[hatari-devel] Proposal for a new disassembler in Hatari (2nd try)

[Eero Tamminen]

Hi,

> I still don't like the weird disassembly output in Hatari, so I wrote my
> own (68000..68060/CPU32/PMMU/FPU) disassembler under my own GPL
> compatible license :-) I've tested the output against other
> disassemblers, noticed they have a bunch of bugs. And tuned it a bit...
> 
> Feedback, please :-)
>
> Here is the first 8k of TOS 3.06 with _NO_ additional info from the
> user, just a plain: disassemble from $e0000 to $e02000. All symbols,
> formatting, etc. is done by the disassembler. The way it's done is by
> text files that define hardware addresses, but also - optional - OS
> specific symbols (i've added a few for test purposes).

Hatari already has support for symbol definition files, please use that
(if you aren't already).  The format is same as used by NM and documented
in the Hatari manual debugger section.  Both GCC binutils and AHCC can
output this format.

For an example:
- on how to access these symbols from code, see ../src/debug/symbols.h
  (and how those functions are called elsewhere in debugger code)
- of this data, see  the *.sym files under tests/ dir
- on how to use that data while in debugger, see tests/debugui/debugger.ini.

When the commands are used interactively in the debugger, you have full
dir- & filename completion in the "dsp/symbols" commands which load the
data.

And if you're working longer on some code, you can put all these debugger 
commands into a file and ask Hatari to load them on startup with
the "--parse <file>" option.


> The beginning for example is just:
> 
>  0xe00000,OSHEADER,1,osroml
>
> which results in:
--------------------------------
$e00000 : 602e                     os_entry: BRA.S $e00030
$e00002 : 0306                     DC.W $0306           ; os_version
$e00004 : 00e0 0030                DC.L osromreset      ; reseth
$e00008 : 00e0 0000                DC.L osroml          ; os_beg
$e0000c : 0000 8316                DC.L $008316         ; os_end
$e00010 : 00e0 0030                DC.L osromreset      ; os_rsvl
$e00014 : 00e3 5fd8                DC.L $e35fd8         ; os_magic
$e00018 : 0924 1991                DC.L $09241991       ; os_date
$e0001c : 0000                     DC.W $0000           ; os_conf
$e0001e : 1738                     DC.W $1738           ; os_dosdate
$e00020 : 0000 5766                DC.L $005766         ; p_root
$e00024 : 0000 10bb                DC.L $0010bb         ; pkbshift
$e00028 : 0000 771e                DC.L $00771e         ; p_run
$e0002c : 0000 0000                DC.L $000000         ; p_rsv2
$e00030 : 46fc 2700                osromreset: MOVE #$2700,SR
$e00034 : 31fc 0100 8606           MOVE.W     #$100,(fifo).w
$e0003a : 31fc 0000 8606           MOVE.W     #0,(fifo).w
...
-----------------------------

The symbols usage looks great to me!

How you've though to include function names and profiler data to that
output?


The current Hatari disassembly output with function/code address symbol
names looks like this:
---
00e0f5fa: 60a2 4e56 0000 342e 0008 BT .B #$ffffffa2 == 00e0f59e (TRUE)
_vdi2ste:
00e0f5fc: 4e56 0000 342e 0008 7003 LINK.W A6,#$0000
00e0f600: 342e 0008 7003 c1c2 48c0 MOVE.W (A6,$0008) == $00010fea,D2
---
(Data symbol addresses aren't currently included into disassembler output
but the way you do it is indeed nice looking.)

This is similar to how the profiler data is added to the output:
---
1065/17040 times/cycles:
00e0f5f6: 4e90 584f 60a2 4e56 0000 JSR.L (A0)
1054/8432 times/cycles:
00e0f5f8: 584f 60a2 4e56 0000 342e ADDA.W #$00000004,A7
1178/11780 times/cycles:
00e0f5fa: 60a2 4e56 0000 342e 0008 BT .B #$ffffffa2 == 00e0f59e (TRUE)
---

(It disrupts the code flow but with current disassembler I couldn't
add it anywhere else than on its own line.)


	- Eero

PS. This is how you use the profiler:
-----------------------
> profile on
Profiling enabled.
> c
Returning to emulation...
<here I return to debugger again>
Allocated CPU profile buffer (18 MB).
Allocated CPU profile address buffer (4 KB).
Normal RAM (0-0x400000):
- no activity
Cartridge ROM (0xFA0000-0xFC0000):
- no activity
ROM TOS (0xE00000-0xE80000):
- active address range:
  0xe0014c-0xe23f02
- active instruction addresses:
  707 (100.00% of all)
- executed instructions:
  629144 (100.00% of all)
- used cycles:
  9408503 (100.00% of all)
- address with most cycles:
  0xe0dcca, 402628 cycles (4.28% of all in area)
- address with most hits:
  0xe004a4, 4900 hits (0.78% of all in area)

CPU=$e176fc, VBL=887, FrameCycles=64, HBL=0, LineCycles=64, DSP=N/A
> profile counts
addr:           count:
0xe004a4        0.78%   4900
0xe004b6        0.78%   4896
0xe0dcbc        0.55%   3448
0xe0dcc6        0.55%   3448
0xe0f59e        0.55%   3446
0xe0dcce        0.55%   3445
0xe0dcd6        0.55%   3445
0xe1657a        0.55%   3445
0xe1af98        0.55%   3445
0xe0f564        0.54%   3390
0xe0f56c        0.54%   3390
0xe0f574        0.54%   3390
0xe0dcb6        0.54%   3389
0xe0dcae        0.54%   3388
0xe0f57e        0.54%   3388
0xe0f584        0.54%   3388
16 CPU addresses listed.
-----------------------