Back to results
US 8,478,808 B1Granted
Approximate calculation of 2D matrix entries via GPU
AssigneeGauda, Inc.
InventorIlhami H. Torunoglu
Life of this patent
Estimating the term…
Abstract
Minimizing memory access by converting a given matrix computation into a set of low-order polynomials. The low-order polynomials can be used by dividing the domain of the polynomials into smaller subregions. If the domain is divided into equal intervals, the low-order polynomial can be used to approximate results from the matrix computation. The set of polynomials is processed using parallel computational hardware such as graphical processing units.
Claims20
A system comprising:
- a computing system comprising at least one central processing unit and at least one graphics processing unit;
- a user interface for interacting with the computer system;
- a computer readable medium comprising data describing size and placement of features to be formed on a photolithography exposure mask used to manufacture semiconductor devices;
- and
- a computer readable medium comprising optical proximity correction calculation procedures for acting upon the data, wherein at least a portion of the optical proximity correction calculation procedures are executed using the graphics processing unit, and the optical proximity correction calculation procedures comprise:
- causing conversion of a matrix of into a set of polynomials;
- and
- causing processing the set of polynomials in parallel using the graphics processing unit.
5 more elements
A method comprising:
- providing a target mask, wherein the target mask is nonpixelated, quantized tone, and in a spatial domain;
- performing a frequency domain transformation on the target mask to obtain a first mask, wherein the first mask is nonpixelated, continuous tone, and in a frequency domain;
- using an electronic processor, computing a first cost function for a first mask to obtain a first value;
- altering the first mask in a frequency domain to obtain a second mask, wherein the second mask is nonpixelated, continuous tone, and in the frequency domain, wherein the altering comprises:
- converting a matrix into a set of polynomials;
- and
- processing the set of polynomials using parallel computational hardware;
- computing the first cost function for the second mask to obtain a second value;
- and
- repeating the altering the first mask and the computing the first cost function for the second mask until the second value is less than the first value.
7 more elements
Dependencies are inferred from the claim text.