Commit Graph

130 Commits

Author SHA1 Message Date
David Greene 63145844c8 Change errs() to dbgs().
llvm-svn: 92571
2010-01-05 01:24:43 +00:00
Nick Lewycky 974e12b2d3 Remove includes of Support/Compiler.h that are no longer needed after the
VISIBILITY_HIDDEN removal.

llvm-svn: 85043
2009-10-25 06:57:41 +00:00
Nick Lewycky 02d5f77d26 Remove VISIBILITY_HIDDEN from class/struct found inside anonymous namespaces.
Chris claims we should never have visibility_hidden inside any .cpp file but
that's still not true even after this commit.

llvm-svn: 85042
2009-10-25 06:33:48 +00:00
Dan Gohman 918ec53c64 The ScheduleDAG framework now requires an AliasAnalysis argument, though
it isn't needed in the ScheduleDAGSDNodes schedulers.

llvm-svn: 83691
2009-10-09 23:33:48 +00:00
Reid Kleckner 8ff5c19ebd Fix integer overflow in instruction scheduling. This can happen if we have
basic blocks that are so long that their size overflows a short.

Also assert that overflow does not happen in the future, as requested by Evan.

This fixes PR4401.

llvm-svn: 83159
2009-09-30 20:15:38 +00:00
Chris Lattner 317dbbcfb1 eliminate uses of cerr()
llvm-svn: 79834
2009-08-23 07:05:07 +00:00
Chris Lattner 4dc3edde9f remove a few DOUTs here and there.
llvm-svn: 79832
2009-08-23 06:35:02 +00:00
Torok Edwin fbcc663cbf llvm_unreachable->llvm_unreachable(0), LLVM_UNREACHABLE->llvm_unreachable.
This adds location info for all llvm_unreachable calls (which is a macro now) in
!NDEBUG builds.
In NDEBUG builds location info and the message is off (it only prints
"UREACHABLE executed").

llvm-svn: 75640
2009-07-14 16:55:14 +00:00
Torok Edwin 56d0659726 assert(0) -> LLVM_UNREACHABLE.
Make llvm_unreachable take an optional string, thus moving the cerr<< out of
line.
LLVM_UNREACHABLE is now a simple wrapper that makes the message go away for
NDEBUG builds.

llvm-svn: 75379
2009-07-11 20:10:48 +00:00
Bill Wendling 026e5d7667 Instead of passing in an unsigned value for the optimization level, use an enum,
which better identifies what the optimization is doing. And is more flexible for
future uses.

llvm-svn: 70440
2009-04-29 23:29:43 +00:00
Bill Wendling 084669a1c9 Second attempt:
Massive check in. This changes the "-fast" flag to "-O#" in llc. If you want to
use the old behavior, the flag is -O0. This change allows for finer-grained
control over which optimizations are run at different -O levels.

Most of this work was pretty mechanical. The majority of the fixes came from
verifying that a "fast" variable wasn't used anymore. The JIT still uses a
"Fast" flag. I'll change the JIT with a follow-up patch.

llvm-svn: 70343
2009-04-29 00:15:41 +00:00
Bill Wendling 56f2987a87 r70270 isn't ready yet. Back this out. Sorry for the noise.
llvm-svn: 70275
2009-04-28 01:04:53 +00:00
Bill Wendling d0ae15946c Massive check in. This changes the "-fast" flag to "-O#" in llc. If you want to
use the old behavior, the flag is -O0. This change allows for finer-grained
control over which optimizations are run at different -O levels.

Most of this work was pretty mechanical. The majority of the fixes came from
verifying that a "fast" variable wasn't used anymore. The JIT still uses a
"Fast" flag. I'm not 100% sure if it's necessary to change it there...

llvm-svn: 70270
2009-04-28 00:21:31 +00:00
Dan Gohman dfaf646c34 When scheduling a block in parts, keep track of the overall
instruction index across each part. Instruction indices are used
to make live range queries, and live ranges can extend beyond
scheduling region boundaries.

Refactor the ScheduleDAGSDNodes class some more so that it
doesn't have to worry about this additional information.

llvm-svn: 64288
2009-02-11 04:27:20 +00:00
Dan Gohman b95434356c Factor out more code for computing register live-range informationfor
scheduling, and generalize is so that preserves state across
scheduling regions. This fixes incorrect live-range information around
terminators and labels, which are effective region boundaries.

In place of looking for terminators to anchor inter-block dependencies,
introduce special entry and exit scheduling units for this purpose.

llvm-svn: 64254
2009-02-10 23:27:53 +00:00
Dan Gohman 483377c639 Move ScheduleDAGSDNodes.h to be a private header. Front-ends
that previously included this header should include
SchedulerRegistry.h instead.

llvm-svn: 63937
2009-02-06 17:22:58 +00:00
Dan Gohman ceac7c34f1 Initial hazard recognizer support in post-pass scheduling. This includes
a new toy hazard recognizier heuristic which attempts to direct the
scheduler to avoid clumping large groups of loads or stores too densely.

llvm-svn: 62291
2009-01-16 01:33:36 +00:00
Dan Gohman 7e105f0b12 Generalize the HazardRecognizer interface so that it can be used
to support MachineInstr-based scheduling in addition to
SDNode-based scheduling.

llvm-svn: 62284
2009-01-15 22:18:12 +00:00
Dan Gohman 619ef48a52 Move a few containers out of ScheduleDAGInstrs::BuildSchedGraph
and into the ScheduleDAGInstrs class, so that they don't get
destructed and re-constructed for each block. This fixes a
compile-time hot spot in the post-pass scheduler.

To help facilitate this, tidy and do some minor reorganization
in the scheduler constructor functions.

llvm-svn: 62275
2009-01-15 19:20:50 +00:00
Dan Gohman 1407484178 The list-td and list-tdrr schedulers don't yet support physreg
scheduling dependencies. Add assertion checks to help catch
this.

It appears the Mips target defaults to list-td, and it has a
regression test that uses a physreg dependence. Such code was
liable to be miscompiled, and now evokes an assertion failure.

llvm-svn: 62177
2009-01-13 20:24:13 +00:00
Dan Gohman 04543e719e Rename BuildSchedUnits to BuildSchedGraph, and refactor the
code in ScheduleDAGSDNodes' BuildSchedGraph into separate functions.

llvm-svn: 61376
2008-12-23 18:36:58 +00:00
Dan Gohman dddc1ac7ea Fix some register-alias-related bugs in the post-RA scheduler liveness
computation code. Also, avoid adding output-depenency edges when both
defs are dead, which frequently happens with EFLAGS defs.

Compute Depth and Height lazily, and always in terms of edge latency
values. For the schedulers that don't care about latency, edge latencies
are set to 1.

Eliminate Cycle and CycleBound, and LatencyPriorityQueue's Latencies array.
These are all subsumed by the Depth and Height fields.

llvm-svn: 61073
2008-12-16 03:25:46 +00:00
Dan Gohman 2d170896ee Rewrite the SDep class, and simplify some of the related code.
The Cost field is removed. It was only being used in a very limited way,
to indicate when the scheduler should attempt to protect a live register,
and it isn't really needed to do that. If we ever want the scheduler to
start inserting copies in non-prohibitive situations, we'll have to
rethink some things anyway.

A Latency field is added. Instead of giving each node a single
fixed latency, each edge can have its own latency. This will eventually
be used to model various micro-architecture properties more accurately.

The PointerIntPair class and an internal union are now used, which
reduce the overall size.

llvm-svn: 60806
2008-12-09 22:54:47 +00:00
Dan Gohman 4ce15e12b9 Factor out the code for verifying the work of the scheduler,
extend it a bit, and make use of it in all schedulers, to
ensure consistent checking.

llvm-svn: 59689
2008-11-20 01:26:25 +00:00
Dan Gohman 60cb69e665 Experimental post-pass scheduling support. Post-pass scheduling
is currently off by default, and can be enabled with
-disable-post-RA-scheduler=false.

This doesn't have a significant impact on most code yet because it doesn't
yet do anything to address anti-dependencies and it doesn't attempt to
disambiguate memory references. Also, several popular targets
don't have pipeline descriptions yet.

The majority of the changes here are splitting the SelectionDAG-specific
code out of ScheduleDAG, so that ScheduleDAG can be moved to
libLLVMCodeGen.a. The interface between ScheduleDAG-using code and
the rest of the scheduling code is somewhat rough and will evolve.

llvm-svn: 59676
2008-11-19 23:18:57 +00:00
Dan Gohman 71b632f905 Update a comment to reflect the current code.
llvm-svn: 59549
2008-11-18 21:14:44 +00:00
Dan Gohman fe1748da07 Fix a typo in a comment.
llvm-svn: 59489
2008-11-18 02:50:01 +00:00
Dan Gohman 22d07b14bc Change SUnit's dump method to take a ScheduleDAG* instead of
a SelectionDAG*.

llvm-svn: 59488
2008-11-18 02:06:40 +00:00
Dan Gohman 5ebdb98a6e Avoid using a loop in ReleasePred and ReleaseSucc methods to compute the
new CycleBound value. Instead, just update CycleBound on each call.
Also, make ReleasePred and ReleaseSucc methods more consistent accross
the various schedulers.

This also happens to make ScheduleDAGRRList's CycleBound computation
somewhat more interesting, though it still doesn't have any noticeable
effect, because no current targets that use the register-pressure
reduction scheduler provide pipeline models.

llvm-svn: 59475
2008-11-18 00:38:59 +00:00
Dan Gohman 92a36d7a78 Eliminate some trivial differences between the ScheduleNodeTopDown
functions in these two schedulers.

llvm-svn: 59465
2008-11-17 21:31:02 +00:00
Dan Gohman a687fd8339 Use SUnit's CycleBound field instead of duplicating it in
a side-car datastructure

llvm-svn: 59458
2008-11-17 19:45:19 +00:00
Dan Gohman 17c226b8ca Don't use the isPending flag to mean what the isAvailable flag means.
llvm-svn: 59445
2008-11-17 16:37:30 +00:00
Dan Gohman 68294c06fe Correct a comment.
llvm-svn: 59341
2008-11-15 00:24:23 +00:00
Dan Gohman d2760c0473 Move ScheduleDAGList's LatencyPriorityQueue class out to a separate file.
llvm-svn: 59340
2008-11-15 00:23:40 +00:00
Dan Gohman 072734ebd6 Remove the FlaggedNodes member from SUnit. Instead of requiring each SUnit
to carry a SmallVector of flagged nodes, just calculate the flagged nodes
dynamically when they are needed.

The local-liveness change is due to a trivial scheduling change where
the scheduler arbitrary decision differently.

llvm-svn: 59273
2008-11-13 23:24:17 +00:00
Dan Gohman 1ddfcba5be Make the Node member of SUnit private, and add accessors.
llvm-svn: 59264
2008-11-13 21:36:12 +00:00
Dan Gohman 5a390b974c Change ScheduleDAG's DAG member from a reference to a pointer, to prepare
for the possibility of scheduling without a SelectionDAG being present.

llvm-svn: 59263
2008-11-13 21:21:28 +00:00
Dan Gohman 5499e89d06 Change the scheduler accessor methods to accept an explicit TargetMachine
argument instead of taking the SelectionDAG's TargetMachine. This is
needed for some upcoming scheduler changes.

llvm-svn: 59055
2008-11-11 17:50:47 +00:00
Dan Gohman 9c4b7d5c4f Fix command-line option printing to print two spaces where needed,
instead of requiring all "short description" strings to begin with
two spaces. This makes these strings less mysterious, and it fixes
some cases where short description strings mistakenly did not
begin with two spaces.

llvm-svn: 57521
2008-10-14 20:25:08 +00:00
Dan Gohman 1705968102 Add a new function, ReplaceAllUsesOfValuesWith, which handles bulk
replacement of multiple values. This is slightly more efficient
than doing multiple ReplaceAllUsesOfValueWith calls, and theoretically
could be optimized even further. However, an important property of this
new function is that it handles the case where the source value set and
destination value set overlap. This makes it feasible for isel to use
SelectNodeTo in many very common cases, which is advantageous because
SelectNodeTo avoids a temporary node and it doesn't require CSEMap
updates for users of values that don't change position.

Revamp MorphNodeTo, which is what does all the work of SelectNodeTo, to
handle operand lists more efficiently, and to correctly handle a number
of corner cases to which its new wider use exposes it.

This commit also includes a change to the encoding of post-isel opcodes
in SDNodes; now instead of being sandwiched between the target-independent
pre-isel opcodes and the target-dependent pre-isel opcodes, post-isel
opcodes are now represented as negative values. This makes it possible
to test if an opcode is pre-isel or post-isel without having to know
the size of the current target's post-isel instruction set.

These changes speed up llc overall by 3% and reduce memory usage by 10%
on the InstructionCombining.cpp testcase with -fast and -regalloc=local.

llvm-svn: 53728
2008-07-17 19:10:17 +00:00
Dan Gohman adec96f438 Reapply 53476 and 53480, with a fix so that it properly updates
the BB member to the current basic block after emitting
instructions.

llvm-svn: 53567
2008-07-14 18:19:29 +00:00
Evan Cheng ef8412c822 Back out 53476 and 53480 for now. Somehow they cause llc to miscompile 179.art.
llvm-svn: 53502
2008-07-12 01:38:51 +00:00
Dan Gohman f4cd404e6f Factor out debugging code into the common base class.
llvm-svn: 53480
2008-07-11 22:36:22 +00:00
Dan Gohman 36a69373dc Add support for putting NamedRegionTimers in TimerGroups, and
use a timer group for the timers in SelectionDAGISel. Also,
Split scheduling out from emitting, to give each their own
timer.

llvm-svn: 53476
2008-07-11 21:54:34 +00:00
Evan Cheng 2c9773155a Do not use computationally expensive scheduling heuristics with -fast.
llvm-svn: 52971
2008-07-01 18:05:03 +00:00
Dan Gohman 0d8a61eb60 Use the new PriorityQueue in ScheduleDAGList too, which also
needs arbitrary-element removal.

llvm-svn: 52654
2008-06-23 23:40:09 +00:00
Dan Gohman 46520a25a4 Remove ScheduleDAG's SUnitMap altogether. Instead, use SDNode's NodeId
field, which is otherwise unused after instruction selection, as an index
into the SUnit array.

llvm-svn: 52583
2008-06-21 19:18:17 +00:00
Dan Gohman e6e1348275 Change ScheduleDAG's SUnitMap from DenseMap<SDNode*, vector<SUnit*> >
to DenseMap<SDNode*, SUnit*>, and adjust the way cloned SUnit nodes are
handled so that only the original node needs to be in the map.
This speeds up llc on 447.dealII.llvm.bc by about 2%.

llvm-svn: 52576
2008-06-21 15:52:51 +00:00
Dan Gohman 4370f26750 Treat EntryToken nodes as "passive" so that they aren't added to the
ScheduleDAG; they don't correspond to any actual instructions so they
don't need to be scheduled.

This fixes a bug where the EntryToken was being scheduled multiple
times in some cases, though it ended up not causing any trouble because 
EntryToken doesn't expand into anything. With this fixed the schedulers
reliably schedule the expected number of units, so we can check this
with an assertion.

This requires a tweak to test/CodeGen/X86/loop-hoist.ll because it
ends up getting scheduled differently in a trivial way, though it was
enough to fool the prcontext+grep that the test does.

llvm-svn: 49701
2008-04-15 01:22:18 +00:00
Dan Gohman 3a4be0fdef Rename MRegisterInfo to TargetRegisterInfo.
llvm-svn: 46930
2008-02-10 18:45:23 +00:00