NetBSD/emips port runs on Xilinx and Beecube FPGA systems and the
Giano system simulator.
eMIPS is a platform developed at Microsoft Research for researching
reconfigurable computing. eMIPS allows dynamic loading and scheduling
of application-specific circuits for the purpose of accelerating
computations based on the current workload.
NetBSD eMIPS support for NetBSD 4.x was written at Microsoft Research
by Alessandro Forin and Neil Pittman. Microsoft Corporation has
donated full copyright to The NetBSD Foundation.
Platform support for eMIPS is the first part of Microsoft's
contribution. The second part includes the hardware accelerator
framework and will be proposed on tech-kern soon.
pointer but those don't use alternative MAC address in reality. So we check
whether the broadcast bit is set or not like Intel's e1000 driver.
Fixes PR kern/44072 reported by Jean-Yves Moulin.
Remove extra Warning for newer cards.
resemble its CardBus attachment very, very closely: slightly more than
24 lines are different. Alas, I cannot commit the CardBus part of this
change, yet, because I have to finish my overhaul of CardBus resource
handling, first.
Basically, KASSERT() should be used for light-weight checks and
KDASSERT() should be used for heavier ones.
Callers should not rely on the side effects of expression because,
depending on the kernel compile options mentioned above, expression might
not be evaluated at all.
- Xr options(4)
- bump date
This is not finished but good enough for others to play with, enable with
options VCONS_DRAW_ASYNC
With this all drawing operations will be posted to a ring buffer instead
of being run directly, and run by a kernel thread. This avoids having to wait
for drawing operations to finish with the kernel lock held ( to a degree at
least ) and scrolling a (slow) framebuffer console should not disrupt other
operations anymore.
Problems:
- we need to switch back to synchronous operations when panicing or entering
ddb, also re-enable async drawing when leaving ddb
- there are still occasional glitches
tested on an SS20 with cg14 and cg6 for dumb and accelerated cases
with adding new calls and makes all existing fd-accepting hijacked
calls dual-kernel. It would be better to autogenerate the code
from syscalls.master, but this is easier for now.
- call bus_dmamap_create(9) after bus_dmamem_alloc(9) and bus_dmamem_map(9)
return successfully
- no need to use PAGE_SIZE for size of descriptor dmamap
- use bus_addr_t to store DMA address
- BUS_DMASYNC_PREWRITE is enough for host-to-device only descriptors
Tested on R5000 Indy.
