Remove redunant optimizations from InstCombine, instead call the appropriate functions from SimplifyInstruction

llvm-svn: 169941

Added a slew of SimplifyInstruction floating-point optimizations, many of which take advantage of fast-math flags. Test cases included.

  fsub X, +0 ==> X
  fsub X, -0 ==> X, when we know X is not -0
  fsub +/-0.0, (fsub -0.0, X) ==> X
  fsub nsz +/-0.0, (fsub +/-0.0, X) ==> X
  fsub nnan ninf X, X ==> 0.0
  fadd nsz X, 0 ==> X
  fadd [nnan ninf] X, (fsub [nnan ninf] 0, X) ==> 0
    where nnan and ninf have to occur at least once somewhere in this expression
  fmul X, 1.0 ==> X

llvm-svn: 169940

m_ConstantFP - match and bind a float constant
m_SpecificConstantFP - match a specific floating point value or vector of floats of that value
m_FPOne - match a floating point 1.0 or vector of 1.0s
m_NegZero - match -0.0
m_AnyZero - match 0 or -0.0

llvm-svn: 169939

llvm-svn: 169938

llvm-svn: 169933

llvm-svn: 169932

Pre-regalloc frame allocation and referencing has been on by default
for ages. No need for the testing option that disables it.

llvm-svn: 169931

Base pointer referencing has been enabled for ages.

llvm-svn: 169930

ScalarTargetTransformInfo::getIntImmCost() instead. "Legal" is a poorly defined
term for something like integer immediate materialization. It is always possible
to materialize an integer immediate. Whether to use it for memcpy expansion is
more a "cost" conceern.

llvm-svn: 169929

llvm-svn: 169928

Adding a validation callback mechanism to OptionValueString (such a feature might theoretically be added to the general OptionValue base class should the need arise)
Using this mechanism, making sure that the options to pass a summary string or a named summary to frame variable do not have invalid values

<rdar://problem/11576143>

llvm-svn: 169927

"self" when those pointers are in registers.
Previously in this case the IRInterpreter would
handle them just as if the user had typed in
"$rdi", which isn't safe because $rdi is passed
in through the argument struct.

Now we correctly break out all three cases (i.e.,
normal variables in registers, $reg, and this/self),
and handle them in a way that's a little bit easier
to read and change.

This results in more accurate printing of "this" and
"self" pointers all around.  I have strengthened the
optimized-code test case for Objective-C to ensure
that we catch regressions in this area reliably in
the future.

<rdar://problem/12693963>

llvm-svn: 169924

llvm-svn: 169923

llvm-svn: 169922

latter is rather a mess to type.

llvm-svn: 169919

llvm-svn: 169917

llvm-svn: 169916

A new backend supporting AMD GPUs: Radeon HD2XXX - HD7XXX

llvm-svn: 169915

Given a thread-local symbol x with global-dynamic access, the generated
code to obtain x's address is:

     Instruction                            Relocation            Symbol
  addis ra,r2,x@got@tlsgd@ha           R_PPC64_GOT_TLSGD16_HA       x
  addi  r3,ra,x@got@tlsgd@l            R_PPC64_GOT_TLSGD16_L        x
  bl __tls_get_addr(x@tlsgd)           R_PPC64_TLSGD                x
                                       R_PPC64_REL24           __tls_get_addr
  nop
  <use address in r3>

The implementation borrows from the medium code model work for introducing
special forms of ADDIS and ADDI into the DAG representation.  This is made
slightly more complicated by having to introduce a call to the external
function __tls_get_addr.  Using the full call machinery is overkill and,
more importantly, makes it difficult to add a special relocation.  So I've
introduced another opcode GET_TLS_ADDR to represent the function call, and
surrounded it with register copies to set up the parameter and return value.

Most of the code is pretty straightforward.  I ran into one peculiarity
when I introduced a new PPC opcode BL8_NOP_ELF_TLSGD, which is just like
BL8_NOP_ELF except that it takes another parameter to represent the symbol
("x" above) that requires a relocation on the call.  Something in the 
TblGen machinery causes BL8_NOP_ELF and BL8_NOP_ELF_TLSGD to be treated
identically during the emit phase, so this second operand was never
visited to generate relocations.  This is the reason for the slightly
messy workaround in PPCMCCodeEmitter.cpp:getDirectBrEncoding().

Two new tests are included to demonstrate correct external assembly and
correct generation of relocations using the integrated assembler.

Comments welcome!

Thanks,
Bill

llvm-svn: 169910

Add -fslp-vectorize (with -ftree-slp-vectorize as an alias for gcc compatibility)
to provide a way to enable the basic-block vectorization pass. This uses the same
acronym as gcc, superword-level parallelism (SLP), also common in the literature,
to refer to basic-block vectorization.

Nadav suggested this as a follow-up to the adding of -fvectorize.

llvm-svn: 169909

byref variable requires extended layout info. to prevent 
a crash involving arrays declared __block. // rdar://12787751

llvm-svn: 169908

llvm-svn: 169907

Loop Vectorize: optimize the vectorization of trunc(induction_var). The truncation is now done on scalars.

llvm-svn: 169904

unavailable due to availability attributes. <rdar://problem/12798237>

llvm-svn: 169903

Instead of doing a binary search over the whole diagnostic table (which weighs
a whopping 48k on x86_64), use the existing enums to compute the index in the
table. This avoids loading any unneeded data from the table and avoids littering
CPU caches with it. This code is in a hot path for code with many diagnostics.

1% speedup on -fsyntax-only gcc.c, which emits a lot of warnings.

llvm-svn: 169890

compatibility with gcc.
rdar://12839978 

llvm-svn: 169888

'file' builtin, and fix up a docstring a little.

Hat tip to Sebastian Kreft Carreno at Google for noticing the bug.

llvm-svn: 169887

because that method is only getting called for MCInstFragment. These
fragments aren't even generated when RelaxAll is set, which is why the
flag reference here is superfluous. Removing it simplifies the code
with no harmful effects.

An assertion is added higher up to make sure this path is never
reached.

llvm-svn: 169886

rdar://12839978 

llvm-svn: 169885

llvm-svn: 169881

llvm-svn: 169880

Since now we have an autogenerated TOC, a manually written table of all passes
was removed.

Patch by Anthony Mykhailenko with small fixes by me.

llvm-svn: 169867

Summary:
A few small coding style changes for StmtDumper, including:
- rename Dump* methods to dump*
- uninline some methods
- comment fixes
- whitespace fixes

Please review and commit if okay.

Reviewers: alexfh

Reviewed By: alexfh

CC: cfe-commits, alexfh, klimek

Differential Revision: http://llvm-reviews.chandlerc.com/D190

llvm-svn: 169866

Summary:
Also rename DumpDeclarator() to dumpDecl(). Once Decl dumping is added, these will be the two main methods of the class, so this is just for consistency in naming.

There was a DumpStmt() method already, but there was no point in having it, so I have merged it into VisitStmt(). Similarly, DumpExpr() is merged into VisitExpr().

Reviewers: alexfh

Reviewed By: alexfh

CC: cfe-commits, alexfh

Differential Revision: http://llvm-reviews.chandlerc.com/D156

llvm-svn: 169865

llvm-svn: 169864

llvm-svn: 169862

llvm-svn: 169861

llvm-svn: 169860

Use explicitely aligned store and load instructions to deal with argument and
retval shadow. This matters when an argument's alignment is higher than
__msan_param_tls alignment (which is the case with __m128i).

llvm-svn: 169859

Initial commit of the MemorySanitizer runtime library.

llvm-svn: 169858