24f200ad7b
Switch this from using an std::map to using a DenseMap. This speeds up
...
phi-elimination from 0.6 to 0.54s on kc++.
llvm-svn: 13454
2004-05-10 19:17:36 +00:00
39a1e0a33e
Use a new VRegPHIUseCount to compute uses of PHI values by other phi values
...
in the basic block being processed. This fixes PhiElimination on kimwitu++
from taking 105s to taking a much more reasonable 0.6s (in a debug build).
llvm-svn: 13453
2004-05-10 19:06:37 +00:00
a2f7b9bddc
Now that we use an ilist of machine instructions, iterators are more robust
...
than before. Because this is the case, we can compute the first non-phi
instruction once when de-phi'ing a block. This shaves ~4s off of
phielimination of _Z7yyparsev in kimwitu++ from 109s -> 105s. There are
still much more important gains to come.
llvm-svn: 13452
2004-05-10 18:47:18 +00:00
72fb3256f8
Fix a fairly serious pessimizaion that was preventing us from efficiently
...
compiling things like 'add long %X, 1'. The problem is that we were switching
the order of the operands for longs even though we can't fold them yet.
llvm-svn: 13451
2004-05-10 15:15:55 +00:00
eefb8a7dd0
Bug fixed
...
llvm-svn: 13450
2004-05-10 14:28:09 +00:00
5eb8094c2d
Patch to fix PR337. Make sure to mark all aliased physical registers as used
...
when we see a read of a register. This is important in cases like:
AL = ...
AH = ...
= AX
The read of AX must make both the AL and AH defs live until the use.
llvm-svn: 13444
2004-05-10 05:12:43 +00:00
60ef22da3d
Testcase for PR337, which was reduced by Reid Spencer.
...
llvm-svn: 13443
2004-05-10 05:11:24 +00:00
45e25d8295
Bug 178 Fixed
...
llvm-svn: 13442
2004-05-09 23:32:39 +00:00
be4fadf1b8
Changes to make the Stacker Stack use 64 bit values. This *should* get
...
around the problem with Stacker on Solaris because the Stack can handle
64-bit entries (pointer sized).
llvm-svn: 13441
2004-05-09 23:20:19 +00:00
a367dd745b
Fix some comments, avoid sign extending booleans when zero extend works fine
...
llvm-svn: 13440
2004-05-09 23:16:33 +00:00
1542a98e7b
Generate more efficient code for casting booleans to integers (no sign extension required)
...
llvm-svn: 13439
2004-05-09 22:28:45 +00:00
f719a52088
syntactically loopify natural loops so that the GCC loop optimizer can find them. This should *dramatically* improve the performance of CBE compiled code on targets that depend on GCC's loop optimizations (like PPC)
...
llvm-svn: 13438
2004-05-09 20:41:32 +00:00
87d036caf2
Do not emit prototypes for setjmp/longjmp, as they are handled specially
...
llvm-svn: 13437
2004-05-09 16:03:29 +00:00
17fcb67a20
Fine grainify namespacification
...
llvm-svn: 13436
2004-05-09 06:22:29 +00:00
9c29991970
Make the floating point constant pools local to each function, split the
...
FindUsedTypes manipulation stuff out to be a seperate pass, and make the
main CWriter be a function pass now!
llvm-svn: 13435
2004-05-09 06:20:51 +00:00
e3e557e553
Bug fixed.
...
Rewrite portability intro.
llvm-svn: 13434
2004-05-09 05:28:35 +00:00
14d328e2a5
Get this looking more like a function pass.
...
llvm-svn: 13433
2004-05-09 04:30:20 +00:00
254677fe0a
Implement the AddPrototypes method
...
llvm-svn: 13432
2004-05-09 04:29:57 +00:00
e7cc64f6e7
Add support for inserting all prototypes up-front
...
llvm-svn: 13431
2004-05-09 04:29:49 +00:00
2484a6336f
Print all PHI copies for successor blocks before the terminator, whether it be a conditional branch or switch.
...
llvm-svn: 13430
2004-05-09 03:42:48 +00:00
8ec5f88c79
Fix stupid bug in my checkin yesterday
...
llvm-svn: 13429
2004-05-08 22:41:42 +00:00
090365b154
Fixed up sched graph.
...
llvm-svn: 13428
2004-05-08 16:14:24 +00:00
f751ca6767
Registering the ModuloScheduling pass.
...
llvm-svn: 13427
2004-05-08 16:14:02 +00:00
8ad6335dfa
Changed CPUResource to allow access to max num users for a resource.
...
Also added ModuloScheduling as a friend.
llvm-svn: 13426
2004-05-08 16:13:26 +00:00
39bd20d6eb
Changed CPUResource to allow access to maxnum users.
...
llvm-svn: 13425
2004-05-08 16:12:50 +00:00
a6820d6704
Updating my versions of ModuloScheduling in cvs. Still not complete.
...
llvm-svn: 13424
2004-05-08 16:12:10 +00:00
1096ba9422
Add support for widening integral casts.
...
Flesh out the SetCC support... which currently ends in a little bit
of unfinished code (which is probably completely hilarious) for
generating the condition value splitting the basic block up into 4
blocks, like this (clearly a better API is needed for this!):
BB
cond. branch
/ / R1=1 R2=0
\ /
\ /
R=phi(R1,R2)
Other minor edits.
llvm-svn: 13423
2004-05-08 06:36:14 +00:00
91bf7cb79b
Add a bunch more branches
...
llvm-svn: 13422
2004-05-08 06:08:29 +00:00
7a66e5fff9
Flesh out GEP support
...
llvm-svn: 13421
2004-05-08 05:27:20 +00:00
d18b330605
Add ADD with immediate
...
llvm-svn: 13420
2004-05-08 05:26:55 +00:00
b56f1c9c10
Add forms of CMP, SUBCC, and a few branches, and some comments.
...
llvm-svn: 13419
2004-05-08 04:21:32 +00:00
f3a479f04a
Add stub support for GEPs.
...
Add support for branches (based loosely on X86/InstSelectSimple).
Add support for not visiting phi nodes in the first pass.
Add support for loading bools.
Flesh out support for stores.
llvm-svn: 13418
2004-05-08 04:21:17 +00:00
2a54b5d3e1
Add required header
...
llvm-svn: 13417
2004-05-08 03:50:03 +00:00
54a89b7c72
Remove unneeded header
...
llvm-svn: 13416
2004-05-08 03:49:35 +00:00
5f667a6f58
Implement folding of GEP's like:
...
%tmp.0 = getelementptr [50 x sbyte]* %ar, uint 0, int 5 ; <sbyte*> [#uses=2]
%tmp.7 = getelementptr sbyte* %tmp.0, int 8 ; <sbyte*> [#uses=1]
together. This patch actually allows us to simplify and generalize the code.
llvm-svn: 13415
2004-05-07 22:09:22 +00:00
9a7b9d07f0
Allow the user to set the LLVMINTERP environment variable as a workaround, for
...
when they have to run a gccld shell script without having lli in their path.
This is intended to address Bug 289.
Also, emit the traditional syntax ${1+"$@"} for passing all of a shell script's
args to a subprocess. If you have arguments that have spaces in them, $* will
not preserve the quoting (i.e., the quoted string "foo bar" as an argument will
end up as two arguments "foo" "bar" to lli.)
llvm-svn: 13414
2004-05-07 21:47:36 +00:00
584308b71a
Add support for copying bool constants to registers.
...
Disable the code that copies long constants to registers - it looks fishy.
Implement some simple casts: integral, smaller than longs, and equal-width
or narrowing only.
llvm-svn: 13413
2004-05-07 21:39:30 +00:00
a2dc6bf6e6
Codegen floating point stores of constants into integer instructions. This
...
allows us to compile:
store float 10.0, float* %P
into:
mov DWORD PTR [%EAX], 1092616192
instead of:
.CPItest_0: # float 0x4024000000000000
.long 1092616192 # float 10
...
fld DWORD PTR [.CPItest_0]
fstp DWORD PTR [%EAX]
llvm-svn: 13409
2004-05-07 21:18:15 +00:00
1edc9637d0
Add more casts. You can never have enough casts.
...
llvm-svn: 13408
2004-05-07 21:10:28 +00:00
cecf3f94a4
Make comparisons against the null pointer as efficient as integer comparisons
...
against zero. In particular, don't emit:
mov %ESI, 0
cmp %ECX, %ESI
instead, emit:
test %ECX, %ECX
llvm-svn: 13407
2004-05-07 19:55:55 +00:00
cf5822a2be
PR implemented
...
llvm-svn: 13406
2004-05-07 19:23:05 +00:00
3553fe3a5b
Bug fixed
...
llvm-svn: 13403
2004-05-07 18:40:38 +00:00
798b5e4a6f
New testcase
...
llvm-svn: 13402
2004-05-07 18:38:59 +00:00
0a8ee61c51
New testcase for PR336
...
llvm-svn: 13401
2004-05-07 15:38:50 +00:00
d9e5813821
Fix PR336: The instcombine pass asserts when visiting load instruction
...
llvm-svn: 13400
2004-05-07 15:35:56 +00:00
e0727fe00d
Regression test for PR#330.
...
llvm-svn: 13399
2004-05-07 13:57:19 +00:00
c6fdf8ddb3
Add the enum corresponding to the source change I made earlier
...
llvm-svn: 13395
2004-05-07 02:27:32 +00:00
6956afa7c5
Bug fixed.
...
llvm-svn: 13392
2004-05-06 22:23:24 +00:00
2af0fd3ca8
Don't call getForwardedType() twice, as recommended by Chris.
...
llvm-svn: 13391
2004-05-06 22:15:47 +00:00
6aee736d1b
Use the new commandline flag to allow us to call bugpoint like this:
...
bugpoint ... --tool-args -enable-correct-eh-support -regalloc=linearscan --args -- -foo
So that tool-args option gets the -enable-correct-eh-support -regalloc=linearscan flags instead of bugpoint.
llvm-svn: 13389
2004-05-06 22:05:35 +00:00
a60f355c87
Implement the new cl::PositionalEatsArgs flag, refactor code a bit
...
llvm-svn: 13388
2004-05-06 22:04:31 +00:00
b68bd76978
Add a new cl::PositionalEatsArgs flag
...
llvm-svn: 13387
2004-05-06 22:03:59 +00:00
f6709bc579
Fix for PR#330.
...
When looking at getelementptr instructions, make sure to use a forwarded
type. We want to do this because a DerivedType may drop its uses and then
refine its users, who may then use another user who hasn't been refined yet.
By getting the forwarded type, we always ensure that we're looking at a
Type that isn't in a halfway refined state.
Now, I should be able to put this stuff in PATypeHandle, but it doesn't work
for some reason. This should do for now.
llvm-svn: 13386
2004-05-06 21:18:08 +00:00
8737169a6a
Remove a really old comment
...
llvm-svn: 13385
2004-05-06 19:29:58 +00:00
432f53b47f
Just testing the "Reid has CVS commit access" system
...
llvm-svn: 13384
2004-05-06 18:06:18 +00:00
afdd8a0608
Give props to Vladimir Prus for the inst_iterator patch.
...
llvm-svn: 13383
2004-05-06 16:53:18 +00:00
991425ae70
numeric_limits::infinity() apparently does not work on all systems. As a
...
workaround, use the C HUGE_VAL macro instead.
llvm-svn: 13377
2004-05-06 16:25:59 +00:00
4c4497e834
Fix for gcc3.4: invalid use of forward delacred class on line 93
...
llvm-svn: 13370
2004-05-06 02:07:42 +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
706bd5bf2b
Apply simplification suggested by Chris: why assign() when operator = will do?
...
llvm-svn: 13364
2004-05-04 22:02:41 +00:00
987ad1945d
Fixed inconsistent indentation.
...
llvm-svn: 13363
2004-05-04 21:46:05 +00:00
d26074eef1
Missing piece of fix for Bug 333
...
llvm-svn: 13362
2004-05-04 21:41:45 +00:00
e307f8a791
Bugs fixed
...
llvm-svn: 13360
2004-05-04 21:22:57 +00:00
48ef3f0fa1
update command guide to have --tool-args
...
llvm-svn: 13359
2004-05-04 21:13:35 +00:00
4a278f0b1b
Add --tool-args flag which lets you pass arguments to llc or lli.
...
This is intended to address Bug 40.
llvm-svn: 13358
2004-05-04 21:09:16 +00:00
f2ff5dd442
Correctly mangle function names when they are used as part of a
...
constant pool member's name.
This is intended to address Bug 333.
Also, fix an anachronistic usage of "M" as a parameter of type Function *.
llvm-svn: 13357
2004-05-04 21:09:02 +00:00
b3a9a0fcbd
Add "Args" optional argument to AbstractInterpreter factory methods, which
...
fills in a ToolArgs vector in the AbstractInterpreter if it is set. This
ToolArgs vector is used to pass additional arguments to LLI and/or LLC.
This is intended to address Bug 40.
Also, make -debug-only=toolrunner work for the LLC and CBE
AbstractInterpreters.
llvm-svn: 13356
2004-05-04 21:09:01 +00:00
c6f601310f
Remove unneeded check
...
llvm-svn: 13355
2004-05-04 19:35:11 +00:00
22df9a59b4
Improve signed division by power of 2 *dramatically* from this:
...
div:
mov %EDX, DWORD PTR [%ESP + 4]
mov %ECX, 64
mov %EAX, %EDX
sar %EDX, 31
idiv %ECX
ret
to this:
div:
mov %EAX, DWORD PTR [%ESP + 4]
mov %ECX, %EAX
sar %ECX, 5
shr %ECX, 26
mov %EDX, %EAX
add %EDX, %ECX
sar %EAX, 6
ret
Note that the intel compiler is currently making this:
div:
movl 4(%esp), %edx #3.5
movl %edx, %eax #4.14
sarl $5, %eax #4.14
shrl $26, %eax #4.14
addl %edx, %eax #4.14
sarl $6, %eax #4.14
ret #4.14
Which has one less register->register copy. (hint hint alkis :)
llvm-svn: 13354
2004-05-04 19:33:58 +00:00
bc1ccfd053
Provide visit(Module&) and visitModule(Module&) functionality.
...
llvm-svn: 13353
2004-05-04 18:30:38 +00:00
600f2045a9
Add stub support for reading BBTraces.
...
llvm-svn: 13352
2004-05-04 17:11:14 +00:00
9a1b8c73a8
Add BBTrace accessor method and data member.
...
llvm-svn: 13351
2004-05-04 17:11:13 +00:00
ea309ff082
Bug fixed
...
llvm-svn: 13350
2004-05-04 17:04:09 +00:00
9490849028
Do not mark instructions in unreachable sections of the function as live.
...
This fixes PR332 and ADCE/2004-05-04-UnreachableBlock.llx
llvm-svn: 13349
2004-05-04 17:00:46 +00:00
1190a4de0b
New testcase for PR332
...
llvm-svn: 13348
2004-05-04 17:00:10 +00:00
ba7b0d7a0e
New header file containing profile info enums shared between the C++ analysis
...
libraries and the C runtime support library
llvm-svn: 13347
2004-05-04 16:57:57 +00:00
34c13fbe94
Share ProfilingType enum with the C profiling runtime libraries.
...
llvm-svn: 13346
2004-05-04 16:53:07 +00:00
495d9340e9
Share the profile info type enum with the C++ analysis libraries.
...
Add a documentation comment for write_profiling_data().
llvm-svn: 13345
2004-05-04 16:51:48 +00:00
875d70b3c0
Constants for profile info type changed names to match the C++ ones.
...
llvm-svn: 13344
2004-05-04 16:51:47 +00:00
6e72e2a6e9
Constants for profile info type changed names to match the C++ ones.
...
Edited comments.
llvm-svn: 13343
2004-05-04 16:51:46 +00:00
8c22ece232
Improve code generated for integer multiplications by 2,3,5,9
...
llvm-svn: 13342
2004-05-04 15:47:14 +00:00
dd1a86d858
Minor efficiency tweak, suggested by Patrick Meredith
...
llvm-svn: 13341
2004-05-04 15:19:33 +00:00
5237476f75
Fix typo
...
llvm-svn: 13340
2004-05-03 23:52:07 +00:00
2157c9a49b
Add initial implementation of basic block tracing runtime
...
llvm-svn: 13339
2004-05-03 23:51:50 +00:00
417d555b6f
Add basic block tracing functions as exported symbols.
...
llvm-svn: 13338
2004-05-03 23:49:17 +00:00
213715bceb
Add basic block tracing information as a type of "profiling" information.
...
llvm-svn: 13337
2004-05-03 23:49:16 +00:00
e96196081e
In InsertProfilingInitCall(), make it legal to pass in a null array, in
...
which case you'll get a null array and zero passed to the profiling function.
llvm-svn: 13336
2004-05-03 22:06:33 +00:00
088dd3e121
Add initial implementation of basic-block tracing instrumentation pass.
...
llvm-svn: 13335
2004-05-03 22:06:32 +00:00
6e6baf6157
Initialize member out of paranoia
...
llvm-svn: 13319
2004-05-02 16:06:40 +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
be6f06818c
Do not clone arbitrary condition instructions.
...
llvm-svn: 13316
2004-05-02 05:19:36 +00:00
51a6dbcb65
Do not infinitely "unroll" single BB loops.
...
llvm-svn: 13315
2004-05-02 05:02:03 +00:00
2ad4878d8e
Add some stuff to the release notes.
...
llvm-svn: 13314
2004-05-02 04:19:15 +00:00
19202cd08b
Chris told me to take these assertions out a few days ago, but I forgot to
...
check this in.
llvm-svn: 13313
2004-05-02 01:04:52 +00:00
1e94ed606e
Dont' merge terminators that are needed to select PHI node values.
...
llvm-svn: 13312
2004-05-02 01:00:44 +00:00
2e93c4275e
Implement SimplifyCFG/branch-cond-merge.ll
...
Turning "if (A < B && B < C)" into "if (A < B & B < C)"
llvm-svn: 13311
2004-05-01 23:35:43 +00:00
c9b34e5648
New testcase:
...
Turn "if (A < B && B < C)" into "if (A < B & B < C)"
llvm-svn: 13310
2004-05-01 23:35:21 +00:00
63d75af920
Make sure to reprocess instructions used by deleted instructions to avoid
...
missing opportunities for combination.
llvm-svn: 13309
2004-05-01 23:27:23 +00:00
b643a9e675
Make sure the instruction combiner doesn't lose track of instructions
...
when replacing them, missing the opportunity to do simplifications
llvm-svn: 13308
2004-05-01 23:19:52 +00:00
4cbd160b45
Fix my missing parens
...
llvm-svn: 13307
2004-05-01 22:41:51 +00:00
88da6f7b52
Implement SimplifyCFG/branch-cond-prop.ll
...
llvm-svn: 13306
2004-05-01 22:36:37 +00:00
380affdc88
New testcase for case that simplifycfg should catch
...
llvm-svn: 13305
2004-05-01 22:36:24 +00:00
7b0a2046d4
Remove unused #include
...
llvm-svn: 13304
2004-05-01 21:29:16 +00:00
acbf0c84de
Iterate over the Machine CFG that Brian added instead of the LLVM CFG.
...
Look at all of the pretty minuses. :)
llvm-svn: 13303
2004-05-01 21:27:53 +00:00
8c0bc68910
Operate on the Machine CFG instead of on the LLVM CFG
...
llvm-svn: 13302
2004-05-01 21:24:39 +00:00
c49a9a5d76
Stop LiveVariables from using BasicBlocks as part of the mapping, instead
...
use MachineBasicBlocks. To do this, we traverse the Machine CFG instead of
the LLVM CFG, which is also *MUCH* more efficient by having fewer levels of
indirections and mappings.
llvm-svn: 13301
2004-05-01 21:24:24 +00:00
5e65763532
Stop LiveVariables from using BasicBlocks as part of the mapping, instead
...
use MachineBasicBlocks.
llvm-svn: 13300
2004-05-01 21:23:35 +00:00
c89208483f
Move the GraphTraits for MachineBasicBlocks to the MachineBasicBlock file.
...
llvm-svn: 13299
2004-05-01 21:05:34 +00:00
15a041ca88
Add comment about optimizations
...
llvm-svn: 13298
2004-05-01 11:40:07 +00:00
490e804a4e
Add a constructor that got lost
...
llvm-svn: 13297
2004-05-01 11:17:13 +00:00
2d7efbbb93
Generalize the strlen size_t hack, for the benefit of the other external
...
functions with wrappers that either take or return size_ts.
llvm-svn: 13296
2004-05-01 06:42:15 +00:00
77a51733e7
Fix broken link, again.
...
llvm-svn: 13295
2004-04-30 22:17:12 +00:00
f0fda31479
Sorry, now friend class name should be right!!
...
llvm-svn: 13294
2004-04-30 20:48:37 +00:00
34f7cfe4d0
Fixed friend class name for ModuloScheduling.
...
llvm-svn: 13293
2004-04-30 20:47:59 +00:00
717f2a51fd
Fixed friend class name for ModuloSched
...
llvm-svn: 13292
2004-04-30 20:46:40 +00:00
6ebf67f8b6
Removing MachineResource class.
...
llvm-svn: 13291
2004-04-30 20:40:38 +00:00
e8b56f27b8
Fix a broken link
...
llvm-svn: 13290
2004-04-30 14:13:16 +00:00
652064e3b8
Fix a major pessimization in the instcombiner. If an allocation instruction
...
is only used by a cast, and the casted type is the same size as the original
allocation, it would eliminate the cast by folding it into the allocation.
Unfortunately, it was placing the new allocation instruction right before
the cast, which could pull (for example) alloca instructions into the body
of a function. This turns statically allocatable allocas into expensive
dynamically allocated allocas, which is bad bad bad.
This fixes the problem by placing the new allocation instruction at the same
place the old one was, duh. :)
llvm-svn: 13289
2004-04-30 04:37:52 +00:00
63c3b3c8d7
Add machine-CFG graph traits specializations.
...
llvm-svn: 13288
2004-04-29 17:43:55 +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
40b4f6eca8
Added `zeroinitializer' keyword.
...
llvm-svn: 13253
2004-04-28 19:36:08 +00:00
da546eac70
class AssemblyWriter:
...
* Make contained ostream pointer, not reference
* Allow setting of that ostream via setStream()
class CachedWriter:
* setStream() in turn calls setStream() on the AssemblyWriter
llvm-svn: 13247
2004-04-28 19:24:28 +00:00
2dc1944f14
* Make contained ostream not public.
...
* Remove various print methods that called the Value* method, just have one that
all subclasses of Value will use anyway.
* Remove template for printing constant references
* Add methods to print char* and strings
* setStream now sets the stream on the contained AssemblyWriter
llvm-svn: 13246
2004-04-28 19:22:58 +00:00
7e66438579
Send text and numbers directly to CachedWriter's contained ostream.
...
llvm-svn: 13243
2004-04-28 18:52:43 +00:00
1846321345
Squelch compile-time warning (profile build).
...
llvm-svn: 13228
2004-04-28 15:32:09 +00:00
4685e26bc7
* Add ability to print out type as symbolic
...
* Add Module accessor to AssemblyWriter
llvm-svn: 13227
2004-04-28 15:31:21 +00:00
d472aff18c
* Add ability to get and set the output stream
...
* New feature: outputting a Type* as symbolic, controlled via the stream
similarly to sending std::hex to change number format
llvm-svn: 13226
2004-04-28 15:30:33 +00:00
4ba46506df
Add pred./succ. list size methods.
...
llvm-svn: 13214
2004-04-28 04:46:35 +00:00
4390e4a7fc
Make RequiresFPRegKill() take a MachineBasicBlock arg.
...
In InsertFPRegKills(), just check the MachineBasicBlock for successors
instead of its corresponding BasicBlock.
llvm-svn: 13213
2004-04-28 04:45:55 +00:00
33ff118407
In InsertFPRegKills(), use the machine-CFG itself rather than the
...
LLVM CFG when trying to find the successors of BB.
llvm-svn: 13212
2004-04-28 04:34:16 +00:00
24ec85689b
Update the machine-CFG edges whenever we see a branch.
...
llvm-svn: 13211
2004-04-28 04:19:37 +00:00
153cb8d18a
Move private methods to end of class decl at Chris's request
...
llvm-svn: 13210
2004-04-28 04:15:06 +00:00
2d875451d2
Fix thinkos that Chris caught for me.
...
Make pred mutators private.
llvm-svn: 13209
2004-04-28 03:59:48 +00:00
dc745e6dc1
Add machine-code CFG support: MachineBasicBlocks may now have their own
...
predecessors and successors
llvm-svn: 13208
2004-04-28 02:16:33 +00:00
444a03b77a
Integrate the rest of my random sparcv9 scribblings into this file
...
llvm-svn: 13204
2004-04-27 22:04:03 +00:00
f011b7dfe6
Fix warning building in optimized mode
...
llvm-svn: 13190
2004-04-27 18:24:38 +00:00
f1c9593a90
Fix warning
...
llvm-svn: 13189
2004-04-27 18:21:56 +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
8bcc86b6ae
Because I like being able to instantiate the cfgprinter from external projects,
...
this header file is born.
llvm-svn: 13176
2004-04-26 16:27:53 +00:00
104341f31e
Add functions that return instances of these printer passes
...
llvm-svn: 13175
2004-04-26 16:27:08 +00:00
1b07de4c2b
Fix a typo in a comment.
...
llvm-svn: 13174
2004-04-26 16:26:21 +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
e20c334e65
Instcombine X/-1 --> 0-X
...
llvm-svn: 13172
2004-04-26 14:01:59 +00:00
e24003a261
Add a new testcase for X/-1, fix bug that prevented tests from running right
...
llvm-svn: 13171
2004-04-26 14:01:47 +00:00
03cac379bd
Fix file header comments and include guards -- many files have been moved or
...
renamed since they were last spiffed up, or they just never had proper comments
in the first place.
llvm-svn: 13148
2004-04-25 07:04:49 +00:00
42d321ab66
Add a getRegisterInfo() accessor just like on the X86 target.
...
llvm-svn: 13147
2004-04-25 06:32:28 +00:00
4ec08db77c
Regularize file header comment and include guard.
...
Include SparcV9RegisterInfo.h.
Add a getRegisterInfo() accessor and SparcV9RegisterInfo instance, just like
on the X86 target.
llvm-svn: 13146
2004-04-25 06:32:16 +00:00
f70544c583
Add MRegisterInfo subclass for the SparcV9 target (containing only stub
...
functions for now). This automatically turns on the printing of machine
registers using their own real names, instead of goofy things like %mreg(42),
and allows us to migrate code incrementally to the new interface as we see fit.
The register file description it uses is hand-written, so that the register
numbers will match the ones that the SparcV9 target already uses.
Perhaps someday we'll tablegen it.
llvm-svn: 13145
2004-04-25 06:32:05 +00:00
7426c893f0
Standardize header comments of top-level Makefiles.
...
llvm-svn: 13143
2004-04-24 00:10:56 +00:00
fc993c39e7
Aggregating function arguments is now an option. Default is `no', as before.
...
llvm-svn: 13142
2004-04-23 23:54:34 +00:00
3596f0a180
* Allow aggregating extracted function arguments (controlled by flag)
...
* Commandline option (for now) controls that flag that is passed in
llvm-svn: 13141
2004-04-23 23:54:17 +00:00
ced65e8792
Fix a typo.
...
llvm-svn: 13136
2004-04-23 21:45:02 +00:00
83cd87efcd
Move the scev expansion code into this pass, where it belongs. There is
...
still room for cleanup, but at least the code modification is out of the
analysis now.
llvm-svn: 13135
2004-04-23 21:29:48 +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
0eab307e3c
Remove the SCEV::expandCodeFor method, add a new SCEVVisitor class.
...
llvm-svn: 13133
2004-04-23 21:28:25 +00:00
02cea4afd3
Teach bugpoint to be a little bit smarter and avoid repeating work
...
llvm-svn: 13132
2004-04-23 20:36:51 +00:00
99e3e94012
Merged this file into the SparcV9 target.
...
llvm-svn: 13128
2004-04-23 18:17:03 +00:00
0b65f8f582
Merge TargetRegInfo.h into SparcV9RegInfo.h, which is its only subclass.
...
This prepares us to be able to de-virtualize and de-abstract it, and
take the register allocator bits out and move them into the register allocator
proper...
llvm-svn: 13127
2004-04-23 18:15:48 +00:00
71509a98c5
Include SparcV9RegInfo.h instead of TargetRegInfo.h.
...
llvm-svn: 13126
2004-04-23 18:15:47 +00:00
a411e07db0
Include SparcV9RegInfo.h instead of TargetRegInfo.h. This serves as a bit of
...
documentation that this module needs to be made independent of the
register file description of the current target.
llvm-svn: 13125
2004-04-23 18:15:46 +00:00
fbd43e43a7
Get rid of the old byte-at-a-time emission code used when the Sparc JIT was
...
being tested on X86, as per Chris's request.
llvm-svn: 13124
2004-04-23 18:10:38 +00:00
ad373c8576
Go back to the interpreter main loop after performing intrinsic lowering,
...
because 1) the first instruction might not be a call site, and
2) CS and SF.Caller were not getting set to point to the new call site
anyway (resulting in a crash on e.g. call %llvm.memset).
llvm-svn: 13122
2004-04-23 18:05:28 +00:00
11550c1deb
mmap of a zero length file returns null on some platforms, so hack around it.
...
llvm-svn: 13121
2004-04-23 17:38:17 +00:00
6e102a7edd
Use emitWordAt() to emit forward-branch fixups.
...
llvm-svn: 13120
2004-04-23 17:11:16 +00:00
fe277c2809
Emit SPARC machine code a word at a time instead of a byte at a time.
...
Use emitWordAt() to emit forward-branch fixups.
llvm-svn: 13119
2004-04-23 17:11:15 +00:00
0c4fb283f3
Implement emitWordAt() for the JIT emitter.
...
llvm-svn: 13118
2004-04-23 17:11:14 +00:00
4aaa9c603a
Implement emitWordAt() for the debug emitter and the file printer emitter. (I
...
am not so sure about the file printer emitter, but the debug emitter change
should be harmless.)
llvm-svn: 13117
2004-04-23 17:11:13 +00:00
1ed0cdbd04
Add emitWordAt() - a quick and dirty interface that the machine-dependent
...
emitters can use to emit "relocations".
llvm-svn: 13116
2004-04-23 17:11:12 +00:00
2ccac826e4
Add command-line option to select whether to isolate or delete function from
...
module. Default is `isolate' as before.
llvm-svn: 13113
2004-04-22 23:07:39 +00:00
98aa516a9c
Clarify the logic: the flag is renamed to `deleteFn' to signify it will delete
...
the function instead of isolating it. This also means the condition is reversed.
llvm-svn: 13112
2004-04-22 23:00:51 +00:00
e0682426f0
Add a flag to choose between isolating a function or deleting the function from
...
the Module. The default behavior keeps functionality as before: the chosen
function is the one that remains.
llvm-svn: 13111
2004-04-22 22:52:22 +00:00
2de9ce4f51
Add a boolean flag to delete this function from module, leaving the rest behind.
...
Useful in manual debugging when bugpoint isn't quite up to snuff.
llvm-svn: 13110
2004-04-22 22:51:37 +00:00
324e9e06d8
Add a space before result for readability on the command line.
...
llvm-svn: 13109
2004-04-22 20:02:09 +00:00
c27302c79f
Disable a previous patch that was causing indvars to loop infinitely :(
...
llvm-svn: 13108
2004-04-22 15:12:36 +00:00
8f02e9c5f7
Add an ugly cast
...
llvm-svn: 13107
2004-04-22 15:00:36 +00:00
c1a682dda0
Fix an extremely serious thinko I made in revision 1.60 of this file.
...
llvm-svn: 13106
2004-04-22 14:59:40 +00:00
334bee87f4
Add a method
...
llvm-svn: 13105
2004-04-22 14:56:51 +00:00
af532f27e7
Implement a todo, rewriting all possible scev expressions inside of the
...
loop. This eliminates the extra add from the previous case, but it's
not clear that this will be a performance win overall. Tommorows test
results will tell. :)
llvm-svn: 13103
2004-04-21 23:36:08 +00:00
fb9a299f68
This code really wants to iterate over the OPERANDS of an instruction, not
...
over its USES. If it's dead it doesn't have any uses! :)
Thanks to the fabulous and mysterious Bill Wendling for pointing this out. :)
llvm-svn: 13102
2004-04-21 22:29:37 +00:00
dc7cc35088
Implement a fixme. The helps loops that have induction variables of different
...
types in them. Instead of creating an induction variable for all types, it
creates a single induction variable and casts to the other sizes. This generates
this code:
no_exit: ; preds = %entry, %no_exit
%indvar = phi uint [ %indvar.next, %no_exit ], [ 0, %entry ] ; <uint> [#uses=4]
*** %j.0.0 = cast uint %indvar to short ; <short> [#uses=1]
%indvar = cast uint %indvar to int ; <int> [#uses=1]
%tmp.7 = getelementptr short* %P, uint %indvar ; <short*> [#uses=1]
store short %j.0.0, short* %tmp.7
%inc.0 = add int %indvar, 1 ; <int> [#uses=2]
%tmp.2 = setlt int %inc.0, %N ; <bool> [#uses=1]
%indvar.next = add uint %indvar, 1 ; <uint> [#uses=1]
br bool %tmp.2, label %no_exit, label %loopexit
instead of:
no_exit: ; preds = %entry, %no_exit
%indvar = phi ushort [ %indvar.next, %no_exit ], [ 0, %entry ] ; <ushort> [#uses=2]
*** %indvar = phi uint [ %indvar.next, %no_exit ], [ 0, %entry ] ; <uint> [#uses=3]
%indvar = cast uint %indvar to int ; <int> [#uses=1]
%indvar = cast ushort %indvar to short ; <short> [#uses=1]
%tmp.7 = getelementptr short* %P, uint %indvar ; <short*> [#uses=1]
store short %indvar, short* %tmp.7
%inc.0 = add int %indvar, 1 ; <int> [#uses=2]
%tmp.2 = setlt int %inc.0, %N ; <bool> [#uses=1]
%indvar.next = add uint %indvar, 1
*** %indvar.next = add ushort %indvar, 1
br bool %tmp.2, label %no_exit, label %loopexit
This is an improvement in register pressure, but probably doesn't happen that
often.
The more important fix will be to get rid of the redundant add.
llvm-svn: 13101
2004-04-21 22:22:01 +00:00
be8bb804c5
Fix an incredibly nasty iterator invalidation problem. I am too spoiled by ilists :)
...
Eventually it would be nice if CallGraph maintained an ilist of CallGraphNode's instead
of a vector of pointers to them, but today is not that day.
llvm-svn: 13100
2004-04-21 20:44:33 +00:00
15572abd14
bug fixed
...
llvm-svn: 13099
2004-04-21 18:52:06 +00:00
ee80c45d46
new testcase for PR326
...
llvm-svn: 13098
2004-04-21 18:49:35 +00:00
0784a601c2
Add doxygenified comments to functions.
...
llvm-svn: 13097
2004-04-21 18:36:43 +00:00
3bcead7201
I'm allergic to the word `stuff'.
...
llvm-svn: 13096
2004-04-21 18:27:56 +00:00
376ac1bed5
Make SparcV9RegInfo::getRegType() return the right answer for registers
...
of IntCC, FloatCC, and Special types.
Make SparcV9RegInfo::getRegClassIDOfRegType() return the right answer
if you ask for the class corresponding to SpecialRegType.
llvm-svn: 13095
2004-04-21 17:53:58 +00:00
f68f40ea42
Include cerrno (gcc-3.4 fix)
...
llvm-svn: 13091
2004-04-21 16:11:40 +00:00
07f7574ba3
Declare iterator as public since it is defined as such (gcc-3.4 fix)
...
llvm-svn: 13090
2004-04-21 16:10:40 +00:00
a9691fe70d
Fix typeo
...
llvm-svn: 13089
2004-04-21 14:23:18 +00:00
c87784f1fc
REALLY fix PR324: don't delete linkonce functions until after the SCC traversal
...
is done, which avoids invalidating iterators in the SCC traversal routines
llvm-svn: 13088
2004-04-20 22:06:53 +00:00
99504890b7
Pass the callgraph not the module
...
llvm-svn: 13087
2004-04-20 21:52:26 +00:00
a45086a45e
Allow getting the module from a call graph
...
llvm-svn: 13086
2004-04-20 21:52:12 +00:00
0a985ac82c
Change it to take a callgraph, from which we can get a module
...
llvm-svn: 13085
2004-04-20 21:52:07 +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
c1aa21f5a7
Fix PR325
...
llvm-svn: 13081
2004-04-20 20:26:03 +00:00
514934051a
Fix PR324 and testcase: Inline/2004-04-20-InlineLinkOnce.llx
...
llvm-svn: 13080
2004-04-20 20:20:59 +00:00
40123bff5b
New testcase, distilled from PR324
...
llvm-svn: 13079
2004-04-20 20:15:01 +00:00
a2f66db6c4
Make it legal to ask for the type of a specialreg
...
llvm-svn: 13078
2004-04-20 20:12:57 +00:00
86a6c75a7e
Add some PRs which were fixed, but not previously mentioned in the rel. notes.
...
llvm-svn: 13077
2004-04-20 18:18:46 +00:00
2b2d7a99be
Add support for the select instruction
...
llvm-svn: 13076
2004-04-20 16:43:21 +00:00
6c272a93bb
Make it legal to request a load or store of %fsr.
...
llvm-svn: 13073
2004-04-19 19:12:12 +00:00
627abf14bb
Fix typo.
...
llvm-svn: 13072
2004-04-19 19:09:24 +00:00
755651df27
Regularize include guards and remove some excess whitespace.
...
llvm-svn: 13071
2004-04-19 18:53:44 +00:00
2fd2536ea8
Tighten up SparcV9FloatCCRegClass::getRegName()'s assertion - if you ask it for
...
the name of %fsr (as the comment in SparcV9RegClassInfo.h used to suggest)
you would walk off the end of the FloatCCRegName array.
llvm-svn: 13070
2004-04-19 18:53:43 +00:00
85cb77737c
Regularize include guards, remove some excess whitespace and fix some comments.
...
Remove the extra %fsr register from SparcV9FloatCCRegClass.
llvm-svn: 13069
2004-04-19 18:53:42 +00:00
f48f777d4c
Initial checkin of a simple loop unswitching pass. It still needs work,
...
but it's a start, and seems to do it's basic job.
llvm-svn: 13068
2004-04-19 18:07:02 +00:00
ceea37590f
New testcase
...
llvm-svn: 13067
2004-04-19 18:06:34 +00:00
1c9fb0007e
Add accessor for a Loop Unswitching pass
...
llvm-svn: 13066
2004-04-19 06:28:37 +00:00
46f888dd80
Eliminate a poorly conceived method
...
llvm-svn: 13065
2004-04-19 06:26:46 +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
ce89b392c6
As per Chris, greatly simplify handling of external functions by using the
...
wrapper idea uniformly: we can use Value::replaceAllUsesWith() instead of
special-casing by class of user.
llvm-svn: 13063
2004-04-19 03:36:47 +00:00
d2493e1565
* Remove note about "UNSUPPORTED: " feature that just became supported
...
* Fix usage of <p> tag and some grammar
llvm-svn: 13062
2004-04-19 03:28:39 +00:00
8a32c6d91b
* Reorder #includes
...
* Wrap a long line
llvm-svn: 13061
2004-04-19 03:12:35 +00:00
4825e215c0
Display sample bugpoint output with the `code' stylesheet class.
...
llvm-svn: 13060
2004-04-19 03:07:29 +00:00
5414021060
* Improve file header comment
...
* Remove #include
* Add some methods to update loop/loopinfo
* Stop explicitly holding the loop depth in the Loop class. Instead, just
dynamically calculate it. This makes it easier to update LoopInfo as a
result of transformations.
llvm-svn: 13059
2004-04-19 03:03:19 +00:00
32447c55fe
Remove code to update loop depths
...
llvm-svn: 13058
2004-04-19 03:02:09 +00:00
bc02177fdc
Add #include
...
llvm-svn: 13057
2004-04-19 03:01:23 +00:00
e70502328e
Add notes about PRs 38 and 290 being fixed.
...
llvm-svn: 13056
2004-04-19 03:01:01 +00:00
017f2f3405
Wrap (some) lines to 80 cols.
...
llvm-svn: 13055
2004-04-19 02:56:49 +00:00
8301864215
Finally implement rewriting global initializers which use external functions
...
by creating an internal wrapper function with same signature as the external
function, and use it instead of the "real" function.
The wrapper then calls the external function using the same JIT function
resolution API that has been used before for rewriting instructions, since the
wrapper has an explicit call instruction which we can rewrite.
llvm-svn: 13054
2004-04-19 01:12:01 +00:00
732b5d5bba
Provide an interface that is more convenient for iterating over the blocks
...
in a loop.
llvm-svn: 13052
2004-04-18 23:37:03 +00:00
fc44a25bcb
Move isLoopInvariant to the Loop class
...
llvm-svn: 13051
2004-04-18 22:46:08 +00:00
26ba2f5fe8
Add new method
...
llvm-svn: 13050
2004-04-18 22:45:27 +00:00
d4d66c86c0
Add new method.
...
llvm-svn: 13049
2004-04-18 22:45:16 +00:00
827826320d
Correct rewriting of exit blocks after my last patch
...
llvm-svn: 13048
2004-04-18 22:27:10 +00:00
d6ce359d3b
Fix computation of exit blocks
...
llvm-svn: 13047
2004-04-18 22:21:41 +00:00
35eaa55cfc
Loop exit sets are no longer explicitly held, they are dynamically computed on demand.
...
llvm-svn: 13046
2004-04-18 22:15:13 +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
bf3c8763c0
Spiff is no longer in our tree
...
llvm-svn: 13041
2004-04-18 18:26:38 +00:00
d15250240c
Reduce the unrolling limit
...
llvm-svn: 13040
2004-04-18 18:06:14 +00:00
30ae18155d
If the preheader of the loop was the entry block of the function, make sure
...
that the exit block of the loop becomes the new entry block of the function.
This was causing a verifier assertion on 252.eon.
llvm-svn: 13039
2004-04-18 17:38:42 +00:00
230bcb6b35
Be much more careful about how we update instructions outside of the loop
...
using instructions inside of the loop. This should fix the MishaTest failure
from last night.
llvm-svn: 13038
2004-04-18 17:32:39 +00:00
3796974563
Another testcase
...
llvm-svn: 13037
2004-04-18 06:55:57 +00:00
e375a4fdc2
Implement method
...
llvm-svn: 13036
2004-04-18 06:54:48 +00:00
5f46b7044c
Add a new method
...
llvm-svn: 13035
2004-04-18 06:54:34 +00:00
4d52e1e401
After unrolling our single basic block loop, fold it into the preheader and exit
...
block. The primary motivation for doing this is that we can now unroll nested loops.
This makes a pretty big difference in some cases. For example, in 183.equake,
we are now beating the native compiler with the CBE, and we are a lot closer
with LLC.
I'm now going to play around a bit with the unroll factor and see what effect
it really has.
llvm-svn: 13034
2004-04-18 06:27:43 +00:00
f2cc841619
Fix a bug: this does not preserve the CFG!
...
While we're at it, add support for updating loop information correctly.
llvm-svn: 13033
2004-04-18 05:38:37 +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
c9de38d316
Allow clients to delete loops, add a new method
...
llvm-svn: 13031
2004-04-18 05:37:42 +00:00
bf9ba24155
Move loop optimization passes up, add loop unroller
...
llvm-svn: 13030
2004-04-18 05:21:01 +00:00
e4ce5dbd1d
Add prototype
...
llvm-svn: 13029
2004-04-18 05:20:32 +00:00
946b255977
Initial checkin of a simple loop unroller. This pass is extremely basic and
...
limited. Even in it's extremely simple state (it can only *fully* unroll single
basic block loops that execute a constant number of times), it already helps improve
performance a LOT on some benchmarks, particularly with the native code generators.
llvm-svn: 13028
2004-04-18 05:20:17 +00:00
c14da9600b
Make the tail duplication threshold accessible from the command line instead of hardcoded
...
llvm-svn: 13025
2004-04-18 00:52:43 +00:00
ca96cee67b
Fix a memory leak. We leaked the vector holding the entries in switch tables.
...
llvm-svn: 13023
2004-04-17 23:49:15 +00:00
bc4118f316
New testcase
...
llvm-svn: 13020
2004-04-17 23:00:51 +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
a814080025
If the loop executes a constant number of times, try a bit harder to replace
...
exit values.
llvm-svn: 13018
2004-04-17 18:44:09 +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
Chris Lattner