44b1b22a2d
int64_t -> int
...
llvm-svn: 11977
2004-02-29 05:07:02 +00:00
e669caafe0
Rename member function to be consistent with the rest.
...
llvm-svn: 11898
2004-02-27 06:11:15 +00:00
941f9310bb
Make spiller push stores right after the definition of a register so
...
that they are as far away from the loads as possible.
llvm-svn: 11895
2004-02-27 04:51:35 +00:00
519e1e8c92
Fix crash caused by passing register 0 to
...
MRegisterInfo::isPhysicalRegister().
llvm-svn: 11894
2004-02-27 01:52:34 +00:00
5a3bab9402
Clear maps right after basic block is processed.
...
llvm-svn: 11892
2004-02-26 23:22:23 +00:00
61719d48d2
Uncomment assertions that register# != 0 on calls to
...
MRegisterInfo::is{Physical,Virtual}Register. Apply appropriate fixes
to relevant files.
llvm-svn: 11882
2004-02-26 22:00:20 +00:00
28a0885929
No need to clear the map here, it will always be empty
...
llvm-svn: 11868
2004-02-26 05:21:21 +00:00
e62ddd405d
Fix bugs found with recent addition of assertions in
...
MRegisterInfo::is{Physical,Virtual}Register.
llvm-svn: 11849
2004-02-25 23:21:52 +00:00
84b76c9be0
Great sparc renaming fallout IV: Sparc --> SparcV9.
...
llvm-svn: 11844
2004-02-25 22:09:36 +00:00
a9f03fba9d
Remove asssert since it is breaking cases that it shouldn't.
...
llvm-svn: 11841
2004-02-25 22:01:06 +00:00
d8bace7f60
Add DenseMap template and actually use it for for mapping virtual regs
...
to objects.
llvm-svn: 11840
2004-02-25 21:55:45 +00:00
f5a393a133
Add an assertion
...
llvm-svn: 11830
2004-02-25 19:37:44 +00:00
10a32da382
FunctionLiveVarInfo.h moved: include/llvm/CodeGen -> lib/Target/Sparc/LiveVar
...
llvm-svn: 11804
2004-02-24 19:46:00 +00:00
1dd872ce94
Move machine code rewriter and spiller outside the register
...
allocator.
The implementation is completely rewritten and now employs several
optimizations not exercised before. For example for 164.gzip we have
997 loads and 699 stores vs the 1221 loads and 880 stores we have
before.
llvm-svn: 11798
2004-02-24 08:58:30 +00:00
63aea0b6a5
Add predicates for checking if a virtual register has a physical
...
register mapping or a stack slot mapping.
llvm-svn: 11795
2004-02-24 06:30:36 +00:00
8b571a64ac
Make enum private as it is an implementation detail.
...
llvm-svn: 11782
2004-02-23 23:49:40 +00:00
ab77b05d8c
Remove '4Virt' from member function names as it is obvious.
...
llvm-svn: 11781
2004-02-23 23:47:10 +00:00
c794a9060f
Refactor VirtRegMap out of RegAllocLinearScan as the first part of bug
...
251 (providing a generic machine code rewriter/spiller).
llvm-svn: 11780
2004-02-23 23:08:11 +00:00
52de9e214c
Add number of spilled registers statistic.
...
llvm-svn: 11759
2004-02-23 18:45:32 +00:00
91a7dc0a6e
Fix bugs in finegrainification
...
llvm-svn: 11758
2004-02-23 18:40:08 +00:00
43df6c268b
Finegrainify namespacification
...
llvm-svn: 11757
2004-02-23 18:38:20 +00:00
b525d54045
Use MachineBasicBlock::getParent().
...
llvm-svn: 11756
2004-02-23 18:36:38 +00:00
d780ce6462
Remove implementation of default constructor as it is useless now.
...
llvm-svn: 11755
2004-02-23 18:28:35 +00:00
af2de4848e
Refactor rewinding code for finding the first terminator of a basic
...
block into MachineBasicBlock::getFirstTerminator().
This also fixes a bug in the implementation of the above in both
RegAllocLocal and InstrSched, where instructions where added after the
terminator if the basic block's only instruction was a terminator (it
shouldn't matter for RegAllocLocal since this case never occurs in
practice).
llvm-svn: 11748
2004-02-23 18:14:48 +00:00
f83cb74b57
Improved PhysRegTracker interface. RegAlloc lazily allocates the register tracker using a std::auto_ptr
...
llvm-svn: 11738
2004-02-23 06:10:13 +00:00
c31ff795d5
Simplify iterator usage now that we have next(). Also don't pass iterators by reference now that MachineInstr* are in an ilist
...
llvm-svn: 11732
2004-02-23 04:12:30 +00:00
887b941fbb
Some code cleanups from Chris
...
llvm-svn: 11724
2004-02-23 01:57:39 +00:00
d42c2f24ec
Fix comments in PhysRegTracker and rename isPhysRegAvail to isRegAvail to be consistent with the other two
...
llvm-svn: 11723
2004-02-23 01:25:05 +00:00
4d7ad6713d
Move LiveIntervals.h up to be the first included header
...
llvm-svn: 11721
2004-02-23 01:01:21 +00:00
1410859484
Pull PhysRegTracker out of RegAllocLinearScan as it can be used by other allocators as well
...
llvm-svn: 11720
2004-02-23 00:53:31 +00:00
486b12b71a
Move LiveIntervals.h to lib/CodeGen since it shouldn't be exposed to other parts of the compiler
...
llvm-svn: 11719
2004-02-23 00:50:15 +00:00
19aef9e757
Another bug fix for empty MBB's
...
llvm-svn: 11716
2004-02-22 19:37:31 +00:00
84b406650e
Fix a bug where we were implicitly assuming that there would be at least
...
one terminator instruction in each basic block.
llvm-svn: 11714
2004-02-22 19:08:15 +00:00
cb352300c9
Print basic block boundaries in machine instruction debug output.
...
llvm-svn: 11704
2004-02-22 05:46:04 +00:00
d890f59f19
Abstract merging of ranges away from number of slots per instruction.
...
Also make it less aggressive as the current implementation breaks in
some cases.
llvm-svn: 11696
2004-02-22 04:05:13 +00:00
334114bdf1
Make 'fold' statistic's description the same in both allocators.
...
llvm-svn: 11687
2004-02-21 18:07:33 +00:00
6ad63f8da2
Some more statistics improvements.
...
llvm-svn: 11676
2004-02-20 20:53:26 +00:00
71751de25c
Disambiguate statistic descriptions.
...
llvm-svn: 11675
2004-02-20 20:46:49 +00:00
f3e77fb7a7
Rename statistic and add another one.
...
llvm-svn: 11674
2004-02-20 20:43:08 +00:00
22de21ea80
Fix crash in debug output.
...
llvm-svn: 11659
2004-02-20 06:41:12 +00:00
bcc99a83ee
Fix instruction numbering in debug output.
...
llvm-svn: 11655
2004-02-20 06:29:51 +00:00
76eca062ea
Too many changes in one commit:
...
1. LiveIntervals now implement a 4 slot per instruction model. Load,
Use, Def and a Store slot. This is required in order to correctly
represent caller saved register clobbering on function calls,
register reuse in the same instruction (def resues last use) and
also spill code added later by the allocator. The previous
representation (2 slots per instruction) was insufficient and as a
result was causing subtle bugs.
2. Fixes in spill code generation. This was the major cause of
failures in the test suite.
3. Linear scan now has core support for folding memory operands. This
is untested and not enabled (the live interval update function does
not attempt to fold loads/stores in instructions).
4. Lots of improvements in the debugging output of both live intervals
and linear scan. Give it a try... it is beautiful :-)
In summary the above fixes all the issues with the recent reserved
register elimination changes and get the allocator very close to the
next big step: folding memory operands.
llvm-svn: 11654
2004-02-20 06:15:40 +00:00
3cf546037e
Fix RA::verifyAssignment()
...
llvm-svn: 11629
2004-02-19 19:24:17 +00:00
f5c5e1f458
Fix problem fusing spill code into instructions: we didn't update the live
...
variable information to take into account the change of instruction
address.
llvm-svn: 11628
2004-02-19 18:34:02 +00:00
4c988391e1
Fix an iterator invalidation problem. :(
...
llvm-svn: 11627
2004-02-19 18:32:29 +00:00
afa9d7eb85
Add method to update livevar when an instruction moves
...
llvm-svn: 11625
2004-02-19 18:28:02 +00:00
63f41abfe6
Fix a __LONG__ term annoyance of mine: symbolic registers weren't being printed
...
by operator<< on MachineInstr's, and looking up what register "24" is all of the
time was greatly annoying.
llvm-svn: 11623
2004-02-19 16:17:08 +00:00
d23a882f2f
Add a MachineBasicBlock::getParent() method
...
llvm-svn: 11622
2004-02-19 16:13:54 +00:00
d0a60b71fc
Rename reloads/spills to loads/stores.
...
llvm-svn: 11619
2004-02-19 06:19:09 +00:00
b490a12501
Implement assignment correctness verification.
...
llvm-svn: 11609
2004-02-18 23:15:23 +00:00
83398e5574
Eliminate operator[] is deprecated warnings
...
llvm-svn: 11578
2004-02-18 16:43:51 +00:00
e99a833c20
Fix deprecated operator[] warnings
...
llvm-svn: 11577
2004-02-18 16:38:18 +00:00
7bef0eafc0
Be more agressive when joining ranges.
...
llvm-svn: 11575
2004-02-18 04:38:37 +00:00
739e5e33d7
Fix overly conservative spill interval computation.
...
llvm-svn: 11574
2004-02-18 03:35:38 +00:00
26583db8b6
Beautify debug output.
...
llvm-svn: 11573
2004-02-18 00:35:06 +00:00
3d894dd3ab
Remove the -disable-kill option. The register allocator is buggy with it,
...
and it was only for debugging in the first place.
llvm-svn: 11557
2004-02-17 17:49:10 +00:00
4e21b2352c
Add support to the local allocator for fusing spill code into the instructions
...
that need them. This is very useful on CISCy targets like the X86 because it
reduces the total spill pressure, and makes better use of it's (large)
instruction set. Though the X86 backend doesn't know how to rewrite many
instructions yet, this already makes a substantial difference on 176.gcc for
example:
Before:
Time:
8.0099 ( 31.2%) 0.0100 ( 12.5%) 8.0199 ( 31.2%) 7.7186 ( 30.0%) Local Register Allocator
Code quality:
734559 asm-printer - Number of machine instrs printed
111395 ra-local - Number of registers reloaded
79902 ra-local - Number of registers spilled
231554 x86-peephole - Number of peephole optimization performed
After:
Time:
7.8700 ( 30.6%) 0.0099 ( 19.9%) 7.8800 ( 30.6%) 7.7892 ( 30.2%) Local Register Allocator
Code quality:
733083 asm-printer - Number of machine instrs printed
2379 ra-local - Number of reloads fused into instructions
109046 ra-local - Number of registers reloaded
79881 ra-local - Number of registers spilled
230658 x86-peephole - Number of peephole optimization performed
So by fusing 2300 instructions, we reduced the static number of instructions
by 1500, and reduces the number of peepholes (and thus the work) by about 900.
This also clearly reduces the number of reload/spill instructions that are
emitted.
llvm-svn: 11542
2004-02-17 08:09:40 +00:00
d1289d0fcd
Fix a bug in my previous refactoring change... arg!
...
llvm-svn: 11535
2004-02-17 07:02:17 +00:00
ba9e3e2fb1
Once we have a way to fold spill code reloads into instructions, we have a way to use it. :)
...
llvm-svn: 11517
2004-02-17 04:08:37 +00:00
4f7c2d9e5e
Fix spilled interval update. It was too conservative.
...
llvm-svn: 11516
2004-02-17 04:04:20 +00:00
ddedac59fe
Refactor code a bit. No functionality changes, though the comment hints at things to come.
...
llvm-svn: 11515
2004-02-17 03:57:19 +00:00
14f3fe81c6
Add LeakDetection to MachineInstr.
...
Move out of line member functions of MachineBasicBlock to
MachineBasicBlock.cpp.
llvm-svn: 11497
2004-02-16 07:17:43 +00:00
e27f33f00e
This pass should not require phi elimination or live variable
...
analysis. It should only preserve them and update LiveVariables if it
already ran.
llvm-svn: 11479
2004-02-15 21:50:32 +00:00
a8d97e4e74
Finegrainify namespacification
...
Remove one of the operands of a two operand instruction
llvm-svn: 11478
2004-02-15 21:38:28 +00:00
bbf53937a4
Make dense maps keyed on physical registers smallerusing
...
MRegisterInfo::getNumRegs() instead of
MRegisterInfo::FirstVirtualRegister.
Also use MRegisterInfo::is{Physical,Virtual}Register where
appropriate.
llvm-svn: 11477
2004-02-15 21:37:17 +00:00
2c5ddd20ba
Eliminate the use of spill (reserved) registers.
...
llvm-svn: 11476
2004-02-15 10:24:21 +00:00
1dfa97034d
Allow for fixed objects to reside in the local area, and if they don't to not
...
clobber them by allocating other objects in the same space!
llvm-svn: 11454
2004-02-15 00:14:20 +00:00
6a35516dcc
Add back machine code deleter pass until we get a MachineCode pass
...
that will be responsible for the creation of MachineFunctions and will
be required by all MachineFunctionPass passes.
llvm-svn: 11453
2004-02-15 00:03:15 +00:00
612ce0c47c
There is no reason to align the stack pointer if there are no callees of this function!
...
llvm-svn: 11449
2004-02-14 20:10:59 +00:00
5a92240270
Use newly added next() and prior() utility functions.
...
llvm-svn: 11430
2004-02-14 01:18:34 +00:00
ddcb4ad53e
Use std::numeric_limits<float>::infinity() instead of
...
std::numeric_limits<float>::max() for weighting preallocated
intervals.
llvm-svn: 11427
2004-02-14 00:44:07 +00:00
8cdd0215bf
Remove getAllocatedRegNum(). Use getReg() instead.
...
llvm-svn: 11393
2004-02-13 21:01:20 +00:00
de6a381691
Use getNumVirtualRegs().
...
Whitespace cleanups.
llvm-svn: 11389
2004-02-13 18:20:47 +00:00
52440fd481
Refactor MachineFunction::print() into MachineBasicBlock::print().
...
Add MachineBasicBlock::dump().
llvm-svn: 11364
2004-02-13 04:39:55 +00:00
e8f7c2f863
Add head-of-file comments and Doxygen comments. Tighten up a lot of whitespace.
...
Rename SetMachineOperandConst's formal parameters to match other methods here.
Mark some methods as being used only by the SPARC back-end.
Fix a missing-paren bug in OutputValue().
llvm-svn: 11363
2004-02-13 04:39:32 +00:00
de8ac749fe
Add parent pointer to MachineInstr that points to owning
...
MachineBasicBlock. Also change opcode to a short and numImplicitRefs
to an unsigned char so that overall MachineInstr's size stays the
same.
llvm-svn: 11357
2004-02-12 18:49:07 +00:00
6a597d6057
Rename the opCode instance variable to Opcode
...
llvm-svn: 11348
2004-02-12 16:09:53 +00:00
6108d9d5ee
This field is never read
...
llvm-svn: 11346
2004-02-12 16:04:49 +00:00
80da865f77
Change MachineBasicBlock's vector of MachineInstr pointers into an
...
ilist of MachineInstr objects. This allows constant time removal and
insertion of MachineInstr instances from anywhere in each
MachineBasicBlock. It also allows for constant time splicing of
MachineInstrs into or out of MachineBasicBlocks.
llvm-svn: 11340
2004-02-12 02:27:10 +00:00
0c3de446f6
getOpCode() --> getOpcode().
...
llvm-svn: 11339
2004-02-12 01:34:05 +00:00
b21ec54e61
Add #include
...
llvm-svn: 11285
2004-02-10 21:18:55 +00:00
5dd5be3ec3
Do not use MachineOperand::isVirtualRegister either!
...
llvm-svn: 11283
2004-02-10 21:12:22 +00:00
373fba5c1d
Eliminate users of MachineOperand::isPhysicalRegister
...
llvm-svn: 11278
2004-02-10 20:41:10 +00:00
2e1cbb735d
Move InstrSchedule's iterator begin/end methods inline.
...
llvm-svn: 11239
2004-02-09 18:42:46 +00:00
312e3a28d2
Make SchedGraph::dump() use SchedGraphNodeCommon's const_iterator
...
instead of randomly groping about inside its outEdges array.
Make SchedGraph::addDummyEdges() use getNumOutEdges() instead of
outEdges.size().
Get rid of ifdefed-out code in SchedGraph::buildGraph().
llvm-svn: 11238
2004-02-09 18:42:05 +00:00
80cbed4f61
Another nice speedup for the register allocator. This time, we replace
...
the Virt2PhysRegMap std::map with an std::vector. This speeds up the
register allocator another (almost) 40%, from .72->.45s in a release build
of LLC on 253.perlbmk.
llvm-svn: 11219
2004-02-09 02:12:04 +00:00
bbd9a43d20
Ugh, perform an optimization that GCC should be able to do itself. This
...
speeds up livevar from .48/.32s -> .45/.31s in LLC on perlbmk
llvm-svn: 11217
2004-02-09 01:43:23 +00:00
26407384ec
Only do stuff for the REAL number of physical registers we have, not 1024.
...
This speeds up live variables a lot, from .60/.39s -> .47/.26s in LLC, for
the first/second pass respectively.
llvm-svn: 11216
2004-02-09 01:35:21 +00:00
490627a404
Change the PhysRegsUsed map into a dense array. Seeing that this is a mapping
...
from physical registers, and they are always dense, it makes sense to not have
a ton of RBtree overhead. This change speeds up regalloclocal about ~30% on
253.perlbmk, from .35s -> .27s in the JIT (in LLC, it goes from .74 -> .55).
Now live variable analysis is the slowest codegen pass. Of course it doesn't
help that we have to run it twice, because regalloclocal doesn't update it,
but even if it did it would be the slowest pass (now it's just the 2x slowest
pass :(
llvm-svn: 11215
2004-02-09 01:26:13 +00:00
e82a70766e
Increase code clarity.
...
llvm-svn: 11151
2004-02-06 18:08:18 +00:00
cd2e7ffcc6
Eliminate uneeded lookups by passing a Virt2PhysMap::iterator instead
...
of the virtual register to certain functions.
llvm-svn: 11143
2004-02-06 03:15:40 +00:00
a97abeb342
Change live interval representation. Machine instructions now have two
...
slots each. As a concequence they get numbered as 0, 2, 4 and so
on. The first slot is used for operand uses and the second for
defs. Here's an example:
0: A = ...
2: B = ...
4: C = A + B ;; last use of A
The live intervals should look like:
A = [1, 5)
B = [3, x)
C = [5, y)
llvm-svn: 11141
2004-02-05 22:55:25 +00:00
1165c1081e
We don't need to scan the blocks that we are live-in on every
...
access. Rather we only have to do it on the creation of the interval.
llvm-svn: 11135
2004-02-05 20:45:40 +00:00
dd0c255d14
Added missing include.
...
llvm-svn: 11129
2004-02-05 05:04:39 +00:00
327426411e
Modify the two address instruction pass to remove the duplicate
...
operand of the instruction and thus simplify the register allocation.
llvm-svn: 11124
2004-02-04 22:17:40 +00:00
3d100ef0ca
When an instruction like: A += B had both A and B virtual registers
...
spilled, A was loaded from its stack location twice. This fixes the bug.
llvm-svn: 11093
2004-02-03 01:13:07 +00:00
74b26f77a3
Revert changes. Will implement this using a different set of primitives
...
llvm-svn: 11091
2004-02-02 23:08:58 +00:00
6e34205653
Fix debugging output.
...
llvm-svn: 11088
2004-02-02 22:00:32 +00:00
6471e66fbb
Correctly update def/use information for modified machine operands.
...
llvm-svn: 11087
2004-02-02 21:56:40 +00:00
86ff30deab
Should be more careful. The previously applied change made all counts
...
outside of loops = 0.
llvm-svn: 11085
2004-02-02 20:29:57 +00:00
2cecace9b2
Create an object for tracking physical register usage. This will look
...
much better when I get rid of the reserved registers.
llvm-svn: 11066
2004-02-02 07:30:36 +00:00
9b9c0485fd
Change weight array into a vector and make it as big as the number of
...
registers (not as the max number of registers).
Change toSpill from a std::set into a std::vector<bool>.
Use the reverse iterator adapter to do a reverse scan of allocatable
registers.
llvm-svn: 11061
2004-02-01 20:13:26 +00:00
5e05a59bf8
Use std::map::count() instead of std::map::find() != std::map::end()
...
where appropriate.
llvm-svn: 11060
2004-02-01 18:39:53 +00:00
a3a56fa041
Change string for joined intervals.
...
llvm-svn: 11057
2004-02-01 16:13:05 +00:00
84a84c38b7
Change xor to ^.
...
llvm-svn: 11054
2004-02-01 08:19:25 +00:00
cae054f59f
Fix a bug in a recent checkin
...
llvm-svn: 11050
2004-02-01 05:25:07 +00:00
8d4f06ed84
Missed one silly assert :-)
...
llvm-svn: 11048
2004-02-01 02:21:31 +00:00
69deb4d236
Simplify joinIntervals() code.
...
llvm-svn: 11047
2004-02-01 02:18:31 +00:00
91dcc03c39
Use MRegisterInfo::isPhysicalRegister and
...
MRegisterInfo::isVirtualRegister.
llvm-svn: 11045
2004-02-01 01:27:01 +00:00
662bb3f495
Apply final part of Chris' patch.
...
llvm-svn: 11040
2004-01-31 23:48:47 +00:00
459a67c309
Be a little smarter on the way we handle physical register defs.
...
llvm-svn: 11038
2004-01-31 23:13:30 +00:00
c330b98c1c
Finegrainify namespacification, use new MRegisterInfo::isVirtualRegister
...
method
llvm-svn: 11037
2004-01-31 21:27:19 +00:00
a2ef296994
Fix, correctly this time, the computation of the return value
...
Fix a spello
Tighten up the assertion checking
No functionality changes.
llvm-svn: 11036
2004-01-31 21:21:43 +00:00
626f431165
* Fix incorrect computation of the runOnMachineFunction return value
...
* Turn a bunch of instance variables into automatic variables
llvm-svn: 11035
2004-01-31 21:14:04 +00:00
d835aa6711
Remove unneeded #includes
...
Move Passes.h (which defines the interface to this file) to the top.
Move statistics to the top of the file.
Add a comment
llvm-svn: 11034
2004-01-31 21:07:15 +00:00
50d97e33b7
Merge safe parts from last night's buggy commit. These do not break
...
any test cases :-)
llvm-svn: 11032
2004-01-31 19:59:32 +00:00
cdf48ab294
Optimize liveAt() and overlaps(). We now use a binary search instead
...
of a linear search to find the first range for comparisons. This cuts
down the linear scan register allocator running time by a factor of 3
in 254.perlbmk and by a factor of 2.2 in 176.gcc.
llvm-svn: 11030
2004-01-31 16:54:54 +00:00
26665e2636
Revert last night's changes as they broke some tests. Will remerge parts of the patch.
...
llvm-svn: 11029
2004-01-31 14:37:41 +00:00
f2fb0fb486
Several performance enhancements and cleanups from Chris.
...
Simplification of LiveIntervals::Interval::overlaps() and addition of
examples to overlaps() and liveAt() to make them clearer.
llvm-svn: 11028
2004-01-31 04:56:07 +00:00
0770862334
Finegrainify namespacification
...
Implement LiveVariables::getIndexMachineBasicBlock
llvm-svn: 11018
2004-01-30 22:08:53 +00:00
845c0dd196
Give clients of MachineFunctionPrinter the ability to specify a banner and
...
choose an ostream.
llvm-svn: 11016
2004-01-30 21:53:46 +00:00
59e3325c26
Fix failing test cases with joined live intervals. It turns out that
...
when joining we need to check if we overlap with the second interval
or any of its aliases.
Also make joining intervals the default.
llvm-svn: 10973
2004-01-23 13:37:51 +00:00
73df113676
Add option to join live intervals. Two intervals are joined if there
...
is a move between two registers, at least one of the registers is
virtual and the two live intervals do not overlap.
This results in about 40% reduction in intervals, 30% decrease in the
register allocators running time and a 20% increase in peephole
optimizations (mainly move eliminations).
The option can be enabled by passing -join-liveintervals where
appropriate.
llvm-svn: 10965
2004-01-22 23:08:45 +00:00
e1c24c06f9
Remove unneeded check. An interval in active, by definition overlaps
...
with the current one.
llvm-svn: 10959
2004-01-22 20:07:18 +00:00
4716d76d22
Improve debugging output. Remove unneeded virtReg->0 mapping when
...
virtReg lives on the stack. Now a virtual register has an entry in the
virtual->physical map or the virtual->stack slot map but never in
both.
llvm-svn: 10958
2004-01-22 19:24:43 +00:00
19638ab2bd
Revert previous change. The code was correct...
...
llvm-svn: 10957
2004-01-22 19:17:52 +00:00
0aa9d0278d
Fix incorrect negatives in LiveIntervals::Interval::liveAt().
...
llvm-svn: 10956
2004-01-22 18:33:50 +00:00
7117b9751d
Moved iterators around.
...
llvm-svn: 10926
2004-01-20 17:51:13 +00:00
72494e04be
Moved iterators to common file.
...
llvm-svn: 10925
2004-01-20 17:49:42 +00:00
d844972574
Handle printing of intervals that are not assign to any physical
...
register yet (2nd try).
llvm-svn: 10896
2004-01-16 20:33:13 +00:00
8672af1cc8
Handle printing of intervals that are not assign to any physical
...
register yet.
llvm-svn: 10895
2004-01-16 20:29:42 +00:00
d86933d21d
Fold open interval ends handling into
...
LiveIntervals::Interval::expiredAt() and simplify regalloc code.
llvm-svn: 10894
2004-01-16 20:17:05 +00:00
e9c6ba891e
Add asserts to previous change.
...
llvm-svn: 10893
2004-01-16 16:23:23 +00:00
3010b3e38e
Use a list instead of a vector to store intervals. This will be needed
...
when we join intervals and one of the two will need to be removed.
llvm-svn: 10892
2004-01-16 16:06:59 +00:00
0680572553
Properly update #intervals statistic.
...
llvm-svn: 10847
2004-01-14 10:44:29 +00:00
2615c98042
Fix bug in LiveIntervals::Interval::overlaps and
...
LiveIntervals::Interval::liveAt. Both were considering the live ranges
closed in the end, when they are actually open.
llvm-svn: 10835
2004-01-14 00:20:09 +00:00
a0865cec3f
Improve debugging output.
...
llvm-svn: 10834
2004-01-14 00:09:36 +00:00
0431621429
Fix miscomputation of live intervals. The catch is that registers can
...
be dead at the defining instruction but can only be killed in
subsequent ones.
llvm-svn: 10833
2004-01-13 22:26:14 +00:00
1a01c80f1b
Remove allocatable registers vector. It is already provided by
...
LiveVariables.
llvm-svn: 10830
2004-01-13 22:10:43 +00:00
13e8623a3f
Cleanup debugging output.
...
llvm-svn: 10824
2004-01-13 21:53:20 +00:00
33f9cb1da7
Fix output of live intervals to show correctly its closed, open
...
ranges, i.e. [a,b)
llvm-svn: 10822
2004-01-13 21:17:47 +00:00
9d0c3d21c8
Remove unneeded check (with the recent change in live variables a use
...
of a physical register is always dominated by a def).
llvm-svn: 10821
2004-01-13 21:16:25 +00:00
65bc990bda
Indentation and whitespace cleanups.
...
llvm-svn: 10820
2004-01-13 20:42:08 +00:00
6b1ec75da9
Fix bug introduced by previous commit: check if fixed intervals
...
overlap before adding their spill weight.
llvm-svn: 10819
2004-01-13 20:37:01 +00:00
ebbd66c042
Correctly compute live variable information for physical registers
...
when an implicitely defined register is later used by an alias. For example:
call foo
%reg1024 = mov %AL
The call implicitely defines EAX but only AL is used. Before this fix
no information was available on AL. Now EAX and all its aliases except
AL get defined and die at the call instruction whereas AL lives to be
killed by the assignment.
llvm-svn: 10813
2004-01-13 06:24:30 +00:00
52a714b45b
Make LiveVariables::HandlePhysRegUse and
...
LiveVariables::HandlePhysRegDef private they use information that is
not in memory when LiveVariables finishes the analysis.
Also update the TwoAddressInstructionPass to not use this interface.
llvm-svn: 10755
2004-01-11 09:18:45 +00:00
8043f8c294
Remove use of llvm/CodeGen/InstrSelection.h
...
llvm-svn: 10749
2004-01-10 19:16:26 +00:00
a4ee66f3ea
Finegrainify namespacification.
...
This should get hunked over to the Sparc backend, along with
MachineCodeForInstruction and a bunch of files in include/llvm/Codegen,
but those battles will have to wait for a later time.
llvm-svn: 10731
2004-01-09 06:30:18 +00:00
83e84166a3
Move InstrSelection into lib/Target/Sparc, as it's sparc specific
...
llvm-svn: 10730
2004-01-09 06:24:06 +00:00
b9c7999881
Move lib/Codegen/RegAlloc into lib/Target/Sparc, as it is sparc specific
...
llvm-svn: 10728
2004-01-09 06:17:12 +00:00
ae5b3d4f74
Add a separate list of fixed intervals. This improves the running time
...
of the register allocator as follows:
before after
mesa 2.3790 1.5994
vpr 2.6008 1.2078
gcc 1.9840 0.5273
mcf 0.2569 0.0470
eon 1.8468 1.4359
twolf 0.9475 0.2004
burg 1.6807 1.3300
lambda 1.2191 0.3764
Speedups range anyware from 30% to over 400% :-)
llvm-svn: 10712
2004-01-07 09:20:58 +00:00
a64d71203e
Minor cleanups.
...
llvm-svn: 10711
2004-01-07 05:31:12 +00:00
9f62b96206
Remove declared but undefined method.
...
llvm-svn: 10710
2004-01-07 02:29:33 +00:00
1e01557c4e
Change implementation of LiveIntervals::overlap(). This results in a
...
30-50% decrease in running time of the linear scan register allocator.
llvm-svn: 10707
2004-01-07 01:45:58 +00:00
2fa1aef707
Remove simple coalescing.
...
llvm-svn: 10695
2004-01-05 08:24:57 +00:00
fc3d0598df
fix warning
...
llvm-svn: 10692
2004-01-05 05:42:17 +00:00
08c5311729
Currently we cannot handle two-address instructions of the form:
...
A = B op C where A == C, but this cannot really occur in practice
because of SSA form. Add an assert to check that just to be safe.
llvm-svn: 10682
2004-01-05 02:25:45 +00:00
5e0e67173d
Update description.
...
llvm-svn: 10681
2004-01-04 23:09:24 +00:00
5d236005b0
Clean up a lot of the code I added yesterday by exposing the IntrinsicLowering
...
implementation from the TargetMachine directly.
llvm-svn: 10636
2003-12-28 21:23:38 +00:00
a158dc8104
Reserve ECX and EDI instead of EBX and EDI. Since EBX is a callee
...
saved register it has a longer free range than ECX (which is defined
every time there is a fnuction call) which makes ECX a better register
to reserve.
llvm-svn: 10635
2003-12-28 18:03:52 +00:00
43b587d94d
Add coalescing to register allocator. A hint is added to each interval
...
which denotes the register we would like to be assigned to (virtual or
physical). In register allocation, if this hint exists and we can map
it to a physical register (it is either a physical register or it is a
virtual register that already got assigned to a physical one) we use
that register if it is available instead of a random one in the free
pool.
llvm-svn: 10634
2003-12-28 17:58:18 +00:00
10cac58d23
Whoops, don't try to lower non intrinsic calls
...
llvm-svn: 10632
2003-12-28 09:53:23 +00:00
4b75e04a60
Use the intrinsic lowering functionality
...
llvm-svn: 10626
2003-12-28 09:43:35 +00:00
24650d2b5e
Move into the VMCore library
...
llvm-svn: 10623
2003-12-28 08:30:20 +00:00
08c196dd4b
Implement the default implementation of the intrinsic lowering class
...
llvm-svn: 10621
2003-12-28 08:19:41 +00:00
27dd64242b
finegrainify namespacification
...
minor cleanups
llvm-svn: 10619
2003-12-28 07:59:53 +00:00
2e8c984c58
Improve debugging output when choosing a register to spill.
...
llvm-svn: 10604
2003-12-24 18:53:31 +00:00
eeea81483f
Do a separate pass to compute spill weights because doing it inline
...
with live intervals was missing registers that were used before they
were defined (in the arbitrary order live intervals numbers
instructions).
llvm-svn: 10603
2003-12-24 15:44:53 +00:00
7d7d7e89d4
Change the way free regusters are computed and perform better
...
allocation in the presence of preallocated intervals.
llvm-svn: 10595
2003-12-23 18:00:33 +00:00
3ec903474f
Fix crash when compiling twolf.
...
llvm-svn: 10584
2003-12-22 13:54:29 +00:00
690e1b8de2
Remove verifyIntervals() since it doesn't actually work right now.
...
llvm-svn: 10570
2003-12-21 20:41:26 +00:00
2665d9c6f9
Change weight into a float so that we can take into account the
...
nesting level when computing it. Right now the allocator uses:
w = sum_over_defs_uses( 10 ^ nesting level );
llvm-svn: 10569
2003-12-21 20:19:10 +00:00
c09b77ea35
Add support for inactive intervals. This effectively reuses registers
...
for live ranges that fall into assigned registers' holes.
llvm-svn: 10566
2003-12-21 05:43:40 +00:00
4cbb97bfdd
Add a new target-independent machine code freeing pass
...
llvm-svn: 10560
2003-12-20 10:20:58 +00:00
9a3478e022
* Finegrainify namespacification
...
* Move sparc specific code out of generic code
* Eliminate the getOffset() method which made INVALID_FRAME_OFFSET
necessary, which made pulling in MAX_INT as a sentinal necessary.
llvm-svn: 10553
2003-12-20 09:17:07 +00:00
7139090fd2
Remove TwoAddressInstruction from the public headers and add an ID
...
instead, since this pass doesn't expose any state to its users.
llvm-svn: 10520
2003-12-18 22:40:24 +00:00
8f88cdfa22
Prune some #includes
...
Add a statistic for # reloads
llvm-svn: 10518
2003-12-18 20:25:31 +00:00
e5e2bd6777
Modify linear scan register allocator to use the two-address
...
instruction pass. This also fixes all remaining bugs for this new
allocator to pass all tests under test/Programs.
llvm-svn: 10515
2003-12-18 13:15:02 +00:00
3bd69eaac5
Fix bug in reserved registers. DH actually aliases DX and EDX which
...
are not reserved registers.
llvm-svn: 10514
2003-12-18 13:12:18 +00:00
c17d57bd82
Modify local register allocator to use the two-address instruction pass.
...
llvm-svn: 10513
2003-12-18 13:08:52 +00:00
725021cb6c
Add TwoAddressInstructionPass to handle instructions that have two or
...
more operands and the two first operands are constrained to be the
same. The pass takes an instruction of the form:
a = b op c
and transforms it into:
a = b
a = a op c
and also preserves live variables.
llvm-svn: 10512
2003-12-18 13:06:04 +00:00
4d87219486
Rename LiveIntervals::expired() to LiveIntervals::expiredAt().
...
llvm-svn: 10511
2003-12-18 08:56:11 +00:00
15b8036b29
When a variable is killed and redifined in a basic block only one
...
killing instruction is tracked. This causes the LiveIntervals to
create bogus intervals. The workaound is to add a range to the
interval from the redefinition to the end of the basic block.
llvm-svn: 10510
2003-12-18 08:53:43 +00:00
40874c708f
Handle multiple virtual register definitions gracefully.
...
Move some of the longer LiveIntervals::Interval method out of the
header and add debug information to them. Fix bug and simplify range
merging code.
llvm-svn: 10509
2003-12-18 08:48:48 +00:00
8870674da7
Change preserve all claim to just preserve live variables and phielimination.
...
llvm-svn: 10469
2003-12-15 04:55:38 +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
4f7f017ab5
When reserving a preallocated register spill the aliases of this
...
register too.
llvm-svn: 10450
2003-12-13 11:58:10 +00:00
7dbdf3839c
Ignore non-allocatable physical registers in live interval analysis.
...
llvm-svn: 10449
2003-12-13 11:11:02 +00:00
ed53b35076
Expire any active intervals left when register allocation is done.
...
llvm-svn: 10448
2003-12-13 05:50:19 +00:00
e6dc614c3c
Add instruction numbers to debugging output.
...
llvm-svn: 10447
2003-12-13 05:48:57 +00:00
06f379a278
Handle explicit physical register defs.
...
llvm-svn: 10445
2003-12-13 05:26:39 +00:00
9bced9455a
Remove unecessary if statements when looping on ImplicitDefs.
...
llvm-svn: 10444
2003-12-13 01:20:58 +00:00
fb55ac24ec
This appears to fix Bug 172 and does not break any other feature tests or
...
regression tests.
llvm-svn: 10388
2003-12-10 22:51:41 +00:00
e59ad4b441
Make assertion stricter. Since the source operands are allocated at
...
this point, the second operand must be a physical register (it cannot
be a virtual one).
llvm-svn: 10292
2003-12-05 11:31:39 +00:00
f8fe176204
Fix bug in register spilling when a preallocated live range overlaps a
...
potential register assignment.
llvm-svn: 10291
2003-12-05 11:17:55 +00:00
2de099e54f
Move operator<<(std::ostream&, const LiveInterval&) out of the header file.
...
llvm-svn: 10290
2003-12-05 10:38:28 +00:00
ed28ca93ef
Sort live intervals by increasing start point.
...
llvm-svn: 10289
2003-12-05 10:32:01 +00:00
5fe0031c68
Improve debugging output and clean up some code.
...
llvm-svn: 10288
2003-12-04 03:57:28 +00:00
26f17e73eb
Print instructions before register allocation is performed. Also fix
...
bug where spill instructions were added to the next basic block
instead of the end of the current one if the instruction that required
the spill was the last in the block.
llvm-svn: 10272
2003-11-30 23:40:39 +00:00
0aa40f8db0
Remove "numReloaded" statistic.
...
llvm-svn: 10268
2003-11-30 05:15:36 +00:00
0e9ded74c0
Merging the linear scan register allocator in trunk. It currently passes most tests under test/Programs/SingleSource/Benchmarks/Shootout so development will continue on trunk. The allocator is not enabled by default. You will need to pass -regallo=linearscan to lli or llc to use it.
...
llvm-svn: 10103
2003-11-20 03:32:25 +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
8f53a89edf
Fix problems linking against the reoptimizer; _llvm_regAllocState must have
...
externally-visible linkage, and SaveStateToModule must default to true for llc.
I don't remember why I made it const; perhaps it should be deconstified.
llvm-svn: 9858
2003-11-10 07:12:01 +00:00
c0c23c29aa
Operand numbers are now ints. Save the register allocation of the value
...
each instruction produces as "operand" -1, and the other operands as 0
.. n, as before. PhyRegAlloc::saveState() is refactored into
PhyRegAlloc::saveStateForValue().
llvm-svn: 9842
2003-11-10 00:05:26 +00:00
2b8724acc6
Let's not forget about our friends -- Constant Pool indices.
...
llvm-svn: 9750
2003-11-06 00:04:11 +00:00
a9e7fe9ea2
Do not use a class before it is defined.
...
Be gcc 3.4 clean
llvm-svn: 9727
2003-11-05 06:25:06 +00:00
69c2841b2a
Update verifySavedState()'s comment, so that it reflects its current
...
status. In doFinalization(), skip over external functions, just like
Anand's mapping info does.
llvm-svn: 9703
2003-11-04 22:42:41 +00:00
2a262ef095
Add comments.
...
llvm-svn: 9697
2003-11-04 18:25:56 +00:00
8541bd9db9
Include llvm/CodeGen/MachineCodeForInstruction.h. Use it to start
...
implementing verifySavedState().
In saveState(), use the new AllocInfo::AllocStateTy enum, and increment
Insn each time through the loop.
llvm-svn: 9617
2003-10-30 21:21:33 +00:00
727ec9bec5
Make AllocState an enum.
...
Move the stringifying method for that enum into class AllocInfo.
llvm-svn: 9616
2003-10-30 21:21:22 +00:00
eca381c7e8
Publicize the type of FnAllocState.
...
Prototype option to save state in a global instead of as a Constant in
the Module. (Turned off, for now, with the on/off switch welded in the off
position. You get the idea.)
llvm-svn: 9500
2003-10-24 21:21:58 +00:00
9ab7fbef60
standardize command line option names
...
llvm-svn: 9496
2003-10-24 20:05:58 +00:00
24fb8f07f3
Move the implementations of ==, != on AllocInfos here, from UnpackTraceFunction.
...
llvm-svn: 9452
2003-10-23 20:39:18 +00:00
390d31cde4
Move AllocInfo structure to a private AllocInfo.h header file.
...
Make FnAllocState contain vectors of AllocInfo, instead of LLVM Constants.
Give doFinalization a method comment, and let it do the work of converting
AllocInfos to LLVM Constants.
llvm-svn: 9451
2003-10-23 20:32:55 +00:00
a6bfe8eb9e
Make FnAllocState contain vectors of AllocInfo, instead of LLVM Constants.
...
llvm-svn: 9450
2003-10-23 20:32:02 +00:00
8e766c7ab3
New file, containing AllocInfo structure.
...
llvm-svn: 9449
2003-10-23 20:31:51 +00:00
88df876012
* Eliminate `using' directive
...
* Fix order of #includes
* Make code layout more consistent
* Eliminate extraneous whitespace and comment-lines
llvm-svn: 9433
2003-10-23 18:10:02 +00:00
dc07775d58
* Fix order of #include files
...
* Doxygen-ify method comments
llvm-svn: 9432
2003-10-23 18:06:27 +00:00
0b624fefb7
* Order #includes as per style guide
...
* Doxygen-ify comments
* Make code layout more consistent
llvm-svn: 9431
2003-10-23 18:03:50 +00:00
28ba80e648
* Use C++ style comments instead of C-style
...
* Make file description more readable
* Make code layout more consistent, include comment in assert so it's visible
during execution if it hits
llvm-svn: 9430
2003-10-23 18:02:47 +00:00
7d56d2c6fb
* Eliminate `using' directive
...
* Make code layout more consistent
llvm-svn: 9427
2003-10-23 17:43:17 +00:00
c7b1bce283
Make code layout more consistent.
...
llvm-svn: 9426
2003-10-23 17:39:37 +00:00
f140b28c16
Change the type of FnAllocState.
...
llvm-svn: 9388
2003-10-22 20:44:29 +00:00
41fe18cbaa
Don't worry about converting each function's reg. alloc. state into One Big
...
Constant early on, because we can do it in doFinalization.
Tighten up a comment.
llvm-svn: 9387
2003-10-22 20:44:23 +00:00
f29231ec3a
Add prototype for verifySavedState().
...
llvm-svn: 9386
2003-10-22 20:23:13 +00:00
e3cf072d9e
Doxygenify method comments.
...
Try to improve method comments a little.
Get rid of some excess whitespace; put braces on previous line when possible.
Add stub for method to verify the work of saveState().
llvm-svn: 9385
2003-10-22 20:22:53 +00:00
9583cfafc5
Added LLVM copyright notice.
...
llvm-svn: 9324
2003-10-21 15:29:18 +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
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
9177d11367
#include vector which we will need here soon
...
llvm-svn: 9144
2003-10-15 22:09:32 +00:00
f95d9b99b3
Decrease usage of use_size()
...
llvm-svn: 9135
2003-10-15 16:48:29 +00:00
44d2c3514a
Regularize header file comments
...
llvm-svn: 9071
2003-10-13 03:32:08 +00:00
80f283b316
* Doxygenified comments
...
* Wrap code at 80 columns
* Ordered includes according to LLVM style guide
llvm-svn: 9020
2003-10-10 17:41:32 +00:00
5f1f337d95
Change MRegisterDesc::AliasSet, TargetInstrDescriptor::ImplicitDefs
...
and TargetInstrDescriptor::ImplicitUses to always point to a null
terminated array and never be null. So there is no need to check for
pointer validity when iterating over those sets. Code that looked
like:
if (const unsigned* AS = TID.ImplicitDefs) {
for (int i = 0; AS[i]; ++i) {
// use AS[i]
}
}
was changed to:
for (const unsigned* AS = TID.ImplicitDefs; *AS; ++AS) {
// use *AS
}
llvm-svn: 8960
2003-10-08 05:20:08 +00:00
5facafaabe
Moved enum and command-line option in separate file. Also added function that returns the user selected register allocator to the caller.
...
llvm-svn: 8819
2003-10-02 16:57:49 +00:00
1f8d21e233
include passes.h which defines the interface this file exposes
...
llvm-svn: 8793
2003-09-30 20:13:59 +00:00
974bc6682e
Update head-of-file comment.
...
llvm-svn: 8699
2003-09-24 18:16:23 +00:00
1542a8b820
Untabify tabs in stuff I've recently added.
...
Check in my register allocator state-saving code.
llvm-svn: 8698
2003-09-24 18:08:54 +00:00
82585e030c
Use getRegClassID() instead of getRegClass()->getID(), since it's there.
...
Shorten the markSuggestedColorUsable method.
Add a switch for saving reg. alloc. state (coming soon).
llvm-svn: 8697
2003-09-24 17:50:28 +00:00
50364030ca
Remove some unused methods of class IGNode.
...
llvm-svn: 8696
2003-09-24 04:29:52 +00:00
6b379da6c3
Move getAnalysisUsage method from header to .cpp file. Add a normal file
...
header comment
llvm-svn: 8679
2003-09-23 15:13:04 +00:00
e383a14960
Use C++ math header instead of C version.
...
llvm-svn: 8648
2003-09-21 03:57:37 +00:00
32cd554853
Erase now-unused prototypes.
...
llvm-svn: 8647
2003-09-21 02:51:00 +00:00
20c888fa9f
Rearrange #includes ... since there are fewer now I guess it's a win.
...
(I also zapped printMachineCode() and printLabel() at the previous checkin,
but forgot to mention it.)
llvm-svn: 8646
2003-09-21 02:50:21 +00:00
3a0a5fc1b3
Standardize the names of include guards.
...
llvm-svn: 8645
2003-09-21 02:31:37 +00:00
58dabb4406
Standardize the names of include guards.
...
Remove more excess whitespace.
llvm-svn: 8644
2003-09-21 02:31:25 +00:00
73d10dc47b
Standardize the names of include guards.
...
Fix typos in file header comment.
llvm-svn: 8643
2003-09-21 02:31:15 +00:00
43593b8cd0
I tried to standardize the formatting and tidy up the huge amount of
...
excess whitespace a little. Also improved some comments.
llvm-svn: 8642
2003-09-21 02:24:09 +00:00
Chris Lattner