5758134505
Add a simple implementation of Andersen's interprocedural pointer analysis
...
llvm-svn: 13666
2004-05-23 21:00:47 +00:00
159ed21b67
Inline both direct and indirect callees in the CBU phase because
...
a direct callee may have indirect callees and so may have changed.
llvm-svn: 13649
2004-05-23 08:00:34 +00:00
17fcb67a20
Fine grainify namespacification
...
llvm-svn: 13436
2004-05-09 06:22:29 +00:00
e330adf842
Move the stuff that fixes the size, orientation & fonts of graphs to
...
the debugging functions that call "dot". These fixed settings have
various problems: for example, the fixed size that is set in the graph
traits classes is not appropriate for turning the dot file into a PNG,
and if TrueType font rendering is being used, the 'Courier' TrueType font
may not be installed. It seems easy enough to specify these things on the
command line, anyhow.
llvm-svn: 13366
2004-05-05 06:10:06 +00:00
600f2045a9
Add stub support for reading BBTraces.
...
llvm-svn: 13352
2004-05-04 17:11:14 +00:00
34c13fbe94
Share ProfilingType enum with the C profiling runtime libraries.
...
llvm-svn: 13346
2004-05-04 16:53:07 +00:00
6942946132
Fix a problem with double freeing memory. For some reason, CallGraph is not
...
acting like a normal pass. :(
llvm-svn: 13318
2004-05-02 16:06:18 +00:00
929291aabb
Plug a minor memory leak
...
llvm-svn: 13317
2004-05-02 07:31:34 +00:00
372d5bc9ab
Wrapped code and comments at 80 cols; doxygenified some comments.
...
llvm-svn: 13264
2004-04-29 04:05:30 +00:00
a70ae90722
Reorder #includes as per style guide.
...
llvm-svn: 13263
2004-04-29 04:04:47 +00:00
7e66438579
Send text and numbers directly to CachedWriter's contained ostream.
...
llvm-svn: 13243
2004-04-28 18:52:43 +00:00
2d3a7a6ff0
Changes to fix up the inst_iterator to pass to boost iterator checks. This
...
patch was graciously contributed by Vladimir Prus.
llvm-svn: 13185
2004-04-27 15:13:33 +00:00
104341f31e
Add functions that return instances of these printer passes
...
llvm-svn: 13175
2004-04-26 16:27:08 +00:00
f374b3f304
If an object is not in the scalar map then it must be a global from another
...
graph.
llvm-svn: 13173
2004-04-26 14:44:08 +00:00
05ef97f994
Eliminate all of the SCEV Expansion code which is really part of the
...
IndVars pass, not part of SCEV *analysis*.
llvm-svn: 13134
2004-04-23 21:29:03 +00:00
99504890b7
Pass the callgraph not the module
...
llvm-svn: 13087
2004-04-20 21:52:26 +00:00
8d0838130c
Add the ability for SCC passes to initialize and finalize themselves
...
llvm-svn: 13084
2004-04-20 21:30:06 +00:00
663ebc3ec6
It's not just a printer, it's actually an analysis too
...
llvm-svn: 13064
2004-04-19 03:42:32 +00:00
32447c55fe
Remove code to update loop depths
...
llvm-svn: 13058
2004-04-19 03:02:09 +00:00
26ba2f5fe8
Add new method
...
llvm-svn: 13050
2004-04-18 22:45:27 +00:00
d6ce359d3b
Fix computation of exit blocks
...
llvm-svn: 13047
2004-04-18 22:21:41 +00:00
d72c3eb54e
Change the ExitBlocks list from being explicitly contained in the Loop
...
structure to being dynamically computed on demand. This makes updating
loop information MUCH easier.
llvm-svn: 13045
2004-04-18 22:14:10 +00:00
e375a4fdc2
Implement method
...
llvm-svn: 13036
2004-04-18 06:54:48 +00:00
1472c63fb9
Add a new method, add a check missing that caused a segfault if a loop didn't
...
have a canonical indvar
llvm-svn: 13032
2004-04-18 05:38:05 +00:00
dd73047673
Add the ability to compute exit values for complex loop using unanalyzable
...
operations. This allows us to compile this testcase:
int main() {
int h = 1;
do h = 3 * h + 1; while (h <= 256);
printf("%d\n", h);
return 0;
}
into this:
int %main() {
entry:
call void %__main( )
%tmp.6 = call int (sbyte*, ...)* %printf( sbyte* getelementptr ([4 x sbyte]* %.str_1, long 0, long 0), int 364 ) ; <int> [#uses=0]
ret int 0
}
This testcase was taken directly from 256.bzip2, believe it or not.
This code is not as general as I would like. Next up is to refactor it
a bit to handle more cases.
llvm-svn: 13019
2004-04-17 22:58:41 +00:00
4021d1af5a
Add the ability to compute trip counts that are only controlled by constants
...
even if the loop is using expressions that we can't compute as a closed-form.
This allows us to calculate that this function always returns 55:
int test() {
double X;
int Count = 0;
for (X = 100; X > 1; X = sqrt(X), ++Count)
/*empty*/;
return Count;
}
And allows us to compute trip counts for loops like:
int h = 1;
do h = 3 * h + 1; while (h <= 256);
(which occurs in bzip2), and for this function, which occurs after inlining
and other optimizations:
int popcount()
{
int x = 666;
int result = 0;
while (x != 0) {
result = result + (x & 0x1);
x = x >> 1;
}
return result;
}
We still cannot compute the exit values of result or h in the two loops above,
which means we cannot delete the loop, but we are getting closer. Being able to
compute a constant trip count for these two loops will allow us to unroll them
completely though.
llvm-svn: 13017
2004-04-17 18:36:24 +00:00
174633b078
Include <cmath> for compatibility with gcc 3.0.x (the system compiler on
...
Debian.)
llvm-svn: 12986
2004-04-16 15:57:32 +00:00
d9dc425376
add some helpful methods. Rearrange #includes to proper order
...
llvm-svn: 12960
2004-04-15 15:16:02 +00:00
b4f681b42b
Factor a bunch of classes out into a public header
...
llvm-svn: 12958
2004-04-15 15:07:24 +00:00
a4e4a63856
Unbreak the build
...
llvm-svn: 12956
2004-04-15 14:17:43 +00:00
d420fe63ea
Implement a FIXME: if we're going to insert a cast, we might as well only
...
insert it once!
llvm-svn: 12955
2004-04-14 22:01:22 +00:00
8a9fd94cfe
This is a trivial tweak to the addrec insertion code: insert the increment
...
at the bottom of the loop instead of the top. This reduces the number of
overlapping live ranges a lot, for example, eliminating a spill in an important
loop in 183.equake with linear scan.
I still need to make the exit comparison of the loop use the post-incremented
version of this variable, but this is an easy first step.
llvm-svn: 12952
2004-04-14 21:11:25 +00:00
55b7ef5a81
Add some methods that are useful for updating loop information.
...
llvm-svn: 12871
2004-04-12 20:26:17 +00:00
8b6db18ac3
Change the call graph class to have TWO external nodes, making call graph
...
SCC passes much more useful. In particular, this should fix the incredibly
stupid missed inlining opportunities that the inliner suffered from.
llvm-svn: 12860
2004-04-12 05:36:32 +00:00
c9e37d7cc9
Hrm, operator new and new[] do not belong here. We should not CSE them! :)
...
llvm-svn: 12859
2004-04-12 05:16:42 +00:00
403abb824f
operator new & operator new[] do not kill any legal memory locations.
...
llvm-svn: 12833
2004-04-11 18:16:34 +00:00
c5fad355f8
Allow clients to be more efficient.
...
llvm-svn: 12831
2004-04-11 16:43:07 +00:00
4a1b03c773
Add a couple of more functions that cannot access memory (the intrinsics) and
...
don't write to memory
llvm-svn: 12808
2004-04-10 06:55:27 +00:00
74498e1066
Fix a bug Brian found.
...
llvm-svn: 12754
2004-04-07 16:16:11 +00:00
d4f78f270b
Sparc don't got not "sqrtl", bum bum bum
...
llvm-svn: 12670
2004-04-05 19:05:15 +00:00
5ebc25c818
Kill warnings during an optimized compile where assert() disappears.
...
llvm-svn: 12669
2004-04-05 19:00:46 +00:00
29153fc2e5
Fix PR312 and IndVarsSimplify/2004-04-05-InvokeCastCrash.llx
...
llvm-svn: 12668
2004-04-05 18:46:55 +00:00
69193f93b6
Support getelementptr instructions which use uint's to index into structure
...
types and can have arbitrary 32- and 64-bit integer types indexing into
sequential types.
llvm-svn: 12653
2004-04-05 01:30:19 +00:00
8ed3c8aa13
Implement test/Regression/Transforms/GCSE/undefined_load.ll
...
llvm-svn: 12641
2004-04-03 00:45:16 +00:00
0defaa1cbc
Add a break in the default case
...
llvm-svn: 12639
2004-04-03 00:43:03 +00:00
6748cca268
Remove obsolete files
...
llvm-svn: 12633
2004-04-02 20:56:24 +00:00
0916921332
Comment out debugging printouts
...
llvm-svn: 12623
2004-04-02 20:26:46 +00:00
d934c70cf8
Add a new analysis
...
llvm-svn: 12619
2004-04-02 20:23:17 +00:00
82d34eb470
Minor efficiency improvement, finegrainify namespacification
...
llvm-svn: 12517
2004-03-25 22:56:03 +00:00
3022b1bc2c
Fix a HORRIBLY NASTY bug that caused siod to stop working last night.
...
llvm-svn: 12479
2004-03-17 23:22:04 +00:00
8ad948ddbd
Add some missing functions. Make sure to handle calls together in case the
...
client has another VN implementation that can VN calls.
llvm-svn: 12427
2004-03-16 03:41:35 +00:00
7f04ebc858
Ok, the assertion was bogus. Calls that do not read/write memory should not
...
have an alias set, just like adds and subtracts don't.
llvm-svn: 12422
2004-03-15 06:28:07 +00:00
f5c8f4cea6
This assertion is bogus now that calls do not necessarily read/write memory
...
llvm-svn: 12421
2004-03-15 06:24:15 +00:00
53a3587997
Implement CSE of call instructions in the most trivial case. This implements
...
GCSE/call_cse.ll
llvm-svn: 12419
2004-03-15 05:44:59 +00:00
ea42c857d6
Fix a minor bug, implementing GCSE/call_pure_function.ll
...
Also, add some stuff I missed before.
llvm-svn: 12417
2004-03-15 04:18:28 +00:00
21c60f1549
Don't be COMPLETELY pessimistic in the face of function calls
...
llvm-svn: 12413
2004-03-15 04:08:36 +00:00
a67dbd02cf
Deinline some virtual methods, provide better mod/ref answers through the
...
use of the boolean queries
llvm-svn: 12410
2004-03-15 04:07:29 +00:00
d441444234
Pass through the boolean queries
...
llvm-svn: 12409
2004-03-15 04:06:46 +00:00
d82256a7a1
Teach basicaa about some stdc functions.
...
llvm-svn: 12408
2004-03-15 03:36:49 +00:00
652eb53dad
Fix a tiny bug that caused an incorrect assertion failure poolallocating
...
boxed-sim.
llvm-svn: 12358
2004-03-13 01:14:23 +00:00
071a5e5649
Rename the intrinsic enum values for llvm.va_* from Intrinsic::va_* to
...
Intrinsic::va*. This avoid conflicting with macros in the stdlib.h file.
llvm-svn: 12356
2004-03-13 00:24:00 +00:00
f9e69b4553
Fix a couple of minor problems. Because PHI nodes can use themselves, this
...
could cause infinite loops. Also, getUnderlyingObject can return null
llvm-svn: 12351
2004-03-12 23:12:55 +00:00
a036253872
Implement mod/ref analysis for a trivial case where locals don't escape.
...
This comes up when you have a local array on the stack and you never pass
the address of elements around.
llvm-svn: 12349
2004-03-12 22:39:00 +00:00
4483c9b864
Simplify code to process CallSites (thanks to Chris).
...
llvm-svn: 12334
2004-03-12 16:20:49 +00:00
bf28cf6b7d
Evaluate ModRef information in addition to regular ol' pointer analysis.
...
llvm-svn: 12331
2004-03-12 06:15:08 +00:00
362841dccc
Implement getModRefInfo() for DSA to calculate whether a function modifies or
...
references a pointer.
llvm-svn: 12330
2004-03-12 06:14:22 +00:00
96d3b0a6de
Make code more readable.
...
llvm-svn: 12305
2004-03-12 00:58:41 +00:00
61f57617d7
Fix PR284: [indvars] Induction variable analysis violates LLVM invariants
...
llvm-svn: 12275
2004-03-10 21:42:19 +00:00
b26b6fe9a7
implement new method
...
llvm-svn: 12264
2004-03-09 19:37:06 +00:00
9bde783c5c
Switch to using edge profiling information as the basic source of profile info
...
from using basic block counts.
llvm-svn: 12242
2004-03-08 22:04:08 +00:00
bd481d588a
Refactor implementations
...
llvm-svn: 12240
2004-03-08 21:30:35 +00:00
63b49d0574
Import the trace class from the reoptimizer
...
llvm-svn: 12236
2004-03-08 20:57:27 +00:00
5302943ff7
If we have edge counts, we can produce block counts. I've verified that
...
using an edge profile to produce block counts gives the exact same numbers
as using a block count directly.
llvm-svn: 12232
2004-03-08 20:03:52 +00:00
8b8a0641a1
Add initial support for reading edge counts. This will be improved to enable
...
translation of edge counts into block/function counts when possible.
llvm-svn: 12229
2004-03-08 18:20:18 +00:00
84c697d520
Fix a bug handling globals that are constants, but are still external
...
llvm-svn: 12208
2004-03-08 03:52:24 +00:00
2cb2905c96
Fix a minor bug
...
llvm-svn: 12169
2004-03-05 22:04:07 +00:00
d2e88a21b7
Unbreak the build on Sparc.
...
llvm-svn: 12161
2004-03-05 20:04:40 +00:00
6c271edcd7
Fix a bug in a previous checkin that broke 175.vpr
...
llvm-svn: 12128
2004-03-04 21:36:57 +00:00
bc3381f267
Add support for strto* and v*printf
...
llvm-svn: 12127
2004-03-04 21:03:54 +00:00
0ed8162c1a
Add non-crappy support for varargs
...
llvm-svn: 12126
2004-03-04 20:33:47 +00:00
3562ea84ab
Implement a FIXME, improving the efficiency of DSA on povray.
...
This reduces CBU time from 145s -> 122s (debug build), reduces # allocated nodes
from 129420 to 116477.
llvm-svn: 12125
2004-03-04 19:47:04 +00:00
30f94a72f4
Speed up the cbu pass from taking somewhere near the age of the universe to about 90s on povray
...
llvm-svn: 12123
2004-03-04 19:16:35 +00:00
536b131c9c
Fix BU datastructures with povray!
...
The problem was that we were merging a field of a node with a value that was
deleted. Thanks to bugpoint for reducing povray to a nice small 3 function
example. :)
llvm-svn: 12116
2004-03-04 17:06:53 +00:00
8db52980b7
Minor changes, remove some debugging code that got checked in somehow.
...
Make sure to scope the NodeMap passed into cloneInto so that it doesn't point
to nodes that are deleted. Add some FIXME's for future performance enhancements.
llvm-svn: 12115
2004-03-04 17:05:28 +00:00
f48eb7519a
Only clone nodes that are needed in the caller, don't clone ALL aux calls. This improves
...
povray from having ~600K nodes and 300K call nodes to 65K nodes and 25K call nodes.
llvm-svn: 12109
2004-03-04 03:57:53 +00:00
559487301f
Fix a minor bug handling incomplete programs
...
llvm-svn: 12105
2004-03-03 23:00:19 +00:00
d9da2678fc
Fix a DSA bug that caused DSA to generate incredibly huge graphs and take forever to
...
do it on povray. The problem is that we were not copying globals from callees to
callers unless the existed in both graphs. We should have copied them in the case
where the global pointed to a node that was copied as well.
llvm-svn: 12104
2004-03-03 22:01:09 +00:00
5fc189157c
Deinline methods, add fast exit
...
llvm-svn: 12102
2004-03-03 20:55:27 +00:00
140b2cbd28
Fix a node mapping problem that was causing the pool allocator to locally allocate
...
nodes that were globally live, thus breaking programs.
llvm-svn: 12094
2004-03-03 05:34:31 +00:00
2e92b469c8
FINALLY be able to get symbolic type names in the globals graph!
...
llvm-svn: 12082
2004-03-02 21:39:43 +00:00
9e100fc3f7
Really, only if reopen
...
llvm-svn: 12080
2004-03-02 20:46:18 +00:00
c8d23b19fb
Correctly add an array marker on a node when appropriate!
...
llvm-svn: 12055
2004-03-01 19:02:54 +00:00
6f6e0f29ad
Expand on my note-to-self
...
llvm-svn: 12029
2004-03-01 02:44:44 +00:00
06a573f63d
Only clone global nodes between graphs if both graphs have the global.
...
llvm-svn: 11928
2004-02-27 20:05:15 +00:00
98f8ca4017
ADD MORE FUNCTIONS!
...
llvm-svn: 11927
2004-02-27 20:04:48 +00:00
5ef1638da2
Be a good little compiler and handle direct calls efficiently, even if there
...
are beastly ConstantPointerRefs in the way...
llvm-svn: 11883
2004-02-26 22:07:22 +00:00
36ab728fe5
Fix typo
...
llvm-svn: 11864
2004-02-26 03:45:03 +00:00
128e84197b
The node doesn't have to be _no_ node flags, it just has to be complete and
...
not have any globals.
llvm-svn: 11863
2004-02-26 03:43:43 +00:00
c8167b0e7e
Add _more_ functions
...
llvm-svn: 11862
2004-02-26 03:43:08 +00:00
71626b8f36
Two changes:
...
1. Functions do not make things incomplete, only variables
2. Constant global variables no longer need to be marked incomplete, because
we are guaranteed that the initializer for the global will be in the
graph we are hacking on now. This makes resolution of indirect calls happen
a lot more in the bu pass, supports things like vtables and the C counterparts
(giant constant arrays of function pointers), etc...
Testcase here: test/Regression/Analysis/DSGraph/constant_globals.ll
llvm-svn: 11852
2004-02-25 23:36:08 +00:00
fab2872b6c
When building local graphs, clone the initializer for constant globals into each
...
local graph that uses the global.
llvm-svn: 11850
2004-02-25 23:31:02 +00:00
6ce59b4a03
Simplify the dead node elimination stuff
...
Make the incompleteness marker faster by looping directly over the globals
instead of over the scalars to find the globals
Fix a bug where we didn't mark a global incomplete if it didn't have any
outgoing edges. This wouldn't break any current clients but is still wrong.
llvm-svn: 11848
2004-02-25 23:08:00 +00:00
5e5e060618
Add a bunch more functions
...
llvm-svn: 11847
2004-02-25 23:06:40 +00:00
17bce88100
Try harder to get symbol info
...
llvm-svn: 11846
2004-02-25 23:06:30 +00:00
864c901444
Add a bunch more functions used by perlbmk
...
llvm-svn: 11824
2004-02-25 17:43:20 +00:00
9ccb1af08f
Add support for 'rename'
...
llvm-svn: 11813
2004-02-24 22:17:00 +00:00
396cdaf067
Add support for remove, fwrite, and fread
...
Also fix problem where we didn't check to see if a node pointer was null.
Though fclose(null) doesn't make a lot of sense, 300.twolf does it.
llvm-svn: 11810
2004-02-24 22:02:48 +00:00
494d510769
Fix a soon-to-be-missing #include
...
llvm-svn: 11707
2004-02-22 06:26:17 +00:00
a376aae246
Use isNull instead of getNode() to test for existence of a node, this is cheaper.
...
FIX MAJOR BUG, whereby we didn't merge null edges correctly. Correcting this
fixes poolallocation on 175.vpr, and possibly others.
llvm-svn: 11695
2004-02-22 00:53:54 +00:00
f3e2a6360c
Fix an iterator invalidation problem which was causing some nodes to not be
...
correctly merged over!
llvm-svn: 11693
2004-02-21 22:28:26 +00:00
eeb69197bf
Use handy method
...
llvm-svn: 11692
2004-02-21 22:27:31 +00:00
ab66f3d143
Instead of cloning the globals for main into the globals graph at the end of
...
BU propagation, clone the globals into the GG of EACH FUNCTION that finishes
processing! The GlobalsGraph *must* include all globals and effects from
all functions in the program. Fixing this makes pool allocation work better
on 175.vpr, but it still ultimately crashes.
llvm-svn: 11686
2004-02-21 00:30:28 +00:00
6a4e39677e
There is no need to merge the globals graph into the function graphs at the
...
end of the BU and CBU passes. The globals will be marked incomplete, so it
doesn't matter if they are missing some info, and merging isn't guaranteed
to bring everything in anyway!
llvm-svn: 11684
2004-02-20 23:52:15 +00:00
7b0368ee16
Add two missing returns, which caused us to be very pessimistic about the
...
printf and scanf families!
llvm-svn: 11683
2004-02-20 23:27:09 +00:00
a061c3b25b
Add support for some string functions, the scanf family, and sprintf
...
llvm-svn: 11673
2004-02-20 20:27:11 +00:00
2a6802ff41
When we complete the bottom-up pass, make sure to merge the globals in 'main' into
...
the globals graph.
llvm-svn: 11562
2004-02-17 19:06:47 +00:00
0cb8855a1c
Only spit out warning for functions that take pointers, not for sin and the like
...
Add more special case handling for stdio functions. I feel dirty, how about you?
llvm-svn: 11506
2004-02-16 22:57:19 +00:00
cedfcf5bca
memset and bcopy and now unified by the llvm.memset intrinsic
...
llvm-svn: 11503
2004-02-16 18:37:40 +00:00
218eb798a2
No need to scan zero initializers. This should make DSA a bit faster.
...
llvm-svn: 11471
2004-02-15 05:53:42 +00:00
d17e15eddd
Add support for a bunch more functions
...
llvm-svn: 11395
2004-02-13 21:21:48 +00:00
e00227248e
Add support for fopen/fclose. Specifically with fopen, we were marking all of the
...
operands as incomplete, though fopen is known to only read them. This just adds
fclose for symmetry, though it doesn't gain anything. This makes the dsgraphs for
181.mcf much more precise.
llvm-svn: 11390
2004-02-13 20:05:32 +00:00
9761271f0f
Restructure code to handle memcpy/memmove
...
llvm-svn: 11374
2004-02-13 16:09:54 +00:00
4b095b937e
Cosmetic improvements to this option.
...
llvm-svn: 11331
2004-02-11 19:14:04 +00:00
f10f6b10f9
Actually load profiling information now! Block layout can use real, live,
...
actual profile info, and works! :)
llvm-svn: 11324
2004-02-11 18:21:05 +00:00
fa9a8f83a5
Fix a typeo
...
llvm-svn: 11323
2004-02-11 18:20:41 +00:00
b8263bb809
Fix copy-and-pastos
...
llvm-svn: 11319
2004-02-11 06:10:18 +00:00
f7a5d9874c
Add skeleton profileinfoloader pass. This will be enhanced to actually LOAD
...
a profile tommorow. :)
llvm-svn: 11318
2004-02-11 06:10:05 +00:00
ddfc5506a7
Factor this code out of llvm-prof
...
llvm-svn: 11314
2004-02-11 05:54:25 +00:00
0c26f0048f
Make sure to register the 'no profile' implementation as the default for ProfileInfo
...
llvm-svn: 11309
2004-02-11 04:47:54 +00:00
6e445dc299
Simplify implementation, and probably speed things up too.
...
llvm-svn: 11308
2004-02-11 03:57:16 +00:00
1f4f70425d
An initial implementation of an LLVM ProfileInfo class which is designed to
...
eventually allow Passes to use profiling information to direct them.
llvm-svn: 11294
2004-02-10 22:11:42 +00:00
ac6db755c3
Adjust to the changed StructType interface. In particular, getElementTypes() is gone.
...
llvm-svn: 11228
2004-02-09 04:37:31 +00:00
fa616e14fc
Add one that I missed
...
llvm-svn: 11179
2004-02-08 01:53:10 +00:00
9ab85275ef
Instead of callign removeTriviallyDeadNodes on the global graph every time
...
removeDeadNodes is called, only call it at the end of the pass being run.
This saves 1.3 seconds running DSA on 177.mesa (5.3->4.0s), which is
pretty big. This is only possible because of the automatic garbage
collection done on forwarding nodes.
llvm-svn: 11178
2004-02-08 01:51:48 +00:00
e1531f3eb1
Remove another unneeded call.
...
llvm-svn: 11177
2004-02-08 01:40:40 +00:00
985282c78e
This call is no longer needed now that merging does not produce garbage
...
llvm-svn: 11176
2004-02-08 01:38:34 +00:00
b0f32183e4
Substantially improve the DSA code by removing 'forwarding' nodes from
...
DSGraphs while they are forwarding. When the last reference to the forwarding
node is dropped, the forwarding node is autodeleted. This should simplify
removeTriviallyDead nodes, and is only (efficiently) possible because we are
using an ilist of dsnodes now.
llvm-svn: 11175
2004-02-08 01:27:18 +00:00
5592af7927
Bugfix for ilist conversion. The ilist wants to make an 'end' node which has
...
G == 0
llvm-svn: 11174
2004-02-08 01:05:37 +00:00
f3861ce4f6
Switch the Nodes list from being an std::vector<DSNode*> to an ilist<DSNode>
...
llvm-svn: 11173
2004-02-08 00:53:26 +00:00
bd16efb88f
Change to use node_iterators instead of direct access to Nodes
...
llvm-svn: 11171
2004-02-08 00:23:16 +00:00
ce74c2f50b
getNodes() is gone, use node_begin/end instead
...
Rename stats from dsnode -> dsa
Add a new stat
llvm-svn: 11167
2004-02-07 23:58:05 +00:00
06834e6d7b
getNodes() is gone
...
llvm-svn: 11166
2004-02-07 23:57:26 +00:00
40862c7578
There is no need to clone over nodes that are going to be dead anyway
...
llvm-svn: 11157
2004-02-07 22:00:03 +00:00
e3f247bcac
In a "seeing the forest through the trees" kinda situation, I realized that a
...
complete rewrite of load-vn will make it a bit faster. This changes speeds up
the gcse pass (which uses load-vn) from 25.45s to 0.42s on the testcase in
PR209.
I've also verified that this gives the exact same results as the old one.
llvm-svn: 11132
2004-02-05 17:20:00 +00:00
d3e4d81ad7
This is a big diff with no functionality change. We just reorder some code,
...
which causes big reindentation. While I'm at it, I fix the fixme by removing
some dead code.
llvm-svn: 11131
2004-02-05 05:56:23 +00:00
90b9043ea5
finegrainify namespacification
...
llvm-svn: 11130
2004-02-05 05:51:40 +00:00
cb435541c4
Fixed Chris' typo.
...
llvm-svn: 11128
2004-02-05 04:45:21 +00:00
5bab6301ac
Implement optimizations for handling large basic blocks.
...
llvm-svn: 11126
2004-02-05 00:36:43 +00:00
2f58ffc571
Avoid referencing deleted DSgraphs when merging an SCC into a larger SCC. This
...
fixes the crash in 176.gcc.
llvm-svn: 11033
2004-01-31 21:02:18 +00:00
729ea9e1d9
Fix thinko
...
llvm-svn: 11027
2004-01-30 22:48:02 +00:00
4ab96324af
Forward method request to chained aa implementation
...
llvm-svn: 11024
2004-01-30 22:20:55 +00:00
f0eac5d0d3
Implement the pointsToConstantMemory() method.
...
llvm-svn: 11022
2004-01-30 22:17:24 +00:00
9605576a21
Improve mod/ref information based on the pointsToConstantMemory method.
...
llvm-svn: 11021
2004-01-30 22:16:42 +00:00
81804b4cb1
Order #includes alphabetically, per style guide.
...
llvm-svn: 11015
2004-01-30 17:26:24 +00:00
6c9861be68
Fix a bug aflicting 265.gap
...
llvm-svn: 11006
2004-01-29 08:36:22 +00:00
c0f1e25c01
Minor bugfixes
...
llvm-svn: 11005
2004-01-29 03:32:15 +00:00
9412b70a4c
Rename DSGraph::ScalarMapTy -> DSScalarMap
...
llvm-svn: 11001
2004-01-28 09:15:42 +00:00
f30ae61683
Fix a bug
...
llvm-svn: 11000
2004-01-28 03:31:34 +00:00
68e4bb2b13
Eliminate the call to removeTriviallyDeadNodes from updateFromGlobals graph,
...
moving it to the start of removeDeadNodes. This speeds up DSA by 2s on perlbmk
from 41s
llvm-svn: 10999
2004-01-28 03:24:41 +00:00
5e14d69d6b
In the TD pass, iterate over globals directly instead of through the whole scalar
...
map. This saves 5s in the TD pass, from 22->17s on perlbmk
llvm-svn: 10998
2004-01-28 03:12:48 +00:00
461310139a
In the TD pass, don't iterate over the scalar map to find the globals, iterate over
...
the globals directly. This doesn't save any substantial time, however, because the
globals graph only contains globals!
llvm-svn: 10997
2004-01-28 03:07:30 +00:00
63c649a9fa
In updateFromGlobalsGraph, instead of iterating over all of the scalars in the
...
function to find the globals, iterate over all of the globals directly. This
speeds the function up from 14s to 6.3s on perlbmk, reducing DSA time from
53->46s.
llvm-svn: 10996
2004-01-28 03:03:06 +00:00
06a40fc140
Minor tweaks, eliminate useless integer pruning optimziation, turn on
...
timers by default
llvm-svn: 10993
2004-01-28 02:41:32 +00:00
9f0f7ca74e
Further reduce the number of nodes cloned with getClonedNH, using merge instead.
...
This reduces the number of nodes allocated, then immediately merged and DNE'd
from 2193852 to 1298049. unfortunately this only speeds DSA up by ~1.5s (of
53s), because it's spending most of its time waddling through the scalar map :(
llvm-svn: 10992
2004-01-28 02:11:49 +00:00
6ee2b154d4
Add a timer, fix a minor bug.
...
Also, use RC::merge when possible, reducing the number of nodes allocated, then immediately merged away from 2985444 to 2193852 on perlbmk.
llvm-svn: 10991
2004-01-28 02:05:05 +00:00
f6f97f9f90
Another bugfix, disable "spurious" output.
...
You gotta love spurious
llvm-svn: 10990
2004-01-28 01:19:52 +00:00
26acc890f9
fix bug in previous checkin
...
llvm-svn: 10989
2004-01-27 22:54:56 +00:00
54a52e53f0
* Add a new commandline argument to control the "global roots hack". Default
...
it to be off. If it looks like it's completely unnecessary after testing, I
will remove it completely (which is the hope).
* Callers of the DSNode "copy ctor" can not choose to not copy links.
* Make node collapsing not create a garbage node in some cases, avoiding a
memory allocation, and a subsequent DNE.
* When merging types, allow two functions of different types to be merged
without collapsing.
* Use DSNodeHandle::isNull more often instead of DSNodeHandle::getNode() == 0,
as it is much more efficient.
*** Implement the new, more efficient reachability cloner class
In addition to only cloning nodes that are reachable from interesting
roots, this also fixes the huge inefficiency we had where we cloned lots
of nodes, only to merge them away immediately after they were cloned.
Now we only actually allocate a node if there isn't one to merge it into.
* Eliminate the now-obsolete cloneReachable* and clonePartiallyInto methods
* Rewrite updateFromGlobalsGraph to use the reachability cloner
* Rewrite mergeInGraph to use the reachability cloner
* Disable the scalar map scanning code in removeTriviallyDeadNodes. In large
SCC's, this is extremely expensive. We need a better data structure for the
scalar map, because we really want to scan the unique node handles, not ALL
of the scalars.
* Remove the incorrect SANER_CODE_FOR_CHECKING_IF_ALL_REFERRERS_ARE_FROM_SCALARMAP code.
* Move the code for eliminating integer nodes from the trivially dead
eliminator to the dead node eliminator.
* removeDeadNodes no longer uses removeTriviallyDeadNodes, as it contains a
superset of the node removal power.
* Only futz around with the globals graph in removeDeadNodes if it is modified
llvm-svn: 10987
2004-01-27 22:03:40 +00:00
070afeb77b
Rewrite to use the reachability cloner interface. Also, make this much more
...
efficient in the case where a function calls into the same graph multiple times
(ie, it either contains multiple calls to the same function, or multiple calls
to functions in the same SCC graph)
llvm-svn: 10986
2004-01-27 21:53:14 +00:00
0e43b44c31
minor cleanups
...
llvm-svn: 10985
2004-01-27 21:51:19 +00:00
b529730af0
Get clone flags right, so we don't build InlinedGlobals only to clear them
...
llvm-svn: 10984
2004-01-27 21:50:41 +00:00
78c36d2112
Initial support for implementing clonePartiallyInto in terms of cloneReachableSubgraph, though this support is currently disabled.
...
llvm-svn: 10970
2004-01-23 01:44:53 +00:00
80f89ea522
Eliminated the CompletedNodes argument to the cloneReachable* methods. This
...
map was only used to implement a marginal GlobalsGraph optimization, and it
actually slows the analysis down (due to the overhead of keeping it), so just
eliminate it entirely.
llvm-svn: 10955
2004-01-22 16:56:13 +00:00
5750d46ccd
Ok, I'm tired of pulling out all my timers to check stuff in, just do it.
...
llvm-svn: 10954
2004-01-22 16:36:28 +00:00
2a0bc1224d
Bug fix: X.mergeWith(Y) was not updating Y if Y was a null node handle!
...
llvm-svn: 10953
2004-01-22 16:31:08 +00:00
e75ae61b47
Start implementing DSGraph::clonePartiallyInto and implement mergeInGraph
...
in terms of it.
Though clonePartiallyInto is not cloning partial graphs yet, this change
dramatically speeds up inlining of graphs with many scalars. For example,
this change speeds up the BU pass on 253.perlbmk from 69s to 36s, because
it avoids iteration over the scalar map, which can get pretty large.
llvm-svn: 10951
2004-01-22 15:30:58 +00:00
45fd9d59c0
Allow disabling of ALL printing overhead when performing timings
...
llvm-svn: 10948
2004-01-22 13:42:43 +00:00
85f325b544
Remove use of ConstantHandling itf
...
llvm-svn: 10781
2004-01-12 18:02:15 +00:00
c99dd898af
Eliminate use of ConstantHandling itf
...
llvm-svn: 10780
2004-01-12 17:57:32 +00:00
97390c21d3
Live var is now in lib/Target/Sparc
...
llvm-svn: 10735
2004-01-09 18:16:20 +00:00
ae36879303
Move sparc-specific code into lib/Target/Sparc
...
llvm-svn: 10734
2004-01-09 18:15:56 +00:00
59d2d7fc33
Improve encapsulation in the Loop and LoopInfo classes by eliminating the
...
getSubLoops/getTopLevelLoops methods, replacing them with iterator-based
accessors.
llvm-svn: 10714
2004-01-08 00:09:44 +00:00
5f4c6f55b6
Add a note
...
llvm-svn: 10617
2003-12-28 04:03:49 +00:00
30513e0a3a
rename ClassifyExpression -> ClassifyExpr
...
llvm-svn: 10592
2003-12-23 08:04:08 +00:00
1e74843e7c
Finegrainify namespacification
...
llvm-svn: 10588
2003-12-23 06:44:41 +00:00
a86c4a3a01
finegrainify namespacification
...
Implement indvar analysis of getelementptr and sub expressions
llvm-svn: 10582
2003-12-22 05:26:29 +00:00
44fea5416d
Add a new AliassetTracker::remove method. Because we need to be able to remove
...
a pointer from an AliasSet, maintain the pointer values on a doubly linked
list instead of a singly linked list, to permit efficient removal from the
middle of the list.
llvm-svn: 10506
2003-12-18 08:11:56 +00:00
aaba4639f8
Change interface of MachineOperand as follows:
...
a) remove opIsUse(), opIsDefOnly(), opIsDefAndUse()
b) add isUse(), isDef()
c) rename opHiBits32() to isHiBits32(),
opLoBits32() to isLoBits32(),
opHiBits64() to isHiBits64(),
opLoBits64() to isLoBits64().
This results to much more readable code, for example compare
"op.opIsDef() || op.opIsDefAndUse()" to "op.isDef()" a pattern used
very often in the code.
llvm-svn: 10461
2003-12-14 13:24:17 +00:00
0a14060ce2
Finegrainify namespacification
...
Add capability to represent volatile AliasSet's
Propagate this information from loads&stores into the aliassets
llvm-svn: 10457
2003-12-14 04:52:11 +00:00
1eed55d6b9
Factor out some duplicated code, implement the rest of the cases in
...
BasicAA/2003-12-11-ConstExprGEP.ll
llvm-svn: 10412
2003-12-11 23:20:16 +00:00
6ea17f77f8
Fix PR86. This makes basicaa _SIGNIFICANLY_ more aggressive with getelementptr's
...
llvm-svn: 10410
2003-12-11 22:44:13 +00:00
62aff8412a
Finegrainify namespacification
...
Add new -print-cfg-only pass
llvm-svn: 10407
2003-12-11 21:48:18 +00:00
17790fbb5b
Fix a glaring bug
...
llvm-svn: 10400
2003-12-11 06:06:28 +00:00
bc1197f292
Realize the gep P, <zeros> must aliases P.
...
This is a partial fix for PR 86
llvm-svn: 10399
2003-12-11 06:02:00 +00:00
abb3bea6ce
With Brian's change to AA.h we can now clean out this uglyness
...
llvm-svn: 10398
2003-12-11 05:44:59 +00:00
b453bc5957
IncludeFile hack to pull in BasicValueNumbering whenever ValueNumbering.h
...
is included.
llvm-svn: 10397
2003-12-11 05:05:56 +00:00
bff5116da6
Finegrainify namespacification
...
llvm-svn: 10367
2003-12-10 15:34:03 +00:00
8dee84159f
Finegrainify namespacification
...
Provide a context module to WriteAsOperand
llvm-svn: 10366
2003-12-10 15:33:59 +00:00
f9f7c2d302
Finegrainify namespacification
...
Move method out of generic dominators construction code
llvm-svn: 10299
2003-12-07 00:35:42 +00:00
567444309a
Do not depend on index type to determine whether it is a structure or sequential index
...
llvm-svn: 10221
2003-11-25 20:19:55 +00:00
a346e64218
Fix file header
...
llvm-svn: 10219
2003-11-25 20:11:47 +00:00
c3de7c4ec2
Remove dead variable
...
llvm-svn: 10218
2003-11-25 20:11:00 +00:00
388bc98954
Use gep_type_begin/end instead of looking for ubytes
...
llvm-svn: 10217
2003-11-25 20:10:07 +00:00
359974801e
Finegrainify namespacification
...
llvm-svn: 10210
2003-11-25 18:33:40 +00:00
ab6895ea76
Implement a small optimization to handling of GEP's that are equivalent to casts.
...
This results in substantially reduced collapsing for some testcases
llvm-svn: 10002
2003-11-14 17:09:46 +00:00
a8150c3f8d
Trying to get the dsgraph for an external function is bad for DSA's health
...
llvm-svn: 9979
2003-11-13 18:48:11 +00:00
baef234ce3
Implement the CompleteBU pass
...
llvm-svn: 9964
2003-11-13 05:05:41 +00:00
71d3377010
Minor code cleanup
...
llvm-svn: 9962
2003-11-13 05:04:19 +00:00
fb4c057c4c
Disable integer tracking by default
...
llvm-svn: 9960
2003-11-13 03:10:49 +00:00
04946381b9
Add an initial version of the CompleteBUDataStructures class, which is currently
...
identical to the BU pass, but has an accurate call graph
llvm-svn: 9956
2003-11-13 01:43:00 +00:00
9e876556fd
Fine-grainify namespaces for this library
...
llvm-svn: 9948
2003-11-12 23:11:14 +00:00
da03a8a792
Add new argument to disable checking
...
llvm-svn: 9922
2003-11-12 17:58:22 +00:00
b740afb2b0
Print return nodes for graphs with multiple functions in them correctly
...
llvm-svn: 9914
2003-11-12 04:58:19 +00:00
12c71c00ef
Don't crash if no gg
...
llvm-svn: 9913
2003-11-12 04:57:58 +00:00
a9450bd4ac
Cleanups
...
llvm-svn: 9907
2003-11-12 00:40:34 +00:00
960707c335
Put all LLVM code into the llvm namespace, as per bug 109.
...
llvm-svn: 9903
2003-11-11 22:41:34 +00:00
da05b05197
Compute node offsets correctly
...
llvm-svn: 9895
2003-11-11 20:12:32 +00:00
fdc8d6547b
Add new method for computing node mappings. This is used by the pool allocator
...
llvm-svn: 9880
2003-11-11 05:08:59 +00:00
9a010038f1
More additions
...
llvm-svn: 9834
2003-11-09 19:54:30 +00:00
950c75f135
Add more functions
...
llvm-svn: 9826
2003-11-09 04:10:41 +00:00
5011b956a9
None of the __llvm_* functions call into the program. This makes the
...
callgraph MUCH simpler for eh using program.
llvm-svn: 9825
2003-11-09 04:00:59 +00:00
cff83a2f24
Add support for memmove
...
llvm-svn: 9824
2003-11-09 03:32:52 +00:00
12dd38aa99
Handle bzero and memset in the local analysis, because we were missing the fact
...
that memset returns its argument!!
llvm-svn: 9811
2003-11-08 21:55:50 +00:00
2cb1db763e
This doesn't use DSCallSiteIterator
...
llvm-svn: 9809
2003-11-08 21:17:37 +00:00
5f6463c041
Fix name collision
...
llvm-svn: 9722
2003-11-05 05:55:45 +00:00
2b9926fc8e
All DSGraphs keep a reference to the targetdata they are created with. This is
...
used to eliminate the hard coded, hacked in, sparc specific, global TargetData.
Changing the TargetData used to actually match the code fixes problems, and
eliminates a crash.
llvm-svn: 9659
2003-11-02 22:27:28 +00:00
53e8fab411
Fix bug: test/Regression/Analysis/DSGraph/2003-11-02-NodeCollapsing.ll
...
llvm-svn: 9658
2003-11-02 21:02:20 +00:00
7949bf0d54
Including the symbol table in the FindUsedTypes analysis was the WRONG way
...
to fix test/Regression/CBackend/2003-10-23-UnusedType.ll. This completely
neutered the deadtypeelim pass.
llvm-svn: 9646
2003-11-02 01:28:41 +00:00
9c4a58b780
Make the call graph more precise despite the hated constantpointerrefs.
...
Do you detect the animosity I feel towards CPRs yet?
llvm-svn: 9640
2003-10-31 21:05:12 +00:00
dd63f9e22b
When someone includes CallGraph.h, make sure that they link in CallGraph.cpp
...
llvm-svn: 9611
2003-10-30 05:17:30 +00:00
54ac910a5c
Fix PR62, and llvm/test/Regression/CBackend/2003-10-28-CastToPtrToStruct.ll
...
ConstantExpr's can use unrelated types, make sure to scan them.
llvm-svn: 9569
2003-10-28 23:09:45 +00:00
fb8e19b7ca
Fix bug: CBackend/2003-10-23-UnusedType.ll and hopefully 252.eon
...
llvm-svn: 9441
2003-10-23 19:30:30 +00:00
5987ab8677
This is a disgusting hack that improves code substantially, by making
...
callgraphSCC passes more effective.
llvm-svn: 9384
2003-10-22 18:53:31 +00:00
ac4238f7bc
Do not add unreachable code to a natural loop!
...
llvm-svn: 9377
2003-10-22 16:41:21 +00:00
aa81dce20c
Make sure to print labels on nodes without names
...
llvm-svn: 9376
2003-10-22 16:30:58 +00:00
9d63afdea5
If the basic block has no name, make sure to print the % number of it
...
llvm-svn: 9375
2003-10-22 16:22:42 +00:00
a93d11bc1d
Implement the Function::viewCFG* methods, for use in a debugger. Also, the
...
-print-cfg pass now lives here.
llvm-svn: 9374
2003-10-22 16:03:49 +00:00
29265fe981
Added LLVM copyright header.
...
llvm-svn: 9321
2003-10-21 15:17:13 +00:00
4436c49787
Added LLVM copyright notice to Makefiles.
...
llvm-svn: 9312
2003-10-20 22:26:57 +00:00
dcd1466a80
Convert this code from using annotations to using a local map
...
llvm-svn: 9310
2003-10-20 20:52:23 +00:00
93fa46476b
Remove using declaration
...
llvm-svn: 9307
2003-10-20 20:38:17 +00:00
482202a601
Added LLVM project notice to the top of every C++ source file.
...
Header files will be on the way.
llvm-svn: 9298
2003-10-20 19:43:21 +00:00
b94550e537
Change the Opcode enum for PHI nodes from "Instruction::PHINode" to "Instruction::PHI" to be more consistent with the other instructions.
...
llvm-svn: 9269
2003-10-19 21:34:28 +00:00
f5d946eced
Make use of "external" depth-first iterators to avoid revisiting nodes
...
multiple times. This reduces the time to construct post-dominance sets a LOT.
For example, optimizing perlbmk goes from taking 12.9894s to 1.4074s.
llvm-svn: 9091
2003-10-13 16:36:06 +00:00
44d2c3514a
Regularize header file comments
...
llvm-svn: 9071
2003-10-13 03:32:08 +00:00
e43629bad6
Make getNumBackEdges more efficient
...
llvm-svn: 9063
2003-10-12 22:14:27 +00:00
6e739d0a3c
Use std::string::size_type for for ColonPos to stop gcc from giving a warning
...
llvm-svn: 8811
2003-10-01 22:49:22 +00:00
66a473371f
DOn't crash if program calls the null pointer
...
llvm-svn: 8708
2003-09-24 23:42:58 +00:00
fd46711e1a
Remove explicit use of BasicBlock::succ_iterator
...
llvm-svn: 8706
2003-09-24 22:18:35 +00:00
3c095c79fa
Functions reachable from the arguments of unresolvable call nodes should
...
not have their arguments marked complete
llvm-svn: 8639
2003-09-20 23:58:33 +00:00
Chris Lattner