The optimization is two-fold:
First, the algorithm now uses SSE instructions to
handle all 4 shadow slots at once. This makes processing
faster.
Second, if shadow contains the same access, we do not
store the event into trace. This increases effective
trace size, that is, tsan can remember up to 10x more
previous memory accesses.
Perofrmance impact:
Before:
[ OK ] DISABLED_BENCH.Mop8Read (2461 ms)
[ OK ] DISABLED_BENCH.Mop8Write (1836 ms)
After:
[ OK ] DISABLED_BENCH.Mop8Read (1204 ms)
[ OK ] DISABLED_BENCH.Mop8Write (976 ms)
But this measures only fast-path.
On large real applications the speedup is ~20%.
Trace size impact:
On app1:
Memory accesses : 1163265870
Including same : 791312905 (68%)
on app2:
Memory accesses : 166875345
Including same : 150449689 (90%)
90% of filtered events means that trace size is effectively 10x larger.
llvm-svn: 209897
This breaks with MSVC.
With IsLateTemplateParsed, FunctionDecl::doesThisDeclarationHaveABody() returns true regardless of Body.
This reinstates what was fixed in r208985.
llvm-svn: 209896
Darwin prologues save their GPRs in two stages: a narrow push of r0-r7 & lr,
followed by a wide push of the remaining registers if there are any. AAPCS uses
a single push.w instruction.
It turns out that, on average, enough registers get pushed that code is smaller
in the AAPCS prologue, which is a nice property for M-class programmers. They
also have other options available for back-traces, so can hopefully deal with
the fact that FP & LR aren't adjacent in memory.
rdar://problem/15909583
llvm-svn: 209895
(clang doesn't complain about this, but gcc does. This is necessary for a
follow-up patch that will enable _LIBCPP_CONSTEXPR for gcc.)
llvm-svn: 209888
The C and C++ semantics for compare_exchange require it to return a bool
indicating success. This gets mapped to LLVM IR which follows each cmpxchg with
an icmp of the value loaded against the desired value.
When lowered to ldxr/stxr loops, this extra comparison is redundant: its
results are implicit in the control-flow of the function.
This commit makes two changes: it replaces that icmp with appropriate PHI
nodes, and then makes sure earlyCSE is called after expansion to actually make
use of the opportunities revealed.
I've also added -{arm,aarch64}-enable-atomic-tidy options, so that
existing fragile tests aren't perturbed too much by the change. Many
of them either rely on undef/unreachable too pervasively to be
restored to something well-defined (particularly while making sure
they test the same obscure assert from many years ago), or depend on a
particular CFG shape, which is disrupted by SimplifyCFG.
rdar://problem/16227836
llvm-svn: 209883
The main problem is in the predicate passed to the `std::stable_sort()`.
This predicate always returns false if **both** section's names do not
start with `.init_array` or `.fini_array` prefixes. In short, it does not
define a strict weak orderng. Suppose we have the following sections:
.A .init_array.1 .init_array.2
The predicate states that:
not .init_array.1 < .A
not .A < .init_array.2
but .init_array.1 < .init_array.2 !!!
The second problem is that `.init_array` section without number should
go last in the list. Not it has the lowest priority '0' and goes first.
The patch fixes both of the problems.
llvm-svn: 209875
http://llvm.org/bugs/show_bug.cgi?id=19876
The following C++1y code results in a crash:
struct X {
int m = 10;
int n = [this](auto) { return m; }(20);
};
When implicitly instantiating the generic lambda's call operator specialization body, Sema is unable to determine the current 'this' type when transforming the MemberExpr 'm' - since it looks for the nearest enclosing FunctionDeclDC - which is obviously null.
I considered two ways to fix this:
1) In InstantiateFunctionDefinition, when the context is saved after the lambda scope info is created, retain the 'this' pointer.
2) Teach getCurrentThisType() to recognize it is within a generic lambda within an NSDMI/default-initializer and return the appropriate this type.
I chose to implement #2 (though I confess I do not have a compelling reason for choosing it over #1).
Richard Smith accepted the patch:
http://reviews.llvm.org/D3935
Thank you!
llvm-svn: 209874
This patch adds support to vectorize intrinsics such as powi, cttz and ctlz in Vectorizer. These intrinsics are different from other
intrinsics as second argument to these function must be same in order to vectorize them and it should be represented as a scalar.
Review: http://reviews.llvm.org/D3851#inline-32769 and http://reviews.llvm.org/D3937#inline-32857
llvm-svn: 209873
The corresponding CFE patch replaces these intrinsics with vector initializers
in avxintrin.h. This patch removes the LLVM intrinsics from the backend.
We now stop lowering at X86ISD::VBROADCAST custom node rather than lowering
that further to the intrinsics.
The patch only changes VBROADCASTS* and leaves VBROADCAST[FI]128 to continue
to use intrinsics. As explained in the CFE patch, the reason is that we
currently don't generate as good code for them without the intrinsics.
CodeGen/X86/avx-vbroadcast.ll already provides coverage for this change. It
checks that for a series of insertelements we generate the appropriate
vbroadcast instruction.
Also verified that there was no assembly change in the test-suite before and
after this patch.
llvm-svn: 209864
They are replaced with the same IR that is generated for the
vector-initializers in avxintrin.h.
The test verifies that we get back the original instruction. I haven't seen
this approach to be used in other auto-upgrade tests (i.e. llc + FileCheck)
but I think it's the most direct way to test this case. I believe this should
work because llc upgrades calls during parsing. (Other tests mostly check
that assembling and disassembling yields the upgraded IR.)
llvm-svn: 209863
original fix would actually trigger the *exact* same crasher as the
original bug for a different reason. Awesomesauce.
Working on test cases now, but wanted to get bots healthier.
llvm-svn: 209860
across PHI nodes. The code was computing the Idxs from the 'GEP'
variable's indices when what it wanted was Op1's indices. This caused an
ASan heap-overflow for me that pin pointed the issue when Op1 had more
indices than GEP did. =] I'll let Louis add a specific test case for
this if he wants.
llvm-svn: 209857
The loop vectorizer instantiates be-taken-count + 1 as the loop iteration count.
If this expression overflows the generated code was invalid.
In case of overflow the code now jumps to the scalar loop.
Fixes PR17288.
llvm-svn: 209854
These tests ensure that a change I will propose in clang works as
expected.
Summary:
Added tests for the generation of blend+immediate instructions from a
shufflevector.
These tests were proposed along with a patch that was dropped. I'm
committing the tests anyway to protect against possible regressions in
codegen.
Reviewers: nadav, bkramer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D3600
llvm-svn: 209853
Changes include:
- ObjectFileMachO can now determine if a binary is "*-apple-ios" or "*-apple-macosx" by checking the min OS and SDK load commands
- ArchSpec now says "<arch>-apple-macosx" is equivalent to "<arch>-apple-ios" since the simulator mixes and matches binaries (some from the system and most from the iOS SDK).
- Getting process inforamtion on MacOSX now correctly classifies iOS simulator processes so they have "*-apple-ios" architectures in the ProcessInstanceInfo
- PlatformiOSSimulator can now list iOS simulator processes correctly instead of showing nothing by using:
(lldb) platform select ios-simulator
(lldb) platform process list
- debugserver can now properly return "*-apple-ios" for the triple in the process info packets for iOS simulator executables
- GDBRemoteCommunicationClient now correctly passes along the triples it gets for process info by setting the OS in the llvm::Triple correctly
<rdar://problem/17060217>
llvm-svn: 209852
Dmitry Vyukov