Enhanced Parallel Processing in Wide Registers

Mitchell, J. L.; Hinds, Arianne T.

doi:10.1109/ipdps.2005.200

Cited by 1 publication

(9 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…See [4] for more details on the techniques to vectorize algorithms to run in parallel in wide registers.…”

Section: Definition Of Dct and Prior Efforts To Mitigate Complexitymentioning

confidence: 99%

“…Use shift right operations (vs. shift left operations) to implement the approximations from step 2. Apply the technique to minimize the average truncation error as described in [4]. A.…”

Section: Fast Hybrid Dctmentioning

confidence: 99%

“…This is accomplished by scaling both the numerator and the denominator of the approximations for any set of simultaneous constants equally such that the numerator is implemented in a power of two less than one. See [4] for a detailed description of the resulting error. Note that because of the resulting truncation error, we choose to negate some of the subtransform flowgraph signals in an effort to control the average truncation error.…”

Section: Step 3 Implementing Approximations With Shifts Rightmentioning

confidence: 99%

“…Note that because of the resulting truncation error, we choose to negate some of the subtransform flowgraph signals in an effort to control the average truncation error. [4] also illustrates that the final outputs of the FDCT (or the initial inputs for the IDCT) can be completed (started) in a negated sense so long as the corresponding scale factor for that particular output/input is also negated, i.e. [4] introduces the concept of a negative scale factor.…”

Section: Step 3 Implementing Approximations With Shifts Rightmentioning

confidence: 99%

“…1180 -1990 [3]. Furthermore, this new DCT requires only 44 additions and 17 shift operations, and can be implemented as a "vectorized" algorithm to execute in parallel in wide registers as described in [4]. Section 2 of this paper reviews the testbed for the IEEE 1180 compliance test for the IDCT.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A fast and accurate inverse discrete cosine transform

Hinds¹,

Mitchell²

IEEE Workshop on Signal Processing Systems Design and Implementation, 2005.

Self Cite

View full text Add to dashboard Cite

The Discrete Cosine Transform (DCT) is a widely used transform in image and video processing applications. By its mathematical definition, it is a computationally complex algorithm defined by cosine multiplications to accomplish the transformation of data to and from the frequency domain. Consequently, the fast implementation of the DCT is an active area of research as engineers endeavor to mitigate its complexity by approximating it with fixed-point algorithms. This paper presents a new design methodology for the design of linear transforms; the application of which introduces an example fixed-point Inverse DCT implementation that is both fast and accurate, and complies with the specification in the IEEE Std. 1180 -1990 (currently withdrawn).

show abstract

“…See [4] for more details on the techniques to vectorize algorithms to run in parallel in wide registers.…”

Section: Definition Of Dct and Prior Efforts To Mitigate Complexitymentioning

confidence: 99%

“…Use shift right operations (vs. shift left operations) to implement the approximations from step 2. Apply the technique to minimize the average truncation error as described in [4]. A.…”

Section: Fast Hybrid Dctmentioning

confidence: 99%

Section: Step 3 Implementing Approximations With Shifts Rightmentioning

confidence: 99%

Section: Step 3 Implementing Approximations With Shifts Rightmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations