a pair of instructions, one for the used pointer and the second for the
user. This simplifies the representation and also makes it more dense.
This was noticed because of the miscompile in PR13926. In that case, we
were running up against a fundamental "bad idea" in the speculation of
PHI and select instructions: the speculation and rewriting are
interleaved, which requires phi speculation to also perform load
rewriting! This is bad, and causes us to miss opportunities to do (for
example) vector rewriting only exposed after PHI speculation, etc etc.
It also, in the old system, required us to insert *new* load uses into
the current partition's use list, which would then be ignored during
rewriting because we had already extracted an end iterator for the use
list. The appending behavior (and much of the other oddities) stem from
the strange de-duplication strategy in the PartitionUse builder.
Amusingly, all this went without notice for so long because it could
only be triggered by having *different* GEPs into the same partition of
the same alloca, where both different GEPs were operands of a single
PHI, and where the GEP which was not encountered first also had multiple
uses within that same PHI node... Hence the insane steps required to
reproduce.
So, step one in fixing this fundamental bad idea is to make the
PartitionUse actually contain a Use*, and to make the builder do proper
deduplication instead of funky de-duplication. This is enough to remove
the appending behavior, and fix the miscompile in PR13926, but there is
more work to be done here. Subsequent commits will lift the speculation
into its own visitor. It'll be a useful step toward potentially
extracting all of the speculation logic into a generic utility
transform.
The existing PHI test case for repeated operands has been made more
extreme to catch even these issues. This test case, run through the old
pass, will exactly reproduce the miscompile from PR13926. ;] We were so
close here!
llvm-svn: 164925
This fixes a regression from r162254, the optimizer has problems reasoning
about the smaller memcpy as it's often not safe to widen a store but making it
smaller is.
llvm-svn: 164917
source of false positives due to globals being declared in a header with some
kind of incomplete (small) type, but the actual definition being bigger.
llvm-svn: 164912
because moden processos can store multiple values in parallel, and preparing the consecutive store requires
some work. We only handle these cases:
1. Consecutive stores where the values and consecutive loads. For example:
int a = p->a;
int b = p->b;
q->a = a;
q->b = b;
2. Consecutive stores where the values are constants. Foe example:
q->a = 4;
q->b = 5;
llvm-svn: 164910
alignment could lose it due to the alloca type moving down to a much
smaller alignment guarantee.
Now SROA will actively compute a proper alignment, factoring the target
data, any explicit alignment, and the offset within the struct. This
will in some cases lower the alignment requirements, but when we lower
them below those of the type, we drop the alignment entirely to give
freedom to the code generator to align it however is convenient.
Thanks to Duncan for the lovely test case that pinned this down. =]
llvm-svn: 164891
buildbots. Original commit message:
A DAGCombine optimization for merging consecutive stores. This optimization is not profitable in many cases
because moden processos can store multiple values in parallel, and preparing the consecutive store requires
some work. We only handle these cases:
1. Consecutive stores where the values and consecutive loads. For example:
int a = p->a;
int b = p->b;
q->a = a;
q->b = b;
2. Consecutive stores where the values are constants. Foe example:
q->a = 4;
q->b = 5;
llvm-svn: 164890
When attaching to a remote system that does not look like a typical vendor system, and no
executable binary was specified to lldb, check a couple of fixed locations where kernels
running in ASLR mode (slid in memory to a random address) store their load addr when booted
in debug mode, and relocate the symbols or load the kernel wholesale from the host computer
if we can find it.
<rdar://problem/7714201>
llvm-svn: 164888
because moden processos can store multiple values in parallel, and preparing the consecutive store requires
some work. We only handle these cases:
1. Consecutive stores where the values and consecutive loads. For example:
int a = p->a;
int b = p->b;
q->a = a;
q->b = b;
2. Consecutive stores where the values are constants. Foe example:
q->a = 4;
q->b = 5;
llvm-svn: 164885
loaded at a random offset).
To get the kernel's UUID and load address I need to send a kdp
packet so I had to implement the kernel relocation (and attempt to
find the kernel if none was provided to lldb already) in ProcessKDP
-- but this code really properly belongs in DynamicLoaderDarwinKernel.
I also had to add an optional Stream to ConnectRemote so
ProcessKDP::DoConnectRemote can print feedback about the remote kernel's
UUID, load address, and notify the user if we auto-loaded the kernel via
the UUID.
<rdar://problem/7714201>
llvm-svn: 164881
runtime, we read method signatures for both class
and instance methods out of the runtime data.
(lldb) fr var str
(NSString *) str = 0x0000000105000180 @"Hello from '/Volumes/Data/projects/lldb/test/lang/objc/foundation/a.out'"
(lldb) expr str.length
(unsigned long long) $0 = 72
(lldb) expr [NSString stringWithCString:"Hello world!" encoding:1]
(id) $1 = 0x0000000105100050
(lldb) po $1
$1 = 0x0000000105100050 Hello world!
(lldb) fr var array1
(NSArray *) array1 = 0x000000010010a6e0 @"3 objects"
(lldb) expr array1.count
(unsigned long long) $0 = 3
(lldb) expr [array1 objectAtIndex:2]
(id) $1 = 0x00000001000025d0
(lldb) po $1
$1 = 0x00000001000025d0 array1 object3
Notice that both regular and property-style notation
work. I still need to add explicit support for
properties with non-default setters/getters.
This information is only queried if an Objective-C
object does not have debug information for a complete
type available. Otherwise we query debug information
as usual.
llvm-svn: 164878
the validation occurred.
The original implementation was pessimistic - we assumed that ivars
which escape are invalidated. This version is optimistic, it assumes
that the ivars will always be explicitly invalidated: either set to nil
or sent an invalidation message.
llvm-svn: 164868
Chandler Carruth