Filter
Top Products
Intel® ICH5 AC ’97 Modem Driver
R
AC ’97 Programmer’s Reference Manual 53
while (tail <= Prefetched_Index)
{
tail++; // Happens IFF Spurious Data was used
}
if ( ((tail <= LastValidIndex) || (tail == free))
&&(((tail+1) <= LastValidIndex) || ((tail+1) == free)) )
{
Descriptor.BufferPtr[tail] = &buffer;
Descriptor.BufferLength[tail] =
length(buffer) – MinBufferLength;
Descriptor.BufferPtr[tail+1] =
&buffer + length(buffer) – MinBufferLength;
Descriptor.BufferLength[tail+1] =
MinBufferLength;
tail += 2;
}
else
{
; //error: no space for this data buffer
}
if ((tail <= LastValid index) || (tail == free))
Descriptor.BufferPtr[tail] =
&(SpuriousBufferList[LastState(buffer)]);
Descriptor.BufferLength[tail] =
SpuriousBufferLength[LastState(buffer)];
LastValidIndex = tail;
//Note: tail is NOT incremented, so next time we’ll overwrite
// this descriptor, which is the whole point of this
// algorithm
}
else
{
LastValidIndex = tail-1; //warning: no space for spurious
// data buffer
}
The above implementation can be improved upon in a number of ways. First, rather than adding a
single (large) spurious buffer, a number of smaller ones could be chained together. In this way, the
amount of spurious data actually transmitted would be reduced while still maintaining a given level
of protection against long latencies for the host-based software. Additionally, the implementation
could be extended to handle multiple buffers at once, inserting several buffers in a row and only
splitting the last one and then appending a spurious buffer or buffers. Finally, the descriptor list is
a circular buffer and a real implementation would have to check tail and tail+1 against
base_address + 31 * 8.