Commit Graph

15 Commits

Author SHA1 Message Date
Greg Clayton ea3e7d5ccf Added more functionality to Range template classes in RangeMap.h and converted remaining DWARF areas that were using ranges over to this class. Also converted lldb_private::Block to use it.
llvm-svn: 141460
2011-10-08 00:49:15 +00:00
Greg Clayton cf0e4f0daf Re-organized the contents of RangeMap.h to be more concise and also allow for a Range, RangeArray, RangeData (range + data), or a RangeDataArray. We have many range implementations in LLDB and I will be converting over to using the classes in RangeMap.h so we can have one set of code that does ranges and searching of ranges.
Fixed up DWARFDebugAranges to use the new range classes.

Fixed the enumeration parsing to take a lldb_private::Error to avoid a lot of duplicated code. Now when an invalid enumeration is supplied, an error will be returned and that error will contain a list of the valid enumeration values.

llvm-svn: 141382
2011-10-07 18:58:12 +00:00
Greg Clayton c26e445403 Since we use address ranges a lot I added a templatized class that allows us to easily control the base address type, the size type, and the data that is stored with each range. It is designed to be populated by appending all needed items, then sorting the resulting list, and optionally minimizing the list when done. I adopted this new list in the DWARFDebugAranges for even further memory savings.
llvm-svn: 141352
2011-10-07 03:58:56 +00:00
Greg Clayton 21f2a4919b Added a new logging channel to the DWARF called "lookups":
(lldb) log enable dwarf lookups

This allows us to see when lookups are being done on functions, addresses,
and types by both name and regular expresssion.

llvm-svn: 141259
2011-10-06 00:09:08 +00:00
Greg Clayton d4a2b37091 Huge memory and performance improvements in the DWARF parser.
Address ranges are now split up into two different tables: 
- one in DWARFDebugInfo that is compile unit specific
- one in each DWARFCompileUnit that has exact function DIE offsets

This helps keep the size of the aranges down since the main table will get
uniqued and sorted and have consecutive ranges merged. We then only parse the
compile unit one on demand once we have determined that a compile unit contains
the address in question. We also now use the .debug_aranges section if there 
is one instead of always indexing the DWARF manually.

NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map.
std::map is very bulky as each node has 3 pointers and the key and value types.
This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves
us a lot of memory when we have very large DWARF.

DWARFDebugAranges now has a smaller footprint for each range it contains to 
save on memory.

llvm-svn: 139557
2011-09-12 23:21:58 +00:00
Johnny Chen a6ca9fb407 Rename some variables, no functionality change.
llvm-svn: 129724
2011-04-18 22:11:31 +00:00
Johnny Chen b57edcab3b Get rid the of set membership test (log(m)) and, instead, use an index variable 'i'
which advances when src collides with a purged slot.
Hi Stephen, you're welcome to overwrite/or improve upon this version.  Thanks.

llvm-svn: 129611
2011-04-15 21:45:12 +00:00
Johnny Chen fec280e750 Update both the src and dst pointers at the end of the loop.
Stephen Wilson is working on a better performing patch in the meantime.

llvm-svn: 129605
2011-04-15 20:59:46 +00:00
Johnny Chen d2ddabac20 Optimize address range coalescing.
DWARFDebugAranges::Sort() calls std::stable_sort() over a set of address ranges
and then proceeds to collapse neighboring ranges together.

One problem with the current implementation is that it does an incomplete job.
When a pair of ranges are merged the next pair considered does not include the
just-merged range.  IOW, three consecutive ranges are never collapsed into one.

Another problem is that for each range merged we are calling
std::vector::erase() which "shifts" all remaining elements of the vector by one
position on every merge.  The end result (in the worst case) is a quadratic
algorithm -- not good when the input vector is large.

The following patch merges all consecutive ranges and removes the quadratic
behavior.  The implementation uses an auxiliary vector of indices in order to
remember all ranges that can be dropped, then performs the coalescing of ranges
in a single pass.

Patch from Stephen Wilson with some minor modification by me.

llvm-svn: 129595
2011-04-15 19:56:24 +00:00
Stephen Wilson 8acdbb8a75 Add missing headers.
Something changed in commit r129112 where a few standard headers vanished from
the include chain when building on Linux.  Fix up by including limits.h for
INT_MAX and PATH_MAX where needed, and stdio.h for printf().

llvm-svn: 129130
2011-04-08 13:36:44 +00:00
Greg Clayton c685f8e540 So we can't use .debug_pubtypes as it, as designed, does not tell us about
all types in all compile units. I added a new kind of accelerator table to
the DWARF that allows us to index the DWARF compile units and DIEs in a way
that doesn't require the data to stay loaded. Currently when indexing the
DWARF we check if the compile unit had parsed its DIEs and if it hasn't we
index the data and free all of the DIEs so we can reparse later when we need
to after using one of our complete accelerator tables to determine we need
to reparse some DWARF. If the DIEs had already been parsed we leave them 
loaded. The new accelerator table uses the "const char *" pointers from our
ConstString class as the keys, and NameToDIE::Info as the value. This info
contains the compile unit index and the DIE index which means we are pointed
right to the DIE we need unlike the other DWARF accelerator tables that often
just point us to the compile unit we would find our answer in. 

llvm-svn: 113933
2010-09-15 04:15:46 +00:00
Greg Clayton 016a95eb04 Looking at some of the test suite failures in DWARF in .o files with the
debug map showed that the location lists in the .o files needed some 
refactoring in order to work. The case that was failing was where a function
that was in the "__TEXT.__textcoal_nt" in the .o file, and in the 
"__TEXT.__text" section in the main executable. This made symbol lookup fail
due to the way we were finding a real address in the debug map which was
by finding the section that the function was in in the .o file and trying to
find this in the main executable. Now the section list supports finding a
linked address in a section or any child sections. After fixing this, we ran
into issue that were due to DWARF and how it represents locations lists. 
DWARF makes a list of address ranges and expressions that go along with those
address ranges. The location addresses are expressed in terms of a compile
unit address + offset. This works fine as long as nothing moves around. When
stuff moves around and offsets change between the remapped compile unit base
address and the new function address, then we can run into trouble. To deal
with this, we now store supply a location list slide amount to any location
list expressions that will allow us to make the location list addresses into
zero based offsets from the object that owns the location list (always a
function in our case). 

With these fixes we can now re-link random address ranges inside the debugger
for use with our DWARF + debug map, incremental linking, and more.

Another issue that arose when doing the DWARF in the .o files was that GCC
4.2 emits a ".debug_aranges" that only mentions functions that are externally
visible. This makes .debug_aranges useless to us and we now generate a real
address range lookup table in the DWARF parser at the same time as we index
the name tables (that are needed because .debug_pubnames is just as useless).
llvm-gcc doesn't generate a .debug_aranges section, though this could be 
fixed, we aren't going to rely upon it.

Renamed a bunch of "UINT_MAX" to "UINT32_MAX".

llvm-svn: 113829
2010-09-14 02:20:48 +00:00
Eli Friedman 07a2437a91 A few more misc warning fixes.
llvm-svn: 108030
2010-07-09 23:04:08 +00:00
Greg Clayton c982c768d2 Merged Eli Friedman's linux build changes where he added Makefile files that
enabled LLVM make style building and made this compile LLDB on Mac OS X. We
can now iterate on this to make the build work on both linux and macosx.

llvm-svn: 108009
2010-07-09 20:39:50 +00:00
Chris Lattner 30fdc8d841 Initial checkin of lldb code from internal Apple repo.
llvm-svn: 105619
2010-06-08 16:52:24 +00:00