Commit Graph

3494 Commits

Author SHA1 Message Date
Owen Anderson
23b0766b47 Fix formatting.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133164 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-16 16:52:24 +00:00
Jakob Stoklund Olesen
2559011a01 Prempt some obnoxious compiler from complaing about signed/unsigned
compares.

2^30 is actually the limit on the number of physical registers per
TargetRegisterInfo.h.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133142 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-16 03:07:40 +00:00
Jakob Stoklund Olesen
0cc0929efc Make sure to pass an unsigned to a printf format that is always %u.
This should unbreak the native ARM testers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133141 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-16 02:55:56 +00:00
Andrew Trick
05c087d893 Add support to lit for build mode requirements. e.g.
REQUIRES: Asserts
REQUIRES: Debug

This required chaining test configuration properties. It seems like a
generally good thing to do.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133131 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-16 01:33:35 +00:00
Owen Anderson
77b4b13c2a Add a new MVT::untyped. This will be used in future work for modelling ISA features like register pairs and lists with "interesting" constraints (such as ARM NEON contiguous register lists or even-odd paired registers). We need to be able to generate these instructions (often from intrinsics), but don't want to have to assign a legal type to them. Instead, we'll use an "untyped" edge to bypass the type-checking and simply ensure that the register classes match.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133106 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-15 23:35:18 +00:00
John McCall
f14bacc862 Update the Clang diagnostic emitter to emit IDs for diagnostic categories.
Patch by Argyrios Kyrtzidis.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133093 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-15 21:43:52 +00:00
John McCall
3a3465b71d Add a new function attribute, nonlazybind, which inhibits lazy-loading
optimizations when emitting calls to the function;  instead those calls may
use faster relocations which require the function to be immediately resolved
upon loading the dynamic object featuring the call.  This is useful when it
is known that the function will be called frequently and pervasively and
therefore there is no merit in delaying binding of the function.

Currently only implemented for x86-64, where it turns into a call through
the global offset table.

Patch by Dan Gohman, who assures me that he's going to add LangRef documentation
for this once it's committed.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133080 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-15 20:36:13 +00:00
Owen Anderson
1e56a2a85f Replace the statically generated hashtables for checking register relationships with just scanning the (typically tiny) static lists.
At the time I wrote this code (circa 2007), TargetRegisterInfo was using a std::set to perform these queries.  Switching to the static hashtables was an obvious improvement, but in reality there's no reason to do anything other than scan.
With this change, total LLC time on a whole-program 403.gcc is reduced by approximately 1.5%, almost all of which comes from a 15% reduction in LiveVariables time.  It also reduces the binary size of LLC by 86KB, thanks to eliminating a bunch of very large static tables.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133051 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-15 06:53:50 +00:00
Jakob Stoklund Olesen
59f26aadce Use a SetTheory instance to expand register lists in register classes.
This prepares tablegen to compute register lists from set theoretic dag
expressions. This doesn't really make any difference as long as
Target.td still declares RegisterClass::MemberList as [Register].

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133043 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-15 05:09:20 +00:00
Jakob Stoklund Olesen
ae1920b1ef Give CodeGenRegisterClass a real sorted member set.
Make the Elements vector private and expose an ArrayRef through
getOrder() instead. getOrder will eventually provide multiple
user-specified allocation orders.

Use the sorted member set for member and subclass tests. Clean up a lot
of ad hoc searches.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133040 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-15 04:50:36 +00:00
Bill Wendling
393c4047c0 Improve the heuristic to emit the alias if the number of hard-coded registers
are also greater than the alias.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133038 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-15 04:31:19 +00:00
Jakob Stoklund Olesen
7b9cafde5e Move the list of register classes into CodeGenRegBank as well.
No functional change intended.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133029 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-15 00:20:40 +00:00
Jakob Stoklund Olesen
952036def9 Fix a compile time regression caused by too small hash tables.
Measure the worst case number of probes for a miss instead of the less
conservative number of probes required for an insertion.

Lower the limit to < 6 probes worst case.

This doubles the size of the ARM and X86 hash tables, other targets are
unaffected. LiveVariables runs 12% faster with this change.

<rdar://problem/9598545>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132999 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-14 16:58:16 +00:00
Bill Wendling
740e5b3586 Heuristic: If the number of operands in the alias are more than the number of
operands in the aliasee, don't print the alias.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132963 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-14 03:17:20 +00:00
Jakob Stoklund Olesen
b95fd2d5fd Tweak hash function and compress hash tables.
Make the hash tables as small as possible while ensuring that all
lookups can be done in less than 8 probes.

Cut the aliases hash table in half by only storing a < b pairs - it
is a symmetric relation.

Use larger multipliers on the initial hash function to ensure that it
properly covers the whole table, and to resolve some clustering in the
very regular ARM register bank.

This reduces the size of most of these tables by 4x - 8x. For instance,
the ARM tables shrink from 48 KB to 8 KB.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132888 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-12 07:04:32 +00:00
Jakob Stoklund Olesen
bf710cc23e Remove now dead code.
These computations have been moved to CodeGenRegisters.cpp.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132887 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-12 07:04:28 +00:00
Jakob Stoklund Olesen
4091b059ec Extract the generateHashTable function.
The constant hash tables for sub-registers and overlaps are generated
the same way, so extract a function to generate and print the hash
table.

Also use the information computed by CodeGenRegisters.cpp instead of the
locally data.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132886 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-12 07:04:26 +00:00
Jakob Stoklund Olesen
026dc223ae Compute lists of sub-regs, super-regs, and overlapping regs.
Besides moving structural computations to CodeGenRegisters.cpp, this
also well-defines the order of these lists:

- Sub-register lists come from a pre-order traversal of the graph
  defined by the SubRegs lists in the .td files.

- Super-register lists are topologically ordered so no register comes
  before any of its sub-registers. When the sub-register graph is not a
  tree, independent super-registers appear in numerical order.

- Lists of overlapping registers are ordered according to register
  number.

This reverses the order of the super-regs lists, but nobody was
depending on that. The previous order of the overlaps lists was odd, and
it may have depended on the precise behavior of std::stable_sort.

The old computations are still there, but will be removed shortly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132881 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-12 03:05:52 +00:00
Jakob Stoklund Olesen
b5923db192 Move the list of registers into CodeGenRegBank.
Also move the sub-register index computations from RegisterInfoEmitter
into CodeGenRegBank.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132865 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-11 00:28:06 +00:00
Jakob Stoklund Olesen
dc29c44713 Move some sub-register index calculations to CodeGenRegisters.cpp
Create a new CodeGenRegBank class that will eventually hold all the code
that computes the register structure from Records.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132849 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-10 18:40:00 +00:00
Jakob Stoklund Olesen
f1e2b23dfa Move TableGen's register bank classes to their own source file.
I'll be moving some more code there to gather all of the
register-specific stuff in one place. Currently it is shared between
CodeGenTarget and RegisterInfoEmitter.

The plan is that CodeGenRegisters can compute the full register bank
structure while RegisterInfoEmitter only will handle the printing part.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132788 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-09 18:42:07 +00:00
Bob Wilson
e450a0046b Add special-case range checking for VCVT_N intrinsic immediate operands.
Radar 9558930.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132782 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-09 16:57:29 +00:00
Chad Rosier
36a300ac02 Fixed a few illegal paths with llvm_unreachable. Patch by Cameron McInally.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132732 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-07 20:41:31 +00:00
Jakob Stoklund Olesen
c017bc1c1e Drop a RecordKeeper reference that wasn't necessary.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132636 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-04 07:49:55 +00:00
Jakob Stoklund Olesen
1023f5a42d Silence compiler warnings.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132624 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-04 05:09:36 +00:00
Jakob Stoklund Olesen
1de99829b6 Teach TableGen to evaluate DAG expressions as set operations.
A TableGen backend can define how certain classes can be expanded into
ordered sets of defs, typically by evaluating a specific field in the
record. The SetTheory class can then evaluate DAG expressions that refer
to these named sets.

A number of standard set and list operations are predefined, and the
backend can add more specialized operators if needed. The -print-sets
backend is used by SetTheory.td to provide examples.

This is intended to simplify how register classes are defined:

  def GR32_NOSP : RegisterClass<"X86", [i32], 32, (sub GR32, ESP)>;

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132621 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-04 04:11:37 +00:00
Nick Lewycky
b2afe8759d Rework the logic to not rely on undefined behaviour (1LL << 64). Also simplify.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132537 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-03 08:25:39 +00:00
Jakob Stoklund Olesen
f462e3fac7 Make it possible to have unallocatable register classes.
Some register classes are only used for instruction operand constraints.
They should never be used for virtual registers. Previously, those
register classes were given an empty allocation order, but now you can
say 'let isAllocatable=0' in the register class definition.

TableGen calculates if a register is part of any allocatable register
class, and makes that information available in TargetRegisterDesc::inAllocatableClass.

The goal here is to eliminate use cases for overriding allocation_order_*
methods.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132508 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-02 23:07:20 +00:00
Joerg Sonnenberger
dd137903e4 Add new -d option to tblgen. It writes a make(1)-style dependency file.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132395 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-01 13:10:15 +00:00
Bruno Cardoso Lopes
895c1e2dee Fix ssat and ssat16 encodings for ARM and Thumb. The bit position value
must be encoded decremented by one. Only add encoding tests for ssat16
because ssat can't be parsed yet.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132324 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-31 03:33:27 +00:00
Rafael Espindola
6e032942cf Use the dwarf->llvm mapping to print register names in the cfi
directives.

Fixes PR9826.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132317 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-30 20:20:15 +00:00
Rafael Espindola
7a067cc6e0 Introduce the DwarfRegAlias class for declaring that two registers have the
same dwarf number. This will be used for creating a dwarf number to register
mapping.

The only case that needs this so far is the XMM/YMM registers that unfortunately
do have the same numbers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132314 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-30 17:49:59 +00:00
John McCall
bd0fa4c00d Change how tblgen generates attributes for intrinsics to use a single
switch.  With this newfound organization, teach tblgen how not to give
all intrinsics the 'nounwind' attribute.  Introduce a new intrinsic,
llvm.eh.resume, which does not have this attribute.  Documentation and uses
to follow.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132252 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-28 06:31:34 +00:00
Rafael Espindola
7bf114c1bd Fix the root cause of the bootstrap failure:
There was no way to check if a given register/mode pair was valid. We now return
an error code (-2) instead of asserting. If anyone thinks that an assert
at this point  is really needed, we can autogen a hasValidDwarfRegNum instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132236 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-28 00:13:01 +00:00
Argyrios Kyrtzidis
804cb23305 [tablegen] A couple of changes to ClangDiagnosticEmmitter.
-Emit an empty warning option as string ("") instead of 0.
-For diagnostic names also emit the size of the string.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132046 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-25 05:04:22 +00:00
Bill Wendling
f415d8b646 Use a more efficient data structure for the "operand map". The number of
operands to an instruction aren't great, so an iterative search is fairly quick
and doesn't have the overhead of std::map.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131886 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-23 00:18:33 +00:00
Cameron Zwarich
3a2d255ce6 Fix PR9947 by placing OPFL_MemRefs on the node using memory operands rather than
the root if there is only one such node. This leaves only 2 verifier failures in
the entire test suite when running "make check".

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131677 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-19 21:13:30 +00:00
Tanya Lattner
2a8eb722c7 In r131488 I misunderstood how VREV works. It splits the vector in half and splits each half. Therefore, the real problem was that we were using a VREV64 for a 4xi16, when we should have been using a VREV32.
Updated test case and reverted change to the PerfectShuffle Table.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131529 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-18 06:42:21 +00:00
Tanya Lattner
c81c9709ef vrev is incorrectly defined in the perfect shuffle table. The ordering is backwards (should be 0x3210 versus 0x1032) which exposed a bug when doing a shuffle on a 4xi16. I've attached a test case.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131488 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-17 20:48:40 +00:00
Jakob Stoklund Olesen
4bfc6fba78 Downgrade a tablegen warning to an error.
Ambiguous sub-register index compositions are OK as long as the backend writer
knows what he is doing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131134 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-10 17:52:59 +00:00
Mon P Wang
183c627d89 Fixed MC encoding for index_align for VLD1/VST1 (single element from one lane) for size 32
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131085 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-09 17:47:27 +00:00
Jakob Stoklund Olesen
123cab9480 Teach TableGen to automatically generate missing SubRegIndex instances.
The RegisterInfo.td file should only specify the indexes that sources need to
refer to. The rest is inferred.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131058 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-07 21:22:39 +00:00
Jim Grosbach
5d754bfa32 Improve diagnostics for some parse errors. Not asserting when a user input
error is detected is a good thing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131005 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-06 18:47:45 +00:00
Jim Grosbach
96883ec700 ParseFile() may throw, so extend the try/catch to handle that.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131004 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-06 18:39:28 +00:00
Mikhail Glushenkov
3a21c55a14 llvmc: Make it possible to provide an argument to (join).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130914 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-05 04:24:58 +00:00
Nick Lewycky
9f1569a3f7 The system suppression file should catch these, but since they *once again* are
not, I'll just add them here and be done with it.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130819 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-04 01:03:02 +00:00
Jim Grosbach
b9d5af05fd Tidy up. Add missing newline to generated file.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130779 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-03 19:09:56 +00:00
Eli Friedman
206a10cf28 Filter out pattterns from the FastISel emitter which it doesn't actually know how to handle. No significant functionality change at the moment, but it's necessary for some changes I'm planning.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130547 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-29 21:58:31 +00:00
Nick Lewycky
bcffb1fbb8 Rename profile_rt.so to libprofile_rt.so under configure+make (it already was
under cmake).

Add libprofile_rt.a so that we can tell clang to link against it in --coverage
mode. Also turn it on by default in cmake builds.

Oscar, this touches a change you made for EXCLUDE_FROM_ALL support -- I think
I've done the right thing, but please let me know (or fix and commit) if not!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130470 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-29 02:12:06 +00:00
Owen Anderson
4cdcb4772d Fix a bug in tblgen that caused incorrect encodings on instructions that specified operands with "bit" instead of "bits<1>".
Unfortunately, my only testcase for this is fragile, and the ARM AsmParser can't round trip the instruction in question.
<rdar://problem/9345702>


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130410 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-28 17:51:45 +00:00