Currently, the custom SGPR spill lowering pass spills
SGPRs into physical VGPR lanes and the remaining VGPRs
are used by regalloc for vector regclass allocation.
This imposes many restrictions that we ended up with
unsuccessful SGPR spilling when there won't be enough
VGPRs and we are forced to spill the leftover into
memory during PEI. The custom spill handling during PEI
has many edge cases and often breaks the compiler time
to time.
This patch implements spilling SGPRs into virtual VGPR
lanes. Since we now split the register allocation for
SGPRs and VGPRs, the virtual registers introduced for
the spill lanes would get allocated automatically in
the subsequent regalloc invocation for VGPRs.
Spill to virtual registers will always be successful,
even in the high-pressure situations, and hence it avoids
most of the edge cases during PEI. We are now left with
only the custom SGPR spills during PEI for special registers
like the frame pointer which is an unproblematic case.
By spilling CSRs into virtual VGPR lanes, we might end up
with broken CFIs that can potentially corrupt the frame
unwinding in the debugger causing either a crash or a
terrible debugging experience. This occurs when regalloc
tries to spill or split the liverange of these virtual VGPRs.
The CFIs should also be inserted at these intermediate
points to correctly propagate the CFI entries. It is not
currently implemented in the compiler. As a short-term fix,
we continue to spill CSR SGPRs into physical VGPR lanes for
the debugger to correctly compute the unwind information.