Summary: This patch adds block frequency analysis to LoopUnswitch pass to recognize hot/cold regions. For cold regions the pass only performs trivial unswitches since they do not increase code size, and for hot regions everything works as before. This helps to minimize code growth in cold regions and be more aggressive in hot regions. Currently the default cold regions are blocks with frequencies below 20% of function entry frequency, and it can be adjusted via -loop-unswitch-cold-block-frequency flag. The entire feature is controlled via -loop-unswitch-with-block-frequency flag and it is off by default.
Reviewers: broune, silvas, dnovillo, reames
Subscribers: davidxl, llvm-commits
Differential Revision: http://reviews.llvm.org/D11605
llvm-svn: 248777
Description.
If the simd clause is specified, the ordered regions encountered by any thread will use only a single SIMD lane to execute the ordered regions in the order of the loop iterations.
Restrictions.
An ordered construct with the simd clause is the only OpenMP construct that can appear in the simd region.
An ordered directive with ‘simd’ clause is generated as an outlined function and corresponding function call to prevent this part of code from vectorization later in backend.
llvm-svn: 248772
Place new and update dbg.declare calls immediately after the
corresponding alloca.
Current code in replaceDbgDeclareForAlloca puts the new dbg.declare
at the end of the basic block. LLVM codegen has problems emitting
debug info in a situation when dbg.declare appears after all uses of
the variable. This usually kinda works for inlining and ASan (two
users of this function) but not for SafeStack (see the pending change
in http://reviews.llvm.org/D13178).
llvm-svn: 248769
This is a bit of an awkward API and I'm not sure what the right solution
is. Having a publicly copy constructible base class makes it easy to
accidentally slice derived objects in a number of contexts.
llvm-svn: 248764
LookupResult should not be copyable, it's not readily copyable and can
only be copied when it's in specific states (in a query state, without
any results, basically). Instead, just extract the /query/ state and
pass that across the copy boundary, then build a new LookupResult on the
other side.
I wonder if a better API (one in which the query state is separate from
the result state - essentialyl making QueryState a first class part of
the Lookup API - pass a QueryState, get a LookupResult, rather than
mutating the LookupResult in place (LookupResult could contain a
QueryState if it's particularly helpful to be able to observe the query
parameters while also examining the result)) might be a good idea here.
Future patches will probably make LookupResult actually non-copyable
(transition the CXXBasePaths to unique_ptr, for example) and hopefully
we'll enable -Wdeprecated in LLVM soon to avoid issues like this.
llvm-svn: 248761
We don't want to filter out the builtins that are present in libSystem like we do for the normal builtins because kexts can't link libSystem, but we should filter out all the builtins that are generally not supported on the OS and architecture.
llvm-svn: 248756
Currently it's 64-bit which will lead to mismatch between host and
device code if we compile for i386.
Differential Revision: http://reviews.llvm.org/D13181
llvm-svn: 248753
For Darwin simulator platforms we shouldn't build the cc_kext builtins at all because they aren't applicable, and we should includ the simulator builtins as slices inside the main platform builtin library.
llvm-svn: 248751
`ScalarEvolution::isImpliedCondOperandsViaNoOverflow` tries to cast the
operand type of the comparison it is given to an `IntegerType`. This is
incorrect because it could actually be simplifying a comparison between
two pointers. Switch it to using `getTypeSizeInBits` instead, which
does the right thing for both pointers and integers.
Fixed PR24956.
llvm-svn: 248743
The splitting of > 4 dword SMRD instructions
if using an offset in an SGPR instead of an immediate
was not setting the destination register,
resulting an an instruction missing an operand
which would assert later.
Test will be included in a following commit
which fixes a related issue.
llvm-svn: 248739
This is a basic initial implementation of the -flat_namespace and
-undefined options for LLD-darwin. It ignores several subtlties,
but the result is close enough that we can now link LLVM (but not
clang) on Darwin and pass all regression tests.
llvm-svn: 248732
Patch by Jake VanAdrighem!
Summary:
Fix the way we sort the llvm.used and llvm.compiler.used members.
This bug seems to have been introduced in rL183756 through a set of improper casts to GlobalValue*. In subsequent patches this problem was missed and transformed into a getName call on a ConstantExpr.
Reviewers: silvas
Subscribers: silvas, llvm-commits
Differential Revision: http://reviews.llvm.org/D12851
llvm-svn: 248728
1. Use a worklist, not a recursive approach, to avoid needless
revisitation and being repeatedly forced to jump back to the
start of the BB if a handle is invalidated.
2. Only insert operands to the worklist if they become unused
after a dead instruction is removed, so we don’t have to
visit them again in most cases.
3. Use a SmallSetVector to track the worklist.
4. Instead of pre-initting the SmallSetVector like in
DeadCodeEliminationPass, only put things into the worklist
if they have to be revisited after the first run-through.
This minimizes how much the actual SmallSetVector gets used,
which saves a lot of time.
llvm-svn: 248727
Summary:
The P5600 is an out-of-order, superscalar implementation of the MIPS32R5
architecture.
The scheduler has a few missing details (see the 'Tricky Instructions'
section and some quirks of the P5600 are deliberately omitted due to
implementation difficulty and low chance of significant benefit (e.g. the
predicate on P5600WriteEitherALU). However, testing on SingleSource is
showing significant performance benefits on some apps (seven in the 10-30%
range) and only one significant regression (12%) when
-pre-RA-sched=linearize is given. Without -pre-RA-sched=linearize the
results are more variable. Some do even better (up to 55% improvement) but
increased numbers of copies are slowing others down (up to 12%).
Overall, the scheduler as it currently stands is a 2.4% win with
-pre-RA-sched=linearize and a 2.7% win without -pre-RA-sched=linearize.
I'm sure we can improve on this further.
For completeness, the FPGA this was tested on shows some failures with and
without the P5600 scheduler. These appear to be scheduling related since
the two test runs have fairly different sets of failing tests even after
accounting for other factors (e.g. spurious connection failures) however
it's not P5600 specific since we also get some for the generic scheduler.
Reviewers: vkalintiris
Subscribers: mpf, llvm-commits, atrick, vkalintiris
Differential Revision: http://reviews.llvm.org/D12193
llvm-svn: 248725
Summary:
compile_commands.json is usually generated in the build directory.
Projects like LLVM/Clang enforce out-of-source builds.
This option allow allow such projects to work out of the box, without
moving the compilation database manually.
The naming of the option is similar to the one use by other tools:
clang-{check,modernize,query,rename,tidy} -p=<build_path> <...>
Reviewers: alexfh
Differential Revision: http://reviews.llvm.org/D13199
llvm-svn: 248723
Chen Li