llvm-svn: 189005

Back in the mists of time (2008), it seems TableGen couldn't handle the
patterns necessary to match ARM's CMOV node that we convert select operations
to, so we wrote a lot of fairly hairy C++ to do it for us.

TableGen can deal with it now: there were a few minor differences to CodeGen
(see tests), but nothing obviously worse that I could see, so we should
probably address anything that *does* come up in a localised manner.

llvm-svn: 188995

llvm-svn: 188524

Added v8f16 to ValueTypes.h, ValueTypes.cpp, ValueTypes.td, 
and CodeGenTarget.cpp

Patch by Daniel Sanders

llvm-svn: 188326

Apparently caused a failure on Darwin

llvm-svn: 188166

clang bootstraps intermittently failed for me due a difference in
the MCK_Reg ordering in ARMGenAsmMatcher.inc.  E.g. in my latest
run the stage 1 and stage 3 versions were the same but the stage 2
one was different (though still functionally correct).  This meant
that the .o comparison failed.

MCK_Regs were assigned by iterating over a std::set< std::set<Record*> >,
and since std::set is sorted lexicographically, the order depended on the
order of the pointer values.  This patch replaces the pointer ordering
with LessRecordByID.

llvm-svn: 188164

LLVM's coding standards recommend raw_ostream and MemoryBuffer for
reading and writing text.

This has the side effect of allowing clang to compile more of Support
and TableGen in the Microsoft C++ ABI.

llvm-svn: 187826

Moving definition of MnemonicContainsDot field from class Instruction to class AsmParser as suggested.

llvm-svn: 187569

The clients of this code have been updated to all support AliasArgs.

This depends on Clang r187538 and lld r187541.

llvm-svn: 187546

This makes option aliases more powerful by enabling them to
pass along arguments to the option they're aliasing.

For example, if we have a joined option "-foo=", we can now
specify a flag option "-bar" to be an alias of that, with the
argument "baz".

This is especially useful for the cl.exe compatible clang driver,
where many options are aliases. For example, this patch enables
us to alias "/Ox" to "-O3" (-O is a joined option), and "/WX" to
"-Werror" (again, -W is a joined option).

Differential Revision: http://llvm-reviews.chandlerc.com/D1245

llvm-svn: 187537

llvm-svn: 187531

This reverts commit 2ca1e4a39c7e0d7a00e66ff5437c6d7ace2404a0.

llvm-svn: 187525

llvm-svn: 187511

The heuristic that merges register pressure sets was bogus for ARM's S/D regs.

llvm-svn: 187479

llvm-svn: 187478

Remove use of sprintf added to X86 disassembler tablegen code. Send message with instruction name to errs() instead and use a generic message for the llvm_unreachable. Consistent with other places in this file.

llvm-svn: 187333

llvm-svn: 187325

Added 512-bit operands printing.
Added instruction formats for KNL instructions.

llvm-svn: 187324

Using a different loop induction variable than the enclosing scope.  No functional changes intended.

llvm-svn: 187159

Fix a bug in TableGen where the intrinsic function name recognizer could mis-identify names if one was a prefix substring of the other

For two intrinsics 'llvm.nvvm.texsurf.handle' and 'llvm.nvvm.texsurf.handle.internal',
TableGen was emitting matching code like:

  if (Name.startswith("llvm.nvvm.texsurf.handle")) ...
  if (Name.startswith("llvm.nvvm.texsurf.handle.internal")) ...

We can never match "llvm.nvvm.texsurf.handle.internal" here because it will
always be erroneously matched by the first condition.

The fix is to sort the intrinsic names and emit them in reverse order.

llvm-svn: 187119

This lets heuristics easily pick the most important set to follow.

llvm-svn: 187108

This removes the need to store the asm variant in each row of the single table that existed before. Shaves ~16K off the size of X86AsmParser.o.

llvm-svn: 187026

llvm-svn: 186929

llvm-svn: 186928

This patch allows targets to define weather the instruction mnemonics in asm matcher tables will contain '.' character.

llvm-svn: 186388

llvm-svn: 186330

llvm-svn: 186312

llvm-svn: 186274

functions. Make the function attributes pass add it to known library functions
and when it can deduce it.

llvm-svn: 185735

Use SmallVectorImpl::iterator/const_iterator instead of SmallVector to avoid specifying the vector size.

llvm-svn: 185606

algorithm when assigning EnumValues to the synthesized registers.

The current algorithm, LessRecord, uses the StringRef compare_numeric
function.  This function compares strings, while handling embedded numbers.
For example, the R600 backend registers are sorted as follows:

  T1
  T1_W
  T1_X
  T1_XYZW
  T1_Y
  T1_Z
  T2
  T2_W
  T2_X
  T2_XYZW
  T2_Y
  T2_Z

In this example, the 'scaling factor' is dEnum/dN = 6 because T0, T1, T2
have an EnumValue offset of 6 from one another.  However, in other parts
of the register bank, the scaling factors are different:

dEnum/dN = 5:
  KC0_128_W
  KC0_128_X
  KC0_128_XYZW
  KC0_128_Y
  KC0_128_Z
  KC0_129_W
  KC0_129_X
  KC0_129_XYZW
  KC0_129_Y
  KC0_129_Z

The diff lists do not work correctly because different kinds of registers have
different 'scaling factors'.  This new algorithm, LessRecordRegister, tries to
enforce a scaling factor of 1.  For example, the registers are now sorted as
follows:

  T1
  T2
  T3
  ...
  T0_W
  T1_W
  T2_W
  ...
  T0_X
  T1_X
  T2_X
  ...
  KC0_128_W
  KC0_129_W
  KC0_130_W
  ...

For the Mips and R600 I see a 19% and 6% reduction in size, respectively.  I
did see a few small regressions, but the differences were on the order of a
few bytes (e.g., AArch64 was 16 bytes).  I suspect there will be even
greater wins for targets with larger register files.

Patch reviewed by Jakob.
rdar://14006013

llvm-svn: 185094

This patch modifies TableGen to generate a function in
${TARGET}GenInstrInfo.inc called getNamedOperandIdx(), which can be used
to look up indices for operands based on their names.

In order to activate this feature for an instruction, you must set the
UseNamedOperandTable bit.

For example, if you have an instruction like:

def ADD : TargetInstr <(outs GPR:$dst), (ins GPR:$src0, GPR:$src1)>;

You can look up the operand indices using the new function, like this:

Target::getNamedOperandIdx(Target::ADD, Target::OpName::dst)  => 0
Target::getNamedOperandIdx(Target::ADD, Target::OpName::src0) => 1
Target::getNamedOperandIdx(Target::ADD, Target::OpName::src1) => 2

The operand names are case sensitive, so $dst and $DST are considered
different operands.

This change is useful for R600 which has instructions with a large number
of operands, many of which model single bit instruction configuration
values.  These configuration bits are common across most instructions,
but may have a different operand index depending on the instruction type.
It is useful to have a convenient way to look up the operand indices,
so these bits can be generically set on any instruction.

llvm-svn: 184879

For decoding, keep the current behavior of always decoding these as their REP
versions. In the future, this could be improved to recognize the cases where
these behave as XACQUIRE and XRELEASE and decode them as such.

llvm-svn: 184207

And add Sandybridge/Haswell resource buffers.

llvm-svn: 184034

Replace the ill-defined MinLatency and ILPWindow properties with
with straightforward buffer sizes:
MCSchedMode::MicroOpBufferSize
MCProcResourceDesc::BufferSize

These can be used to more precisely model instruction execution if desired.

Disabled some misched tests temporarily. They'll be reenabled in a few commits.

llvm-svn: 184032

llvm-svn: 183690

PR16281.

llvm-svn: 183630

llvm-svn: 183465

The element passed to push_back is not copied before the vector reallocates.
The client needs to copy the element first before passing it to push_back.

No test case, will be tested by follow-up swift scheduler model change (it
segfaults without this change).

llvm-svn: 183459

llvm-svn: 183426