Fix off by one error. I misunderstood the comment about killedAt.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134229 91177308-0d34-0410-b5e6-96231b3b80d8

Check the liveinterval, not the kill flag.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134228 91177308-0d34-0410-b5e6-96231b3b80d8

Don't inflate register classes used by inline asm.

The constraints are represented by the register class of the original
virtual register created for the inline asm. If the register class were
included in the operand descriptor, we might be able to do this.

For now, just give up on regclass inflation when inline asm is involved.

No test case, this bug hasn't happened yet.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134226 91177308-0d34-0410-b5e6-96231b3b80d8

Improve Mips back-end's handling of DBG_VALUE.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134224 91177308-0d34-0410-b5e6-96231b3b80d8

Improve constant folding of undef for cmp and select operators.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134223 91177308-0d34-0410-b5e6-96231b3b80d8

Add support for the 'j' immediate constraint. This is conditionalized on
supporting the instruction that the constraint is for 'movw'.

Part of rdar://9119939

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134222 91177308-0d34-0410-b5e6-96231b3b80d8

Improve constant folding of undef for binary operators.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134221 91177308-0d34-0410-b5e6-96231b3b80d8

Add support for the ARM 't' register constraint. And another testcase
for the 'x' register constraint.

Part of rdar://9119939

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134220 91177308-0d34-0410-b5e6-96231b3b80d8

Switch SubtargetFeatures from std::string to StringRef.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134219 91177308-0d34-0410-b5e6-96231b3b80d8

We'll return a null RC by default if we can't match.

Part of rdar://9119939

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134217 91177308-0d34-0410-b5e6-96231b3b80d8

Add 134199 back, but disable the optimization when the second copy is a kill.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134216 91177308-0d34-0410-b5e6-96231b3b80d8

Add support for the 'x' constraint.

Part of rdar://9307836 and rdar://9119939

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134215 91177308-0d34-0410-b5e6-96231b3b80d8

Remove tabs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134212 91177308-0d34-0410-b5e6-96231b3b80d8

Capitalize the unsigned part of the initializer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134211 91177308-0d34-0410-b5e6-96231b3b80d8

Rename Pair to RCPair lacking any better naming ideas.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134210 91177308-0d34-0410-b5e6-96231b3b80d8

Improve comment: Show the register the DWARF label is added to.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134209 91177308-0d34-0410-b5e6-96231b3b80d8

Use the correct registers on X86_64.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134208 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a problem with fast-isel return values introduced in r134018.

We would put the return value from long double functions in the wrong
register.

This fixes gcc.c-torture/execute/conversion.c

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134205 91177308-0d34-0410-b5e6-96231b3b80d8

Refact ARM Thumb1 tMOVr instruction family.

Merge the tMOVr, tMOVgpr2tgpr, tMOVtgpr2gpr, and tMOVgpr2gpr instructions
into tMOVr. There's no need to keep them separate. Giving the tMOVr
instruction the proper GPR register class for its operands is sufficient
to give the register allocator enough information to do the right thing
directly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134204 91177308-0d34-0410-b5e6-96231b3b80d8

Add support for the 'h' constraint.

Part of rdar://9119939

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134203 91177308-0d34-0410-b5e6-96231b3b80d8

Add target a target hook to get the register number used by the compact unwind
encoding for the registers it knows about. Return -1 if it can't handle that
register.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134202 91177308-0d34-0410-b5e6-96231b3b80d8

Revert my previous patch while I debug llvm-gcc bootstrap.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134201 91177308-0d34-0410-b5e6-96231b3b80d8

Add one more comment to the FDE verbose asm output.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134200 91177308-0d34-0410-b5e6-96231b3b80d8

Don't give up on coalescing A and B when we find

A = X
B = X

Instead, proceed as if we had found

A = X
B = A

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134199 91177308-0d34-0410-b5e6-96231b3b80d8

Add a convenience typedef for std::pair<unsigned, const TargetRegisterClass*>.

No functional change.

Part of rdar://9119939

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134198 91177308-0d34-0410-b5e6-96231b3b80d8

Thumb1 register to register MOV instruction is predicable.

Fix a FIXME and allow predication (in Thumb2) for the T1 register to
register MOV instructions. This allows some better codegen with
if-conversion (as seen in the test updates), plus it lays the groundwork
for pseudo-izing the tMOVCC instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134197 91177308-0d34-0410-b5e6-96231b3b80d8

Add comments to the FDE.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134196 91177308-0d34-0410-b5e6-96231b3b80d8

Add more comments to the ASM output for the CIE's "moves".

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134194 91177308-0d34-0410-b5e6-96231b3b80d8

Tweak error messages to match GCC. Should fix gcc.target/i386/pr30848.c

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134193 91177308-0d34-0410-b5e6-96231b3b80d8

Add comments to the ASM output to help understand the compact unwind and CIE tables.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134191 91177308-0d34-0410-b5e6-96231b3b80d8

Create a isFullCopy predicate.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134189 91177308-0d34-0410-b5e6-96231b3b80d8

Add r134057 back, but splice the predecessor after the successors phi
nodes.

Original message:
Let simplify cfg simplify bb with only debug and lifetime intrinsics.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134182 91177308-0d34-0410-b5e6-96231b3b80d8

Pseudo-ize the Thumb tTPsoft instruction.

It's just a call to a special helper function. Get rid of the T2 variant
entirely, as it's identical to the Thumb1 version.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134178 91177308-0d34-0410-b5e6-96231b3b80d8

indvars -disable-iv-rewrite: handle cloning binary operators that cannot overflow.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134177 91177308-0d34-0410-b5e6-96231b3b80d8

Pseudo-ize the t2LDMIA_RET instruction.

It's just a t2LDMIA_UPD instruction with extra codegen properties, so it
doesn't need the encoding information. As a side-benefit, we now correctly
recognize for instruction printing as a 'pop' instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134173 91177308-0d34-0410-b5e6-96231b3b80d8

Pseudo-ize the Thumb tPOP_RET instruction.

It's just a tPOP instruction with additional code-gen properties, so it
doesn't need encoding information.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134172 91177308-0d34-0410-b5e6-96231b3b80d8

Remove dead code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134148 91177308-0d34-0410-b5e6-96231b3b80d8

The enum was moved to ISDOpcodes.h.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134134 91177308-0d34-0410-b5e6-96231b3b80d8

Kill dead code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134131 91177308-0d34-0410-b5e6-96231b3b80d8

Size reducing SP adjusting t2ADDri needs to check predication.

tADDrSPi is not predicable, so we can't size-reduce a t2ADDri to it if the
predicate is anything other than "always."

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134130 91177308-0d34-0410-b5e6-96231b3b80d8

Fix ARMSubtarget feature parsing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134129 91177308-0d34-0410-b5e6-96231b3b80d8

Fix the ridiculous SubtargetFeatures API where it implicitly expects CPU name to
be the first encoded as the first feature. It then uses the CPU name to look up
features / scheduling itineray even though clients know full well the CPU name
being used to query these properties.

The fix is to just have the clients explictly pass the CPU name!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134127 91177308-0d34-0410-b5e6-96231b3b80d8

Recognize the xstorerng alias for VIA PadLock's xstore instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134126 91177308-0d34-0410-b5e6-96231b3b80d8

Reapply r134047 now that the world is ready for it.

This patch will sometimes choose live range split points next to
interference instead of always splitting next to a register point. That
means spill code can now appear almost anywhere, and it was necessary
to fix code that didn't expect that.

The difficult places were:

- Between a CALL returning a value on the x87 stack and the
  corresponding FpPOP_RETVAL (was FpGET_ST0). Probably also near x87
  inline assembly, but that didn't actually show up in testing.

- Between a CALL popping arguments off the stack and the corresponding
  ADJCALLSTACKUP.

Both are fixed now. The only place spill code can't appear is after
terminators, see SplitAnalysis::getLastSplitPoint.

Original commit message:

Rewrite RAGreedy::splitAroundRegion, now with cool ASCII art.

This function has to deal with a lot of special cases, and the old
version got it wrong sometimes. In particular, it would sometimes leave
multiple uses in the stack interval in a single block. That causes bad
code with multiple reloads in the same basic block.

The new version handles block entry and exit in a single pass. It first
eliminates all the easy cases, and then goes on to create a local
interval for the blocks with difficult interference. Previously, we
would only create the local interval for completely isolated blocks.

It can happen that the stack interval becomes completely empty because
we could allocate a register in all edge bundles, and the new local
intervals deal with the interference. The empty stack interval is
harmless, but we need to remove a SplitKit assertion that checks for
empty intervals.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134125 91177308-0d34-0410-b5e6-96231b3b80d8

indvars -disable-iv-rewrite: handle an edge case involving identity phis.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134124 91177308-0d34-0410-b5e6-96231b3b80d8

Remove getRegClassForInlineAsmConstraint and all dependencies.

Fixes rdar://9643582

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134123 91177308-0d34-0410-b5e6-96231b3b80d8

Make sure we use the correct register class here since we'll need to
care about spill values.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134122 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a small thinko for constant i64 lock/orq optimization where we
we didn't have an opcode for 64-bit constant or expressions.

Fixes rdar://9692967

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134121 91177308-0d34-0410-b5e6-96231b3b80d8

* Use the proper size to output the range size.
* Rough in the compact encoding part.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134119 91177308-0d34-0410-b5e6-96231b3b80d8

Stupid error: If the LSDA and Personality functions aren't there, emit 0 instead
of the encoding.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134117 91177308-0d34-0410-b5e6-96231b3b80d8

Revert r133953 for now.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134116 91177308-0d34-0410-b5e6-96231b3b80d8

We don't want to use relocations inside the compact unwind section. Just use the
symbols instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134115 91177308-0d34-0410-b5e6-96231b3b80d8

Remove redundant Thumb2 ADD/SUB SP instruction definitions.

Unlike Thumb1, Thumb2 does not have dedicated encodings for adjusting the
stack pointer. It can just use the normal add-register-immediate encoding
since it can use all registers as a source, not just R0-R7. The extra
instruction definitions are just duplicates of the normal instructions with
the (not well enforced) constraint that the source register was SP.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134114 91177308-0d34-0410-b5e6-96231b3b80d8

Always adjust the stack pointer immediately after the call.

Some x86-32 calls pop values off the stack, and we need to readjust the
stack pointer after the call. This happens when ADJCALLSTACKUP is
eliminated.

It could happen that spill code was inserted between the CALL and
ADJCALLSTACKUP instructions, and we would compute wrong stack pointer
offsets for those frame index references.

Fix this by inserting the stack pointer adjustment immediately after the
call instead of where the ADJCALLSTACKUP instruction was erased.

I don't have a test case since we don't currently insert code in that
position. We will soon, though. I am testing a regalloc patch that
didn't work on Linux because of this.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134113 91177308-0d34-0410-b5e6-96231b3b80d8

indvars -disable-iv-rewrite: insert new trunc instructions carefully.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134112 91177308-0d34-0410-b5e6-96231b3b80d8

Added IRBuilder::SetInsertPoint(Use) to find a valid insertion point
that dominates the given Use.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134111 91177308-0d34-0410-b5e6-96231b3b80d8

whitespace

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134110 91177308-0d34-0410-b5e6-96231b3b80d8

In the ARM global merging pass, allow extraneous alignment specifiers. This pass
already makes the assumption, which is correct on ARM, that a type's alignment is
less than its alloc size. This improves codegen with Clang (which inserts a lot of
extraneous alignment specifiers) and fixes <rdar://problem/9695089>.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134106 91177308-0d34-0410-b5e6-96231b3b80d8

ARM RSCS* don't need explicit TableGen decoder checks.

They've been pseudos for a while now, so the decoder will never see
them in the first place.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134101 91177308-0d34-0410-b5e6-96231b3b80d8

Indentation

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134100 91177308-0d34-0410-b5e6-96231b3b80d8

Remove getRegClassForInlineAsmConstraint from the ARM port.

Part of rdar://9643582

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134095 91177308-0d34-0410-b5e6-96231b3b80d8

Remove todo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134094 91177308-0d34-0410-b5e6-96231b3b80d8

make compose and isMoveInstr static functions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134093 91177308-0d34-0410-b5e6-96231b3b80d8

Refactor away tSpill and tRestore pseudos in ARM backend.

The tSpill and tRestore instructions are just copies of the tSTRspi and
tLDRspi instructions, respectively. Just use those directly instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134092 91177308-0d34-0410-b5e6-96231b3b80d8

Add a TODO for the Alpha port inline asm constraints.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134089 91177308-0d34-0410-b5e6-96231b3b80d8

Move Alpha from getRegClassForInlineAsmConstraint to
getRegForInlineAsmConstraint.

Part of rdar://9643582

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134088 91177308-0d34-0410-b5e6-96231b3b80d8

Update comment for getRegForInlineAsmConstraint for Mips.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134087 91177308-0d34-0410-b5e6-96231b3b80d8

Move the Blackfin port away from getRegClassForInlineAsmConstraint by
creating a few specific register classes.

Part of rdar://9643582

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134086 91177308-0d34-0410-b5e6-96231b3b80d8

Remove getRegClassForInlineAsmConstraint from MBlaze. Add a TODO comment
for the port.

Part of rdar://9643582

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134085 91177308-0d34-0410-b5e6-96231b3b80d8

Remove getRegClassForInlineAsmConstraint for Mips.

Part of rdar://9643582

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134084 91177308-0d34-0410-b5e6-96231b3b80d8

Remove getRegClassForInlineAsmConstraint from sparc.

Part of rdar://9643582

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134083 91177308-0d34-0410-b5e6-96231b3b80d8

Move XCore from getRegClassForInlineAsmConstraint to
getRegForInlineAsmConstraint.

Part of rdar://9643582

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134080 91177308-0d34-0410-b5e6-96231b3b80d8

Use getRegForInlineAsmConstraint instead of custom defining regclasses
via vectors.

Part of rdar://9643582

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134079 91177308-0d34-0410-b5e6-96231b3b80d8

Add missing operand. rdar://9694169

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134078 91177308-0d34-0410-b5e6-96231b3b80d8

Temporarily revert r134057: "Let simplify cfg simplify bb with only debug and
lifetime intrinsics" due to buildbot failures.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134071 91177308-0d34-0410-b5e6-96231b3b80d8

Asm parser range checking on .<size> <value> directives.

For example, ".byte 256" would previously assert() when emitting an object
file. Now it generates a diagnostic that the literal value is out of range.

rdar://9686950

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134069 91177308-0d34-0410-b5e6-96231b3b80d8

Don't depend on the optimization reverted in r134067.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134068 91177308-0d34-0410-b5e6-96231b3b80d8

Revert a part of r126557 which could create unschedulable DAGs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134067 91177308-0d34-0410-b5e6-96231b3b80d8

Change AsmName's type from StringRef to std::string. AsmName was pointing to a temporary string object that was destroyed. This is undefined behavior and MSVC didn't like it.

This fixes over 300+ failing tests on MSVC.
Credit for this fix goes to chapuni.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134064 91177308-0d34-0410-b5e6-96231b3b80d8

Let simplify cfg simplify bb with only debug and lifetime intrinsics.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134057 91177308-0d34-0410-b5e6-96231b3b80d8

Fix CMake build.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134055 91177308-0d34-0410-b5e6-96231b3b80d8

indvars -disable-iv-rewrite: just because SCEV ignores casts doesn't
mean they can be removed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134054 91177308-0d34-0410-b5e6-96231b3b80d8

Revert r134047 while investigating a llvm-gcc-i386-linux-selfhost
miscompile.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134053 91177308-0d34-0410-b5e6-96231b3b80d8

Sink SubtargetFeature and TargetInstrItineraries (renamed MCInstrItineraries) into MC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134049 91177308-0d34-0410-b5e6-96231b3b80d8

Trim include

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134048 91177308-0d34-0410-b5e6-96231b3b80d8

Rewrite RAGreedy::splitAroundRegion, now with cool ASCII art.

This function has to deal with a lot of special cases, and the old
version got it wrong sometimes. In particular, it would sometimes leave
multiple uses in the stack interval in a single block. That causes bad
code with multiple reloads in the same basic block.

The new version handles block entry and exit in a single pass. It first
eliminates all the easy cases, and then goes on to create a local
interval for the blocks with difficult interference. Previously, we
would only create the local interval for completely isolated blocks.

It can happen that the stack interval becomes completely empty because
we could allocate a register in all edge bundles, and the new local
intervals deal with the interference. The empty stack interval is
harmless, but we need to remove a SplitKit assertion that checks for
empty intervals.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134047 91177308-0d34-0410-b5e6-96231b3b80d8

build/Darwin: Fix epic typo fail.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134036 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a CMake warning, following up to r134008.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134033 91177308-0d34-0410-b5e6-96231b3b80d8

Unbreak every backend.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134031 91177308-0d34-0410-b5e6-96231b3b80d8

Move CallFrameSetupOpcode and CallFrameDestroyOpcode to TargetInstrInfo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134030 91177308-0d34-0410-b5e6-96231b3b80d8

Hide more details in tablegen generated MCRegisterInfo ctor function.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134027 91177308-0d34-0410-b5e6-96231b3b80d8

Add MCInstrInfo registeration machinery.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134026 91177308-0d34-0410-b5e6-96231b3b80d8

Merge XXXGenRegisterNames.inc into XXXGenRegisterInfo.inc

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134024 91177308-0d34-0410-b5e6-96231b3b80d8

- Rename TargetInstrDesc, TargetOperandInfo to MCInstrDesc and MCOperandInfo and
sink them into MC layer.
- Added MCInstrInfo, which captures the tablegen generated static data. Chang
TargetInstrInfo so it's based off MCInstrInfo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134021 91177308-0d34-0410-b5e6-96231b3b80d8

Clean up the handling of the x87 fp stack to make it more robust.

Drop the FpMov instructions, use plain COPY instead.

Drop the FpSET/GET instruction for accessing fixed stack positions.
Instead use normal COPY to/from ST registers around inline assembly, and
provide a single new FpPOP_RETVAL instruction that can access the return
value(s) from a call. This is still necessary since you cannot tell from
the CALL instruction alone if it returns anything on the FP stack. Teach
fast isel to use this.

This provides a much more robust way of handling fixed stack registers -
we can tolerate arbitrary FP stack instructions inserted around calls
and inline assembly. Live range splitting could sometimes break x87 code
by inserting spill code in unfortunate places.

As a bonus we handle floating point inline assembly correctly now.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134018 91177308-0d34-0410-b5e6-96231b3b80d8

Reinstate r133516 "Remove some unnecessary uses of c_str()."  A trailing null
character in std::string was causing failures for a few ObjC and Obj-C++ tests
when -flto was enabled.  Revision 133999 resolved this issue. Thanks Jay!
rdar://9685235
PR10210

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134017 91177308-0d34-0410-b5e6-96231b3b80d8

Remove warning: 'c0' may be used uninitialized in this function.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134014 91177308-0d34-0410-b5e6-96231b3b80d8

Print registers by name instead of by number.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134013 91177308-0d34-0410-b5e6-96231b3b80d8

cleanup: misleading comment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134010 91177308-0d34-0410-b5e6-96231b3b80d8

cmake: Our MSVC build does not support config-time build mode.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134008 91177308-0d34-0410-b5e6-96231b3b80d8