Patch by Ivan Llopard.

llvm-svn: 158025

Minimum latency determines per-cycle scheduling groups.
Expected latency determines critical path and cost.

llvm-svn: 158021

expression (a * b + c) that can be implemented as a fused multiply-add (fma)
if the target determines that this will be more efficient. This intrinsic
will be used to implement FP_CONTRACT support and an aggressive FMA formation
mode.

If your target has a fast FMA instruction you should override the
isFMAFasterThanMulAndAdd method in TargetLowering to return true.

llvm-svn: 158014

llvm-svn: 158013

llvm-svn: 158007

llvm-svn: 158004

llvm-svn: 157981

llvm-svn: 157980

This allows a subtarget to explicitly specify the issue width and
other properties without providing pipeline stage details for every
instruction.

llvm-svn: 157979

Use ILP heuristics for long latency instrs if no scoreboard exists.

llvm-svn: 157978

valid itinerary but no pipeline stages.

An itinerary can contain useful scheduling information without specifying pipeline stages for each instruction.

llvm-svn: 157977

llvm-svn: 157976

llvm-svn: 157975

Rematerialization is handled by LiveRangeEdit now.

llvm-svn: 157974

It is an old function that does a lot more than required by
CalcSpillWeights, which was the only remaining caller.

The isRematerializable() function never actually sets the isLoad
argument, so don't try to compute that.

llvm-svn: 157973

llvm-svn: 157972

when a compile time constant is known.  This occurs when implicitly zero 
extending function arguments from 16 bits to 32 bits.

<rdar://problem/11481151>

llvm-svn: 157966

llvm-svn: 157963

replacement to make it at least as generic as the instruction being replaced.
This includes:
* dropping nsw/nuw flags
* getting the least restrictive tbaa and fpmath metadata
* merging ranges

Fixes PR12979.

llvm-svn: 157958

No functional change.

llvm-svn: 157957

Remove the same pointers from all the other LiveRangeCalc functions,
simplifying the interface.

llvm-svn: 157941

inserted after the shift-left-logical node.

llvm-svn: 157937

llvm-svn: 157935

This was mostly done already in r156162, but I missed one place.

llvm-svn: 157929

llvm-svn: 157925

llvm-svn: 157920

llvm-svn: 157917

Add VFMADDSUB and VFMSUBADD FMA instructions to folding tables. Also add 213 forms of scalar FMA instructions.

llvm-svn: 157914

llvm-svn: 157912

It seems that this no longer causes test suite failures on PPC64 (after r157159),
and often gives a performance benefit, so it can be enabled by default.

llvm-svn: 157911

vector. No functionality change.
Extracted from a patch by the PaX Team.

llvm-svn: 157909

Part of a patch by the PaX Team.

llvm-svn: 157908

llvm-svn: 157903

llvm-svn: 157902

llvm-svn: 157899

Rename fma4 intrinsics to just fma since they are now used for both FMA4 and FMA3. Autoupgrade support coming in a separate commit.

llvm-svn: 157898

llvm-svn: 157896

Use sse_load_f32/64 for scalar FMA3 intrinsic patterns instead of 128-bit loads to match instruction behavior.

llvm-svn: 157895

llvm-svn: 157894

Apart from being slightly cheaper, this fixes a real bug that hits 32 bit
linux systems. When passing a file larger than 2G to be linked (which isn't
that uncommon with large projects such as WebKit), clang's driver checks
if the file exists but the file size doesn't fit in an off_t and stat(2)
fails with EOVERFLOW. Clang then says that the file doesn't exist instead
of passing it to the linker.

llvm-svn: 157891