Commit Graph

8183 Commits

Author SHA1 Message Date
Chris Lattner
69091be83b This changes the PromoteMemToReg function to create "pruned" SSA form, not
"minimal" SSA form (in other words, it doesn't insert dead PHIs).  This
speeds up the mem2reg pass very significantly because it doesn't have to
do a lot of frivolous work in many common cases.

In the 252.eon function I have been playing with, this doesn't even insert
the 120 PHI nodes that it used to which were trivially dead (in the process
of promoting 356 alloca instructions overall).  This speeds up the mem2reg
pass from 1.2459s to 0.1284s.  More significantly, the DCE pass used to take
2.4138s to remove the 120 dead PHI nodes that mem2reg constructed, now it
takes 0.0134s (which is the time to scan the function and decide that there
is nothing dead).  So overall, on this one function, we speed things up a
total of 3.5179s, which is a 24.8x speedup!  :)

This change is tested by the Mem2Reg/2003-10-05-DeadPHIInsertion.ll test,
which now passes.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8884 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 22:19:20 +00:00
Chris Lattner
43f820d1f7 Change the interface to PromoteMemToReg to also take a DominatorTree
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8883 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 21:20:13 +00:00
Chris Lattner
5b15deedc0 new testcase
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8882 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 21:11:57 +00:00
Chris Lattner
24011be956 Speed up the mem2reg transform for allocas which are only read/written in a single
basic block.  This is amazingly common in code generated by the C/C++ front-ends.
This change makes it not have to insert ANY phi nodes, whereas before it would insert
a ton of dead ones which DCE would have to clean up.

Thus, this fix improves compile-time performance of these trivial allocas in two ways:
  1. It doesn't have to do the walking and book-keeping for renaming
  2. It does not insert dead phi nodes for them which would have to
     subsequently be cleaned up.

On my favorite testcase from 252.eon, this special case handles 305 out of
356 promoted allocas in the function.  It speeds up the mem2reg pass from 7.5256s
to 1.2505s.  It inserts 677 fewer dead PHI nodes, which speeds up a subsequent
-dce pass from 18.7524s to 2.4806s.

There are still 120 trivially dead PHI nodes being inserted for variables used
in multiple basic blocks, but they are not handled by this patch.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8881 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 20:54:03 +00:00
Chris Lattner
55e3459eda The support directory no longer exists
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8880 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:33:27 +00:00
Chris Lattner
3fd064c384 Oops, missed the . files
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8879 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:32:56 +00:00
Chris Lattner
5516347535 Move support/lib into lib/Support
Move support/tools into utils


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8878 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:32:12 +00:00
Chris Lattner
091bbbada3 Revert back to keeping Burg and TableGen in the utils directory
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8876 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:28:27 +00:00
Chris Lattner
0a92ebf557 Move support/tools/* back into utils
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8875 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:27:59 +00:00
Chris Lattner
f708c7901d Move support/lib/Support back to lib/Support
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8874 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:24:38 +00:00
Chris Lattner
35fccb7e42 The sparc backend now support invoke, but still doesn't support unwind
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8873 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:17:10 +00:00
Chris Lattner
d324e2501c Add support for the Invoke instruction by using the LowerInvoke pass
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8872 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:16:09 +00:00
Chris Lattner
c58c169eda Instead of hacking in custom support for Invoke/Unwind, use the LowerInvoke pass
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8871 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:15:47 +00:00
Chris Lattner
5e5252b4c9 Add new prototype for createLowerInvokePass(). Make simplifycfg be a
functionpass


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8870 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:15:13 +00:00
Chris Lattner
86e4445ec3 Initial checkin of the LLVM->LLVM transform to support code generators which
do not support stack unwinding yet


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8869 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:14:42 +00:00
Chris Lattner
b55803a9f4 simplify-cfg is really a function pass
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8868 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 19:14:16 +00:00
Chris Lattner
bede31ff3a Be more careful handling PHI nodes, which might be of potentially high degree.
This reduces the time to verify a function from eon with a large number of
large PHI nodes from 22996s (6.38 hours) to 10.5499s


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8866 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 17:44:18 +00:00
Chris Lattner
92e4975af4 The first PHI node may be null, scan for the first non-null one
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8865 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 05:34:39 +00:00
Chris Lattner
3c881cb4ce The VersionNumbers vector is only used during PHI placement. Turn it into an argument, allowing us to get rid of the vector.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8864 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 04:33:22 +00:00
Chris Lattner
afa060ea3f * Update file header comment
*** Revamp the code which handled unreachable code in the function.  Now the
    code is much more efficient for high-degree basic blocks, such as those
    that occur in the 252.eon SPEC benchmark.

For the interested, the time to promote a SINGLE alloca in _ZN7mrScene4ReadERSi
function used to be > 3.5s.  Now it is < .075s.  The function has a LOT of
allocas in it, so it appeared to be infinite looping, this should make it much
nicer.  :)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8863 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 04:26:39 +00:00
Chris Lattner
521c16aadd Simplify the loop a bit
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8862 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 03:45:44 +00:00
Chris Lattner
92581c24a3 There is no need for separate WriteSets and PhiNodeBlocks lists. It is just a
work-list of value definitions.  This allows elimination of the explicit
'iterative' step of the algorithm, and also reuses temporary memory better.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8861 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 03:39:10 +00:00
Chris Lattner
036f1e7478 The PhiNodes 2D vector is only used during PHI node placement. It doesn't
need to be an instance variable!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8860 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 03:26:25 +00:00
Chris Lattner
9e38fbf57f * Document instance vars better
* Fuse two parallel loops
* Use a more specific type for AllocaLookup


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8859 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 03:16:07 +00:00
Chris Lattner
9157f041ab Two small cleanups/speedups:
* Do not insert a new entry into NewPhiNodes during the rename pass if there are no PHIs in a block.
 * Do not compute WriteSets in parallel


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8858 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 02:37:36 +00:00
Chris Lattner
0fa157127f * Minor cleanups
* Eliminate the KillList instance variable, instead, just delete loads and
  stores as they are "renamed", and delete allocas when they are done
* Make the 'visited' set an instance variable to avoid passing it on the stack.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8857 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 01:52:53 +00:00
Chris Lattner
27db7e0472 Output a very high-precision number
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8856 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 00:41:07 +00:00
Chris Lattner
dd76acac3f A couple of minor code cleanups.
Print literal doubles using ftostr instead of <<, because it yields higher precision numbers.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8855 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 00:40:51 +00:00
Chris Lattner
ed468e37a1 Type tables are now AbstractTypeUsers. This allows them to merge together
constants as necessary due to type resolution.  With this change, the
following spec benchmarks now link: 176.gcc, 177.mesa, 252.eon,
253.perlbmk, & 300.twolf.  IOW, all SPEC INT and FP benchmarks now link.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8853 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 00:17:43 +00:00
Chris Lattner
5133a5cf2e Add some new methods to forward to
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8852 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-05 00:13:28 +00:00
Chris Lattner
00413e3d63 Rename AbstractModuleProvider -> ModuleProvider, to match the header file name,
and because, while the class used by the interface is abstract, the actual
concept is not.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8850 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-04 20:14:59 +00:00
Chris Lattner
9e460f23bb Transform two methods to return pointers directly instead of returning them
as 'by reference' arguments.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8849 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-04 20:00:03 +00:00
Chris Lattner
51ca860bda Use V for values, not D.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8848 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-04 19:29:21 +00:00
Chris Lattner
a983359f59 Do not leak the ModuleProvider if releaseModule() throws.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8847 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-04 19:19:37 +00:00
Chris Lattner
ded46e27cc Constants are not getting merged together correctly when types are resolved
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8846 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-04 18:27:51 +00:00
Chris Lattner
70b06fc717 Minor random speedup: make Constant::getNullValue only call ConstantFOO::get
once!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8845 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-03 19:34:51 +00:00
Chris Lattner
23c3288ce9 Explicit copy ctors are no longer needed
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8844 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-03 19:02:41 +00:00
Chris Lattner
af6f93ce35 Minor cleanups
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8843 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-03 18:57:54 +00:00
Chris Lattner
7685ac8d35 This checkin basically amounts to a complete rewrite of the type-resolution
machinery.  This dramatically simplifies how things works, removes irritating
little corner cases, and overall improves speed and reliability.

Highlights of this change are:

1. The exponential algorithm built into the code is now gone.  For example
   the time to disassemble one bytecode file from the mesa benchmark went
   from taking 12.5s to taking 0.16s.
2. The linker bugs should be dramatically reduced.  The one remaining bug
   has to do with constant handling, which I actually introduced in
   "union-find" checkins.
3. The code is much easier to follow, as a result of fewer special cases.
   It's probably also smaller.  yaay.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8842 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-03 18:46:24 +00:00
John Criswell
4797826e17 Changed all of these tests to be TestRunner tests (or, at least they can be
TestRunner tests).
This makes creating the new test database class easier to implement.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8841 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-03 18:42:25 +00:00
Chris Lattner
9ad4dac582 These methods are dead, remove them. Because the bodies will soon be
ressurected, just ifdef them out for now.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8840 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-03 18:39:57 +00:00
Chris Lattner
1ad37b2b59 These methods are dead, remove them
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8839 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-03 18:39:40 +00:00
John Criswell
cf21d4fe70 Renamed these tests so that they don't contain extra periods.
This interferes with QMTest and the new test database.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8838 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-03 18:24:47 +00:00
John Criswell
aa5ff246e9 Added links to the C and C++ frontends.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8836 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-03 13:48:27 +00:00
John Criswell
700c82ddc4 Adding basic documentation for the LLVM C and C++ frontends.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8835 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-03 13:45:55 +00:00
Chris Lattner
8ef852f4ea Dramatically simplify DerivedType::refineAbstractTypeToInternal
This makes use of the new PATypeHolder's to keep types from being deleted
prematurely, instead of the wierd "self reference" garbage.  This is easier
to understand and more efficient as well.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8834 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-03 04:48:21 +00:00
Chris Lattner
1c5164e9cf Make the PATypeHolder use a simple union-find implementation to handle
merging of types.  This makes it MUCH more efficient than before, also
making things simpler.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8833 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-02 23:35:57 +00:00
Chris Lattner
c7b6f03501 There is no need for BytecodeParser to be an AbstractTypeUser. Instead, it
can just use PATypeHolders


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8832 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-02 20:26:18 +00:00
Chris Lattner
3200219843 There is no reason for Value to be an AbstractTypeUser. This just makes things
significantly more complete.  Instead, just make DerivedType's AbstractTypeUser's,
and make Value contain a PATypeHolder.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8828 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-02 19:44:40 +00:00
Chris Lattner
0dde36c4c7 There is no reason for Value to be an AbstractTypeUser. This just makes things
significantly more complete.  Instead, just make DerivedType's AbstractTypeUser's,
and make Value contain a PATypeHolder.  This will also be more efficient in the
future.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@8827 91177308-0d34-0410-b5e6-96231b3b80d8
2003-10-02 19:44:23 +00:00