Commit Graph

53184 Commits

Author SHA1 Message Date
James Molloy
1038b57cac Fix a codegen fault in which log2 or exp2 could be dead-code eliminated even though they could have sideeffects.
Only allow log2/exp2 to be converted to an intrinsic if they are declared "readnone".

llvm-svn: 151807
2012-03-01 14:32:18 +00:00
Benjamin Kramer
44c3c88cb7 Make TargetRegisterClasses non-virtual by making the only virtual function a function pointer.
This allows us to make TRC non-polymorphic and value-initializable, eliminating a huge static
initializer and a ton of cruft from the generated code.

Shrinks ARMBaseRegisterInfo.o by ~100k.

llvm-svn: 151806
2012-03-01 13:37:55 +00:00
Benjamin Kramer
c57a6d4a2a Emit the "is an intrinsic overloaded" table as a bitfield.
llvm-svn: 151792
2012-03-01 02:16:57 +00:00
Akira Hatanaka
7964a20cf6 Pass endian information to constructors. Define separate functions to create
objects for big endian and little endian targets.

Patch by Jack Carter.

llvm-svn: 151788
2012-03-01 01:53:15 +00:00
Jakob Stoklund Olesen
2055fc1331 Make InlineSpiller bundle-aware.
Simply treat bundles as instructions. Spill code is inserted between
bundles, never inside a bundle.  Rewrite all operands in a bundle at
once.

Don't attempt and memory operand folding inside bundles.

llvm-svn: 151787
2012-03-01 01:43:25 +00:00
David Meyer
44201a2d17 [Object]
* Add begin_dynamic_table() / end_dynamic_table() private interface to ELFObjectFile.
* Add begin_libraries_needed() / end_libraries_needed() interface to ObjectFile, for grabbing the list of needed libraries for a shared object or dynamic executable.
* Implement this new interface completely for ELF, leave stubs for COFF and MachO.
* Add 'llvm-readobj' tool for dumping ObjectFile information.

llvm-svn: 151785
2012-03-01 01:36:50 +00:00
Jakob Stoklund Olesen
1885bf0ac7 Move getBundleStart() into MachineInstrBundle.h.
This allows the function to be inlined, and makes it suitable for use in
getInstructionIndex().

Also provide a const version. C++ is great for touch typing practice.

llvm-svn: 151782
2012-03-01 01:26:01 +00:00
Lang Hames
6cd018d0bc Don't redundantly copy implicit operands when rematerializing.
While we're at it - don't copy vreg implicit operands while rematerializing.
This fixes PR12138.

llvm-svn: 151779
2012-03-01 00:41:17 +00:00
Sean Callanan
171b2ddfbd Fixed the 32-bit runtime dynamic loader to allocate
code sections when needed.  It just had a conditional
the wrong way around.

llvm-svn: 151777
2012-03-01 00:15:29 +00:00
Kevin Enderby
46c639cf8a Added annotations for x86 pc relative loads to llvm's 'C' disassembler.
So with darwin's otool(1) an x86_64 hello world .o file will print:
leaq	L_.str(%rip), %rax ## literal pool for: Hello world

llvm-svn: 151769
2012-02-29 22:58:34 +00:00
Daniel Dunbar
ce00dc437f BitstreamWriter: Change primary output buffer to be a SmallVector instead of an
std::vector.
 - Good for 1-2% speedup on writing PCH for Cocoa.h.
 - Clang side API match to follow shortly, there wasn't an easy way to make this
   non-breaking.

llvm-svn: 151750
2012-02-29 20:31:09 +00:00
Daniel Dunbar
769ebcfda1 BitcodeWriter: Expose less implementation details -- make BackpatchWord private
and remove getBuffer().

llvm-svn: 151748
2012-02-29 20:31:01 +00:00
Daniel Dunbar
b887c34acf Bitcode: Don't expose WriteBitcodeToStream to clients.
llvm-svn: 151747
2012-02-29 20:30:56 +00:00
Andrew Trick
cf63a3d03a Intel Atom instruction itineraries for mov sign extension and mov zero extension.
Patch by Tyler Nowicki!

llvm-svn: 151743
2012-02-29 19:44:41 +00:00
Benjamin Kramer
1b5aa9f5cd LegalizeIntegerTypes: Reorder operations in the "big shift by small amount" optimization, making the lives of later passes easier.
llvm-svn: 151722
2012-02-29 13:27:00 +00:00
Duncan Sands
207ee17589 Have GVN also do condition propagation when the right-hand side is not
a constant.  This fixes PR1768.

llvm-svn: 151713
2012-02-29 11:12:03 +00:00
David Meyer
e9705e2949 [Object] Add symbol attribute flags: ST_ThreadLocal, ST_Common, and ST_Undefined. Implement these completely for ELF.
Rename ST_External to ST_Unknown, and slightly change its semantics. It now only indicates that the symbol's type
is unknown, not that the symbol is undefined. (For that, use ST_Undefined).

llvm-svn: 151696
2012-02-29 02:11:55 +00:00
Bill Wendling
7786b1f9dc Restrict this transformation to equality conditions.
This transformation is not correct for not-equal conditions:

(trunc x) != C1 & (and x, CA) != C2 -> (and x, CA|CMAX) != C1|C2

Let
  C1 == 0
  C2 == 0
  CA == 0xFF0000
  CMAX == 0xFF
and truncating to i8.

The original truth table:

    x   | A: trunc x != 0 | B: x & 0xFF0000 != 0 | A & B != 0
--------------------------------------------------------------
0x00000 |        0        |          0           |     0
0x00001 |        1        |          0           |     0
0x10000 |        0        |          1           |     0
0x10001 |        1        |          1           |     1

The truth table of the replacement:

    x   | x & 0xFF00FF != 0
----------------------------
0x00000 |        0
0x00001 |        1
0x10000 |        1
0x10001 |        1

So they are different.

llvm-svn: 151691
2012-02-29 01:46:50 +00:00
Jakob Stoklund Olesen
99c9b19c27 Add an analyzeVirtReg() function.
This function does more or less the same as
MI::readsWritesVirtualRegister(), but it supports bundles as well.

It also determines if any constraint requires reading and writing
operands to use the same register.  Most clients want to know.

Use the more modern MO.readsReg() instead of trying to sort out undefs
and partial redefines.  Stop supporting the extra full <imp-def> operand
as an alternative to <def,undef> sub-register defines.

llvm-svn: 151690
2012-02-29 01:40:37 +00:00
Derek Schuff
db82ff1e09 Make MemoryObject accessor members const again
llvm-svn: 151687
2012-02-29 01:09:06 +00:00
Daniel Dunbar
fd297dc564 Support/PathV2: Fix namespace qualifier in make_absolute(), for Win32.
llvm-svn: 151685
2012-02-29 00:46:46 +00:00
Jakob Stoklund Olesen
55ff251cc1 Move the operand iterator into MachineInstrBundle.h where it belongs.
Extract a base class and provide four specific sub-classes for iterating
over const/non-const bundles/instructions.

This eliminates the mystery bool constructor argument.

llvm-svn: 151684
2012-02-29 00:33:41 +00:00
Daniel Dunbar
4fd53dbd0e Support/PathV2: Fix make_absolute() to match is_absolute() and not expect to
find root names on Unix.
 - This fixes make_absolute to not basically always call current_path() on
   Unix systems.
 - I think the API probably needs cleanup in this area, but I'll let Michael
   handle that.

llvm-svn: 151681
2012-02-29 00:20:37 +00:00
Derek Schuff
c344aa4c7f Fix PR12080 by ensuring that MaterializeModule actually reads all the bitcode
in the streaming case.

llvm-svn: 151676
2012-02-29 00:07:09 +00:00
Michael J. Spencer
d29a314efb [PathV2] Fix bug in relative_path.
llvm-svn: 151675
2012-02-29 00:06:24 +00:00
Jim Grosbach
cb853fbdc9 ARM implement TargetInstrInfo::getNoopForMachoTarget()
Without this hook, functions w/ a completely empty body (including no
epilogue) will cause an MCEmitter assertion failure.

For example,
define internal fastcc void @empty_function() {
  unreachable
}

rdar://10947471

llvm-svn: 151673
2012-02-28 23:53:30 +00:00
David Meyer
31e23de700 In the ObjectFile interface, replace isInternal(), isAbsolute(), isGlobal(), and isWeak(), with a bitset of flags.
llvm-svn: 151670
2012-02-28 23:47:53 +00:00
Lang Hames
61e76ce0cf Kill off LiveRangeEdit::getNewVRegs and LiveRangeEdit::getUselessVRegs. These
methods are no longer needed now that LinearScan has gone away.

(Contains tweaks trivialSpillEverywhere to enable the removal of getNewVRegs).

llvm-svn: 151658
2012-02-28 22:07:24 +00:00
Rafael Espindola
646dff508a On ELF, create relocations to the abbreviation and line sections when producing
debug info for assembly files. We were already doing the right thing when
producing debug info for C/C++.

ELF linkers don't know dwarf, so they depend on these relocations to produce
valid dwarf output.

llvm-svn: 151655
2012-02-28 21:13:05 +00:00
Jim Grosbach
329e745b30 ARM vbit/vbif/vbsl assembly optional size suffix.
These instructions accept but do not require a size suffix.

rdar://10947225

llvm-svn: 151646
2012-02-28 19:11:07 +00:00
Evan Cheng
c5ead6c49e Re-commit r151623 with fix. Only issue special no-return calls if it's a direct call.
llvm-svn: 151645
2012-02-28 18:51:51 +00:00
Benjamin Kramer
9bff93fd22 Fix off-by one in comment.
llvm-svn: 151644
2012-02-28 18:37:06 +00:00
Roman Divacky
0398d1fddc Properly MCize the section switch, removing a FIXME.
llvm-svn: 151639
2012-02-28 18:15:25 +00:00
Benjamin Kramer
daa291f4fd LegalizeIntegerTypes: Reenable the large shift with small amount optimization.
To avoid problems with zero shifts when getting the bits that move between words
we use a trick: first shift the by amount-1, then do another shift by one. When
amount is 0 (and size 32) we first shift by 31, then by one, instead of by 32.

Also fix a latent bug that emitted the low and high words in the wrong order
when shifting right.

Fixes PR12113.

llvm-svn: 151637
2012-02-28 17:58:00 +00:00
Daniel Dunbar
b448d31a6b Revert r151623 "Some ARM implementaions, e.g. A-series, does return stack prediction. ...", it is breaking the Clang build during the Compiler-RT part.
llvm-svn: 151630
2012-02-28 15:36:07 +00:00
Daniel Dunbar
001b372f9a Remove stray semi-colon.
llvm-svn: 151629
2012-02-28 15:35:52 +00:00
Nadav Rotem
9c1789a96c Code cleanup following CR by Duncan.
llvm-svn: 151627
2012-02-28 14:13:19 +00:00
Nadav Rotem
75b36e6716 Fix a bug in the code that builds SDNodes from vector GEPs.
When the GEP index is a vector of pointers, the code that calculated the size
of the element started from the vector type, and not the contained pointer type.
As a result, instead of looking at the data element pointed by the vector, this
code used the size of the vector. This works for 32bit members (on 32bit
systems), but not for other types. Added code to peel the vector type and
added a test.

llvm-svn: 151626
2012-02-28 11:54:05 +00:00
Jia Liu
bdcd314be3 remove blanks, and some code format
llvm-svn: 151625
2012-02-28 07:46:26 +00:00
Evan Cheng
d29a22e4b0 Some ARM implementaions, e.g. A-series, does return stack prediction. That is,
the processor keeps a return addresses stack (RAS) which stores the address
and the instruction execution state of the instruction after a function-call
type branch instruction.

Calling a "noreturn" function with normal call instructions (e.g. bl) can
corrupt RAS and causes 100% return misprediction so LLVM should use a
unconditional branch instead. i.e.
mov lr, pc
b _foo
The "mov lr, pc" is issued in order to get proper backtrace.

rdar://8979299

llvm-svn: 151623
2012-02-28 06:42:03 +00:00
Pete Cooper
ab5f2302dc Reverted r152620 - DSE: Shorten memset when a later store overwrites the start of it. There were all sorts of buildbot issues
llvm-svn: 151621
2012-02-28 05:06:24 +00:00
Pete Cooper
93352dcd53 DSE: Shorten memset when a later store overwrites the start of it
llvm-svn: 151620
2012-02-28 04:27:10 +00:00
Bill Wendling
5841e676f3 Oops...Don't commit the other stuff..
llvm-svn: 151618
2012-02-28 04:01:21 +00:00
Bill Wendling
993b2b234e Modify comment to reflect the importance of this code.
llvm-svn: 151617
2012-02-28 03:47:09 +00:00
Akira Hatanaka
df24c0c218 Add comments.
llvm-svn: 151615
2012-02-28 03:18:43 +00:00
Akira Hatanaka
247d834415 Do not reserve $gp as a dedicated global base register if the target ABI is not O32.
llvm-svn: 151614
2012-02-28 03:17:38 +00:00
Akira Hatanaka
0934449dd8 Add support for floating point base register + offset register addressing mode
load and store instructions.

llvm-svn: 151611
2012-02-28 02:55:02 +00:00
Jakob Stoklund Olesen
c74b7b271e Handle regmasks in MachineCSE.
Don't attempt to extend physreg live ranges across calls.

<rdar://problem/10942095>

llvm-svn: 151610
2012-02-28 02:08:50 +00:00
Jakob Stoklund Olesen
e3a308c116 Handle regmasks in the machine code verifier.
llvm-svn: 151607
2012-02-28 01:42:41 +00:00
Jakob Stoklund Olesen
c6377253f7 Enable ARM base pointer when calling functions with large arguments.
When an outgoing call takes more than 2k of arguments on the stack, we
don't allocate that call frame in the prolog, but adjust the stack
pointer immediately before the call instead.

This causes problems with the emergency spill slot because PEI can't
track stack pointer adjustments on the second pass, and if the outgoing
arguments are too big, SP can't be used to reach the emergency spill
slot at all.

Work around these problems by ensuring there is a base or frame pointer
that can be used to access the emergency spill slot.

<rdar://problem/10917166>

llvm-svn: 151604
2012-02-28 01:15:01 +00:00