S/390 Assembler---how much new for Application programmers? 
Author Message
 S/390 Assembler---how much new for Application programmers?

I see the new S/390 instruction set is greatly expanded, but how many
instructions are of interest to run-of-the-mill application programmers
(as opposed to system or specialty product programmers)?



Mon, 30 Aug 1999 03:00:00 GMT  
 S/390 Assembler---how much new for Application programmers?

Quote:

> I see the new S/390 instruction set is greatly expanded, but how many
> instructions are of interest to run-of-the-mill application programmers
> (as opposed to system or specialty product programmers)?

"Nothing", as far as our shop is concerned.  I'd have liked to see LHL
(Load Halfword Logical) and AHL/SHL (Add/Subtract Halfword Logical)
added, because once in a while a table element's offset is more than 32K
away.  :-)

John Chase
Systems Engineer
NOTIS Support Center



Mon, 30 Aug 1999 03:00:00 GMT  
 S/390 Assembler---how much new for Application programmers?

Quote:


> > I see the new S/390 instruction set is greatly expanded, but how many
> > instructions are of interest to run-of-the-mill application programmers
> > (as opposed to system or specialty product programmers)?

> "Nothing", as far as our shop is concerned.  I'd have liked to see LHL
> (Load Halfword Logical) and AHL/SHL (Add/Subtract Halfword Logical)
> added, because once in a while a table element's offset is more than 32K
> away.  :-)

> John Chase
> Systems Engineer
> NOTIS Support Center

--

Just use an ICM to ensure sign bit isn't carried.

As far as new instructions, there a few cool ones, especially in the
area of handling strings, and built it tree support. It's a pitty there
implemented in firmware versus hardwired.

Check out the following site and you can see for yourself.
http://www3.pok.ibm.com/os390/bkserv/index.shtml



Tue, 31 Aug 1999 03:00:00 GMT  
 S/390 Assembler---how much new for Application programmers?

I wonder whther IBM itself uses them, for example in the code emitted by
their compilers?

Jonathan


Quote:
> I see the new S/390 instruction set is greatly expanded, but how many
> instructions are of interest to run-of-the-mill application programmers
> (as opposed to system or specialty product programmers)?



Tue, 31 Aug 1999 03:00:00 GMT  
 S/390 Assembler---how much new for Application programmers?

Quote:
>As far as new instructions, there a few cool ones, especially in the
>area of handling strings, and built it tree support. It's a pitty there
>implemented in firmware versus hardwired.

I'm not sure I understand this comment.  All IBM mainframes have been
microcoded for decades.  Am I missing something?


Tue, 31 Aug 1999 03:00:00 GMT  
 S/390 Assembler---how much new for Application programmers?

The implementation of assembler instructions in firmware was one of the
original features of the IBM 360 series. None of them (with the possible
exception of the 44) was hardwired.

Jonathan


Quote:
> >As far as new instructions, there a few cool ones, especially in the
> >area of handling strings, and built it tree support. It's a pitty there
> >implemented in firmware versus hardwired.

> I'm not sure I understand this comment.  All IBM mainframes have been
> microcoded for decades.  Am I missing something?



Tue, 31 Aug 1999 03:00:00 GMT  
 S/390 Assembler---how much new for Application programmers?

Quote:


> > "Nothing", as far as our shop is concerned.  I'd have liked to see LHL
> > (Load Halfword Logical) and AHL/SHL (Add/Subtract Halfword Logical)
> > added, because once in a while a table element's offset is more than 32K
> > away.  :-)

> Just use an ICM to ensure sign bit isn't carried.

Still need an additional instruction to ensure the "top half" of the
register is zeroed out; e.g.:

       SR    Ra,Ra
       ICM   Ra,3,HWORD
or
       ICM   Ra,12,HWORD  (Could also use L   Ra,HWORD here*)
       SRL   Ra,16
(* Provided that the two bytes following HWORD are not fetch-protected
:-)  )

Even more useful would be the AHL/SHL when HWORD contains a value GT
+32767:

NOW    LA    Ra,ORIGIN
       SR    Rb,Rb
       ICM   Rb,3,OFFSET
       ALR   Ra,Rb
       ...
vs.
NEW    LA    Ra,ORIGIN
       AHL   Ra,OFFSET
       ...

Presently need two extra instructions AND an extra register, and it's
not getting any easier to FIND that extra register.  I realize that,
nowadays, storage is "cheap", but re-architecting a mature,
sophisticated software system to use fullword offsets isn't.

John Chase
Systems Engineer
NOTIS Support Center



Tue, 31 Aug 1999 03:00:00 GMT  
 S/390 Assembler---how much new for Application programmers?


Quote:
>The implementation of assembler instructions in firmware was one of the
>original features of the IBM 360 series. None of them (with the possible
>exception of the 44) was hardwired.

I think the 44, 75, 91, 95 and 195 were hardware implementations.  


Tue, 31 Aug 1999 03:00:00 GMT  
 S/390 Assembler---how much new for Application programmers?

Quote:
> As far as new instructions, there a few cool ones, especially in the
> area of handling strings, and built it tree support. It's a pitty there
> implemented in firmware versus hardwired.

What difference does it make between firmware and hardware?  I thought the
whole machine was done in firmware, going back to S/360 days and microcode
on metallic punch cards.


Wed, 01 Sep 1999 03:00:00 GMT  
 S/390 Assembler---how much new for Application programmers?

No chance - at least at this time, since the compilers support iron back to 3083 / 43xx technology.

Rex Widmer
Builder of software archeology tools and other strange programs to help survive in a legacy based world.



Fri, 03 Sep 1999 03:00:00 GMT  
 S/390 Assembler---how much new for Application programmers?


: > > "Nothing", as far as our shop is concerned.  I'd have liked to see LHL
: > > (Load Halfword Logical) and AHL/SHL (Add/Subtract Halfword Logical)
: > > added, because once in a while a table element's offset is more than 32K
: > > away.  :-)
: > Just use an ICM to ensure sign bit isn't carried.
: Still need an additional instruction... SR    Ra,Ra
That's all.

So I don't see much need for an "add unsigned short" instruction.
Especially since it could not be an RX, only RS spots are left in
the opcode space!

Futhermore, there is no immutable law that says the index to a
table element has to be an unsigned short int (16 bits).
An index defined as a signed short int (16 bits) is just fine.
Just a little more messy at init time.  However, all the bulk
of the runtime code can be left as-is when upgrading from
the halfword indexes which had been restricted to a 32KB range,
now expanded to a 64KB range.

Historical case c. 1980:  We were changing the memory manager in
a TP monitor, so init was changing a lot anyway.  This monitor
uses pools of hot 16- and 20-byte control blocks for IO events
and dataset buffer tracking.  There is a SYS header block of
maybe a hundred bytes, followed by the pool spaces for the
volatile elements.  There are several chain headers in SYS
and various single- and double-linked lists thru the elements.
Stess the word "various".

All these pointers are HW in all directions.  The kernel code
(we didn't call it that ;-)  is riddled with indexed LH/STH
instructions.  (The SYS base is implicite, and many instructions
also specify an explicite index reg.)

[brag flame shield on] I changed init to calculate the desired
size of the SYS pool before initing it.  If it exceded 32K,
the SYS header was built exactly 32K from the high end.  Up to
32K of pool space was inited in front of SYS.  [brag off ;-)]
The extra free elements were chained with the rest into the free
pools at init time.  Runtime code was not changed.  +$$$

Going back to messing with LH versus ICM 3, or whatever:
Note well that doing such an effort only gives you approx
a 2X increase in capacity.  I think a 3X factor of scaling
is an engineering rule of thumb... else chances of justifying
development costs are poor.  The 2X will just fill up a year
later. ;-]
--                                  __  __

:-)    Search Service Development  \_\/_/\/  Office:  B5/F4/MA/54S05
The opinions I assert are opinion.  \_\_\/\  (513) 865-6800 x-4733
Home:  682 Deerfield Dr.            /_/_/\/  Lexis-Nexis, POB 933
Harrison, OH 45030  (513) 367-5653  \_\_\/   Dayton, OH  45401



Sat, 04 Sep 1999 03:00:00 GMT  
 
 [ 17 post ]  Go to page: [1] [2]

 Relevant Pages 

1. Need title for new OS/390 applications programmer course

2. Need title for new OS/390 applications programmer course

3. Help Application Development in IBM 390 Assemblers.

4. New addresssing method in S/390 assembler

5. Need title for new OS/390 applications programming class

6. For all Systems/390 PL/I programmers and system-programmers

7. S/390 assembler instruction reference

8. Palm Pilot app for Assembler 370/390 programemrs

9. AS/400 Assembler like S/360-390

10. Position available: s/390 Assembler prog/analyst

11. Definitive 370/390 assembler book?

12. BNF for 370/390 assembler ?

 

 
Powered by phpBB® Forum Software