files. It was reading non-initialized global vars when the flag said it was
initialized and vice versa. Causes mis-alignment since initialized and
non-initialized constants have different bytecode lengths.
llvm-svn: 14057
must always coexist. Cleaned up the documentation on these interfaces
significantly. This is in preparation for moving Parser.h to the include
directories to make it a public interface.
llvm-svn: 14054
will (eventually) provide statistical analysis of bytecode files as well
as the ability to dump them in a low level format (slot numbers not
resolved). The purpose of this is to aid in the Type!=Value change of
bug 122. With this initial release, llvm-abcd merely dumps out the
bytecode. However, the infrastructure for separating bytecode parsing from
handling the parsing events is in place. The style chosen is similar to
SAX XML parsing where a handler object is called to handlign the parsing
events. This probably isn't useful to anyone but me right now as there is
no analysis yet, and the dumper doesn't work on every bytecode file. It
will probably be useful by the end of this week. Note that there is some
duplication of code from the bytecode reader. This was done to eliminate
errors from being introduced in the reader and to minimize the impact to
other LLVM developers. At some point, the Analyzer and the Reader will be
integrated to use the same infrastructure. Also, sorry for the minor change
to Instruction.h but I just couldn't bring myself to write code that
depends on Instruction internals.
llvm-svn: 14048
planes. A SymbolTable could still have types in it! This fixes problems
with two regression tests that failed because a symbol table that only
contained types was being omitted from bytecode files. Thanks to Chris
for the reduced test case that helped find this immediately.
llvm-svn: 13842
This change removes the BuildBytecodeInfo flag from the SlotCalculator
class. This flag was needed to distinguish between the Bytecode/Writer
and the AsmWriter. Now that AsmWriter doesn't use SlotCalculator, we can
remove this flag and simplify some code. Also, some minor name changes
to CachedWriter.h needed to be committed (missed in previous commit).
llvm-svn: 13785
the Abstract Data Type that holds slot number values and associates them
with Type* and Value*. The SlotTable is simply the holder of the slot
numbers and provides a controlled interface for building the table. It does
not enforce any particular idiom or functionality for manipulating the slot
numbers.
This is part of bug_122. The SlotCalculator and SlotMachine classes will
follow.
llvm-svn: 13764
This shrinks the bytecode file for 176.gcc by about 200K (10%), and 254.gap by
about 167K, a 25% reduction. There is still a lot of room for improvement in
the encoding of the compaction table.
llvm-svn: 10915
This shrinks the bytecode file for 176.gcc by about 200K (10%), and 254.gap by
about 167K, a 25% reduction. There is still a lot of room for improvement in
the encoding of the compaction table.
llvm-svn: 10914
bytecode files when compiling 176.gcc, but more importantly will make it
easier to eliminate CPR's in the future (no new .bc revision will be
required to support them)
llvm-svn: 10884
of forcing them to go through ConstantPointerRef's. This allows bytecode
files to mirror .ll files, allows more efficient encoding, and makes it easier
to eventually eliminate CPR's.
llvm-svn: 10883
returning error codes. Because they don't return an error code, they can
return the value read, which simplifies the code and makes the reader more
efficient (yaay!).
Also eliminate the special case code for little endian machines.
llvm-svn: 10871
intended to save size (and does on small programs), but on big programs it
actually increases the size of the program slightly. The deal is that many
functions end up using the characters that the string contained, and the
characters are no longer in the global constant table, so they have to be
emitted in function specific constant pools.
This pessimization will be fixed in subsequent patches.
llvm-svn: 10864
occurs when the symbol table for a module has been stripped, making all of the
function local symbols go away.
This saves 6728 bytes in the stripped bytecode file of 254.gap (which obviously
has 841 functions), which isn't a ton, but helps and was easy.
llvm-svn: 10750
* Refactor reader stuff out of include/llvm/Bytecode/Primitives.h. This is
internal implementation details for the reader, not public interfaces!
llvm-svn: 10739
Modified ReadArchiveBuffer() so that it dynamically allocates the
std::string object used to hold the bytecode object file's name. This is
necessary because it is passed by reference to the new Module that is
allocated to represent the bytecode object, and previously we were
using a std::string that disappeared on function exit.
llvm-svn: 10565
beginning of the archive member data as an argument.
Get rid of ParseLongFilenameSection(), which is dead.
In ReadArchiveBuffer(), implement support for 4.4BSD/MacOSX long filenames.
This is kind of invasive, because they prepend the long filename to the archive
member data, and then lie about the size. So we have to keep track of the real
size.
llvm-svn: 10392
each basic block in function. Instead, just emit a stream of instructions,
chopping up basic blocks based on when we find terminator instructions. This
saves a fairly substantial chunk of bytecode space. In stripped, sample
cases, for example, we get this reduction in size:
197.parser: 163036 -> 137180: 18.8% reduction
254.gap : 844936 -> 689392: 22.6%
255.vortex: 621724 -> 528444: 17.7%
...
Not bad for something this simple. :) Note that this doesn't require a new
bytecode version number at all, though version 1.1 should not need to support
the old format.
llvm-svn: 10280
* Strength reduce several data structures which were left over from the
"bad old days"
* Minor efficiency improvements
* Major efficiency improvement: In BytecodeParser::insertValue, do not allocate
a new ValueTab entry just because some value exists with a large type. This
dramatically reduces the number of allocations/deallocations performed by the
bytecode reader, and speeds up parsing of Kimwitu++ from 34s to 17s. This is
to help address PR127
llvm-svn: 10085
Correctly parse the Long Filename section of the archive.
When reading in archive members, set their ModuleIDs to
"ARCHIVENAME(MEMBERNAME)", as is traditional.
llvm-svn: 10043
Rename SlotCalculator::getValSlot() to SlotCalculator::getSlot(),
SlotCalculator::insertValue() to SlotCalculator::getOrCreateSlot(),
SlotCalculator::insertVal() to SlotCalculator::insertValue(), and
SlotCalculator::doInsertVal() to SlotCalculator::doInsertValue().
llvm-svn: 9190
For now, we translate linkonce into weak linkage in the bytecode format because
we don't have enough bits to represent it. We will rev the bytecode version
soon anyways, so this will be fixed in the near future.
llvm-svn: 9170
are ordered by name, not by slot, so the previous solution wasn't any good.
On a large testcase, this reduces time to parse from 2.17s to 1.58s.
llvm-svn: 9002
changes:
* BytecodeReader::getType(...) used to return a null pointer
on error. This was only checked about half the time. Now we convert
it to throw an exception, and delete the half that checked for error.
This was checked in before, but psmith crashed and lost the change :(
* insertValue no longer returns -1 on error, so callers don't need to
check for it.
* Substantial rewrite of InstructionReader.cpp, to use more efficient,
simpler, data structures. This provides another 5% speedup. This also
makes the code much easier to read and understand.
llvm-svn: 8984
new, simpler, ForwardReferences data structure. This is just the first
simple replacement, subsequent changes will improve the code more.
This simple change improves the performance of loading a file from HDF5
(contributed by Bill) from 2.36s to 1.93s, a 22% improvement. This
presumably has to do with the fact that we only create ONE placeholder for
a particular forward referenced values, and also may be because the data
structure is much simpler.
llvm-svn: 8979
in the bytecode parser. Before we tried to shoehorn basic blocks into the
"getValue" code path with other types of values. For a variety of reasons
this was a bad idea, so this patch separates it out into its own data structure.
This simplifies the code, makes it fit in 80 columns, and is also much faster.
In a testcase provided by Bill, which has lots of PHI nodes, this patch speeds
up bytecode parsing from taking 6.9s to taking 2.32s. More speedups to
follow later.
llvm-svn: 8977
* Instead of a #define, use inline function
* Fix the name on the #define, errr... now inline function to be more logical:
it doesn't CHECK the alignment, it PERFORMS the alignment
* To get string name of a Type*, use getDescription(), not getName()
llvm-svn: 8683
* Make sure we align the buffer we're given
* Do not let exceptions propagate when the caller asks for a Module*
* Add doxygenified comments to wrapper functions
llvm-svn: 8682
- no more passing around a string pointer to set errors
- no more returning booleans and checking for errors, we use C++ exceptions
* Broke functionality into 2 new classes, one reads from file, one from a stream
* Implemented lazy function streaming - the parser can read in a function at-a-time
llvm-svn: 8671
- Null values are implicitly encoded instead of explicitly, this makes
things more compact!
- More compactly represent ConstantPointerRefs
- Bytecode files are represented as:
Header|GlobalTypes|GlobalVars/Function Protos|Constants|Functions|SymTab
instead of
Header|GlobalTypes|Constants|GlobalVars/Function Protos|Functions|SymTab
which makes a lot of things simpler.
Writer changes:
- We now explictly encode versioning information in the bytecode files.
- This allows new code to read bytecode files produced by old code, but
new bytecode files can have enhancements such as the above. Although this
makes the reader a bit more complex (having to deal with old formats), the
writer only needs to be able to produce the most recent version.
llvm-svn: 5749
- Fix problems where the constant table would not get updated when
resolving constants causes other constants to change.
Changes to the V2 bytecode format
- Null values are implicitly encoded instead of explicitly, this makes
things more compact!
- More compactly represent ConstantPointerRefs
- Bytecode files are represented as:
Header|GlobalTypes|GlobalVars/Function Protos|Constants|Functions|SymTab
instead of
Header|GlobalTypes|Constants|GlobalVars/Function Protos|Functions|SymTab
which makes a lot of things simpler.
Changes to the reader:
- Function loading code is much simpler. We now no longer make function
PlaceHolderHelper objects to be replaced with real functions.
llvm-svn: 5748
reading bytecode files with > 255 types in them, but only when optimization
is enabled. This was caused by GCC shrinking an enum to a single byte
instead of a whole word.
llvm-svn: 4266
Handle forward referenced constants in a general way. This fixes bug:
Assembler/2002-10-13-ConstantEncodingProblem.llx and allows the SPEC
197.parser benchmark to be built
llvm-svn: 4161
* Renamed StatisticReporter.h/cpp to Statistic.h/cpp
* Broke constructor to take two const char * arguments instead of one, so
that indendation can be taken care of automatically.
* Sort the list by pass name when printing
* Make sure to print all statistics as a group, instead of randomly when
the statistics dtors are called.
* Updated ProgrammersManual with new semantics.
llvm-svn: 4002
- ParseConstantPool was resolving reference to value using the function
slot # instead of the global slot #.
- Bytecode reader changes:
- Remove the failure<> template from Bytecode Reader
- Remove extraneous #includes
- s/method/function/ a bit
- Eliminate the fwdRefs class that just added abstraction where it was not
needed, making things more complex.
- Use a vector instead of a list for function signatures.
llvm-svn: 3366
* Correctly delete TypeHandles in AsmParser. In addition to not leaking
memory, this prevents a bug that could have occurred when a type got
resolved that the constexpr was using
* Check for errors in the AsmParser instead of hitting assertion failures
deep in the code
* Simplify the interface to the ConstantExpr class, removing unneccesary
parameters to the ::get* methods.
* Rename the 'getelementptr' version of ConstantExpr::get to
ConstantExpr::getGetElementPtr
llvm-svn: 3161
* Correctly delete TypeHandles in AsmParser. In addition to not leaking
memory, this prevents a bug that could have occurred when a type got
resolved that the constexpr was using
* Check for errors in the AsmParser instead of hitting assertion failures
deep in the code
* Simplify the interface to the ConstantExpr class, removing unneccesary
parameters to the ::get* methods.
* Rename the 'getelementptr' version of ConstantExpr::get to
ConstantExpr::getGetElementPtr
llvm-svn: 3160
may be constructed by expressions of other types (and so the
contents of the primitive type planes must come after all types).
Use a helper function outputConstantsInPlane in outputConstants to
do this.
llvm-svn: 2898
Method::inst_* is now in llvm/Support/InstIterator.h
GraphTraits specializations for BasicBlock and Methods are now in llvm/Support/CFG.h
llvm-svn: 1746
