Fix contradictory description of extension requirements for int in int+fp structs for the FP ABI (#89)

Pull request #74 clarified that when an int+fp struct is passed in a
GPR+FPR in the floating point ABI, the integer isn't extended to XLEN
(i.e. it is anyext, just as if it were passed on the stack).
Unfortunately, a previous part of the sentence which mandated
sign/zero-extension was left in. This patch removes the contradictory
information.

Author: asb

riscv-elf.md

@@ -33,7 +33,7 @@ This RISC-V ELF psABI specification document is
  &copy; 2016 Kito Cheng <[email protected]>
  &copy; 2016-2017 Andrew Waterman <[email protected]>
  &copy; 2016-2017 Michael Clark <[email protected]>
- &copy; 2017-2018 Alex Bradbury <[email protected]>
+ &copy; 2017-2019 Alex Bradbury <[email protected]>
  &copy; 2017 David Horner <[email protected]>
  &copy; 2017 Max Nordlund <[email protected]>
  &copy; 2017 Karsten Merker <[email protected]>
@@ -208,12 +208,11 @@ floating-point reals.
 
 A struct containing one floating-point real and one integer (or bitfield), in
 either order, is passed in a floating-point register and an integer register,
-with the integer zero- or sign-extended as though it were a scalar, provided
-the floating-point real is no more than FLEN bits wide and the integer is no
-more than XLEN bits wide, and at least one floating-point argument register
-and at least one integer argument register is available, with the integer
-placed in its argument register without extension to XLEN bits.  Otherwise,
-it is passed according to the integer calling convention.
+without extending the integer to XLEN bits, provided the floating-point real
+is no more than FLEN bits wide and the integer is no more than XLEN bits wide,
+and at least one floating-point argument register and at least one integer
+argument register is available. Otherwise, it is passed according to the
+integer calling convention.
 
 Unions are never flattened and are always passed according to the integer
 calling convention.