RE: [oglbase-discuss] OK, back to Re: 2nd draft:GL_EXT_get_proc_address spec

[John R. Dennis]

> I'll defer to the consensus view on this matter, but my preference
> is for the static offsets.

> Second, I'd quibble with John's cost estimates.

> I'm willing to pay the extra one or two instructions...

> As I said, although I prefer the static offset implementation...

I think my proposal did indeed have the property of static offset for ALL
functions, there would be no use of a run-time offset. My digression over the
minor cost of a dynamic offset was perhaps a red-herring and unneccessarily
confusing to the discussion. I only wanted to make the case that if a dynamic
offset was used it would not be a terrible cost.

Allen's pseudo assembly code closely matches what I thought would be needed,
where we differ is the overall performance impact of the extra 1-2 machine
instructions and memory reference per OGL function. My belief is that any given
OGL function will require enough machine instructions to execute that the
overhead of an additional 1-2 instructions would be amortized and negligble. Of
course the possible exception are the vertex functions which are heavily
utilized and often perform very little work, for these the cost would be
measureable. But like I said, I don't even think you need a dynamic offset.

John