Back to results
US 6,275,243 B1Granted
Method and apparatus for accelerating the transfer of graphical images
AssigneeNVIDIA Corporation
InventorsCurtis Priem, Raymond Lim
Life of this patent
Estimating the term…
Abstract
A graphics accelerator including an address remapping memory which straddles slow address spaces and fast address spaces.
Claims9
An engine for use with a memory having storage locations and memory lines, said engine comprising:
- circuitry configured to translate addresses to the memory to generate translated addresses to place data in locations in the memory where it is less likely that access to adjacent locations will require boundary crossings that have longer access times than non-boundary crossing access times;
- and
- a first set of outputs which assert values of a first subset of original address bits, wherein the first set of outputs are configured to be coupled to a first subset of the memory lines,
- wherein the circuitry configured to translate addresses includes lookup table circuitry having a first set of inputs coupled to receive a second subset of original address bits, and a second set of outputs configured to be coupled to a second subset of the memory lines, wherein the lookup table circuitry is configured to assert a different output value to the second set of outputs in response to assertion of each input value to the first set of inputs, wherein no two input values at the first set of inputs produce the same output value at the second set of inputs, the first subset of original address bits and the second subset of original address bits together determine an address, and each said output value together with the first subset of original address bits determines a translated address corresponding to said address.
1 more element
A graphics system, comprising:
- a memory having storage locations and memory lines;
- and
- a graphics accelerator coupled to the memory, wherein the graphics accelerator is configured to translate addresses to the memory to generate translated addresses to place data in locations in the memory where it is less likely that access to adjacent locations will require boundary crossings that have longer access times than non-boundary crossing access times, and wherein the graphics accelerator comprises:
- a first set of outputs coupled to assert values of a first subset of original address bits to a first subset of the memory lines, and
- lookup table circuitry having a first set of inputs coupled to receive a second subset of original address bits, and a second set of outputs coupled to a second subset of the memory lines, wherein the lookup table circuitry is configured to assert a different output value to the second set of outputs in response to assertion of each input value to the first set of inputs, wherein no two input values at the first set of inputs produce the same output value at the second set of inputs, the first subset of original address bits and the second subset of original address bits together determine an address, and each said output value together with the first subset of original address bits determines a translated address corresponding to said address.
2 more elements
A method for asserting address bits from an engine to memory lines, wherein the engine includes lookup table circuitry having inputs, said method comprising the steps of:
- (a) asserting values of a first subset of original address bits, as a first set of output values, to a first subset of the memory lines;
- and
- (b) translating input values of a second subset of original address bits to generate a second set of output values and asserting the second set of output values to a second subset of the memory lines, by asserting the input values of the second subset of original address bits to the inputs of the lookup table circuitry and in response to said input values of the second subset of original address bits, asserting the second set of output values from the lookup table circuitry to the second subset of the memory lines, such that no two of the input values of the second subset of original address bits produce the same output value, and wherein the first set of output values and the second set of output values determine translated addresses to place data in locations in a memory where it is less likely that access to adjacent locations will require boundary crossings that have longer access times than non-boundary crossing access times.
Dependencies are inferred from the claim text.