METHOD OF ANALYZING NON-PREEMPTIVE DRAM TRANSACTIONS IN REAL-TIME UNIFIED MEMORY ARCHITECTURES
Systems and methods are disclosed for non-preemptive DRAM transactions. More specifically, the present invention relates to improvements in non-preemptive DRAM transactions in real-time unified memory architectures. One embodiment of the present invention relates to a method for determining access to non-preemptive DRAM devices. This method comprises determining real time need for access to the device and prioritizing access using a rate monotonic scheduling.
1 - 33 . (canceled)
34 . A method of determining an effective service deadline of a client in a system that employs rate monotonic scheduling to determine access priority of a plurality of clients to a common resource, the method comprising:
(a) determining an actual service deadline of the client;
(b) determining a total communication delay between the client and the common resource; and
(c) subtracting the total communication delay from the actual service deadline to obtain an effective service deadline.
35 . The method of claim 34 , further comprising:
(d) determining a rate monotonic scheduling priority of the client based upon the effective service deadline.
36 . The method of claim 34 wherein the total communication delay is the sum of the communication delay from the client to the shared resource and the communication delay from the common resource to the client.
37 . The method of claim 34 wherein a service deadline of a given client is the point in time by which the client needs an operation requested of the common resource to be completed.
38 . The method of claim 34 wherein the clients are hardware devices.
39 . The method of claim 38 wherein the common resource is a memory element.
40 . The method of claim 34 wherein the clients are software clients.
41 . The method of claim 40 wherein the common resource is a CPU.
42 - 50 . (canceled)
51 . A machine-readable storage, having stored thereon a computer program having a plurality of code sections for implementing a method of determining an effective service deadline of a client in a system that employs rate monotonic scheduling, the code sections executable by a machine for causing the machine to perform operations comprising:
(a) determining an actual service deadline of the client;
(b) determining a total communication delay between the client and a shared resource; and
(c) subtracting the total communication delay from the actual service deadline to obtain an effective service deadline.
52 . The machine-readable storage of claim 51 , the code sections executable by a machine for causing the machine to perform operations further comprising:
(d) determining a rate monotonic scheduling priority of the client based upon the effective service deadline.
53 . The machine-readable storage of claim 51 wherein the total communication delay is the sum of the communication delay from the client to the shared resource and the communication delay from the shared resource to the client.
54 . The machine-readable storage of claim 51 wherein a service deadline of a given client is the point in time by which the client needs an operation requested of the shared resource to be completed.
55 . The machine-readable storage of claim 51 wherein the clients are hardware devices.
56 . The machine-readable storage of claim 55 wherein the shared resource is a memory element.
57 . The machine-readable storage of claim 51 wherein the clients are software clients.
58 . The machine-readable storage of claim 57 wherein the shared resource is a CPU.