A successful port to Cray T3E.

[Randy Terbush <ra...@Covalent.NET>]

This one is for Dean. :-)

Re: A successful port to Cray T3E.

[Dean Gaudet <dg...@arctic.org>]

On Wed, 26 Aug 1998, Brian Behlendorf wrote:

> This looks fine to me, +1, except that I don't know if the change to "short
> int" is valid (it well could be, I just don't know).

No, the short int change is invalid.  It would be 16-bits on pretty much
every other architecture out there with that change... and that's wrong,
it's supposed to be 32-bits. 

Dean

Re: A successful port to Cray T3E.

[Brian Behlendorf <br...@hyperreal.org>]

This looks fine to me, +1, except that I don't know if the change to "short
int" is valid (it well could be, I just don't know).

	Brian

At 04:42 PM 8/14/98 -0500, Randy Terbush wrote:
>
>This one is for Dean. :-)
>
>
>From: John Van Bockel <jv...@server.ppco.com>
>Message-Id: <19...@server.ppco.com>
>Subject: A successful port to Cray T3E.
>To: Apache@Apache.Org
>Date: Fri, 14 Aug 1998 15:31:19 -0500 (CDT)
>Content-Type: text
>
>Greetings,
>
>I recently compiled and installed Apache 1.3.1 on
>a Cray T3E running UNICOSmk 2.0.3.  Apache is running
>like a champ and we are very pleased.
>
>It's entirely possible that a more capable person than
>I could tweek the pre-compilation options and get
>better performance.  However, as is, it's very fast
>and our needs are very low volume.
>
>The following is a collection of diffs for the four
>files which were altered.  The ease by which Apache
>was built on this unusual OS is a testimony to the
>outstanding efforts of the Apache group.  Many thanks.
>
>                                John J. Van Bockel
>                            JVBockel@SprintParanet.com
>
>
>#
>#   Pretty standard stuff here.
>#
>*** src/helpers/GuessOS
>63a64,66
>>     *:*:unicosmk:CRAY\ T3E)
>> 	echo "T3E-Cray-UNICOSmk"; exit 0
>> 	;;
>
>
>#
>#   UNICOSmk has neither the BSD mmap or SysV IPC shmget.  I
>#   would guess that this is do to a lack of cache coherency.
>#
>#   Also, the offsetof used to create XtOffsetOf is not
>#   defined appropriately in stddef.h.
>#
>*** src/include/ap_config.h
>108a109,119
>> #elif defined(UNICOSmk)
>> #define HAVE_CRYPY_H
>> #define HAVE_SYS_SELECT_H
>> #define NO_MMAP
>> #define NO_SHMGET
>> #define NO_WRITEV
>> #define SYS_SIGLIST _sys_siglist
>> #define USE_FLOCK_SERIALIZED_ACCEPT
>> #include <stddef.h>
>> #undef HAVE_SYS_RESOURCE_H
>> 
>1047a1059,1062
>> #endif
>> 
>> #if defined(UNICOSmk)
>> #define XtOffsetOf(s_type,field) (size_t)(&(((s_type *)0)->field))
>
>
>#
>#   An "unsigned int" is 8 bytes under UNICOSmk.  UINT4
>#   must be 4 bytes as provided by "unsigned short int".
>#
>*** src/include/ap_md5.h
>98c98
>< typedef unsigned int UINT4;
>---
>> typedef unsigned short int UINT4;
>
>
>#
>#   Pretty standard stuff here.
>#
>*** src/Configure
>286a287,291
>>     T3E-Cray-UNICOSmk)
>> 	OS='UNICOSmk'
>> 	CFLAGS="$CFLAGS -DUNICOSmk"
>> 	DEF_WANTHSREGEX=yes
>> 	;;
> 
--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
"Common sense is the collection of prejudices  |     brian@apache.org
acquired by the age of eighteen." - Einstein   |  brian@hyperreal.org