F
Fabio de Matos Gon?alves
Guest
Hi,
I have been in troubles when configuring Xilinx FPGA's. It seems to
end up not properly configuring the boards in some scenarios.
My system is a linux 2.2.x and 2.4.x, and i use HOT2 boards, by VCC -
featuring Spartan 2 and XCS40 XLA with PCI interfaces - in 8 machines
of a cluster. Configuration happens by PCI access. After they are off
for a while, on first boot, almost none of the machines can succeed
configuring the boards. When i hard reboot the ones that didn't, some
of it can succeed configuring. If i need all machines working, this
process goes on for quite a long time. What happens is that all memory
positions contain 0xff, as far as i look.
Quoting Steve Casselman on earlier messages ( back in 1999 ) , there
is a mechanism that brings these boards to a known state when
configuration errors are detected.
This behaviour seems to happen only in some motherboard models. On
others, i have never seen it happening - although never tried it hard
enough.
This problem asks far more observation than i can manage. Because
failures take a long time to happen ( altough too often to be
tolerable ), only after a few months of errors I could really start
understanding what was going on.
Any related problems should help.
Anyway, what is happening to vcc.com site?
I have been in troubles when configuring Xilinx FPGA's. It seems to
end up not properly configuring the boards in some scenarios.
My system is a linux 2.2.x and 2.4.x, and i use HOT2 boards, by VCC -
featuring Spartan 2 and XCS40 XLA with PCI interfaces - in 8 machines
of a cluster. Configuration happens by PCI access. After they are off
for a while, on first boot, almost none of the machines can succeed
configuring the boards. When i hard reboot the ones that didn't, some
of it can succeed configuring. If i need all machines working, this
process goes on for quite a long time. What happens is that all memory
positions contain 0xff, as far as i look.
Quoting Steve Casselman on earlier messages ( back in 1999 ) , there
is a mechanism that brings these boards to a known state when
configuration errors are detected.
This behaviour seems to happen only in some motherboard models. On
others, i have never seen it happening - although never tried it hard
enough.
This problem asks far more observation than i can manage. Because
failures take a long time to happen ( altough too often to be
tolerable ), only after a few months of errors I could really start
understanding what was going on.
Any related problems should help.
Anyway, what is happening to vcc.com site?