[PATCH 0/3] RISC-V: gdb.base/gnu_vector fixes.

[Jim Wilson <jimw@sifive.com>]

This set of patches fixes these infcall related failures on a rv64gc
linux system.
FAIL: gdb.base/gnu_vector.exp: call add_many_charvecs
FAIL: gdb.base/gnu_vector.exp: call add_various_floatvecs

The first one fails because we have 2 4-byte vector args passed in the
same 8-byte stack slot.  The code is failing round up the address of a
stack slot when an arg is smaller than the stack slot size.

The second one fails because of two problems.  The 16-byte vector of
floats is given 4 byte alignment when it should have 16 byte
alignment. This is because there is no support for vector types.  And
unnamed args that require twice XLEN alignment are supposed to be
passed in aligned register pairs.  The support for this is also
missing.

This was tested on a rv64gc linux system with the gdb testsuite, and
it fixes 2 failures without causing any regressions.

Jim