RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-22 04:22:57 +02:00
Code Issues Projects Releases Wiki Activity
4,906 Commits 4 Branches 5 Tags 704 MiB
8f7c7a40c5
Commit Graph

412 Commits

Author SHA1 Message Date
Misha Brukman
1264a03c8a 'graph' is spelled without a 'c'. 
Also added Statistic counters for NoAlias and MayAlias.

llvm-svn: 4972
 2002-12-12 05:34:10 +00:00
Chris Lattner
8d4b4b020e Remove #includes 
llvm-svn: 4968
 2002-12-12 03:47:27 +00:00
Vikram S. Adve
c9617fde54 Iterator that enumerates the ProgramDependenceGraph (PDG) for a function, 
i.e., enumerates all data and control dependences for the function.

llvm-svn: 4958
 2002-12-08 14:13:19 +00:00
Vikram S. Adve
90bc3110b6 An explicit representation of dependence graphs, and a pass that 
computes a dependence graph for data dependences on memory locations
using interprocedural Mod/Ref information.

llvm-svn: 4957
 2002-12-08 13:26:29 +00:00
Chris Lattner
5f4d7d2801 Add total instruction, bb, & function counts 
llvm-svn: 4954
 2002-12-07 23:24:24 +00:00
Vikram S. Adve
773fe59c79 Two bug fixes: 
(1) Make entries for Constant values in the ScalarMap.
(2) Set MOD bit for the node pointed to by the
    argument of a free instruction.

llvm-svn: 4948
 2002-12-06 21:17:10 +00:00
Vikram S. Adve
cb7daf77e8 Fix several related bugs in DSNode::mergeWith() caused by the 
fact that the incoming nodes may be merged away at intermediate
steps.  Use an extra level of indirection via DSNodeHandles
to track the nodes being merged.  All this now happens in a
static helper function MergeNodes().

llvm-svn: 4947
 2002-12-06 21:15:21 +00:00
Vikram S. Adve
ab3a68a7c3 Cute bug fix: when moving links from N to this, some links could have 
been missed if node *this got merged away due to recursive merging!
Also, links were not moved correctly if a node is collapsed.

llvm-svn: 4933
 2002-12-05 17:17:26 +00:00
Chris Lattner
18b9a192e1 Simplify code 
llvm-svn: 4886
 2002-12-03 19:40:16 +00:00
Vikram S. Adve
b0371788d9 Fix logical error in TD pass: we should clear Mod/Ref bits of each caller 
before inlining their graphs into a function.  To support this,
added flags to CloneFlags to strip/keep Mod/Ref bits.

llvm-svn: 4836
 2002-11-27 17:41:13 +00:00
Vikram S. Adve
8788221a72 (1) Bug fix that was causing nodes with dangling references to be freed. 
We run removeDeadNodes() on the TD graph up front before using it.
(2) Major enhancement to printing of results: now we list the actual objects
    that are mod/ref instead of just printing the bit vectors.
Also an important bug fix in TDDataStructures pass (no change here):
clear Mod/Ref bits of callers before inlining into a function.

llvm-svn: 4833
 2002-11-27 17:37:46 +00:00
Vikram S. Adve
7bbda247e2 Keep global nodes in each DS Graph (by forcing them to be marked live). 
llvm-svn: 4831
 2002-11-25 18:21:25 +00:00
Chris Lattner
ad3955d6a5 Add peak memory usage measurement stuff 
Add structure padding optimizations

llvm-svn: 4749
 2002-11-18 21:44:46 +00:00
Chris Lattner
f179a40a00 Add peak memory usage support 
llvm-svn: 4748
 2002-11-18 21:44:19 +00:00
Chris Lattner
25a8814af0 Add stats 
llvm-svn: 4747
 2002-11-18 21:42:45 +00:00
Chris Lattner
38bc372a49 Add hack to only consider indirect calls indirect if they do more than cast 
their source function

llvm-svn: 4723
 2002-11-17 22:17:12 +00:00
Chris Lattner
322901a840 Add MaxSCC statistics 
llvm-svn: 4722
 2002-11-17 22:16:28 +00:00
Chris Lattner
acfc11664a Count CallInsts correctly, remove unneccesary initializers 
S: ----------------------------------------------------------------------

llvm-svn: 4721
 2002-11-17 22:15:40 +00:00
Dinakar Dhurjati
9da9f0534e Gives the count for various instructions. 
llvm-svn: 4706
 2002-11-13 18:22:13 +00:00
Vikram S. Adve
2dfc11dc3b Pass to compute various statisics related to DSGraphs. 
For now, this just computes the #indirect call sites and
the avg. #callees per indirect call site (actually it prints
both totals and the average is their ratio).

llvm-svn: 4705
 2002-11-13 15:41:00 +00:00
Chris Lattner
b257e87db5 Inline graphs from outside the SCC into the SCC before SCC resolution starts 
llvm-svn: 4701
 2002-11-12 15:58:08 +00:00
Chris Lattner
cd37857ad0 Fix two bugs: 
* The globals vector was getting broken and unsorted, this caused vortex
   to get badly pessimized
 * Node offset handling was being handled really poorly, and in particular
   we were not merging types with offsets right.  This causes several graphs
   to be non-merged.

llvm-svn: 4699
 2002-11-12 07:20:45 +00:00
Chris Lattner
3b303e5ab8 Fix bug 
llvm-svn: 4697
 2002-11-11 22:23:56 +00:00
Chris Lattner
f56a5282d2 Handle a mismatch between # function args and call site args 
llvm-svn: 4696
 2002-11-11 21:36:05 +00:00
Chris Lattner
6fcb6f1472 Elimiante calls to a node with nothing in it. 
llvm-svn: 4695
 2002-11-11 21:35:38 +00:00
Chris Lattner
0052b69a32 Complete rewrite of BU code to use Tarjan's SCC finding algorithm to drive 
the algorithm instead of hand coded depth first iteration

llvm-svn: 4694
 2002-11-11 21:35:13 +00:00
Chris Lattner
fb44e1ff28 Mark stuff reachable by _AUX_ calls as incomplete in the BU graph 
llvm-svn: 4690
 2002-11-11 03:36:55 +00:00
Chris Lattner
f05d126a68 Fix infinite loop in the BU algorithm. Unfortunately this dies a serious 
death when handling moderately sized SCC's, but what can you do

llvm-svn: 4689
 2002-11-11 00:01:34 +00:00
Chris Lattner
a7302da4e8 Print the right call set size 
llvm-svn: 4688
 2002-11-11 00:01:02 +00:00
Chris Lattner
41bf6e1bc6 Use call site mergeWith method to simplify code 
llvm-svn: 4687
 2002-11-11 00:00:46 +00:00
Chris Lattner
0fe1af497d Fix a bug that could trigger when varargs call sites had non-matching number of arguments 
llvm-svn: 4683
 2002-11-10 07:46:08 +00:00
Chris Lattner
928e9c012b * Dramatically rework liveness evaluation. 
* Implement the first step of the Globals graph: Deleting nodes from function
  graphs.  In practice, these nodes need to be moved to the globals graph, but
  this will be taken care of later.  Note that the graphs computed right now are
  not strictly correct!

llvm-svn: 4681
 2002-11-10 06:59:55 +00:00
Chris Lattner
f5667969d1 Honor the shouldPrintAuxCalls flag 
llvm-svn: 4678
 2002-11-10 06:53:59 +00:00
Chris Lattner
2597c1158f Initialize PrintAuxCalls member 
llvm-svn: 4677
 2002-11-10 06:53:38 +00:00
Chris Lattner
a30663385f * Bottom-Up graphs print the Aux call vector 
* Significantly improve DEBUG output
* Aggressively fold calls together if we inlined a graph that provides
  call nodes.
* Add a bailout if the current graph has over 200 call nodes in it, this is
  a really whacky case that should never happen.

llvm-svn: 4675
 2002-11-10 06:52:47 +00:00
Chris Lattner
1fe76d2385 eliminate the ability to remove global nodes from deadNodeElminate... for now. 
This slows stuff down a bit, but it should get much better before it gets any
worse.

llvm-svn: 4666
 2002-11-09 22:07:02 +00:00
Chris Lattner
d9c732c581 Remove obsolete code 
llvm-svn: 4665
 2002-11-09 21:41:44 +00:00
Chris Lattner
20d264e16b Fix logic 
llvm-svn: 4664
 2002-11-09 21:40:58 +00:00
Chris Lattner
0752756a6a Add globals graphs to all three passes 
llvm-svn: 4663
 2002-11-09 21:12:07 +00:00
Chris Lattner
f6ea37ce99 Move maskNodeTypes to header file 
llvm-svn: 4661
 2002-11-09 21:02:26 +00:00
Chris Lattner
e726eb5ea7 Clean up DSGraph::removeDeadNodes interface 
llvm-svn: 4660
 2002-11-09 21:00:49 +00:00
Chris Lattner
3e54e8f3ca Make removeTriviallyDeadNodes a private interface of DSGraph 
llvm-svn: 4659
 2002-11-09 20:55:24 +00:00
Chris Lattner
082bbe3069 Don't put constants into the scalar map! 
llvm-svn: 4657
 2002-11-09 20:14:03 +00:00
Chris Lattner
1c5c03d6f7 Add initial support for a globals graph 
llvm-svn: 4656
 2002-11-09 20:01:01 +00:00
Chris Lattner
5a11bf70c3 Tighten up array handling 
llvm-svn: 4655
 2002-11-09 19:25:27 +00:00
Chris Lattner
6172544663 #include Datastructure.h first 
llvm-svn: 4654
 2002-11-09 19:25:10 +00:00
Chris Lattner
0141841bc4 Fix warning 
llvm-svn: 4649
 2002-11-09 00:49:43 +00:00
Chris Lattner
b3580d5111 Fix warning & release build 
llvm-svn: 4648
 2002-11-09 00:49:05 +00:00
Chris Lattner
608bb157f7 Fold arrays down to a single element. This causes huge wins on some benchmarks 
for example: 197.parser (64M->14M), 164.gzip (14M->2.7M).  The actual graphs
represented should not change at all.

llvm-svn: 4643
 2002-11-08 22:49:57 +00:00
Chris Lattner
4baa01ba61 ONLY merge in the aux call sites, this causes a HUGE speedup 
llvm-svn: 4640
 2002-11-08 22:27:25 +00:00