From:	MERC::"uunet!WKUVX1.BITNET!MacroMan" 14-APR-1993 17:53:14.76
To:	MACRO32@WKUVX1.BITNET
CC:	
Subj:	Re: more alpha macro problems.


In article <01GX07X8BFHY9352KJ@kopc.hhs.dk>, Arne Vajhxj <ARNE@kopc.hhs.dk>
 writes:
|>>     .ascid  /SHARE_DATA/
|>>
|>> map_range:
|>>     .address share_data
|>>     .address share_data+<512*3>
|>                           ^^^
|>> screen_buf:
|>>     .blkb   508     ; buffer containing screen update
|>>  = . + <512*4> - < . - share_data >
|>          ^^^
|>
|>>     $crmpsc_s inadr=map_range, flags=#sec$m_gbl!sec$m_wrt!sec$m_pagfil, -
|>>         gsdnam=map_name, -
|>>         pagcnt=#4       ; note no use of retaddr?
|>          ^^^^^^
|>
|>
|>Pages are not 512 bytes on an ALPHA !
|>
|>On the current versions of the ALPHA a page is 8192 byte, but DEC has
|>announced, that they may increase this later on !
|>
|>So a carefull programmer will retrieve the appropriate value (any methods
|>smarter than $GETSYI with SYI$_PAGE_SIZE ?).
|>
|>

While it is true that an Alpha page is not 512 bytes, that's not the
source of his ACCVIO.  On Alpha, to quote from the system service
reference manual:

"On Alpha VMS, the addresses must be aligned on CPU-specific pages
(unless the SEC$M_EXPREG flag is set); no rounding to CPU-specific
pages occurs.  The lower address of the _inadr_ argument must be on
a CPU-specific page boundary and the higher address of the _inadr_
argument must be 1 less than a CPU-specific boundary, thus forming
a range from lowest to highest address bytes.  You may use the
SYI$_PAGE_SIZEitem code in the $GETSYIsystem service to set the
_inadr_ argument to the proper values."

Further, the pagcnt argument is a count of pagelets (512-byte chunks)
on Alpha so his usage in the $CRMPSC call is correct.

Basically, to fix the problem, you need to to align SHARE_DATA on a
CPU-specific page boundary.  You also really want to read the whole
description of the $CRMPSC system service.  If you have it no other
place, it is on the V1.0 CD in the directory [DECW$BOOK].

-R
--
-----------------------------------------------------------------------------
Richard Critz                                         critz@star.enet.dec.com
      -- I speak only for myself; I make no commitments for Digital. --
-----------------------------------------------------------------------------