A Domain-Specific Architecture for Elementary Function Evaluation
We propose a Domain-Specific Architecture for elementary function computation to improve throughput while reducing power consumption as a model for more general applications: support fine-grained parallelism by eliminating branches, and eliminate the duplication required by coprocessors by decomposi...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2015-01-01
|
| Series: | International Journal of Mathematics and Mathematical Sciences |
| Online Access: | http://dx.doi.org/10.1155/2015/843851 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832561420000034816 |
|---|---|
| author | Anuroop Sharma Christopher Kumar Anand |
| author_facet | Anuroop Sharma Christopher Kumar Anand |
| author_sort | Anuroop Sharma |
| collection | DOAJ |
| description | We propose a Domain-Specific Architecture for elementary function computation to improve throughput while reducing power consumption as a model for more general applications: support fine-grained parallelism by eliminating branches, and eliminate the duplication required by coprocessors by decomposing computation into instructions which fit existing pipelined execution models and standard register files. Our example instruction architecture (ISA) extension supports scalar and vector/SIMD implementations of table-based methods of calculating all common special functions, with the aim of improving throughput by (1) eliminating the need for tables in memory, (2) eliminating all branches for special cases, and (3) reducing the total number of instructions. Two new instructions are required, a table lookup instruction and an extended-precision floating-point multiply-add instruction with special treatment for exceptional inputs. To estimate the performance impact of these instructions, we implemented them in a modified Cell/B.E. SPU simulator and observed an average throughput improvement of 2.5 times for optimized loops mapping single functions over long vectors. |
| format | Article |
| id | doaj-art-432193044e7d49adaaa77ec33c689654 |
| institution | Kabale University |
| issn | 0161-1712 1687-0425 |
| language | English |
| publishDate | 2015-01-01 |
| publisher | Wiley |
| record_format | Article |
| series | International Journal of Mathematics and Mathematical Sciences |
| spelling | doaj-art-432193044e7d49adaaa77ec33c6896542025-02-03T01:25:04ZengWileyInternational Journal of Mathematics and Mathematical Sciences0161-17121687-04252015-01-01201510.1155/2015/843851843851A Domain-Specific Architecture for Elementary Function EvaluationAnuroop Sharma0Christopher Kumar Anand1Department of Computing and Software, McMaster University, Hamilton, ON, L8S 4K1, CanadaDepartment of Computing and Software, McMaster University, Hamilton, ON, L8S 4K1, CanadaWe propose a Domain-Specific Architecture for elementary function computation to improve throughput while reducing power consumption as a model for more general applications: support fine-grained parallelism by eliminating branches, and eliminate the duplication required by coprocessors by decomposing computation into instructions which fit existing pipelined execution models and standard register files. Our example instruction architecture (ISA) extension supports scalar and vector/SIMD implementations of table-based methods of calculating all common special functions, with the aim of improving throughput by (1) eliminating the need for tables in memory, (2) eliminating all branches for special cases, and (3) reducing the total number of instructions. Two new instructions are required, a table lookup instruction and an extended-precision floating-point multiply-add instruction with special treatment for exceptional inputs. To estimate the performance impact of these instructions, we implemented them in a modified Cell/B.E. SPU simulator and observed an average throughput improvement of 2.5 times for optimized loops mapping single functions over long vectors.http://dx.doi.org/10.1155/2015/843851 |
| spellingShingle | Anuroop Sharma Christopher Kumar Anand A Domain-Specific Architecture for Elementary Function Evaluation International Journal of Mathematics and Mathematical Sciences |
| title | A Domain-Specific Architecture for Elementary Function Evaluation |
| title_full | A Domain-Specific Architecture for Elementary Function Evaluation |
| title_fullStr | A Domain-Specific Architecture for Elementary Function Evaluation |
| title_full_unstemmed | A Domain-Specific Architecture for Elementary Function Evaluation |
| title_short | A Domain-Specific Architecture for Elementary Function Evaluation |
| title_sort | domain specific architecture for elementary function evaluation |
| url | http://dx.doi.org/10.1155/2015/843851 |
| work_keys_str_mv | AT anuroopsharma adomainspecificarchitectureforelementaryfunctionevaluation AT christopherkumaranand adomainspecificarchitectureforelementaryfunctionevaluation AT anuroopsharma domainspecificarchitectureforelementaryfunctionevaluation AT christopherkumaranand domainspecificarchitectureforelementaryfunctionevaluation |