[En-Nut-Discussion] App piper porting on H8/300H

tyou tyou at i-da.co.jp
Thu Aug 7 03:34:01 CEST 2003


Hi, Harald,
For corrupted TCP stack, I just mean the TCP/IP stack
stop working but Nut/OS (threads/timers) still works.
Maybe it is called incorrectly.
For H8 execution speed, the following is quoted from
H8/3069F datasheet:
...the number of instruction fetch, data read/write, and other
cycles occurring in each instruction.
The number of states required for execution of an instruction
can be calculated from these two tables as follows:
Number of states =
I • SI + J • SJ + K • SK + L • SL + M • SM + N • SN
Examples of Calculation of Number of States Required for
Execution
Examples: Advanced mode, stack located in external
address space, on-chip supporting modules accessed
with 8-bit bus width, external devices accessed in three
states with one wait state and 16-bit bus width.
BSET #0, @FFFFC7:8
Number of states = 2 ´ 4 + 2 ´ 3 = 14
one state needs two clock cycles.
So this instruction will take 28 cycles.
Although RISC machine may need several instructions
to make the same function, it is still slow for H8 to
manipulate memory in networking.

How ever , H8/S only need half time of H8/300H to
execute same instruction generally.

> I've almost finished release 2.3 and want to start with
> 3.0 next. This should compile with ARM ...

Sounds great, which ARM will be used? ATMEL's
AT91FR401XX?


Regards,
Tyou





More information about the En-Nut-Discussion mailing list