3bf6f80588
make fp div trick dependend on flag
...
llvm-svn: 20967
2005-03-31 22:02:25 +00:00
dec53920b4
PCMarker support for DAG and Alpha
...
llvm-svn: 20965
2005-03-31 21:24:06 +00:00
21ce5f7a66
Assorted fixes:
...
* Stop being pessimistic about output register allocation
* Start to handle function descriptors: compute target GPs and so on
when doing indirect calls etc. Not there yet, though. For the moment,
we try to use _indirect_ branches wherever possible, to stress test
function descriptors.
* FP divide-by-zero should work now
* add varargs (it doesn't work, though)
At this point, all of SingleSource passes (modulo C++ tests that are due
to issues with the CFE, see note in the README.) Much of MultiSource also
passes although there's still a ton of bugs around. Something for me to
work on tomorrow, then. :)
llvm-svn: 20960
2005-03-31 12:31:11 +00:00
ffb9d14bb3
* declare local common symbols as such (.lcomm, not merely .common)
...
* begin great adventure into correct function descriptor materialization
llvm-svn: 20956
2005-03-31 07:40:24 +00:00
8106f12d33
daintyness
...
llvm-svn: 20955
2005-03-31 07:36:43 +00:00
73c2777a0e
add what we need to fudge a 'floating point conditional move', this is
...
used to get FP div-by-zero working properly (shunt the right answer
depending on how frcpa sets its predicate output)
llvm-svn: 20954
2005-03-31 07:32:32 +00:00
eddfff338a
Pass the correct values to the chain argument for node construction during
...
LowerCallTo.
Handle ISD::ADD in SelectAddr, allowing us to have nonzero immediates for
loads and stores, amazing!
llvm-svn: 20946
2005-03-31 02:05:53 +00:00
9de7ddf7d5
Rewrite LowerCallTo and Select(ISD::CALL) to properly handle float varargs
...
Tell the SelectionDAG ISel to expand SEXTLOAD of i1 and i8, rather than
complicate the code in ISD::SEXTLOAD to do it by hand
Combine the FP and Int ISD::LOAD codegen
Generate better code for constant pool loads
As a result, all of Shootout, and likely many other programs are now
working.
llvm-svn: 20945
2005-03-31 00:15:26 +00:00
4c0780cf42
Fix calls whose arguments fit entirely in registers to not break the Chain.
...
Implement SINT_TO_FP and UINT_TO_FP
Remove some dead code from the simple ISel
llvm-svn: 20944
2005-03-30 19:38:35 +00:00
e3d9f8c4f9
yea, fine Duraid
...
llvm-svn: 20942
2005-03-30 18:22:52 +00:00
a8114a0f48
Fix frame index code to generate legal PowerPC instructions. About half of
...
Shootout now works.
llvm-svn: 20940
2005-03-30 02:23:08 +00:00
5851a66128
Fix external symbol printing in the AsmPrinter. Tell the ISel that we
...
don't support things like memcpy directly. This allows a handful of the
Shootout programs to work, yay!
llvm-svn: 20939
2005-03-30 01:45:43 +00:00
5ca31d9831
Instead of setting up the CFG edges at selectiondag construction time, set
...
them up after the code has been emitted. This allows targets to select one
mbb as multiple mbb's as needed.
llvm-svn: 20937
2005-03-30 01:10:47 +00:00
472a265ef6
don't depend on the cfg being set up yet
...
llvm-svn: 20936
2005-03-30 01:10:00 +00:00
9c40e9c61f
Fix BranchCC (it's still dumb), and implement FP select (also dumb)
...
llvm-svn: 20935
2005-03-29 22:48:55 +00:00
28145edd30
Implement integer select and i1 sign extend
...
llvm-svn: 20934
2005-03-29 22:24:51 +00:00
28c5ac9ff4
Implement SetCC, fix ZERO_EXTEND_INREG
...
llvm-svn: 20933
2005-03-29 21:54:38 +00:00
af4d93cd39
import all of the rest of the stubs that dsa uses for direct comparison
...
llvm-svn: 20932
2005-03-29 20:36:05 +00:00
91964490e2
Compare dependence analysis with llvm instructions versus machine instrutions. the problem with using machine instructions and alias analysis is that aa does not handle tmp instructions.
...
llvm-svn: 20931
2005-03-29 20:35:10 +00:00
6e19b21e30
Dependence analyzer that just determines dependences within a loop for loads and stores using alias analysis.
...
llvm-svn: 20930
2005-03-29 20:33:42 +00:00
4dba2ee974
learn about some more functions.
...
llvm-svn: 20929
2005-03-29 20:04:24 +00:00
93d008990e
Fix up some types and constants
...
llvm-svn: 20928
2005-03-29 19:24:04 +00:00
2e1971c19d
Fix a problem where we not marking incoming arguments to functions with
...
external linkage as incomplete.
llvm-svn: 20927
2005-03-29 19:16:59 +00:00
db45f7d763
Fix a bug that andrew noticed where we do not correctly sign/zero extend
...
returned integer values all of the way to 64-bits (we only did it to 32-bits
leaving the top bits undefined). This causes problems for targets like alpha
whose ABI's define the top bits too.
llvm-svn: 20926
2005-03-29 19:09:56 +00:00
8de8b7bf40
there is no point comparing against null pointer.
...
llvm-svn: 20925
2005-03-29 17:44:52 +00:00
ca85130b0a
Fix a major problem with global variable initializers. This could cause
...
us to have stuff pointing to the null pointer, which makes no sense
(the null ptr is an ssa value, not the null object)
llvm-svn: 20922
2005-03-29 17:21:53 +00:00
1ce9aacf2e
fix a warning in the optimized build
...
llvm-svn: 20920
2005-03-29 15:13:27 +00:00
dd94b2d01e
add some more functions, ignore setcc for constraints!
...
llvm-svn: 20917
2005-03-29 06:52:20 +00:00
ecfa9b5810
disable this transformation in the one obscure case that really pessimizes
...
pointer analysis.
llvm-svn: 20916
2005-03-29 06:37:47 +00:00
7617e886c1
Handle "known" external calls context sensitively, add support for realloc
...
and a couple of other functions that are important.
Handle aggregate undef values for gv initializers
llvm-svn: 20914
2005-03-29 06:09:07 +00:00
60956dd52f
Implement div, rem, and frameindex
...
llvm-svn: 20907
2005-03-29 00:03:27 +00:00
731bed10c7
Pattern ISel: fix argument loading for i64s (thanks chris)
...
Simple ISel: fix i64 subtract
llvm-svn: 20903
2005-03-28 23:08:54 +00:00
66acad746f
Remove fake instruction 'subc' (mnemonic for subfc).
...
More pattern isel updates
llvm-svn: 20902
2005-03-28 22:28:37 +00:00
32e08b7c06
implement legalization of build_pair for nate
...
llvm-svn: 20901
2005-03-28 22:03:13 +00:00
b442791124
First step in adding pcmarker intrinsic. Second step (soon) is adding backend support.
...
llvm-svn: 20900
2005-03-28 20:05:49 +00:00
5bdfc63516
Implement proper loads and zero-extends of all types
...
llvm-svn: 20897
2005-03-28 19:36:43 +00:00
bdbf8ba2d0
Emit .global @function and .global @object entries so the Intel ias
...
assembler may be used; identify LLVM output as such.
llvm-svn: 20892
2005-03-28 15:21:43 +00:00
81b6d86a3a
Teach andersens that non-escaping memory cannot be mod/ref'd by external fn calls.
...
llvm-svn: 20891
2005-03-28 06:21:17 +00:00
009585d2ac
Fix grammar
...
llvm-svn: 20890
2005-03-28 04:32:12 +00:00
a8ad8b2826
Make anders-aa much more precise by not being completely pessimistic about
...
external functions. Teach it about a few important ones.
llvm-svn: 20889
2005-03-28 04:03:52 +00:00
9ead0d7b4c
Rename createPromoteMemoryToRegister() to
...
createPromoteMemoryToRegisterPass() to be consistent with other pass
creation functions.
llvm-svn: 20885
2005-03-28 02:01:12 +00:00
147f8b789c
wrap some long lines
...
llvm-svn: 20884
2005-03-27 22:03:46 +00:00
2db10ba297
remove ...
...
llvm-svn: 20883
2005-03-27 21:57:09 +00:00
f290da862d
speed up steens by using spliceFrom, improve its precision by realizing that
...
an incomplete node cannot alias a complete node.
llvm-svn: 20882
2005-03-27 21:56:55 +00:00
ea69f1f9e8
teach andersens about undef
...
llvm-svn: 20881
2005-03-27 18:58:23 +00:00
82166467c3
Don't give up completely, maybe other AA can say something about this.
...
llvm-svn: 20873
2005-03-27 00:02:33 +00:00
3f08e7899f
Factor out percentage printing into its own function. Make two changes to
...
the function: print more precision XX.X% instead of XX%, and cast to ULL
before scaling by 100/1000 to avoid wrap around for large numbers of queries
(such as occur for 253.perlbmk and 176.gcc)
llvm-svn: 20872
2005-03-26 23:56:33 +00:00
2ba37389c9
Cache mapping information for a call site after computing it for a mod/ref
...
query. If the next mod/ref query happens to be for the same call site
(which is extremely likely), use the cache instead of recomputing the
callee/caller mapping. This makes -aa-eval ***MUCH*** faster with
ds-aa
llvm-svn: 20871
2005-03-26 23:29:03 +00:00
139f43b340
Remove more long dead code: dsa doesn't provide must alias info
...
llvm-svn: 20870
2005-03-26 22:54:46 +00:00
1a02d15dab
remove some unsafe code that has long been dead
...
llvm-svn: 20869
2005-03-26 22:48:42 +00:00
9cc932d3b3
slightly improve mod/ref for DSAA by checking the globals graph for fallback
...
llvm-svn: 20868
2005-03-26 22:47:03 +00:00
b3baff1830
Teach steens-aa two things about mod/ref information:
...
1. If memory never escapes the program, it cannot be mod/ref'd by external
functions.
2. If memory is global never mod/ref'd in the program, it cannot be mod/ref'd
by any call.
llvm-svn: 20867
2005-03-26 22:43:20 +00:00
68ee8f572c
Interchange this loop so that we test all pointers against one call site
...
before moving on to the next call site. This will be a more efficient way
to compute the mod/ref set for AA implementations like DSA.
llvm-svn: 20866
2005-03-26 22:16:44 +00:00
518a994b7d
Fix that pesky floats in integer regs problem by assigning the f32 type to
...
the correct register class. Also remove the loading of float data into int
regs part of varargs; it will need to be implemented differently later.
llvm-svn: 20857
2005-03-26 08:25:22 +00:00
dcad53e7bd
Get closer to having varargs working. There's still something strange
...
going on with copies between floating point and integer register files
being generated. Once that is solved, varargs will be done.
llvm-svn: 20856
2005-03-26 07:46:36 +00:00
c95bf7b226
Make 64bit args and float args work correct with calls. Thanks to Chris
...
for explaining EXTRACT_ELEMENT to me.
llvm-svn: 20847
2005-03-26 02:17:46 +00:00
f656525cb6
Change interface to LowerCallTo to take a boolean isVarArg argument.
...
llvm-svn: 20842
2005-03-26 01:29:23 +00:00
e70b5c79f8
Next round of pattern isel changes, mostly dealing with calls.
...
llvm-svn: 20841
2005-03-26 01:28:53 +00:00
5d7610de65
Correct a documention link
...
llvm-svn: 20840
2005-03-26 01:28:05 +00:00
d21e220035
no really, don't double count these nodes either!
...
llvm-svn: 20837
2005-03-25 20:54:45 +00:00
b7ffd50537
Don't count all of the nodes in the SCC once for each function in the SCC.
...
llvm-svn: 20836
2005-03-25 20:37:32 +00:00
5ecec08ab5
Grow the EQ classes for globals at the end of the BU pass. This shrinks
...
memory usage in the TD pass for 254.gap from 31.3MB to 3.9MB.
llvm-svn: 20834
2005-03-25 16:45:43 +00:00
0ad33afe22
Support global addresses and fix call returns. Varargs still aren't
...
handled correctly for floating point arguments, or more than 8 arguemnts.
This does however, allow hello world to run.
llvm-svn: 20832
2005-03-25 08:34:25 +00:00
514e843e89
Enhance loopsimplify to preserve alias analysis instead of clobbering it.
...
This prevents crashes on some programs when using -ds-aa -licm.
llvm-svn: 20831
2005-03-25 06:37:22 +00:00
8d9f3735cd
Treat free operations as volatile, since they cannot be moved. This fixes
...
Transforms/LICM/2005-03-24-LICM-Aggregate-Crash.ll
llvm-svn: 20830
2005-03-25 05:49:37 +00:00
faf7791fea
Fix a bug where LICM was not updating AA information properly when sinking
...
a pointer value out of a loop causing it to be duplicated.
llvm-svn: 20828
2005-03-25 00:22:36 +00:00
3322e28f75
remove a debugging timer.
...
llvm-svn: 20827
2005-03-25 00:06:09 +00:00
278bac843c
Two changes here:
...
1. Instead of copying Local graphs to the BU graphs to start with, use
spliceFrom to do the job (which is constant time in this case). On
176.gcc, this chops off .17s from the bu pass.
2. When building SCC graphs, simplify the logic and use spliceFrom to
do the heavy lifting, instead of cloneInto/delete. This slices
another .14s off 176.gcc.
llvm-svn: 20826
2005-03-25 00:05:04 +00:00
a396eb8ae9
Make the spliceFrom case where one graph is completely empty be constant time.
...
llvm-svn: 20825
2005-03-25 00:02:41 +00:00
127dd12427
add a new DSGraph::spliceFrom method, which violently takes the content of
...
one graph and plops it into another, without breaking a sweat.
llvm-svn: 20824
2005-03-24 23:46:04 +00:00
ad5151d785
Implement next round of Pattern ISel fixes
...
1. void returns
2. multiplies
3. calls
llvm-svn: 20822
2005-03-24 23:35:30 +00:00
3a9a2f9ec7
Fix an incorrect argument being passed to BuildMI for indirect calls.
...
llvm-svn: 20821
2005-03-24 23:34:38 +00:00
301764e28c
This replaces the correct but slow code with a more aggressive scc-finder
...
based approach to find globals and call sites that need to be copied. This
speeds up the BU pass on 176.gcc from 22s back up to 2.3s. Not as good
as 1.5s, but at least it's correct :)
llvm-svn: 20820
2005-03-24 23:06:02 +00:00
1b5b559011
only look at successors of globals. This gets us down to "only" 22s in the
...
bu pass for 176.gcc
llvm-svn: 20818
2005-03-24 21:17:27 +00:00
189d7bb9ee
Unfortunately, a previous patch was not safe. Revert it, reimplement
...
something correct. Unfortunately this takes 176.gcc's BU phase back
up to 29s from 1.5. This fixes DSGraph/2005-03-24-Global-Arg-Alias.ll
llvm-svn: 20817
2005-03-24 21:07:47 +00:00
d73c87a424
Commit Gabor Greif's patch to use iterators in lowering intrinsics.
...
llvm-svn: 20816
2005-03-24 20:07:16 +00:00
923d925953
don't bother |'ing in 0's
...
llvm-svn: 20815
2005-03-24 18:42:51 +00:00
ba6006d84b
be more aggressive about incompleteness marking
...
llvm-svn: 20814
2005-03-24 18:42:28 +00:00
b15317b74a
eliminate dead variables, patch contributed by Gabor Greif!
...
llvm-svn: 20812
2005-03-24 17:32:20 +00:00
7ec19b4d5a
Implement more of the PPC32 Pattern ISel:
...
1) dynamic stack alloc
2) loads
3) shifts
4) subtract
5) immediate form of add, and, or, xor
6) change flag from -pattern-isel to -enable-ppc-pattern-isel
Remove dead arguments from getGlobalBaseReg in the simple ISel
llvm-svn: 20810
2005-03-24 06:28:42 +00:00
a3abb4c01a
Fix silly "no newline at end of file" warning
...
llvm-svn: 20809
2005-03-24 06:16:18 +00:00
d6888355fa
Fix compilation errors, patch contributed by the fabulous Bill Wendling!
...
llvm-svn: 20808
2005-03-24 05:13:53 +00:00
fed2a42895
fix a compilation error, patch contributed by Bill Wendling!
...
llvm-svn: 20807
2005-03-24 05:12:48 +00:00
20e7e28ee7
Addition of the PPC32 Pattern ISel. While it is far from complete, it will
...
be brought up to parity with the current simple ISel in the coming days.
Currently, -pattern-isel is required to trigger it.
llvm-svn: 20805
2005-03-24 04:41:43 +00:00
952105220e
Remove comments that are now meaningless from the pattern ISels, at Chris's
...
request.
llvm-svn: 20804
2005-03-24 04:39:54 +00:00
19d7f2883d
Fix a crash while promoting a value out of a loop from a global variable
...
when using ds-aa
llvm-svn: 20802
2005-03-24 04:22:04 +00:00
f0d919ece9
teach ds-aa about mod/ref for external function calls.
...
llvm-svn: 20801
2005-03-24 03:04:50 +00:00
aff85ac552
Simplify dead code into a fixme :)
...
llvm-svn: 20800
2005-03-24 02:41:19 +00:00
85a1d92323
fix a compiler crash in runtime/libprofile
...
llvm-svn: 20799
2005-03-24 01:22:52 +00:00
6a3f5fd623
wrap a long line
...
llvm-svn: 20797
2005-03-23 23:51:12 +00:00
2a7358064e
If we are calling an external function, chain to another AA to potentially
...
decide, don't just immediately give up.
This implements GlobalsModRef/chaining-analysis.ll
llvm-svn: 20796
2005-03-23 23:49:47 +00:00
8424b0f82e
Make this more efficient by only making one virtual method call.
...
llvm-svn: 20793
2005-03-23 23:26:58 +00:00
07eac1e698
Make this a bit more aggressive
...
llvm-svn: 20792
2005-03-23 22:06:41 +00:00
9f9da538d6
a hack to allow count-aa to work with ds-aa :(
...
llvm-svn: 20791
2005-03-23 21:59:34 +00:00
ccb6faad7e
Add two options to allow -count-aa to print queries either (1) all queries,
...
or (2) only queries that are not successful (e.g. return may alias)
llvm-svn: 20790
2005-03-23 21:59:07 +00:00
1c790bf656
enable -debug-only=licm
...
llvm-svn: 20788
2005-03-23 21:00:12 +00:00
61130216ec
turn a dead conditional into an assert.
...
llvm-svn: 20787
2005-03-23 20:12:08 +00:00
01f0879c48
Totally gut mergeInGraph. There is absolutely no reason to be merging
...
global roots in from callees to callers. The BU graphs do not have accurate
globals information and all of the clients know it. Instead, just make sure
the GG is up-to-date, and they will be perfectly satiated.
This speeds up the BU pass on 176.gcc from 5.5s to 1.5s, and Loc+BU+TD
from 7s to 2.7s.
llvm-svn: 20786
2005-03-23 20:08:59 +00:00
048ae36b93
wrap a long line
...
llvm-svn: 20785
2005-03-23 16:43:11 +00:00
82f364515c
don't lie to the register allocator
...
llvm-svn: 20784
2005-03-23 15:20:01 +00:00
e36d77b726
Make -steens-aa more conservative (aka correct) by making sure to obey
...
incompleteness flags.
Make it more aggressive by taking field sensitive information into
account.
llvm-svn: 20781
2005-03-23 01:48:09 +00:00
13417b5be3
Added alias analysis.
...
Fixed many many bugs.
This now works on almost all Singlesource , and most of MultiSource.
llvm-svn: 20780
2005-03-23 01:47:20 +00:00
7d3df410ab
implement Analysis/DSGraph/field-sensitive.ll
...
llvm-svn: 20779
2005-03-23 01:47:19 +00:00
dcb82f5d0a
don't crash in some bad cases.
...
llvm-svn: 20776
2005-03-23 01:29:26 +00:00
310d1a3882
Several changes here:
...
1. Increase max node size from 64->256 to avoid collapsing an important
structure in 181.mcf
2. If we have multiple calls to an indirect call node with an indirect
callee, fold these call nodes together, to avoid DSA turning apoc into
a flaming fireball of death when analyzing 176.gcc.
With this change, 176.gcc now takes ~7s to analyze for loc+bu+td, with
5.7s of that in the BU pass.
llvm-svn: 20775
2005-03-22 23:54:52 +00:00
ca174306f0
Mark external globals incomplete in the BU Globals graph, fixing
...
Analysis/DSGraph/2005-03-22-IncompleteGlobal.ll
llvm-svn: 20773
2005-03-22 22:10:22 +00:00
8180e87e51
hum, it is good to use real instructions
...
llvm-svn: 20769
2005-03-22 16:42:52 +00:00
f00684079a
Directly count the number of memory instructions.
...
llvm-svn: 20766
2005-03-22 03:55:10 +00:00
24922c7a2b
Timers SHOULD NOT record the time taken to count the bytes allocated in the heap!
...
llvm-svn: 20765
2005-03-22 03:20:38 +00:00
495f00946e
Remove an iteration pass over the entire scalarmap for each function created
...
by not allowing integer constants to get into the scalar map in the first
place.
llvm-svn: 20764
2005-03-22 02:45:13 +00:00
250726ec25
When making a clone of a DSGraph from the BU pass, make sure to remember that
...
this clone is supposed to be used for *ALL* of the functions in the SCC.
This fixes the memory explosion problem the TD pass was having, reducing the
memory growth from 24MB -> 3.5MB on povray and 270MB ->8.3MB on perlbmk!
This obviously also speeds up the TD pass *a lot*.
llvm-svn: 20763
2005-03-22 01:50:42 +00:00
e05c633099
Don't use operator[], use the new method instead, which is faster. This speeds
...
up the TD pass about 30% for povray and perlbmk. It's still not clear why
copying a 5MB set of graphs turns into a 25MB set of graphs though :(
llvm-svn: 20762
2005-03-22 01:42:59 +00:00
fcae88f280
Now that the dead ctor is gone, nothing uses the old node mapping exported by
...
cloneInto: make it an internally used mapping.
llvm-svn: 20760
2005-03-22 00:36:51 +00:00
8016c975c9
remove a dead ctor
...
llvm-svn: 20759
2005-03-22 00:33:35 +00:00
ea2c02bcca
now that the second argument is always this->ReturnNodes, don't bother passing it.
...
llvm-svn: 20758
2005-03-22 00:29:44 +00:00
571578143c
instead of using a local RetValMap, just use the graph we are cloning into
...
for the return node map.
llvm-svn: 20757
2005-03-22 00:25:52 +00:00
6294dee913
two things: 1)evilness reduction patch, reduces the number of instructions hiding in the small jump CC moving code. 2)implement div of small data types by FP DIV (also reduces evilness and should speed things up)
...
llvm-svn: 20756
2005-03-22 00:24:07 +00:00
d05ab9cab4
now that the valuemapping is always the local scalar map, we can eliminate
...
this identity merge.
llvm-svn: 20755
2005-03-22 00:22:45 +00:00
19944fade3
remove the second argument to cloneInto
...
llvm-svn: 20754
2005-03-22 00:21:05 +00:00
03b7f99338
add some timers, don't clone aux nodes
...
llvm-svn: 20752
2005-03-22 00:12:00 +00:00
410da35a1f
move this out of line
...
llvm-svn: 20751
2005-03-22 00:09:45 +00:00
5f10e6dda1
don't generate temporary scalarmaps.
...
llvm-svn: 20749
2005-03-22 00:04:21 +00:00
848ef879c5
allow passing clone flags down to cloneInto
...
llvm-svn: 20748
2005-03-21 22:49:53 +00:00
c1439d539f
Enhance the TD pass to build composite graphs when we have indirect call
...
sites that target multiple callees. If we have a function table, for
example, with N callees, and M callers call through it, we used to have
to perform O(M*N) graph inlinings. Now we perform O(M+N) inlinings.
This speeds up the td pass on perlbmk from 36.26s to 25.75s.
llvm-svn: 20743
2005-03-21 20:31:29 +00:00
d90f5bab82
make this const correct
...
llvm-svn: 20741
2005-03-21 20:28:50 +00:00
040426aee0
Fix a major problem where we didn't add call graph edges for call sites with
...
more than 1 callee. This fixes Analysis/DSGraph/FunctionPointerTable-const.ll
llvm-svn: 20740
2005-03-21 20:20:49 +00:00
82ae9e1dda
We may be adding functions to the Module during initialization, so
...
conservatively, it's modified
llvm-svn: 20735
2005-03-21 19:22:14 +00:00
e8a43291cd
Ugh, for some reason, I can't call this unless the reference is const!?!?!?
...
llvm-svn: 20732
2005-03-21 10:00:45 +00:00
13429dc5af
The reachability cloner should add arguments to merged calls when the RHS of
...
the merge has more operands than the LHS.
llvm-svn: 20731
2005-03-21 09:39:51 +00:00
7269f15e8f
'note to self'
...
llvm-svn: 20727
2005-03-21 08:43:32 +00:00
a5f4b773d9
Remove a bunch of cruft and dead code for handling the case when types were
...
defined in function constant pools. The assembler grammar has long
disallowed functions from having constant pools, so all of this stuff is
dead.
This makes it an immediate error for functions to refer to nonexisting
types, fixing Regression/Verifier/2005-03-21-UndefinedTypeReference.ll.
Before, references to non-existing types in functions would only be
detected when the subsequent function was parsed, due to the call to
"ResolveTypes". "ResolveTypes" has not resolved any types for a long time,
instead it emitted an error message if no resolved types are left. Since
the only caller of this method is in the module code, just inline it.
llvm-svn: 20726
2005-03-21 06:27:42 +00:00
87ee2ba48d
Make this method non-const. Making it const causes the const version of
...
getIGroup to be called instead of the non-const one. These two methods
have (??) different behavior, so this change fixes bugs.
llvm-svn: 20724
2005-03-21 06:07:43 +00:00
98f7e1c370
Change the way that the TD pass inlines graphs. Instead of inlining each
...
graph into all of the functions it calls when we visit a graph, change it so
that the graph visitor inlines all of the callers of a graph into the current
graph when it visits it.
While we're at it, inline global information from the GG instead of from each
of the callers. The GG contains a superset of the info that the callers do
anyway, and this way we only need to do it one time (not one for each caller).
This speeds up the TD pass substantially on several programs, and there is
still room for improvement. For example, the TD pass used to take 147s
on perlbmk, it now takes 36s. On povray, we went from about 5s to 1.97s.
134.perl is down from ~1s for Loc+BU+TD to .6s.
The TD pass needs a lot of improvement though, which will occur with later
patches.
llvm-svn: 20723
2005-03-21 04:55:35 +00:00
c5cdc2260d
Don't strip modref bits when inlining down the call graph. This fixes
...
the DSGraph/2003-06-30-TopDownResolve.ll regression from last night.
llvm-svn: 20717
2005-03-20 18:02:56 +00:00
21ef7678d6
Remove the ability to keep track of inlined globals, which is always dead.
...
llvm-svn: 20716
2005-03-20 04:30:30 +00:00
571c62a008
This call is always a noop, remove it.
...
llvm-svn: 20714
2005-03-20 04:29:54 +00:00
b1cabb6541
remove some pointless asserts
...
llvm-svn: 20713
2005-03-20 04:29:39 +00:00
ded489ec3b
-steens doesn't use the inlined globals facility.
...
llvm-svn: 20712
2005-03-20 04:23:57 +00:00
ff634f4e32
Build EC's for globals twice. The first is after constructing the initial
...
Globals Graph for the local pass, the second is after all of the locals
graphs have been constructed. This allows for many additional global EC's
to be recognized that weren't before. This speeds up analysis of programs
like 177.mesa, where it changes DSA from taking 0.712s to 0.4018s.
llvm-svn: 20711
2005-03-20 03:32:35 +00:00
36a810ac40
Add a new DSNode::removeGlobal method.
...
llvm-svn: 20710
2005-03-20 03:29:54 +00:00
5ae215248f
Transform BU pass to not use the horrible DSCallSiteIterator class.
...
llvm-svn: 20708
2005-03-20 02:42:07 +00:00
a550a5b173
Implement new methods for dealing with DSNode global lists, rename uses of
...
old methods.
llvm-svn: 20707
2005-03-20 02:41:38 +00:00
e68e808bcd
use simpler methods.
...
llvm-svn: 20706
2005-03-20 02:41:16 +00:00
1ffdd57e0a
method was renamed.
...
llvm-svn: 20705
2005-03-20 02:40:27 +00:00
5635a88fcb
print out equiv class info.
...
llvm-svn: 20704
2005-03-20 02:40:11 +00:00
e7a8e8677e
some methods got renamed
...
llvm-svn: 20703
2005-03-20 02:40:04 +00:00
711e083c26
Make use of simpler DSNode methods.
...
llvm-svn: 20702
2005-03-20 02:39:49 +00:00
61f3befdc8
#ifdef out a function only used by #ifdef'd code.
...
llvm-svn: 20700
2005-03-20 02:14:15 +00:00
e965bd16bb
comment cleanup
...
llvm-svn: 20699
2005-03-20 01:18:00 +00:00
9379e69af9
Create an equivalence class of global variables that DSA will never be able
...
to tell apart anyway, and only track the leader for of these equivalence
classes in our graphs.
This dramatically reduces the number of GlobalValue*'s that appear in scalar
maps, which A) reduces memory usage, by eliminating many many scalarmap entries
and B) reduces time for operations that need to execute an operation for each
global in the scalar map.
As an example, this reduces the memory used to analyze 176.gcc from 1GB to
511MB, which (while it's still way too much) is better because it doesn't hit
swap anymore. On eon, this shrinks the local graphs from 14MB to 6.8MB,
shrinks the bu+td graphs of povray from 50M to 40M, shrinks the TD graphs of
130.li from 8.8M to 3.6M, etc.
This change also speeds up DSA on large programs where this makes a big
difference. For example, 130.li goes from 1.17s -> 0.56s, 134.perl goes
from 2.14 -> 0.93s, povray goes from 15.63s->7.99s (!!!).
This also apparently either fixes the problem that caused DSA to crash on
perlbmk and gcc, or it hides it, because DSA now works on these. These
both take entirely too much time in the TD pass (147s for perl, 538s for
gcc, vs 7.67/5.9s in the bu pass for either one), but this is a known
problem that I'll deal with later.
llvm-svn: 20696
2005-03-19 22:23:45 +00:00
Andrew Lenharth