- Created a new class to do post-analysis
- Updated several test cases with unreachable code to expect a warning
- Added some general tests
llvm-svn: 109286
comments explaining why it was wrong. 8225024.
Fix the real problem in 8213383: the code that splits very large
blocks when no other place to put constants can be found was not
considering the case that the block contained a Thumb tablejump.
llvm-svn: 109282
it's too late to start backing off aggressive latency scheduling when most
of the registers are in use so the threshold should be a bit tighter.
- Correctly handle live out's and extract_subreg etc.
- Enable register pressure aware scheduling by default for hybrid scheduler.
For ARM, this is almost always a win on # of instructions. It's runtime
neutral for most of the tests. But for some kernels with high register
pressure it can be a huge win. e.g. 464.h264ref reduced number of spills by
54 and sped up by 20%.
llvm-svn: 109279
- When we JIT an expression, we print the disassembly
of the generated code
- When we put the structure into the target, we print
the individual entries in the structure byte for
byte.
llvm-svn: 109278
class is a templatized class that allows you to have a cleanup function called
on a data value of type T when the value is set or when the object goes
out of scope. It has support for very rudimentary invalid value detection that
can be enabled by using the appropriate constructor.
Anyone with template experience that can see ways of improving this class
please let me know. The example code shows a few typical scenarios in which
I would like to use it. It is currently coded with simple type T values
in mind (integer file descriptors, pointers, etc), but I am sure some
specialization might help out the class for more complex types.
There is a lot of documentation including examples in the CleanUp.h header
file.
llvm-svn: 109239
them as such. Type::is(Signed|Unsigned|)IntegerType() now return false
for vector types, and new functions
has(Signed|Unsigned|)IntegerRepresentation() cover integer types and
vector-of-integer types. This fixes a bunch of latent bugs.
Patch from Anton Yartsev!
llvm-svn: 109229
ThreadCancel in Host::WillTerminate g_monitor_thread may be blocked on a call
to pthread_cond_wait (for example, line 640). Now, by default, when a
cancellation request is serviced g_monitor_thread will again own the mutex
guarding the condition variable it was waiting on. This causes the call to
SetValue in Host::WillTerminate to hit a deadlock.
The call to SetValue does not appear to be needed, so removing it solves
the issue.
Patch from Stephen Wilson.
llvm-svn: 109228
a segfault when calling pthread_cancel. Also, sets m_read_thread_enabled if
the thread is actually spawned.
Patch from Stephen Wilson.
llvm-svn: 109227
is not a good idea. The codebase does not depend
in this any more, and it may introduce hidden
runtime cost. If you get compile errors, please
dereference your iterator before passing to cast<>
(and friends).
Also: please consider caching the result of
operator* and reusing that instead of dereferencing
many times.
llvm-svn: 109220
with an existing allocator. The interesting use case of this
is that it allows "StringMap<whatever, BumpPtrAllocator&>" for
when you want to allocate out of a preexisting bump pointer
allocator owned by someone else.
llvm-svn: 109213