Sample request mechanism for supplying a filtering engine
View Patent ↗A graphics system comprising a scheduling network, a sample buffer and a plurality of filtering units. The sample buffer is configured to store sample generated by a rendering engine. The plurality of filtering units are coupled in a linear series. Each filtering unit of the linear series is configured to send a request for a scanline of sample bins to a first filtering unit of the linear series. The first filtering unit is configured to service the scanline requests by sending burst requests to a scheduling network and coordinating the flow of samples forming the bursts from the sample buffer to the filtering units.
1. A system comprising:
a sample request interface configured to initiate the transfer of slots onto a cyclic request bus, to receive filled slots indicating scanline requests from a plurality of filtering units, to store the scanline requests in request buffers corresponding to the filtering units, and to withhold the initiation of the transfer of slots for a first of the filtering units while the corresponding request buffer is full;
a plurality of arbiters, wherein each arbiter is configured to read the request buffers and to arbitrate among scanline requests addressing a corresponding memory group, resulting in burst requests being stored in corresponding forwarder buffers; and
a plurality of forwarder buffers, wherein each forwarder buffer is configured to read a burst request from the corresponding forwarder buffer and to transfer the burst request to one or more scheduling units to induce transfer of a corresponding sample burst to a requesting filtering unit of the plurality of filtering units.
2. A method comprising:
a plurality of filtering units generating scanline requests;
the filtering units transferring the scanline requests to a first of the filtering units through a cyclic request bus;
storing the scanline requests in request buffers which correspond to the filtering units;
servicing first scanline requests from the request buffers which address a first group of memory devices by generating first burst requests, and storing the first burst requests in a first forwarder buffer;
servicing second scanline requests from the request buffers which address a second group of memory devices by generating second burst requests, and storing the second burst requests in a second forwarder buffer;
forwarding the burst requests from the first forwarder buffer to one or more scheduling devices to induce transfer of sample bursts from the first group of memory devices to the filtering units;
forwarding the burst requests from the second forwarder buffer to the one or more scheduling devices to induce transfer of sample bursts from the second group of memory devices to the filtering units; and
wherein said servicing the first scanline requests from the request buffers which address a first group of memory devices is performed according to an arbitration scheme which incorporates a measure of the number of arbitration cycles each scanline request has waited without being serviced.