International Business Machines Corporation

Globalfoundries Inc.

Yoav Avraham Katz

Roy Emek

Avi Ziv

Itai Jaeger

Michal Rimon

Elad Yom-Tov

Anatoly Koyfman

Shady Copty

Shai Lubliner

Allon Adir

Oz Dov Hershkovitz

Michael Vinov

Elena Tsanko

Yoav Kantor

Ron Maharik

Naftali Liberman

Combining predictions made by multiple different prediction systems, including: obtaining a new input sample; obtaining a combiner which comprises optimal selection values that are configured to maximize a predefined performance measure; automatically applying multiple different prediction systems to the new input sample, to generate predictions; and automatically selectively combining the generated predictions based on the optimal selection values, to generate a combined prediction whose predefined performance measure is improved compared to individual usage of each of the prediction systems. The selective combination includes splitting the generated predictions into an intersection subset, fully-disjoint subsets, and optionally one or more partially-disjoint subsets; selecting predictions, from the subsets, at a ratio corresponding to the optimal selection values associated with the subsets, and modifying the subsets to remove all non-selected predictions; and merging all predictions in the modified subsets, to produce the combined prediction.

Automated combination of predictions made by different prediction systems

A computer-implemented method, an apparatus and a computer program for automatically extracting useful information for functional verification. The method comprising performing repeatedly both operating an instruction generator associated with a Design Under Test (DUT), whereby a generated instruction is determined, the generated instruction having one or more instruction attributes; and collecting information relating to the generated instruction. Based on the generated instruction and the collected information, a classification technique is utilized to classify the information based on the instruction attributes.

Automatic identification of information useful for generation-based functional verification

A computerized apparatus, method and computer product for generating tests. The apparatus comprises: a processor; an interface for obtaining a test template associated with a computerized system that comprises a template segment comprising instructions and directives or related control constructs; a test generator for generating a test associated with the template segment, comprising: a simulator for determining a state of the system associated with an execution of the test; a selector for selecting a template instruction or segment from the test template based on the state of the system; and a generator configured to generate a multiplicity of instructions based on system's state and on the selected template segment, wherein the test generator further comprises a verifier configured to verify that a previously generated instruction is in line with the current state of the system and with the selected template instruction or segment.

Dynamic generation of test segments

A method, apparatus and product for generating elements based on generation streams. The method comprises: obtaining one or more generation streams, wherein the streams comprise elements, wherein each element is a formal specification of an operation that stimulates a system, wherein based on each of the generation streams one or more alternative stimuli for the system can be generated, which stimuli comprises operations according to the elements; and generating a stimuli in accordance with the one or more generation streams, wherein the stimuli comprises at least one hybrid operation, wherein the hybrid operation complies simultaneously with two or more elements of the one or more generation stream, whereby the stimuli is comprised of a number of operations that is smaller than a sum of the numbers of elements of the one or more generation streams.

Stream generation

A method, apparatus and product. The method comprising automatically determining an abstract CSP based on a formally defined problem having interconnected sub-problems, wherein the abstract CSP corresponds to the problem, wherein the abstract CSP has a reduced complexity in comparison to a CSP representing the problem, wherein the abstract CSP captures the interconnection between the sub-problems and reduces the details of each sub-problem, wherein the abstract CSP comprises constraints over variables, wherein each variable having an associated domain; and repeatedly: (1) propagating constraints of the abstract CSP to reduce domains of the abstract CSP; (2) selecting a sub-problem to solve; (3) solving the sub-problem; and (4) updating the abstract CSP with values in accordance with the solution of the sub-problem. Whereby, a solution to the formally defined problem is determined based on the solutions to the sub-problems.

Solving constraint satisfaction problems having loosely interconnected sub-problems

Generation of a test based on a test template comprising of branch instructions. The test template may be a layout test template, defining a set of possible control flows possibilities between template instructions in the layout test template. The test is generated by a test generator which may simulate a state of a target computerized system executing the test. The simulation may be performed during generation of the test. The test generator may further verify previously generated instructions. The test generator may further generate instructions associated with leftover template instructions.

Dynamic generation of tests

Device, system and method of verification of address translation mechanisms. For example, an apparatus for testing an address translation mechanism of a design-under-test, the apparatus including: a test generator to receive a specification of at least one address translation table, and to generate one or more constraint-satisfaction-problem projectors over a plurality of attributes of said address translation table.

Method of verification of address translation mechanisms

Systems and methods for solving a constraint satisfaction problem (CSP) defined by one or more variables, constraints and probabilistic requests (PRs) are provided. In one embodiment, each PR is mapped onto a set of constraints that implement the PR. The CSP solver then finds a series of solutions to the CSP. For each solution, if a constraint associated with a PR is satisfied, then a success counter for that PR is incremented; else a failure counter for that PR is incremented. If the constraint was not considered for the purpose of solving the CSP, then the failure counter for the PR associated with the constraint is incremented.

Determining compliance rates for probabilistic requests

A computer-implemented method for verification of a hardware design includes specifying requests to allocate regions in a memory of the hardware design, such that at least two of the requests are specified independently of one another. The requests indicate respective allocation types. Overlap restrictions are specified between at least some of the allocation types. The requests and the overlap restrictions are automatically converted to a constraint satisfaction problem (CSP), which includes CSP constraints based on the requests, the allocation types and the overlap restrictions. The CSP is solved to produce a random test program, which includes a memory map that allocates the regions in the memory while complying with the requests and the overlap restrictions. The test program is applied to the hardware design.

Random stimuli generation of memory maps and memory allocations

Methods and systems are disclosed that enhance the ability of a test generator to automatically deal with address translation in a processor design, and without need for creating specific code. A model of the address translation mechanism of a design-under-test is represented as a directed acyclic graph and then converted into a constraint satisfaction problem. The problem is solved by a CSP engine, and the solution used to generate test cases for execution. Using the model, testing knowledge can be propagated to models applicable to many different designs to produce extensive coverage of address translation mechanisms.

Modeling language and method for address translation design mechanisms in test generation

A computer-implemented method for modeling a target system includes defining a cloned constraint satisfaction problem (CSP) that characterizes the target system in terms of a set of variables and constraints applicable to the variables. The cloned CSP includes a non-predetermined number of duplicate sub-problems corresponding to instances of a repeating feature of the target system. The variables are partitioned so as to define an abstract CSP containing a subset of the variables relating to the duplicate sub-problems. The abstract CSP is solved to generate an abstract solution indicating the number of duplicate sub-problems to use in the cloned CSP. A concrete solution to the cloned CSP is found using the abstract solution.

Growing community of inventors

Zichron Yaacov, Israel

Yoav Avraham Katz