s390 TLS incorporated

s390 TLS incorporated

[Ulrich Drepper]

I've added Martin's TLS stuff to the TLS document.  Version 0.18
includes it.  Please check it out.  A _lot_ of editing was necessary so
I might have screwed something up.

And on more thing: it would be good if the code sequences would get
pseudo addresses just for the other archs.  Otherwise it's not
necessarily obvious how long an instruction has to be.

-- 
--------------.                        ,-.            444 Castro Street
Ulrich Drepper \    ,-----------------'   \ Mountain View, CA 94041 USA
Red Hat         `--' drepper at redhat.com `---------------------------

Re: s390 TLS incorporated

[Ulrich Drepper]

Martin Schwidefsky wrote:

> Sorry to hear that. I had hoped to make your life easier by using
> latex for the writeup. Seems like I was mistaken.

Not really.  It always is  a lot of work.


> I proof-read the s390 stuff. I found some minor things:

Thanks, I've changed fixed these problems.  The new version is already
uploaded.


> I thought about using address but decided against them because the
> four/three code pieces of the examples are supposed to get scheduled
> by the compiler.

Yes, but this could be explained in a comment.  This is why I left the
horizontal lines in.  These would not be real addresses, just some guidance.

-- 
--------------.                        ,-.            444 Castro Street
Ulrich Drepper \    ,-----------------'   \ Mountain View, CA 94041 USA
Red Hat         `--' drepper at redhat.com `---------------------------

Re: s390 TLS incorporated

[Martin Schwidefsky]

> I've added Martin's TLS stuff to the TLS document.  Version 0.18
> includes it.  Please check it out.  A _lot_ of editing was necessary so
> I might have screwed something up.
Sorry to hear that. I had hoped to make your life easier by using
latex for the writeup. Seems like I was mistaken.

I proof-read the s390 stuff. I found some minor things:
* on page 41 the ear %r7,%a0 isn't indented correctly
* on page 42 a vertical separation line is missing
* on page 49 there is some irregular spacing in the code example
* on page 64 there is more irregular spacing in the GD->IE transition
* and finally on page 79 in the revision history another nice misspelling
  of my name ;-)
I found no problem with the content of the s390 pieces.

> And on more thing: it would be good if the code sequences would get
> pseudo addresses just for the other archs.  Otherwise it's not
> necessarily obvious how long an instruction has to be.
I thought about using address but decided against them because the
four/three code pieces of the examples are supposed to get scheduled
by the compiler. They in fact are never consecutive in memory because
the literal pool can not directly follow the instructions. As alternative
the size of the instructions could be added in a seperate column.

blue skies,
   Martin