xfail

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55914 91177308-0d34-0410-b5e6-96231b3b80d8

Remove OptimizeIVType()

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55913 91177308-0d34-0410-b5e6-96231b3b80d8

Didn't mean to commit this change to how the
callgraph is printed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55912 91177308-0d34-0410-b5e6-96231b3b80d8

Avoid redefinition and nnbreak windows build.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55911 91177308-0d34-0410-b5e6-96231b3b80d8

Unbreak and fix indentation

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55904 91177308-0d34-0410-b5e6-96231b3b80d8

Testcase for 64-bit lowering of eh_return & unwind_init

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55903 91177308-0d34-0410-b5e6-96231b3b80d8

Add test for checking proper lowering of eh_return & unwind init intrinsics on 32bit x86 targets

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55902 91177308-0d34-0410-b5e6-96231b3b80d8

Drop unused variable

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55901 91177308-0d34-0410-b5e6-96231b3b80d8

We do support EH on x86-64!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55900 91177308-0d34-0410-b5e6-96231b3b80d8

First draft of EH support on x86/64-linux

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55899 91177308-0d34-0410-b5e6-96231b3b80d8

Implement FRAME_TO_ARGS_OFFSET for x86-64

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55898 91177308-0d34-0410-b5e6-96231b3b80d8

Add testcase from bug 2770.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55897 91177308-0d34-0410-b5e6-96231b3b80d8

Update the callgraph correctly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55896 91177308-0d34-0410-b5e6-96231b3b80d8

Update the callgraph correctly in ArgumentPromotion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55895 91177308-0d34-0410-b5e6-96231b3b80d8

Reapply 55859.  This doesn't change anything as
long as the callgraph is correct.  It checks
for wrong callgraphs more strictly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55894 91177308-0d34-0410-b5e6-96231b3b80d8

Correctly handle physical register inputs. They are not explicit input operands in the resulting machine instrs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55893 91177308-0d34-0410-b5e6-96231b3b80d8

Add fast isel physical register definition support.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55892 91177308-0d34-0410-b5e6-96231b3b80d8

Add support to extend call operands when needed. Enable x86 fastisel call support.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55891 91177308-0d34-0410-b5e6-96231b3b80d8

Revert my previous change -- the subtraction of two constants was a no-op
before. This is taken care of in the selection DAG pass. In my opinion, this
should be in one place or the other. I.e., it should probably be removed from
the DAG combiner (along with the other arithmetic transformations on constants
that are essentially no-ops).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55889 91177308-0d34-0410-b5e6-96231b3b80d8

Convert

    // fold (sub c1, c2) -> c1-c2

from a no-op into an actual transformation.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55886 91177308-0d34-0410-b5e6-96231b3b80d8

Clean up CSS validator errors: 'Property align doesn't exist : center'

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55885 91177308-0d34-0410-b5e6-96231b3b80d8

Formatting changes. to make the information cleaner.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55884 91177308-0d34-0410-b5e6-96231b3b80d8

Correct a comment and strip trailing whitespace.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55883 91177308-0d34-0410-b5e6-96231b3b80d8

Initial fastisel call support for C, Fast, and X86_FastCall calling conventions. It's meant to handle "simple" calls, i.e. no byval, structret, etc. It doesn't support multi-result returns either.

Not yet turned on, it needs to support sext / zext of arguments and result.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55882 91177308-0d34-0410-b5e6-96231b3b80d8

Some code clean up.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55881 91177308-0d34-0410-b5e6-96231b3b80d8

Indentation.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55880 91177308-0d34-0410-b5e6-96231b3b80d8

- Doh. Pass vector by value is bad.
- Add a AnalyzeCallResult specialized for calls which produce a single value. This is used by fastisel.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55879 91177308-0d34-0410-b5e6-96231b3b80d8

Eliminate a compile time warning.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55878 91177308-0d34-0410-b5e6-96231b3b80d8

Handle x86 truncate to i8 with target hook for now.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55877 91177308-0d34-0410-b5e6-96231b3b80d8

Fix indentation of generated code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55876 91177308-0d34-0410-b5e6-96231b3b80d8

Ignore multi-instruction patterns. e.g.
def : Pat<(i8 (trunc GR32:$src)),
          (i8 (EXTRACT_SUBREG (MOV32to32_ GR32:$src), x86_subreg_8bit))>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55875 91177308-0d34-0410-b5e6-96231b3b80d8

fix crash when the malloc/free function is defined or is a declaration with 0 parameters.
this pass doesnt seem to be used, but still it's now a little more correct

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55873 91177308-0d34-0410-b5e6-96231b3b80d8

When PruneEH turned an invoke into an ordinary
call (thus changing the call site) it didn't
inform the callgraph about this.  But the
call site does matter - as shown by the testcase,
the callgraph become invalid after the inliner
ran (with an edge between two functions simply
missing), resulting in wrong deductions by
GlobalsModRef.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55872 91177308-0d34-0410-b5e6-96231b3b80d8

Fix constant pool loads, and remove broken versions of addConstantPoolReference.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55868 91177308-0d34-0410-b5e6-96231b3b80d8

Fix the X86 addConstantPoolReference, which had the operands in the wrong order.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55867 91177308-0d34-0410-b5e6-96231b3b80d8

Next limited float precision expansion (log2 12 bits)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55866 91177308-0d34-0410-b5e6-96231b3b80d8

Revert r55859.  This is breaking the build in the abscence of its companion commit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55865 91177308-0d34-0410-b5e6-96231b3b80d8

Fix for PR2687: Add patterns to match sint_to_fp and fp_to_sint for <2 x
i32>.  This is a little messy, but it works.

We should really get rid of the intrinsics, though, since they map
perfectly well to standard LLVM instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55864 91177308-0d34-0410-b5e6-96231b3b80d8

Move the code that inserts copies for function livein registers
out of ScheduleDAGEmit.cpp and into SelectionDAGISel.cpp. This
allows it to be run exactly once per function, even if multiple
SelectionDAG iterations happen in the entry block, as may happen
with FastISel.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55863 91177308-0d34-0410-b5e6-96231b3b80d8

Remove unused map.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55861 91177308-0d34-0410-b5e6-96231b3b80d8

Mark math lib intrinsics readonly, on the grounds
that they read the rounding mode.  This is conservatively
correct, which they weren't before.  We can do more
optimization on these if we actually model the rounding
mode.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55860 91177308-0d34-0410-b5e6-96231b3b80d8

Delete the removeCallEdgeTo callgraph method,
because it does not maintain a correct list
of callsites.  I discovered (see following
commit) that the inliner will create a wrong
callgraph if it is fed a callgraph with
correct edges but incorrect callsites.  These
were created by Prune-EH, and while it wasn't
done via removeCallEdgeTo, it could have been
done via removeCallEdgeTo, which is an accident
waiting to happen.  Use removeCallEdgeFor
instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55859 91177308-0d34-0410-b5e6-96231b3b80d8

Prevent our own passes from promoting this to
readonly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55858 91177308-0d34-0410-b5e6-96231b3b80d8

Fix X86FastISel's shift and select code to reject illegal types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55857 91177308-0d34-0410-b5e6-96231b3b80d8

Add the next limited-precision expansion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55856 91177308-0d34-0410-b5e6-96231b3b80d8

Fix the opcodes used by X86FastISel for shifts and conditional moves.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55855 91177308-0d34-0410-b5e6-96231b3b80d8

Factor out code that emits load and store instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55854 91177308-0d34-0410-b5e6-96231b3b80d8

Rename method.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55853 91177308-0d34-0410-b5e6-96231b3b80d8

Fix test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55849 91177308-0d34-0410-b5e6-96231b3b80d8

FastISel support for AND and OR with type i1.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55846 91177308-0d34-0410-b5e6-96231b3b80d8

Add hooks for other intrinsics to get low-precision expansions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55845 91177308-0d34-0410-b5e6-96231b3b80d8

X86FastISel support for shifts and conditional moves.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55844 91177308-0d34-0410-b5e6-96231b3b80d8

FastISel support for ConstantExprs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55843 91177308-0d34-0410-b5e6-96231b3b80d8

Revert r55817. It broke PIC. FastISel will need to find a different
approach here.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55842 91177308-0d34-0410-b5e6-96231b3b80d8

If SSE2 is available, x86 should pass first 3 f32/f64 arguments in XMM registers for fastcc calls.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55840 91177308-0d34-0410-b5e6-96231b3b80d8

Add a variant of AnalyzeCallOperands that can be used by fast isel.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55838 91177308-0d34-0410-b5e6-96231b3b80d8

Use removeAllCalledFunctions rather than removing
edges one by one by hand.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55836 91177308-0d34-0410-b5e6-96231b3b80d8

Remove trailing whitespace.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55835 91177308-0d34-0410-b5e6-96231b3b80d8

Strip trailing whitespace.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55834 91177308-0d34-0410-b5e6-96231b3b80d8

Make this pass return that it made a change if
it modifies a functions attributes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55831 91177308-0d34-0410-b5e6-96231b3b80d8

"Fix" PR2762.  The testcase now crashes codegen
elsewhere due to a missing pattern for
v2f64 = sint_to_fp v2i32.  That is PR2687.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55828 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a search+replace-o.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55824 91177308-0d34-0410-b5e6-96231b3b80d8

Add -flimit-float-precision to enable some faster,
but less accurate (non-IEEE) code sequences for
certain math library functions.  Add the first of
several such expansions.  Don't worry, if you don't
turn it on it won't affect you.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55823 91177308-0d34-0410-b5e6-96231b3b80d8

Check a comparion's operand type for legality before
expanding its operands.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55820 91177308-0d34-0410-b5e6-96231b3b80d8

Fix X86FastISel code for comparisons and conditional branches
to check the result of getRegForValue before using it, and
to check for illegal operand types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55819 91177308-0d34-0410-b5e6-96231b3b80d8

FastISel support for unreachable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55818 91177308-0d34-0410-b5e6-96231b3b80d8

In FastISel mode, the scheduler may be invoked multiple times
in the same block. Fix the entry-block handling to only run at
at the beginning of the entry block, and not any other times.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55817 91177308-0d34-0410-b5e6-96231b3b80d8

X86FastISel support for conditional branches.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55816 91177308-0d34-0410-b5e6-96231b3b80d8

Add initial support for selecting constant materializations that require constant
pool loads on X86 in fast isel.  This isn't actually used yet.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55814 91177308-0d34-0410-b5e6-96231b3b80d8

X86FastISel support for ICmpInst and FCmpInst.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55811 91177308-0d34-0410-b5e6-96231b3b80d8

update.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55809 91177308-0d34-0410-b5e6-96231b3b80d8

Document function notes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55808 91177308-0d34-0410-b5e6-96231b3b80d8

For whatever the reason, x86 CallingConv::Fast (i.e. fastcc) was not passing scalar arguments in registers. This patch defines a new fastcc CC which is slightly different from the FastCall CC. In addition to passing integer arguments in ECX and EDX, it also specify doubles are passed in 8-byte slots which are 8-byte aligned (instead of 4-byte aligned). This avoids a potential performance hazard where doubles span cacheline boundaries.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55807 91177308-0d34-0410-b5e6-96231b3b80d8

A loop may be unswitched multiple times. Reconstruct dom info. at the end.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55806 91177308-0d34-0410-b5e6-96231b3b80d8

Adjust tests to expect new math intrinsics to be
emitted only when errno is not in use.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55797 91177308-0d34-0410-b5e6-96231b3b80d8

If function notes say optimize for size, then adjust alignment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55794 91177308-0d34-0410-b5e6-96231b3b80d8

Add an include of SmallSet.h.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55793 91177308-0d34-0410-b5e6-96231b3b80d8

Initialize loop data first.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55792 91177308-0d34-0410-b5e6-96231b3b80d8

Neaten this up a bit.  No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55789 91177308-0d34-0410-b5e6-96231b3b80d8

Do not unswitch if the function notes say we're optimizing this function for size.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55786 91177308-0d34-0410-b5e6-96231b3b80d8

try to seperate the mechanism into something others can use

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55785 91177308-0d34-0410-b5e6-96231b3b80d8

Adjust libcalls tests to expect intrinsic for exp2

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55784 91177308-0d34-0410-b5e6-96231b3b80d8

Add intrinsic forms of pow and exp2.  The non-intrinsic
forms remain to handle older IR files, but will go away soon.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55781 91177308-0d34-0410-b5e6-96231b3b80d8

Tidy up several unbeseeming casts from pointer to intptr_t.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55779 91177308-0d34-0410-b5e6-96231b3b80d8

Fix the ordering of operands to the store (inverted relative to LLVM IR), and fix the testcase.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55777 91177308-0d34-0410-b5e6-96231b3b80d8

Clean up uses of TargetLowering::getTargetMachine.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55769 91177308-0d34-0410-b5e6-96231b3b80d8

cleanup as per Duncan's review

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55766 91177308-0d34-0410-b5e6-96231b3b80d8

Add a first attempt at implementing stores for X86 fast isel using target hooks.
Dan or Evan, please review.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55764 91177308-0d34-0410-b5e6-96231b3b80d8

Load from GV stub should be locally CSE'd.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55763 91177308-0d34-0410-b5e6-96231b3b80d8

Fix an overly strict assertion. Source register of a copy may not be killed, it may be killed by an implicit super-register use.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55762 91177308-0d34-0410-b5e6-96231b3b80d8

Remove code that pad number of bytes to pop for X86_FastCall CC. The code doesn't do the "aligning" for Cygwin, Mingw, and Windows. But aligning it on Darwin and Linux breaks gcc compatibility. That ruled out all the platforms we support!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55756 91177308-0d34-0410-b5e6-96231b3b80d8

Add intrinsics for log, log2, log10, exp, exp2.
No functional change (and no FE change to generate them).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55753 91177308-0d34-0410-b5e6-96231b3b80d8

Do trivial local CSE for constants and other non-Instruction values
in FastISel.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55748 91177308-0d34-0410-b5e6-96231b3b80d8

Put RegsForValue in the llvm namespace to avoid warnings about
classes in the llvm namespace having members with types from
anonymous namespaces.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55747 91177308-0d34-0410-b5e6-96231b3b80d8

Create HandlePHINodesInSuccessorBlocksFast, a version of
HandlePHINodesInSuccessorBlocks that works FastISel-style. This
allows PHI nodes to be updated correctly while using FastISel.

This also involves some code reorganization; ValueMap and
MBBMap are now members of the FastISel class, so they needn't
be passed around explicitly anymore. Also, SelectInstructions
is changed to SelectInstruction, and only does one instruction
at a time.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55746 91177308-0d34-0410-b5e6-96231b3b80d8

Update inline threshold for current function if the notes say, optimize for size.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55745 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a bug that prevented PRE from applying in some cases.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55744 91177308-0d34-0410-b5e6-96231b3b80d8

Avoid extra comma.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55742 91177308-0d34-0410-b5e6-96231b3b80d8

Parse and print opt_size note.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55740 91177308-0d34-0410-b5e6-96231b3b80d8

Initial version of a Partial Specialization IPO pass.  It triggers a couple hundred times on 176.gcc.  I don't know the performance impact yet, the heuristic is quite simple still.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55734 91177308-0d34-0410-b5e6-96231b3b80d8