Commit Graph

55 Commits

Author SHA1 Message Date
Evan Cheng
7da8f399bf Bug fix. Passive nodes are not in SUnitMap.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@43922 91177308-0d34-0410-b5e6-96231b3b80d8
2007-11-09 01:27:11 +00:00
Evan Cheng
1fd15ba961 Add pseudo dependency to force two-address instruction to be scheduled after
other uses. There was a overly restricted check that prevented some obvious
cases.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@43762 91177308-0d34-0410-b5e6-96231b3b80d8
2007-11-06 08:44:59 +00:00
Chris Lattner
01d029b82c One mundane change: Change ReplaceAllUsesOfValueWith to *optionally*
take a deleted nodes vector, instead of requiring it.

One more significant change:  Implement the start of a legalizer that
just works on types.  This legalizer is designed to run before the 
operation legalizer and ensure just that the input dag is transformed
into an output dag whose operand and result types are all legal, even
if the operations on those types are not.

This design/impl has the following advantages:

1. When finished, this will *significantly* reduce the amount of code in
   LegalizeDAG.cpp.  It will remove all the code related to promotion and
   expansion as well as splitting and scalarizing vectors.
2. The new code is very simple, idiomatic, and modular: unlike 
   LegalizeDAG.cpp, it has no 3000 line long functions. :)
3. The implementation is completely iterative instead of recursive, good
   for hacking on large dags without blowing out your stack.
4. The implementation updates nodes in place when possible instead of 
   deallocating and reallocating the entire graph that points to some 
   mutated node.
5. The code nicely separates out handling of operations with invalid 
   results from operations with invalid operands, making some cases
   simpler and easier to understand.
6. The new -debug-only=legalize-types option is very very handy :), 
   allowing you to easily understand what legalize types is doing.

This is not yet done.  Until the ifdef added to SelectionDAGISel.cpp is
enabled, this does nothing.  However, this code is sufficient to legalize
all of the code in 186.crafty, olden and freebench on an x86 machine.  The
biggest issues are:

1. Vectors aren't implemented at all yet
2. SoftFP is a mess, I need to talk to Evan about it.
3. No lowering to libcalls is implemented yet.
4. Various operations are missing etc.
5. There are FIXME's for stuff I hax0r'd out, like softfp.

Hey, at least it is a step in the right direction :).  If you'd like to help,
just enable the #ifdef in SelectionDAGISel.cpp and compile code with it.  If
this explodes it will tell you what needs to be implemented.  Help is 
certainly appreciated.

Once this goes in, we can do three things:

1. Add a new pass of dag combine between the "type legalizer" and "operation
   legalizer" passes.  This will let us catch some long-standing isel issues
   that we miss because operation legalization often obfuscates the dag with
   target-specific nodes.
2. We can rip out all of the type legalization code from LegalizeDAG.cpp,
   making it much smaller and simpler.  When that happens we can then 
   reimplement the core functionality left in it in a much more efficient and
   non-recursive way.
3. Once the whole legalizer is non-recursive, we can implement whole-function
   selectiondags maybe...



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42981 91177308-0d34-0410-b5e6-96231b3b80d8
2007-10-15 06:10:22 +00:00
Evan Cheng
32dfbeada7 EXTRACT_SUBREG coalescing support. The coalescer now treats EXTRACT_SUBREG like
(almost) a register copy. However, it always coalesced to the register of the
RHS (the super-register). All uses of the result of a EXTRACT_SUBREG are sub-
register uses which adds subtle complications to load folding, spiller rewrite,
etc.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42899 91177308-0d34-0410-b5e6-96231b3b80d8
2007-10-12 08:50:34 +00:00
Dan Gohman
2eb4ebd953 Fix a typo in a comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42635 91177308-0d34-0410-b5e6-96231b3b80d8
2007-10-05 14:11:58 +00:00
Evan Cheng
d5cb5a462b Chain producing nodes cannot be moved, not chain reading nodes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42627 91177308-0d34-0410-b5e6-96231b3b80d8
2007-10-05 01:42:35 +00:00
Evan Cheng
117c366219 Oops. Didn't mean to leave this in.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42626 91177308-0d34-0410-b5e6-96231b3b80d8
2007-10-05 01:39:40 +00:00
Evan Cheng
f10c973797 If a node that defines a physical register that is expensive to copy. The
scheduler will try a number of tricks in order to avoid generating the
copies. This may not be possible in case the node produces a chain value
that prevent movement. Try unfolding the load from the node before to allow
it to be moved / cloned.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42625 91177308-0d34-0410-b5e6-96231b3b80d8
2007-10-05 01:39:18 +00:00
Evan Cheng
22a529990b If two instructions are both two-address code, favors (schedule closer to
terminator) the one that has a CopyToReg use. This fixes
2006-05-11-InstrSched.ll with -new-cc-modeling-scheme.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42453 91177308-0d34-0410-b5e6-96231b3b80d8
2007-09-28 22:32:30 +00:00
Evan Cheng
ddde335cb7 Remove a poor scheduling heuristic.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42443 91177308-0d34-0410-b5e6-96231b3b80d8
2007-09-28 19:37:35 +00:00
Evan Cheng
74d2fd8dd8 Trim some unneeded fields.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42442 91177308-0d34-0410-b5e6-96231b3b80d8
2007-09-28 19:24:24 +00:00
Evan Cheng
cd1c00cc65 Avoid inserting a live register more than once.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42410 91177308-0d34-0410-b5e6-96231b3b80d8
2007-09-27 18:46:06 +00:00
Evan Cheng
2dc7a0e075 Boogs.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42388 91177308-0d34-0410-b5e6-96231b3b80d8
2007-09-27 07:29:27 +00:00
Evan Cheng
a2ee2756f7 Be smarter about which node to force schedule. Reduce # of duplications + copies; Added statistics.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42387 91177308-0d34-0410-b5e6-96231b3b80d8
2007-09-27 07:09:03 +00:00
Evan Cheng
6e4c46cea5 Backtracking only when it won't create a cycle.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42384 91177308-0d34-0410-b5e6-96231b3b80d8
2007-09-27 00:25:29 +00:00
Evan Cheng
42d60274ea - Move getPhysicalRegisterRegClass() from ScheduleDAG to MRegisterInfo.
- Added ability to emit cross class register copies to the BBRU scheduler.
- More aggressive backtracking.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42375 91177308-0d34-0410-b5e6-96231b3b80d8
2007-09-26 21:36:17 +00:00
Evan Cheng
a6fb1b6743 Added major new capabilities to scheduler (only BURR for now) to support physical register dependency. The BURR scheduler can now backtrace and duplicate instructions in order to avoid "expensive / impossible to copy" values (e.g. status flag EFLAGS for x86) from being clobbered.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42284 91177308-0d34-0410-b5e6-96231b3b80d8
2007-09-25 01:54:36 +00:00
Evan Cheng
713a98dee8 Use struct SDep instead of std::pair for SUnit pred and succ lists. First step
in tracking physical register output dependencies.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42125 91177308-0d34-0410-b5e6-96231b3b80d8
2007-09-19 01:38:40 +00:00
Evan Cheng
6600377546 Bug fixes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@41900 91177308-0d34-0410-b5e6-96231b3b80d8
2007-09-13 00:06:00 +00:00
Dan Gohman
8d1bfad00b Minor cleanups to reduce some spurious differences between different
scheduler implementations.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@41191 91177308-0d34-0410-b5e6-96231b3b80d8
2007-08-20 19:28:38 +00:00
Dale Johannesen
e7e7d0d7e3 Skeleton of post-RA scheduler; doesn't do anything yet.
Change name of -sched option and DEBUG_TYPE to
pre-RA-sched; adjust testcases.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@39816 91177308-0d34-0410-b5e6-96231b3b80d8
2007-07-13 17:13:54 +00:00
David Greene
821262953a Remove unused variables.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37816 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-29 21:42:03 +00:00
David Greene
a2a488594d Remove unnecessary attributions in comments.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37799 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-29 03:42:23 +00:00
David Greene
a4ab2e8c72 Remove the "special tie breaker" because it resulted in inconsistent
ordering and thus violated the strict weak ordering requirement of
priority_queue.  Uncovered by _GLIBCXX_DEBUG.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37794 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-29 02:48:09 +00:00
Evan Cheng
0b2ce1fc19 std::set is really really terrible. Switch to SmallPtrSet to reduce compile time. For Duraid's example. The overall isel time is reduced from 0.6255 sec to 0.1876 sec.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37701 91177308-0d34-0410-b5e6-96231b3b80d8
2007-06-22 01:35:51 +00:00
Dan Gohman
edc1d15984 Fix a typo in a comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@36485 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-26 19:40:56 +00:00
Evan Cheng
c6deb3d447 Estimate a cost using the possible number of scratch registers required and use
it as a late BURR scheduling tie-breaker.
Intuitively, it's good to push down instructions whose results are liveout so
their long live ranges won't conflict with other values which are needed inside
the BB. Further prioritize liveout instructions by the number of operands which
are calculated within the BB.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35109 91177308-0d34-0410-b5e6-96231b3b80d8
2007-03-14 22:43:40 +00:00
Evan Cheng
61230d18d2 Try schedule def + use closer whne Sethi-Ullman numbers are the same.
e.g.
t1 = op t2, c1
t3 = op t4, c2
and the following instructions are both ready.
t2 = op c3
t4 = op c4

Then schedule t2 = op first.
i.e.
t4 = op c4
t2 = op c3
t1 = op t2, c1
t3 = op t4, c2

This creates more short live intervals which work better with the register
allocator.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35089 91177308-0d34-0410-b5e6-96231b3b80d8
2007-03-13 23:25:11 +00:00
Chris Lattner
2ba528b3a7 switch the sched unit map over to use a DenseMap instead of std::map. This
speeds up isel as a whole time by 2.6%.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@33810 91177308-0d34-0410-b5e6-96231b3b80d8
2007-02-03 01:34:13 +00:00
Chris Lattner
fea997aac5 Fit in 80 columns
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@33745 91177308-0d34-0410-b5e6-96231b3b80d8
2007-02-01 04:55:59 +00:00
Evan Cheng
c8edc64188 Naming consistency.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@33026 91177308-0d34-0410-b5e6-96231b3b80d8
2007-01-08 23:55:53 +00:00
Evan Cheng
c62d4bb695 Fix for PR1075: bottom-up register-reduction scheduling actually increases register pressure.
- Fixed bugs in sethi-ullman number computation and priority queue comparison
functions.
- Separate code that handles priority computation special cases from SU number computation.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@33025 91177308-0d34-0410-b5e6-96231b3b80d8
2007-01-08 23:50:38 +00:00
Bill Wendling
832171cb97 Removing even more <iostream> includes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@32320 91177308-0d34-0410-b5e6-96231b3b80d8
2006-12-07 20:04:42 +00:00
Evan Cheng
ba59a1e453 Match TargetInstrInfo changes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@32098 91177308-0d34-0410-b5e6-96231b3b80d8
2006-12-01 21:52:58 +00:00
Evan Cheng
d5ad440f43 Remove dead code; added a missing null ptr check.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@31478 91177308-0d34-0410-b5e6-96231b3b80d8
2006-11-06 21:33:46 +00:00
Evan Cheng
95f6edeff5 Changes to use operand constraints to process two-address instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@31453 91177308-0d34-0410-b5e6-96231b3b80d8
2006-11-04 09:44:31 +00:00
Chris Lattner
02cb49ee67 silence warning
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@31397 91177308-0d34-0410-b5e6-96231b3b80d8
2006-11-03 01:28:29 +00:00
Reid Spencer
3ed469ccd7 For PR786:
Turn on -Wunused and -Wno-unused-parameter. Clean up most of the resulting
fall out by removing unused variables. Remaining warnings have to do with
unused functions (I didn't want to delete code without review) and unused
variables in generated code. Maintainers should clean up the remaining
issues when they see them. All changes pass DejaGnu tests and Olden.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@31380 91177308-0d34-0410-b5e6-96231b3b80d8
2006-11-02 20:25:50 +00:00
Evan Cheng
1dabb68ab4 Clean up.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@31359 91177308-0d34-0410-b5e6-96231b3b80d8
2006-11-01 22:39:30 +00:00
Evan Cheng
93467e7fe3 CopyFromReg starts a live range so its use should not be considered a floater.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@31356 91177308-0d34-0410-b5e6-96231b3b80d8
2006-11-01 22:17:06 +00:00
Evan Cheng
d42a5238a9 Debug tweak.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@30959 91177308-0d34-0410-b5e6-96231b3b80d8
2006-10-14 08:34:06 +00:00
Chris Lattner
a4f0b3a084 s|llvm/Support/Visibility.h|llvm/Support/Compiler.h|
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29911 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-27 12:54:02 +00:00
Chris Lattner
228a18e0f2 switch the SUnit pred/succ sets from being std::sets to being smallvectors.
This reduces selectiondag time on kc++ from 5.43s to 4.98s (9%).  More
significantly, this speeds up the default ppc scheduler from ~1571ms to 1063ms,
a 33% speedup.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29743 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-17 00:09:56 +00:00
Jim Laskey
eb577ba3b8 Final polish on machine pass registries.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29471 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-02 12:30:23 +00:00
Jim Laskey
9ff542f2cc 1. Change use of "Cache" to "Default".
2. Added argument to instruction scheduler creators so the creators can do
special things.
3. Repaired target hazard code.
4. Misc.

More to follow.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29450 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-01 18:29:48 +00:00
Jim Laskey
13ec702c43 Introducing plugable register allocators and instruction schedulers.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29434 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-01 14:21:23 +00:00
Jim Laskey
60f09928a0 Use an enumeration to eliminate data relocations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29249 91177308-0d34-0410-b5e6-96231b3b80d8
2006-07-21 20:57:35 +00:00
Chris Lattner
9525528a7d Use hidden visibility to make symbols in an anonymous namespace get
dropped.  This shrinks libllvmgcc.dylib another 67K


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28975 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-28 23:17:24 +00:00
Chris Lattner
f8c68f694c Shave another 27K off libllvmgcc.dylib with visibility hidden
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28973 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-28 22:17:39 +00:00
Evan Cheng
6b8e5a9318 Make sure the register pressure reduction schedulers work for non-uniform
latency targets, e.g. PPC32.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28561 91177308-0d34-0410-b5e6-96231b3b80d8
2006-05-30 18:05:39 +00:00