note that the runtime changes will be sent separately (hopefully this
week, need to add some tests).
This patch includes the LLVM pass to instrument memory accesses with
either inline sequences to increment the access count in the shadow
location, or alternatively to call into the runtime. It also changes
calls to memset/memcpy/memmove to the equivalent runtime version.
The pass is modeled on the address sanitizer pass.
The clang changes add the driver option to invoke the new pass, and to
link with the upcoming heap profiling runtime libraries.
Currently there is no attempt to optimize the instrumentation, e.g. to
aggregate updates to the same memory allocation. That will be
implemented as follow on work.
Differential Revision: https://reviews.llvm.org/D85948
Apparently, we don't do this, neither in EarlyCSE, nor in InstSimplify,
nor in (old) GVN, but do in NewGVN and SimplifyCFG of all places..
While i could teach EarlyCSE how to hash PHI nodes,
we can't really do much (anything?) even if we find two identical
PHI nodes in different basic blocks, same-BB case is the interesting one,
and if we teach InstSimplify about it (which is what i wanted originally,
https://reviews.llvm.org/D86530), we get EarlyCSE support for free.
So i would think this is pretty uncontroversial.
On vanilla llvm test-suite + RawSpeed, this has the following effects:
```
| statistic name | baseline | proposed | Δ | % | \|%\| |
|----------------------------------------------------|-----------|-----------|-------:|---------:|---------:|
| instsimplify.NumPHICSE | 0 | 23779 | 23779 | 0.00% | 0.00% |
| asm-printer.EmittedInsts | 7942328 | 7942392 | 64 | 0.00% | 0.00% |
| assembler.ObjectBytes | 273069192 | 273084704 | 15512 | 0.01% | 0.01% |
| correlated-value-propagation.NumPhis | 18412 | 18539 | 127 | 0.69% | 0.69% |
| early-cse.NumCSE | 2183283 | 2183227 | -56 | 0.00% | 0.00% |
| early-cse.NumSimplify | 550105 | 542090 | -8015 | -1.46% | 1.46% |
| instcombine.NumAggregateReconstructionsSimplified | 73 | 4506 | 4433 | 6072.60% | 6072.60% |
| instcombine.NumCombined | 3640264 | 3664769 | 24505 | 0.67% | 0.67% |
| instcombine.NumDeadInst | 1778193 | 1783183 | 4990 | 0.28% | 0.28% |
| instcount.NumCallInst | 1758401 | 1758799 | 398 | 0.02% | 0.02% |
| instcount.NumInvokeInst | 59478 | 59502 | 24 | 0.04% | 0.04% |
| instcount.NumPHIInst | 330557 | 330533 | -24 | -0.01% | 0.01% |
| instcount.TotalInsts | 8831952 | 8832286 | 334 | 0.00% | 0.00% |
| simplifycfg.NumInvokes | 4300 | 4410 | 110 | 2.56% | 2.56% |
| simplifycfg.NumSimpl | 1019808 | 999607 | -20201 | -1.98% | 1.98% |
```
I.e. it fires ~24k times, causes +110 (+2.56%) more `invoke` -> `call`
transforms, and counter-intuitively results in *more* instructions total.
That being said, the PHI count doesn't decrease that much,
and looking at some examples, it seems at least some of them
were previously getting PHI CSE'd in SimplifyCFG of all places..
I'm adjusting `Instruction::isIdenticalToWhenDefined()` at the same time.
As a comment in `InstCombinerImpl::visitPHINode()` already stated,
there are no guarantees on the ordering of the operands of a PHI node,
so if we just naively compare them, we may false-negatively say that
the nodes are not equal when the only difference is operand order,
which is especially important since the fold is in InstSimplify,
so we can't rely on InstCombine sorting them beforehand.
Fixing this for the general case is costly (geomean +0.02%),
and does not appear to catch anything in test-suite, but for
the same-BB case, it's trivial, so let's fix at least that.
As per http://llvm-compile-time-tracker.com/compare.php?from=04879086b44348cad600a0a1ccbe1f7776cc3cf9&to=82bdedb888b945df1e9f130dd3ac4dd3c96e2925&stat=instructions
this appears to cause geomean +0.03% compile time increase (regression),
but geomean -0.01%..-0.04% code size decrease (improvement).
This patch optionally replaces the CRT allocator (i.e., malloc and free) with rpmalloc (mixed public domain licence/MIT licence) or snmalloc (MIT licence) or mimalloc (MIT licence). Please note that the source code for these allocators must be available outside of LLVM's tree.
To enable, use `cmake ... -DLLVM_INTEGRATED_CRT_ALLOC=D:/git/rpmalloc -DLLVM_USE_CRT_RELEASE=MT` where `D:/git/rpmalloc` has already been git clone'd from `https://github.com/mjansson/rpmalloc`. The same applies to snmalloc and mimalloc.
When enabled, the allocator will be embeded (statically linked) into the LLVM tools & libraries. This currently only works with the static CRT (/MT), although using the dynamic CRT (/MD) could potentially work as well in the future.
When enabled, this changes the memory stack from:
new/delete -> MS VC++ CRT malloc/free -> HeapAlloc -> VirtualAlloc
to:
new/delete -> {rpmalloc|snmalloc|mimalloc} -> VirtualAlloc
The goal of this patch is to bypass the application's global heap - which is thread-safe thus inducing locking - and instead take advantage of a modern lock-free, thread cache, allocator. On a 6-core Xeon Skylake we observe a 2.5x decrease in execution time when linking a large scale application with LLD and ThinLTO (12 min 20 sec -> 5 min 34 sec), when all hardware threads are being used (using LLD's flag /opt:lldltojobs=all). On a dual 36-core Xeon Skylake with all hardware threads used, we observe a 24x decrease in execution time (1 h 2 min -> 2 min 38 sec) when linking a large application with LLD and ThinLTO. Clang build times also see a decrease in the range 5-10% depending on the configuration.
Differential Revision: https://reviews.llvm.org/D71786
Currently we bail out early for strlen calls with a GEP operand, if none
of the GEP specific optimizations fire. But there could be later
optimizations that still apply, which we currently miss out on.
An example is that we do not apply the following optimization
strlen(x) == 0 --> *x == 0
Unless I am missing something, there seems to be no reason for bailing
out early there.
Fixes PR47149.
Reviewed By: lebedev.ri, xbolva00
Differential Revision: https://reviews.llvm.org/D85886
This applies the same fix that D84748 did for macro definitions.
Appropriate include path is now automatically set for all libraries
which link against gtest targets, which avoids the need to set
include_directories in various parts of the project.
Differential Revision: https://reviews.llvm.org/D86616
This is the first of a set of DAGCombiner changes enabling strictfp
optimizations. I want to test to waters with this to make sure changes
like these are acceptable for the strictfp case- this particular change
should preserve exception ordering and result precision perfectly, and
many other possible changes appear to be able to as well.
Copied from regular fadd combines but modified to preserve ordering via
the chain, this change allows strict_fadd x, (fneg y) to become
struct_fsub x, y and strict_fadd (fneg x), y to become strict_fsub y, x.
Differential Revision: https://reviews.llvm.org/D85548
This reverts commit b9d977b0ca60c54f11615ca9d144c9f08b29fd85.
This cutoff is no longer required. The commit 34ffa7fc501 (D86153) introduces a
performance improvement which was tested against the motivating case for this
patch.
Discussed in differential revision: https://reviews.llvm.org/D86153
Almost NFC (see end).
The backwards scan in validThroughout significantly contributed to compile time
for a pathological case, causing the 'X86 Assembly Printer' pass to account for
roughly 70% of the run time. This patch guards the loop against running
unnecessarily, bringing the pass contribution down to 4%.
Almost NFC: There is a hack in validThroughout which promotes single constant
value DBG_VALUEs in the prologue to be live throughout the function. We're more
likely to hit this code path with this patch applied. Similarly to the parent
patches there is a small coverage change reported in the order of 10s of bytes.
Reviewed By: aprantl
Differential Revision: https://reviews.llvm.org/D86153
With the changes introduced in D86151 we can now check for single locations
which span multiple blocks for inlined scopes and blocks.
D86151 introduced the InstructionOrdering parameter, replacing a scan through
MBB instructions. The functionality to compare instruction positions across
blocks was add there, and this patch just removes the exit checks that were
previously (but no longer) required.
CTMark shows a geomean binary size reduction of 2.2% for RelWithDebInfo builds.
llvm-locstats (using D85636) shows a very small variable location coverage
change in 5 of 10 binaries, but just like in D86151 it is only in the order of
10s of bytes.
Reviewed By: djtodoro
Differential Revision: https://reviews.llvm.org/D86152
With this patch we're now accounting for two more cases which should be
considered 'valid throughout': First, where RangeEnd is ScopeEnd. Second, where
RangeEnd comes before ScopeEnd when including meta instructions, but are both
preceded by the same non-meta instruction.
CTMark shows a geomean binary size reduction of 1.5% for RelWithDebInfo builds.
`llvm-locstats` (using D85636) shows a very small variable location coverage
change in 2 of 10 binaries, but it is in the order of 10s of bytes which lines
up with my expectations.
I've added a test which checks both of these new cases. The first check in the
test isn't strictly necessary for this patch. But I'm not sure that it is
explicitly tested anywhere else, and is useful for the final patch in the
series.
Reviewed By: aprantl
Differential Revision: https://reviews.llvm.org/D86151
Group the map and methods used to query instruction ordering for trimVarLocs
(D82129) into a class. This will make it easier to reuse the functionality
upcoming patches.
Reviewed By: aprantl
Differential Revision: https://reviews.llvm.org/D86150
This patch adds code to recognize vector shuffles which can be
represented as VDUP (splat) of a vector lane with of a different
(wider) type than the original vector lane type.
For example:
shufflevector <4 x i16> %v, <4 x i16> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
is essentially:
shufflevector <2 x i32> %v, <2 x i32> undef, <2 x i32> <i32 0, i32 0>
Such patterns are generated by the SelectionDAG machinery in some cases
(see DAGCombiner::visitBITCAST in DAGCombiner.cpp, the "Remove double
bitcasts from shuffles" part).
Reviewed By: dmgreen
Differential Revision: https://reviews.llvm.org/D86225
This patch adds type information for SVE ACLE vector types,
by describing them as vectors, with a lower bound of 0, and
an upper bound described by a DWARF expression using the
AArch64 Vector Granule register (VG), which contains the
runtime multiple of 64bit granules in an SVE vector.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D86101
Since the canonical floatig-point move is fsgnj rd, rs, rs, we should
handle this case in RISCVInstrInfo::isAsCheapAsAMove().
Reviewed By: lenary
Differential Revision: https://reviews.llvm.org/D86518
