Only the most recent cpus support really 1cy 64-bit multiplies, and the X64 cost table represents a realistic worst case. The 1cy value was also discouraging vectorization when most vXi64 PMULDQ expansions aren't actually slower than scalarization.
Noticed while investigating PR51436.
Clang regression tests should not break when changes are made to
the LLVM optimizer. This file broke on the 1st attempt at D110170,
so I'm trying to prevent that on another try.
Similar to other files in this directory, we make a compromise and
run -mem2reg to reduce noise by about 1000 lines out of 5000+ CHECK lines.
Remove nmissed_expected variable.
It's a leftover from removed "expected race" feature and is never incremented.
Reviewed By: melver
Differential Revision: https://reviews.llvm.org/D110321
At the moment, DSE only considers whether a pointer may be captured at
all in a function. This leads to cases where we fail to remove stores to
local objects because we do not check if they escape before potential
read-clobbers or after.
Doing context-sensitive escape queries in isReadClobber has been removed
a while ago in d1a1cce5b1 to save compile-time. See PR50220 for more
context.
This patch introduces a new capture tracker, which keeps track of the
'earliest' capture. An instruction A is considered earlier than instruction
B, if A dominates B. If 2 escapes do not dominate each other, the
terminator of the common dominator is chosen. If not all uses cannot be
analyzed, the earliest escape is set to the first instruction in the
function entry block.
If the query instruction dominates the earliest escape and is not in a
cycle, then pointer does not escape before the query instruction.
This patch uses this information when checking if a load of a loaded
underlying object may alias a write to a stack object. If the stack
object does not escape before the load, they do not alias.
I will share a follow-up patch to also use the information for call
instructions to fix PR50220.
In terms of compile-time, the impact is low in general,
NewPM-O3: +0.05%
NewPM-ReleaseThinLTO: +0.05%
NewPM-ReleaseLTO-g: +0.03
with the largest change being tramp3d-v4 (+0.30%)
http://llvm-compile-time-tracker.com/compare.php?from=1a3b3301d7aa9ab25a8bdf045c77298b087e3930&to=bc6c6899cae757c3480f4ad4874a76fc1eafb0be&stat=instructions
Compared to always computing the capture information on demand, we get
the following benefits from the caching:
NewPM-O3: -0.03%
NewPM-ReleaseThinLTO: -0.08%
NewPM-ReleaseLTO-g: -0.04%
The biggest speedup is tramp3d-v4 (-0.21%).
http://llvm-compile-time-tracker.com/compare.php?from=0b0c99177d1511469c633282ef67f20c851f58b1&to=bc6c6899cae757c3480f4ad4874a76fc1eafb0be&stat=instructions
Overall there is a small, but noticeable benefit from caching. I am not
entirely sure if the speedups warrant the extra complexity of caching.
The way the caching works also means that we might miss a few cases, as
it is less precise. Also, there may be a better way to cache things.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D109844
The NFC commit e5692a564a changed the logic for
DomTreeUpdates to use the range [succ_begin, succ_begin) when
looking for SuccsOfPredBB rather than using [succ_begin, succ_end).
As the commit was NFC this is identified as a typo (it has been
discussed briefly in phabricator).
The typo was found when inspecting the code, so I've got no idea if
changing back to the old range has any significant impact (such as
solving any PR:s or causing some new problems). But at least this
restores the code to the originally indented behavior.
getpeername will return addrlen = 2 (sizeof sa_family_t) for unnamed
sockets (those not assigned a name with bind(2)). This is typically true
for client sockets as well as those created by socketpair(2).
This GetSocketName used to crash for sockets which were connected to
these kinds of sockets. Now it returns an empty string.
This patch adds codegen support for lowering the vector-predicated
reduction intrinsics to RVV instructions. The process is similar to that
of the other reduction intrinsics, save for the fact that every VP
reduction has a start value. We reuse the existing custom "VL" nodes,
adding extra patterns where required to handle non-true masks.
To support these nodes, the `RISCVISD::VECREDUCE_*_VL` nodes have been
given an explicit "merge" operand. This is to faciliate the VP
reductions, where we must be careful to ensure that even if no operation
is performed (when VL=0) we still produce the start value. The RVV
reductions don't update the destination register under these conditions,
so we tie the splatted start value to the output register.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D107657
This patch is part of the upstreaming effort from fir-dev branch and sync changes. Inline trival `isa_<type>` functions.
Co-authored-by: schweitzpgi
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D110233
tsan_rtl.h is very huge and contains too many things.
Move FastState and Shadow types into a new tsan_shadow.h file.
This also allows to use FastState/Shadow in other header files
without creating circular dependencies (which most likely will
happen today).
Reviewed By: melver
Differential Revision: https://reviews.llvm.org/D110272
Fix few remaining cases where we use u64 instead of the new RawShadow type.
Depends on D110265.
Reviewed By: melver
Differential Revision: https://reviews.llvm.org/D110266
Add a test for a trace corner case that lead to a bug
in experimental runtime replacement.
Since it passes with the current runtime it makes sense
to submit it on its own.
Depends on D110264.
Reviewed By: melver
Differential Revision: https://reviews.llvm.org/D110265
DontDumpShadow is used only in InitializeShadowMemory which is Go-only.
Depends on D110263.
Reviewed By: melver
Differential Revision: https://reviews.llvm.org/D110264
Remove unnecessary enum values in the memory profiler.
There is no point in spelling them, it can only lead to bugs
and larger diffs when values are added/removed.
Reviewed By: melver
Differential Revision: https://reviews.llvm.org/D110263
When statically linking C++ standard library, we shouldn't add -Bdynamic
after including the library on the link line because that might override
user settings like -static and -static-pie. Rather, we should surround
the library with --push-state/--pop-state to make sure that -Bstatic
only applies to C++ standard library and nothing else. This has been
supported since GNU ld 2.25 (2014) so backwards compatibility should
no longer be a concern.
Differential Revision: https://reviews.llvm.org/D110128
This simplifies the API and addresses a FIXME in
TwoAddressInstructionPass::convertInstTo3Addr.
Differential Revision: https://reviews.llvm.org/D110229
Bisecting and reducing opt pipelines that includes the
ModuleInlinerWrapperPass has turned out to be a bit problematic.
This is far from perfect (it still lacks information about inline
advisor params etc.), but it should give some kind of hint to what
the wrapped pipeline looks like when using -print-pipeline-passes.
Reviewed By: aeubanks, mtrofin
Differential Revision: https://reviews.llvm.org/D109878
Current warning message in method `addAffineForOpDomain` of mlir/lib/Analysis/AffineStructures.cpp is being printed to the stdout/stderr.
This patch redirects the warning with LLVM_DEBUG following standard llvm practice.
Reviewed By: bondhugula
Differential Revision: https://reviews.llvm.org/D108340
Without such wrapping, linking lld fails with missing symbols because of
C++ symbol mangling with older versions of the MacOSX SDK, in which
xar.h doesn't have an extern "C" block itself.
Reviewed By: #lld-macho, thakis
Differential Revision: https://reviews.llvm.org/D110224
This patch is to support transform something like
_mm512_add_ph(acc, _mm512_fmadd_pch(a, b, _mm512_setzero_ph()))
to _mm512_fmadd_pch(a, b, acc).
Differential Revision: https://reviews.llvm.org/D109953
This patch uses a different command-line arguments to test `clang::tooling::ToolInvocation` that are not specific to Darwin.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D110160
clang-cl errors out while handling the templated version of tgfmt. This
patch works around the issue by explicitly choosing the non-templated
version of tgfmt, which takes an ArrayRef<std::string>.
More details in this thread:
https://lists.llvm.org/pipermail/cfe-dev/2021-September/068936.html
Thanks @Mehdi Amini for suggesting the fix :)
Differential Revision: https://reviews.llvm.org/D110223
Implements parts of P1614, including synth-three-way and three way comparison for std::pair.
Reviewed By: #libc, Quuxplusone, Mordante
Differential Revision: https://reviews.llvm.org/D107721
This patch fixes a problem when the AAKernelInfo state was invalidated,
e.g., due to `optnone` for a kernel, but not all parts indicated the
invalidation properly. We further eliminate most full state
invalidations as they should never be necessary.
Differential Revision: https://reviews.llvm.org/D109468
Enables putting types and attributes in sets and in dicts as keys.
Reviewed By: stellaraccident
Differential Revision: https://reviews.llvm.org/D110301
It seems we missed one spot to persist `SampleContextFrameVector` into the global table (CSProfileGenerator::populateFunctionBoundarySamples:340) which causes a crash.
This change tried to fix it in a centralized way i. e. where we generate the `FunctionSamples`.
Reviewed By: hoy, wenlei
Differential Revision: https://reviews.llvm.org/D110275
Christian Kandeler