This is needed to allow an InstAlias for an instruction with an "OptionalDef"
result register (like ARM's cc_out) where you want to set the optional register
to reg0.
llvm-svn: 123490
analyzer -cc1 options that are tailored to the
input type. If the input type is "C++", we should
only run the dead stores checker (for now). Similarly,
checks specific to Objective-C should only run
on Objective-C Code.
llvm-svn: 123481
disabled in this checkin. Sorry for the large diffs due to
refactoring. New functionality is all guarded by EnableSchedCycles.
Scheduling the isel DAG is inherently imprecise, but we give it a best
effort:
- Added MayReduceRegPressure to allow stalled nodes in the queue only
if there is a regpressure need.
- Added BUHasStall to allow checking for either dependence stalls due to
latency or resource stalls due to pipeline hazards.
- Added BUCompareLatency to encapsulate and standardize the heuristics
for minimizing stall cycles (vs. reducing register pressure).
- Modified the bottom-up heuristic (now in BUCompareLatency) to
prioritize nodes by their depth rather than height. As long as it
doesn't stall, height is irrelevant. Depth represents the critical
path to the DAG root.
- Added hybrid_ls_rr_sort::isReady to filter stalled nodes before
adding them to the available queue.
Related Cleanup: most of the register reduction routines do not need
to be templates.
llvm-svn: 123468
when handling one-liner commands that contain escaped characters. In
order to deal with the new namespace/dictionary stuff, the command was
being embedded within a second string, which messed up the escaping.
This fixes the problem by handling one-liners in a different manner, so they
no longer need to be embedded within another string, and can still be
processed in the proper namespace/dictionary context.
llvm-svn: 123467
Debuggers on ELF platforms hook into the runtime linker by monitoring a special
"rendezvous" embedded in the address space of the inferior process. The exact
location of this structure is filled in by the runtime linker and can be
resolved by locating the DT_DEBUG entry in the processes .dynamic section. The
new GetImageInfoAddress() method (morally equivalent to
Process::GetImageInfoAddress) provides the mechanism to locate this information.
GetEntryPoint() simply returns the address of the start symbol in the executable
if present. It is useful to the dynamic loader plugin for ELF systems as this
is the earliest point where LLDB can break and probe the inferiors .dynamic
section and rendezvous structure. Also, this address can be used in the
computation of the virtual base address for position independent executables.
llvm-svn: 123466
Setting m_private_state_thread to an invalid value when the child thread exits
results in a race condition between calls to ThreadCancel and ThreadJoin.
llvm-svn: 123465
the case where the called function has fewer
formal arguments than actual arguments. This
fixes a crash in the analyzer when doing
function call inlining.
Patch by Zhenbo Xu!
llvm-svn: 123458
Instead, it should perform a textual replacement of $(CC) from "clang" to "clang++". The same is true
for "llvm-gcc" to "llvm-g++" and for "gcc" to "g++". This way, we keep the path component of the $(CC)
passed in from the user and do not end up with a mixed toolchains with different paths.
Ditto for a newly added function called cxx_linker.
llvm-svn: 123451
expansion in it, we may end up instantiating to an empty
expression-list. In this case, the variable is uninitialized; tweak
the instantiation logic to handle this case. Fixes PR8977.
llvm-svn: 123449
expansion, when it is known due to the substitution of an out
parameter pack. This allows us to properly handle substitution into
pack expansions that involve multiple parameter packs at different
template parameter levels, even when this substitution happens one
level at a time (as with partial specializations of member class
templates and the signatures of member function templates).
Note that the diagnostic we provide when there is an arity mismatch
between an outer parameter pack and an inner parameter pack in this
case isn't as clear as the normal diagnostic for an arity
mismatch. However, this doesn't matter because these cases are very,
very rare and (even then) only typically occur in a SFINAE context.
The other kinds of pack expansions (expression, template, etc.) still
need to support optional tracking of the number of expansions, and we
need the moral equivalent of SubstTemplateTypeParmPackType for
substituted argument packs of template template and non-type template
parameters.
llvm-svn: 123448
simplification present in fully optimized code (I think instcombine fails to
transform some of these when "X-Y" has more than one use). Fires here and
there all over the test-suite, for example it eliminates 8 subtractions in
the final IR for 445.gobmk, 2 subs in 447.dealII, 2 in paq8p etc.
llvm-svn: 123442
threading of shifts over selects and phis while there. This fires here and
there in the testsuite, to not much effect. For example when compiling spirit
it fires 5 times, during early-cse, resulting in 6 more cse simplifications,
and 3 more terminators being folded by jump threading, but the final bitcode
doesn't change in any interesting way: other optimizations would have caught
the opportunity anyway, only later.
llvm-svn: 123441
replace all uses of the entry with the predecessor. There are no cleanups
relying on this right now, but if we ever want a cleanup with a phi inside
it, this will be important.
llvm-svn: 123438
early in the cleanup code and one late interlaced with the inliner. The second one is
important because inlining and other scalar optzns can unpin allocas, allowing them to
be split up and promoted. While important for performance, this is also relatively
rare, and we would previously force a (non-lazy) computation of DomFrontiers, which
happened even if nothing became unpinned.
With this patch, the first pass of scalarrepl still promotes the vast bulk of allocas
in programs, but hte second pass has changed to use SSAUpdater, which is more "sparse"
and lazy. This speeds up opt -O3 time on kimwitu++ (a c++ app) by about 1%. The
numbers are interesting: the first pass promotes ~17500 allocas. The second pass
promotes about 1600. For non-C++ codes, the compile time win should be greater,
because the second pass of scalarrepl does less.
llvm-svn: 123437
Ted Kremenek