the type is analyzed. Instead, only compute it when requested (with
getDescription), and cached for reuse later.
This dramatically speeds up LLVM in general because these descriptions almost
_never_ need to be constructed. The only time they are used is when a type is
<<'d. Printing of modules by themselves uses other code to print symbolic
types when possible, so these descriptions are really only used for debugging.
Also, this fixes the particularly bad case when lots of types get resolved to
each other, such as during linking of large programs. In these cases, the type
descriptions would be repeatedly recomputed and discarded even though: A. noone
reads the description before it gets resolved, and B. many many resolutions
happen at intermediate steps, causing a HUGE waste of time.
Overall, this makes the getTypeDesc function much more light-weight, and fixes
bug: Assembler/2002-07-08-HugePerformanceProblem.llx, which went from taking
1048.770u/19.150s (which is 17.5 MINUTES, on apoc), to taking 0.020u/0.000s,
which is a nice little speedup. :)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8320 91177308-0d34-0410-b5e6-96231b3b80d8
* Remove ambiguous code that caused problems with understanding how stuff
worked.
* Implement DerivedType::typeIsRefined right. Add a big comment talking
about it.
* The removeUserFromConcrete method already checks to see if the type is
not abstract, so no need to duplicate the test all over this code
This checkin makes all of the type parsing related failures work. All
Assembler Regression tests now pass!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@2127 91177308-0d34-0410-b5e6-96231b3b80d8
That was due to the StructType object not getting called multiple times
to remove itself from the ATU list for an object, because it stopped
getting queried due to this wrong test
* Improve debug output when DEBUG_MERGE_TYPES is on
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@2124 91177308-0d34-0410-b5e6-96231b3b80d8
* Refactor a bunch of code. Types now only remove one subelement at a time
when they are told that they have changed
* Improve debugging output, add more assertions...
* FIX Bugs:
* test/Regression/Assembler/2002-04-04-PureVirtMethCall.ll
* test/Regression/Assembler/2002-04-04-PureVirtMethCall2.ll
* The change to fix these bugs was the to ValTypeBase::refineAbstractType
method. Basically we #if 0'd out the chunk of code there and make the
table get reorganized EVEN IF the type has the same pointer. Merging
opportunities were being missed!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@2121 91177308-0d34-0410-b5e6-96231b3b80d8
* Fix a bug in DerivedType::refineAbstractTypeTo that caused
test/Regression/Assembler/2002-04-05-TypeParsing.ll to fail.
The added comment describes the fix
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@2118 91177308-0d34-0410-b5e6-96231b3b80d8
1. Forward propogate a loaded constant to make debugging slightly less annoying
2. Fix an extremely nasty hard to find bug that really irritated me for the better
part of 6 hours and was causing linking to die and bizarre and mysterious ways.
Things should be much more stable now that this one liner has been fixed. GRR
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@971 91177308-0d34-0410-b5e6-96231b3b80d8
Remove lockty, remove fillerty
Make type lookup more efficient
Support shared generic factory code
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@465 91177308-0d34-0410-b5e6-96231b3b80d8