More discussion on the PC-DOS 1.10 boot sector and how it compares to the COMPAQ 1.11 boot sector
This commit is contained in:
parent
97ee1cf73d
commit
6d16f582c7
7 changed files with 474 additions and 307 deletions
|
|
@ -302,9 +302,15 @@ look at the code, using the PCjs Debugger to stop at the first instruction:
|
|||
&7DE4 00 00 00 E7 01 00 00 00-00 00 00 00 00 00 00 00 ................
|
||||
&7DF4 00 00 00 00 00 00 00 00-00 00 00 00
|
||||
|
||||
In terms of raw code (90 instructions), it's shorter than the PC-DOS 1.00 boot sector (99 instructions),
|
||||
even though it supports twice as many disk formats. This is because it uses tables to describe both the
|
||||
different disk formats *and* the discrete read operations, relieving the code from dealing with track boundaries.
|
||||
In terms of raw code (90 instructions), it's shorter than both the
|
||||
[PC-DOS 1.00 Boot Sector](/disks/pcx86/dos/ibm/1.00/#pc-dos-100-boot-sector) (99 instructions) and
|
||||
[PC-DOS 1.10 Boot Sector](/disks/pcx86/dos/ibm/1.10/#pc-dos-110-boot-sector) (116 instructions),
|
||||
because it uses tables to describe both the different disk formats *and* the discrete read operations,
|
||||
relieving the code from dealing with head and track size calculations.
|
||||
|
||||
It's arguably an improvement over the [PC-DOS 1.10 Boot Sector](/disks/pcx86/dos/ibm/1.10/#pc-dos-110-boot-sector),
|
||||
because it didn't need to be altered to support both 160Kb and 320Kb diskettes. On the other hand, that flexibility
|
||||
also makes it slightly slower to boot, because it must read the first FAT sector to determine the diskette format.
|
||||
|
||||
### Mounting the COMPAQ MS-DOS 1.11 Diskette
|
||||
|
||||
|
|
|
|||
Loading…
Reference in a new issue