No new test case is added.
This patch makes test JITTest.FunctionIsRecompiledAndRelinked pass on mips
platform.

Patch by Petar Jovanovic.

llvm-svn: 161098

llvm-svn: 161097

llvm-svn: 161095

instructions that decrement and increment the stack pointer before and after a
call when the function does not have a reserved call frame.

llvm-svn: 161093

MipsSERegisterInfo.

llvm-svn: 161092

MipsSEFrameLowering.

Implement MipsSEFrameLowering::hasReservedCallFrame. Call frames will not be
reserved if there is a call with a large call frame or there are variable sized
objects on the stack.

llvm-svn: 161090

llvm-svn: 161085

llvm-svn: 161083

llvm-svn: 161082

and MipsSEInstrInfo (for mips32/64).

llvm-svn: 161081

of them.

llvm-svn: 161080

computing it in MipsFrameLowering::emitPrologue.

llvm-svn: 161078

The frame object which points to the dynamically allocated area will not be
needed after changes are made to cease reserving call frames.

llvm-svn: 161076

Use stable_sort instead of sort. Follow-up to r161062.

rdar://11980766

llvm-svn: 161075

Assuming infinite issue width, compute the earliest each instruction in
the trace can issue, when considering the latency of data dependencies.
The issue cycle is record as a 'depth' from the beginning of the trace.

This is half the computation required to find the length of the critical
path through the trace. Heights are next.

llvm-svn: 161074

llvm-svn: 161072

or not it is in mips16 mode. Define MipsPseudo (mode-independant pseudo) and
PseudoSE (mips32/64 pseudo) classes.

llvm-svn: 161071

instruction class. SE stands for standard encoding.

llvm-svn: 161069

arguments to the stack in MipsISelLowering::LowerCall, use stack pointer and
integer offset operands rather than frame object operands.

llvm-svn: 161068

Previously, we were using EBX, but PIC requires the GOT to be in EBX before 
function calls via PLT GOT pointer.

llvm-svn: 161066

llvm-svn: 161065

single-precision load and store.

Also avoid selecting LUXC1 and SUXC1 instructions during isel. It is incorrect
to map unaligned floating point load/store nodes to these instructions.

llvm-svn: 161063

One motivating example is to sink an instruction from a basic block which has
two successors: one outside the loop, the other inside the loop. We should try
to sink the instruction outside the loop.

rdar://11980766

llvm-svn: 161062

llvm-svn: 161061

llvm-svn: 161053

re-use allocated vectors as much as possible.

llvm-svn: 161041

llvm-svn: 161037

llvm-svn: 161034

Tidy up. Move for loop index declarations into for statements. Use unsigned instead of uint16_t for loop indices. Use unsigned instead of uint32_t for arguments to raw_ostream.indent.

llvm-svn: 161033

llvm-svn: 161032

llvm-svn: 161031

llvm-svn: 161030

llvm-svn: 161029

llvm-svn: 161027

llvm-svn: 161026

llvm-svn: 161025

for this class. These tests exercise most of the basic properties, but
the API for TinyPtrVector is very strange currently. My plan is to start
fleshing out the API to match that of SmallVector, but I wanted a test
for what is there first.

Sadly, it doesn't look reasonable to just re-use the SmallVector tests,
as this container can only ever store pointers, and much of the
SmallVector testing is to get construction and destruction right.

Just to get this basic test working, I had to add value_type to the
interface.

While here I found a subtle bug in the combination of 'erase', 'begin',
and 'end'. Both 'begin' and 'end' wanted to use a null pointer to
indicate the "end" iterator of an empty vector, regardless of whether
there is actually a vector allocated or the pointer union is null.
Everything else was fine with this except for erase. If you erase the
last element of a vector after it has held more than one element, we
return the end iterator of the underlying SmallVector which need not be
a null pointer. Instead, simply use the pointer, and poniter + size()
begin/end definitions in the tiny case, and delegate to the inner vector
whenever it is present.

llvm-svn: 161024

We are extending live ranges, so kill flags are not accurate. They
aren't needed until they are recomputed after RA anyway.

<rdar://problem/11950722>

llvm-svn: 161023

We branch to the successor with higher edge weight first.
Convert from
     je    LBB4_8  --> to outer loop
     jmp   LBB4_14 --> to inner loop
to
     jne   LBB4_14
     jmp   LBB4_8

PR12750
rdar: 11393714

llvm-svn: 161018

llvm-svn: 161010