公开(公告)号：US20190095303A1

公开(公告)日：2019-03-28

申请号：US16198299

申请日：2018-11-21

Applicant: The MathWorks, Inc.

Inventor： Kiran K. Kintali ,  Shomit Dutta ,  E. Mehran Mestchian ,  Pieter J. Mosterman

IPC: G06F11/26 ,  G06F17/50 ,  G06F7/483

Abstract: Systems and methods evaluate simulation models and measure floating point arithmetic errors in terms of Unit in Last Place (ULP). The simulation model may include model elements that perform numerical computations using Native Floating Point (NFP) arithmetic. The model elements may be arranged to implement a procedure. A data store may include local ULP errors predetermined for the model elements. The systems and methods may retrieve the local ULP errors for the model elements included in the model, and may apply a rules-based analysis to compute an overall ULP error of the simulation model. The systems and methods may present the overall ULP computed for the model. The systems and methods may also present intermediate ULP errors determined for portions of the simulation model. Changes may be made to the model to reduce the overall ULP error.

公开(公告)号：US20170351493A1

公开(公告)日：2017-12-07

申请号：US15398176

申请日：2017-01-04

Applicant: The MathWorks, Inc.

Inventor： Kiran K. Kintali ,  Shomit Dutta ,  Anand S. Krishnamoorthi ,  Ebrahim Mehran Mestchian

IPC: G06F9/44

CPC classification number: G06F8/35 ,  G06F7/483 ,  G06F8/20 ,  G06F17/5027 ,  G06F17/5054 ,  G06F2207/3824 ,  H03M7/24 ,  H03M7/28

Abstract: Systems and methods generate code from an executable model. The model may operate on variables having floating point data types. The systems and methods may unpack the sign, exponent, and mantissa components of the floating point variables, and interpret them as boolean, integer, or fixed-point data types. The systems and methods may include operators that operate on the extracted sign, exponent, and mantissa components, and that produce sign, exponent, and mantissa outputs having boolean, integer or fixed-point data types. The systems and methods may pack the sign, exponent, and mantissa components of the output into an integer and reinterpret the integer as a floating point data type. Having replaced the floating point data types with boolean, integer or fixed-point data types, the generated code may be suitable for programmable logic devices and/or microcontrollers that lack Floating Point Units (FPUs).

公开(公告)号：US11354463B1

公开(公告)日：2022-06-07

申请号：US16580048

申请日：2019-09-24

Applicant: The MathWorks, Inc.

Inventor： Mohamed Babaali ,  Wurigen Bo ,  Kiran K. Kintali ,  Shomit Dutta ,  Ebrahim M. Mestchian ,  Naman Saraf

IPC: G06F30/20 ,  G06F8/35 ,  G06F17/16 ,  G06F17/13 ,  G06F111/10

Abstract: A solver may generate a system of equations for an acausal model. A partitioning engine may transform at least some of the equations into groups of equations whose inputs/outputs are connected directly. The partitioning engine may transform at least some of the equations into groups of linear equations and/or groups of switched linear equations that are connected through nonlinear functions. The solver may determine input-output relationships of the groups of equations. A simulation model generator that may include a library of types of model elements may construct a causal simulation model.

公开(公告)号：US10936769B2

公开(公告)日：2021-03-02

申请号：US16409165

申请日：2019-05-10

Applicant: The MathWorks, Inc.

Inventor： Kiran K. Kintali ,  Shomit Dutta ,  E. Mehran Mestchian ,  Pieter J. Mosterman

IPC: G06F9/44 ,  G06F30/20 ,  H03M7/28 ,  G06F8/35 ,  H03M7/24 ,  G06F7/499

Abstract: Systems and methods evaluate simulation models and measure floating point arithmetic errors in terms of Unit in Last Place (ULP). The simulation model may include model elements that perform numerical computations using Native Floating Point (NFP) arithmetic. The model elements may be arranged to implement a procedure. A data store may include local ULP errors predetermined for the model elements. The systems and methods may retrieve the local ULP errors for the model elements included in the model, and may apply a rules-based analysis to compute an overall ULP error of the simulation model. The systems and methods may present the overall ULP computed for the model. The systems and methods may also present intermediate ULP errors determined for portions of the simulation model. Changes may be made to the model to reduce the overall ULP error.

公开(公告)号：US20190332732A1

公开(公告)日：2019-10-31

申请号：US16409165

申请日：2019-05-10

Applicant: The MathWorks, Inc.

Inventor： Kiran K. Kintali ,  Shomit Dutta ,  E. Mehran Mestchian ,  Pieter J. Mosterman

IPC: G06F17/50 ,  H03M7/28 ,  G06F7/499 ,  H03M7/24 ,  G06F8/35

Abstract: Systems and methods evaluate simulation models and measure floating point arithmetic errors in terms of Unit in Last Place (ULP). The simulation model may include model elements that perform numerical computations using Native Floating Point (NFP) arithmetic. The model elements may be arranged to implement a procedure. A data store may include local ULP errors predetermined for the model elements. The systems and methods may retrieve the local ULP errors for the model elements included in the model, and may apply a rules-based analysis to compute an overall ULP error of the simulation model. The systems and methods may present the overall ULP computed for the model. The systems and methods may also present intermediate ULP errors determined for portions of the simulation model. Changes may be made to the model to reduce the overall ULP error.

公开(公告)号：US10140099B2

公开(公告)日：2018-11-27

申请号：US15398176

申请日：2017-01-04

Applicant: The MathWorks, Inc.

Inventor： Kiran K. Kintali ,  Shomit Dutta ,  Anand S. Krishnamoorthi ,  Ebrahim Mehran Mestchian

IPC: G06F9/44 ,  G06F8/35 ,  G06F8/20

Abstract: Systems and methods generate code from an executable model. The model may operate on variables having floating point data types. The systems and methods may unpack the sign, exponent, and mantissa components of the floating point variables, and interpret them as boolean, integer, or fixed-point data types. The systems and methods may include operators that operate on the extracted sign, exponent, and mantissa components, and that produce sign, exponent, and mantissa outputs having boolean, integer or fixed-point data types. The systems and methods may pack the sign, exponent, and mantissa components of the output into an integer and reinterpret the integer as a floating point data type. Having replaced the floating point data types with boolean, integer or fixed-point data types, the generated code may be suitable for programmable logic devices and/or microcontrollers that lack Floating Point Units (FPUs).