| Line 1035... |
Line 1035... |
transfer mechanism accesses physical memory and not virtual memory
|
transfer mechanism accesses physical memory and not virtual memory
|
(this is typically the case with DMA). So the kernel will no longer
|
(this is typically the case with DMA). So the kernel will no longer
|
rely on the disk driver (and possibly the disk hardware) to transfer
|
rely on the disk driver (and possibly the disk hardware) to transfer
|
multiple blocks in a single call, but split them up into single block
|
multiple blocks in a single call, but split them up into single block
|
transfers. Done - the "expansion swap" is now running.
|
transfers. Done - the "expansion swap" is now running.
|
|
|
|
----------------------------------------------------------------
|
|
|
|
In the meantime...
|
|
|
|
- experiments with porting ECO32 to another FPGA board
|
|
- splitting the project (separating the OS part)
|
|
- experiments with a pipelined version of ECO32
|
|
|
|
----------------------------------------------------------------
|
|
|
|
03-Feb-2014
|
|
Project "ECO32" created on OpenCores, based on version eco32-0.22:
|
|
http://opencores.org/project,eco32
|
|
|
|
18-Feb-2014
|
|
Several changes in the simulator:
|
|
a) There are two independent identical timer/counters
|
|
available, of course with different interrupts and I/O
|
|
addresses (see below).
|
|
b) The timer/counters are readable, so that they can
|
|
be used for short-time measurements. They count clock
|
|
cycles (no pre-scaling any longer). This change will
|
|
affect all programs which use the timer/counters.
|
|
c) The simulation timing model is completely based on
|
|
clock cycles (and does no longer try to function in
|
|
some sort of "real-time"). As there is no real clock
|
|
within the simulator, but the natural time unit is
|
|
one instruction, the simulation time is incremented
|
|
by the CPI value (clock cycles per instruction) every
|
|
instruction. I measured the CPI value in the real ECO32
|
|
implementation: it's a horrible 18 cycles per instruction,
|
|
yielding an instruction rate of about 2.8 MIPS. There is
|
|
so much room for improvements...
|
|
d) I changed the resolution of the simulation timers to
|
|
microseconds. All timing constants had to be adapted and
|
|
can now be specified more precisely. They are automatically
|
|
scaled to clock cycles (see above).
|
|
e) The addressing scheme for peripherals of the same sort
|
|
has changed slightly. Virtual addresses of I/O devices have
|
|
the form 0xFxxyyrrr, where xx is the device type, yy is the
|
|
device number, and rrr is the register within the device
|
|
(must be a word address). This change could affect existent
|
|
programs which use more than one terminal.
|
|
f) The ECO32 simulation got a new peripheral, called the
|
|
"shutdown device". A write to address 0xFF100000 results
|
|
in terminating the simulation run, with the lower 8 bits
|
|
of the data written supplied as exit status.
|
