release_33 branch: proposing r185594, r185615, r185616 and r185617

[Tom Stellard]

On Mon, Jul 08, 2013 at 09:55:47PM +0200, Dimitry Andric wrote:
> Hi,
> 
> For the 3.3 release branch, I would like to propose merging r185594, r185615, r185616 and r185617:
> 
> r185594 from llvm trunk (by Jakob Stoklund Olesen):
> 
>   Add MachineBasicBlock::addLiveIn().
> 
>   This function adds a live-in physical register to an MBB and ensures
>   that it is copied to a virtual register immediately.
> 

This patch adds a public function which I think will break the ABI, and
it was decided that stable releases shouldn't break the ABI.

> r185615 from llvm trunk (by Jakob Stoklund Olesen):
> 
>   Live-in copies go *after* EH_LABELs.
> 
>   This will soon be tested by exception handling working at all.
> 
> r185616 from llvm trunk (by Jakob Stoklund Olesen):
> 
>   FastISel can only append to basic blocks.
> 
>   Compute the insertion point from the end of the basic block instead of
>   skipping labels from the front.
> 
>   This caused failures in landing pads when live-in copies where inserted
>   before instruction selection.
> 
> r185617 from llvm trunk (by Jakob Stoklund Olesen):
> 
>   Simplify landing pad lowering.
> 
>   Stop using the ISD::EXCEPTIONADDR and ISD::EHSELECTION when lowering
>   landing pad arguments. These nodes were previously legalized into
>   CopyFromReg nodes, but that never worked properly because the
>   CopyFromReg node weren't guaranteed to be  scheduled at the top of the
>   basic block.
> 
>   This meant the exception pointer and selector registers could be
>   clobbered before being copied to a virtual register.
> 
>   This patch copies the two physical registers to virtual registers at
>   the beginning of the basic block, and lowers the landingpad instruction
>   directly to two CopyFromReg nodes reading the *virtual* registers. This
>   is safe because virtual registers don't get clobbered.
> 
>   A future patch will remove the ISD::EXCEPTIONADDR and ISD::EHSELECTION
>   nodes.
> 

This patch adds two new public members which I also think will break the
ABI.  I don't think I can accept r185594 and r185617 unless you can either:
1. Rework them in a way that doesn't break the ABI.
2. Revive the discussion about ABI compatibility of stable release and
   get a general consensus that breaking the ABI is OK.  The original
   discussion is here:
   https://groups.google.com/forum/#!msg/llvm-dev/NLxx_8SIaHI/5aNecjnpeXEJ

-Tom

> Together, these changes fix llvm PR 16038 ('qt4 webcore file results in "Bad machine code: Using an undefined physical register"'), and should make it possible again to compile the FreeBSD www/qt4-webkit port again on the i386 arch.
> 
> Only r185594, r185615 and r185616 apply to the release_33 branch without modification, so I have attached a modified version of r186617 that does apply.
> 
> -Dimitry