From:	SMTP%"MACRO32@WKUVX1.BITNET" 29-JUL-1993 11:14:40.16
To:	EVERHART
CC:	
Subj:	Re: AMAC vs. computed JMP

X-ListName: "VMS Internals, MACRO, & BLISS Discussions" <MACRO32@WKUVX1.BITNET>
Warnings-To: <>
Errors-To: MacroMan@WKUVX1.BITNET
Sender: MacroMan@WKUVX1.BITNET
Date: Thu, 29 Jul 1993 17:07:45 +0100
From: Arne Vajhxj <ARNE@kopc.hhs.dk>
Reply-To: MACRO32@WKUVX1.BITNET
Subject: Re: AMAC vs. computed JMP
To: MACRO32@WKUVX1.BITNET, moeller@gwdgv1.dnet.gwdg.de
Message-ID: <01H14EJQMSFM9LV9EB@kopc.hhs.dk>
X-VMS-To: KOPC::IN%"MACRO32@WKUVX1.BITNET"
X-VMS-Cc: IN::"moeller@gwdgv1.dnet.gwdg.de"
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; CHARSET=ISO-8859-1
Content-Transfer-Encoding: 8BIT

> The program is tranferring control via a table of (among other things)
> jump adresses, like
>
>                 ...
>                 .psect  data
>         table:
>                 .long   code1 - table
>                 ...
>                 .psect  code
>                 .entry  ...
>                 ...
>                 moval   table,r10
>                 movl    (r10)[index],r11
>                 jmp     (r11)[r10]
>                 ...
>         code1:
>                 ...
>                 ret
>
> How do I convince Alpha-MACRO32 that
>         (1)     'code1' is reachable at all,
>         (2)     I do want the "PIC" address difference stored in 'table',
>         (3)     I do want to get to 'code1' via JMP only (no prologue!) ??

I guess the MACRO32-compiler on AXP is DEC's revenge for all the poor
MACRO32 they have seen through the years. It is excellent in giving
mysterious error-messages whenever something unusually is done.

I tried the following small test-version:

        .title  z
        .psect  $PDATA quad,pic,con,lcl,shr,noexe,nowrt
table:  .long   code1 - table
        .long   code2 - table
        .psect  $CODE quad,pic,con,lcl,shr,exe,nowrt
        .entry  z,^m<r9,r10,r11>
        movl    #0,r9
        moval   table,r10
        movl    (r10)[r9],r11
        jmp     (r11)[r10]
code1:  movl    #44,r0
        ret
code2:  movl    #1,r0
        ret
        .end    z

It do give informational error-messages during compilation and
chrashes at run-time.

I have also tried the following small variant:

        .title  z
        .psect  $PDATA quad,pic,con,lcl,shr,noexe,nowrt
table:  .long   code1 - table
        .long   code2 - table
        .psect  $CODE quad,pic,con,lcl,shr,exe,nowrt
        .entry  z,^m<r9,r10,r11>
        movl    #0,r9
        moval   table,r10
        movl    (r10)[r9],r11
        jmp     (r11)[r10]
code1:  .jsb32_entry
        movl    #44,r0
        ret
code2:  .jsb32_entry
        movl    #1,r0
        ret
        .end    z

It also gives informational error-messages during compilation, but it
runs as it is supposed to.

Conclusion: you could try with the .JSB32_ENTRY directive and see, if
your code (just by pure luck) will work too.

                                                          Arne

Arne Vajhxj                             local DECNET:  KO::ARNE
Computer Department                     PSI:           PSI%238310013040::ARNE
Business School of Southern Denmark     Internet:      ARNE@KO.HHS.DK