I don't really see this as being needed, but there is little harm from doing
it.

llvm-svn: 58385

 - No functionality change.

llvm-svn: 58355

 - No functionality change.

llvm-svn: 58352

llvm-svn: 58351

that need it to require it by ID.

llvm-svn: 58238

LargeBlockInfo, we can now dramatically simplify their implementation
and speed them up at the same time.  Now the code has time proportional
to the number of uses of the alloca, not the size of the block.

This also eliminates code that tried to batch up different allocas which
are used in the same blocks, and eliminates the 'retry list' logic which
was baroque and no unneccesary.  In addition to being a speedup for crazy
cases, this is also a nice cleanup:

PromoteMemoryToRegister.cpp |  270 +++++++++++++++-----------------------------
 1 file changed, 96 insertions(+), 174 deletions(-)

llvm-svn: 58229

a trivial dense map.  Use this in RewriteSingleStoreAlloca to
avoid aggressively rescanning blocks over and over again.  This
fixes PR2925, speeding up mem2reg on the testcase in that bug
from 4.56s to 0.02s in a debug build on my machine.

llvm-svn: 58227

Understand that mul %x, 1 = %x.

llvm-svn: 58069

LoopPass*.
 - Although less precise, this means they can be used in clients
   without RTTI (who would otherwise need to include LoopPass.h, which
   eventually includes things using dynamic_cast). This was the
   simplest solution that presented itself, but I am happy to use a
   better one if available.

llvm-svn: 58010

llvm-svn: 57870

llvm-svn: 57864

llvm-svn: 57660

to find opportunities for store-to-load forwarding or load CSE,
in the same way that visitStore scans back to do DSE. Also, define
a new helper function for testing whether the addresses of two
memory accesses are known to have the same value, and use it in
both visitStore and visitLoad.

These two changes allow instcombine to eliminate loads in code
produced by front-ends that frequently emit obviously redundant
addressing for memory references.

llvm-svn: 57608

llvm-svn: 57515

- Renumber fcmp predicates to match their icmp counterparts.
- Try swapping operands to expose more optimization opportunities.

llvm-svn: 57513

     ord && olt -> olt
     ord && ueq -> oeq

llvm-svn: 57507

constant expression with all zero indices as being the same as a bitcast.

llvm-svn: 57442

a couple other cases for clarity, but shouldn't affect correctness.

Patch by Eli Friedman!

llvm-svn: 57387

This patch fixes PR 2869

llvm-svn: 57369

llvm-svn: 57353

and APFloat::convertToInteger.  Restore return value to
IEEE754.  Adjust all users accordingly.

llvm-svn: 57329

llvm-svn: 57320

llvm-svn: 57305

Patch by Samuel Tardieu.

llvm-svn: 57291

patch by Samuel Tardieu!

llvm-svn: 57288

llvm-svn: 57254

llvm-svn: 57221

This includes not marking a GEP involving a vector as unsafe, but only when it
has all zero indices. This allows scalarrepl to work in a few more cases.

llvm-svn: 57177

shifting and masking inside a bswap expr.  This allows it to handle
the cases from PR2842, which involve the intermediate 'or' 
expressions being shifted, not just the input value.

llvm-svn: 57095

ashr.  It should only apply to lshr.

llvm-svn: 57089

when deciding whether to mark a function readnone/readonly.
Since the pass is currently run before SROA, this may be
quite helpful.  Requested by Chris on IRC.

llvm-svn: 57050

applicable.

llvm-svn: 57033

While hosting instruction check PHI node.

llvm-svn: 57025

Why?  Because it was there!

llvm-svn: 56996

llvm-svn: 56994

pointer bitcasts and GEP's", and centralize the
logic in Value::getUnderlyingObject.  The
difference with stripPointerCasts is that
stripPointerCasts only strips GEPs if all
indices are zero, while getUnderlyingObject
strips GEPs no matter what the indices are.

llvm-svn: 56922

llvm-svn: 56917

instruction, not after. This fixes some uses of free'd memory.

llvm-svn: 56908

llvm-svn: 56887

llvm-svn: 56868