documentation for operand constraints/patterns?

[David Mosberger <davidm at napali dot hpl dot hp dot com>]

Is anybody maintaining documentation for the operand constraints and
patterns available for inline assembly?  The reason I'm asking is that
I just had another person fall into the trap of not using the %P
pattern on ia64 when the "m" constraint was used.  Such errors are
rather silly and timeconsuming.

If there is no interest in maintaining such a document in general,
would someone at least be interested in doing so for ia64?  As a
starting point, I listed the stuff I'm aware of below.

	--david

* Operand constraints:

`b'	branch register
`c'	predicate register ('c' as in "conditional")
`d'	application register residing in M-unit
`e'	application register residing in I-unit
`f'	floating-point register
`r'	general register
`m'	memory address (must %Pn to get post-increment/decrement!)
`o'	offsetable-only memory address (use when post-increment/decrement
	cannot be used)
'G'	float constant 0.0 or 1.0
`I'	14-bit signed integer constant
`J'	22-bit signed integer constant
`K'	8-bit signed integer constant for logical instructions
`L'	8-bit adjusted signed integer constant for compare pseudo-ops
`M'	6-bit unsigned integer constant for shift counts
`N'	9-bit signed integer constant for load/store post-increments
`O'	the constant zero.
`P'	0 or -1 for dep instruction

* Operand printing formats:

C	Swap and print a comparison operator.
D	Print an FP comparison operator.
E	Print 32 - constant, for SImode shifts as extract.
e	Print 64 - constant, for DImode rotates.
F	A floating point constant 0.0 emitted as f0, or 1.0 emitted as f1, or
	a floating point register emitted normally.
I	Invert a predicate register by adding 1.
J	Select the proper predicate register for a condition.
j	Select the inverse predicate register for a condition.
O	Append .acq for volatile load.
P	Postincrement of a MEM.
Q	Append .rel for volatile store.
S	Shift amount for shladd instruction.
T	Print an 8-bit sign extended number (K) as a 32-bit unsigned number
	for Intel assembler.
U	Print an 8-bit sign extended number (K) as a 64-bit unsigned number
	for Intel assembler.
r	Print register name, or constant 0 as r0.  HP compatibility for
	Linux kernel.

+	Print branch-prediction qualifier (.spnt, .dpnt, .sptk, spnt) based
	on branch likelihood.

,	Print parenthesized predicate of a predicated instruction.