International Business Machines Corporation

Globalfoundries Inc.

Christopher M Barton

Raul Esteban Silvera

Ehsan Amiri

Denis Palmeiro

Steven Wayne White

Yaoqing Gao

Satish Kumar Sadasivam

Shimin Cui

Ettore Tiotto

Jose Nelson Amaral

Taylor J Lloyd

Al Chakra

Sumit Patel

Madhavi Gopal Valluri

Arie Tal

Denis M Palmeiro

Mahavi G Valluri

A method, computer system, and a computer program product for high-speed data compression is provided. The present invention may include receiving an input stream. The present invention may include selecting a header based on the received input stream, wherein the header includes a base, a scheme and a delta count. The present invention may include determining whether there are any remaining values in an uncompressed input stream. The present invention may include reading a first next value from the input stream. The present invention may include determining whether the read first next value is representable with a current base scheme. The present invention may include calculating the delta count based on determining that the read first next value is representable with the current base scheme. The present invention may include writing the calculated delta count to the selected header. The present invention may include incrementing the written delta count.

Run-length base-delta encoding for high-speed compression

Utilizing problem insights based on the entire environment as inputs to drive a static compiler. A decision engine receives inputs associated with applications to be compiled. The decision engine also receives optimization constraints based on available resources. A decision learning model is applied to the inputs to predict compiler performance and the results are provided to the decision engine. The decision engine determines a profile that comprises an order of execution and an optimization level for use during compilation of the plurality of applications. The profile is then used to schedule compiling and optimization of the applications.

Analytics driven compiler

A computer identifies one or more pairs of scalar statements and performs a cost analysis of operations of each of the one or more pairs of scalar statements to determine both a benefit and a cost of operations. The computer determines, based, at least in part, on the cost analysis, a gain for each of the one or more pairs of scalar statements. The computer creates based, at least in part, on the gain, a sorted list of each of the one or more pairs of scalar statements and selects a first pair from the sorted list. The computer issues a query to a hash table using a statement of the first pair and selects from results received from the query, a second pair. The computer then extends, based, at least in part, on the second pair, the first pair to create a pack.

Extending superword level parallelism

Generating SIMD code from code statements that include non-isomorphic code statements. Code statements are received, each code statement has one or more operators in a respective operator order and each operator has a type and associated operands. At least two code statements among the code statements received have an operator of the same type in a different operator order position. A first operator order position is identified for the operators of the same type in each of the code statements. For each of the code statements, code is generated for operators and their associated operands having operator order positions preceding the first operator order positions. SIMD code is generated at least based on the identified first operator order positions, the corresponding operator type, and the operands associated with the operator type at the identified operator order positions.

Generating SIMD code from code statements that include non-isomorphic code statements

A computer processor collects information for a dominant data access loop and reference code patterns based on data reference pattern analysis, and for pointer aliasing and data shape based on pointer escape analysis. The computer processor selects a candidate array for data splitting wherein the candidate array is referenced by a dominant data access loop. The computer processor determines a data splitting mode by which to split the data of the candidate array, based on the reference code patterns, the pointer aliasing, and the data shape information, and splits the data into two or more split arrays. The computer processor creates a software cache that includes a portion of the data of the two or more split arrays in a transposed format, and maintains the portion of the transposed data within the software cache and consults the software cache during an access of the split arrays.

Optimizing memory bandwidth consumption using data splitting with software caching

A method, apparatus, and computer instructions are provided by the present invention for a programming framework for easy pattern matching and transformation of intermediate language expression trees. With expression matching and transformation framework (EMTF) of the present invention, pattern matchers may be defined and embedded in a C++ program code in a manner that resembles their intermediate representation. EMTF also provides capabilities for unification, unbinding, transformation, intra-pattern referencing, common operators, creation of pattern objects in a storage pool. With EMTF, less common patterns may be applied against input expressions without repeatedly implementing an algorithm.

Method and apparatus for a programming framework for pattern matching and transformation of intermediate language expression trees

Growing community of inventors

Stouffville, Canada

Christopher M Barton