HP3000-L Archives

April 1997, Week 1

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Stan Sieler <[log in to unmask]>
Reply To:
Stan Sieler <[log in to unmask]>
Date:
Thu, 3 Apr 1997 10:15:21 -0800
Content-Type:
text/plain
Parts/Attachments:
text/plain (152 lines)
Reid asks:

>         I am wondering if anybody has information about the caching
> configuration and strategy in an HP 9000 -715/100.  Or any pointers to
> where this information can be found.  I've looked around on the HP site
> and haven't found anything but I definitely could have missed it.
>
>         I am looking for all details including L1, L2 cache sizes,
> replacement strategy, associativity, write-back strategy, etc.
>
>         Would there be a better newsgroup to post this to?

Yes, comp.sys.hp.hpux

However, here's an old script I got from comp.sys.hp.hpux in 1995, which
seems to report some of the information you wanted, but I don't have
a 9000 715 to run it on.

------------------------------------cut here------------------------------------
#! /bin/sh

# Script produces cache size/type info.

# User must be logged on as root

# SYS=/hp-ux   for pre HP-UX 10.x systems
SYS=/stand/vmunix

MEM=/dev/kmem
TMP=/tmp/$$.prevpanic
trap 'trap "" 0 1 2 15; rm -f $TMP' 0 1 2 15

if [ ! -r $SYS ]
then
        echo >&2 $0: cannot read $SYS
        exit 1
fi

if [ ! -r $MEM ]
then
        echo >&2 $0: cannot read $MEM
        exit 1
fi

cat <<EOF | adb $SYS $MEM 2> /dev/null | sed -n -e '2,$s/.*:    //p' > $TMP
cache_tlb_parms+0d0/D
cache_tlb_parms+0d4/D
cache_tlb_parms+0d24/D
cache_tlb_parms+0d28/D
cache_tlb_parms+0d48/D
cache_tlb_parms+0d52/D
cache_tlb_parms+0d84/D
cache_tlb_parms+0d88/D
EOF

if [ ! -s $TMP ]
then
        echo >&2 $0: info not present in this kernel
        exit 1
fi

exec < $TMP

echo ICACHE:
read ISIZE
echo "  $ISIZE bytes"

read ICONF
BLOCK=`expr $ICONF / 16777216 % 256`
LINE=`expr $ICONF / 2097152 % 8`
LINESIZE=`expr $BLOCK \* $LINE \* 16`
echo "  $LINESIZE byte cache line"
SH=`expr $ICONF / 65536 % 2`
case $SH in
0)      echo "  separate I and D caches" ;;
1)      echo "  shared I and D caches" ;;
esac
ASSOC=`expr $ICONF % 256`
echo "  $ASSOC way associative"

echo DCACHE:
read DSIZE
echo "  $DSIZE bytes"

read DCONF
BLOCK=`expr $DCONF / 16777216 % 256`
LINE=`expr $DCONF / 2097152 % 8`
LINESIZE=`expr $BLOCK \* $LINE \* 16`
echo "  $LINESIZE byte cache line"
WT=`expr $DCONF / 131072 % 2`
case $WT in
0)      echo "  write-to" ;;
1)      echo "  write-through" ;;
esac
SH=`expr $DCONF / 65536 % 2`
case $SH in
0)      echo "  separate I and D caches" ;;
1)      echo "  shared I and D caches" ;;
esac
ASSOC=`expr $DCONF % 256`
echo "  $ASSOC way associative"

echo ITLB:
read ITSIZE
case $ITSIZE in
0)      echo "  no instruction TLB" ;;
*)      echo "  $ITSIZE entries"
        read ITCONF
        HW=`expr $ITCONF / 131072 % 2`
        case $HW in
        0)      echo "  software instruction TLB" ;;
        1)      echo "  hardware instruction TLB" ;;
        esac
        SH=`expr $ITCONF / 65536 % 2`
        case $SH in
        0)      echo "  separate I and D TLBs" ;;
        1)      echo "  shared I and D TLBs" ;;
        esac
        ASSOC=`expr $ITCONF % 256`
        echo "  $ASSOC way associative"
        ;;
esac

echo DTLB:
read DTSIZE
case $DTSIZE in
0)      echo "  no data TLB" ;;
*)      echo "  $DTSIZE entries"
        read DTCONF
        HW=`expr $DTCONF / 131072 % 2`
        case $HW in
        0)      echo "  software data TLB" ;;
        1)      echo "  hardware data TLB" ;;
        esac
        SH=`expr $DTCONF / 65536 % 2`
        case $SH in
        0)      echo "  separate I and D TLBs" ;;
        1)      echo "  shared I and D TLBs" ;;
        esac
        ASSOC=`expr $DTCONF % 256`
        echo "  $ASSOC way associative"
        ;;
esac

exit 0

------------------------------------cut here------------------------------------

--
Stan Sieler                                          [log in to unmask]
                                     http://www.allegro.com/sieler.html

ATOM RSS1 RSS2