forked from M-Labs/artiq
rtio: fix guard cycle computation
This commit is contained in:
parent
39c4b5416f
commit
c78c5a2b4f
@ -61,10 +61,9 @@ class _RTIOCounter(Module):
|
|||||||
#
|
#
|
||||||
# The buffer must be transferred to the FIFO soon enough to account for:
|
# The buffer must be transferred to the FIFO soon enough to account for:
|
||||||
# * transfer of counter to sys domain: Tio + 2*Tsys + Tsys
|
# * transfer of counter to sys domain: Tio + 2*Tsys + Tsys
|
||||||
# * guard time detection latency: Tsys
|
|
||||||
# * FIFO latency: Tsys + 2*Tio
|
# * FIFO latency: Tsys + 2*Tio
|
||||||
# Therefore we must choose:
|
# Therefore we must choose:
|
||||||
# guard_io_cycles > (3*Tio + 5*Tsys)/Tio
|
# guard_io_cycles > (3*Tio + 4*Tsys)/Tio
|
||||||
#
|
#
|
||||||
# We are writing to the FIFO from the buffer when the guard time has been
|
# We are writing to the FIFO from the buffer when the guard time has been
|
||||||
# reached without checking the FIFO's writable status. If the FIFO is full,
|
# reached without checking the FIFO's writable status. If the FIFO is full,
|
||||||
|
Loading…
Reference in New Issue
Block a user