C_BLAS 123
by Sinectonalysis, Inc.

Software Overview

C_BLAS 123 is a library of optimized functions for vector-vector, vector-matrix, and matrix-matrix 
operations common to a large number of engineering tasks. Though mathematically simple, efficient 
linear algebra functions are particularly difficult to write, and unless written well, can consume 
thousands of wasted CPU cycles. Callable libraries for linear algebra functions have been available 
to Fortran programmers for many years. Now C_BLAS 123 makes these features and functions available 
to C  programmers writing code for the TMS320C30 and TMS320C40 processors. C_BLAS can save the high 
performance possible only from optimized algorithms and algorithms specially coded for the TMS320 
architecture.
Library uses a proprietary indexing scheme and includes functions which have been optimized to 
exploit specific matrix structures, such as symmetric, triangular, banded, hermitian, and 
symmetric-banded, redundant values are not stored. The result is that C_BLAS 123 can handle larger 
matrices in less memory with fewer cycles, C compilers will produce more efficient user code.

Features and Benefits

•	Highest possible performance for TMS320, callable from C, C++, ADA, Assembler.
•	Universal—works with all environments and `C3x/'C4x boards.
•	Full functionality of BLAS+, interruptible, relocatable, re-entrant.
•	Small linkable modules keep programs small. Data memory minimized.
•	Matrix algorithms specially optimized for efficient memory access.

Processor and System Specifications

•	Devices supported: TMS320C3x, TMS320C4x, Linear Algebra Run-Time Support Libraries
•	Languages supported: C, C++, ADA, Assember
•	Host platforms: PC, Sun, HP, DEC, RS6000, Mac, VAX, PowerPC

Algorithm Verification

•	The European community-supported ESPRIT project for Neural Network  ased Vision and Sensor Signal 
System for Industrial Quality Control (NEUROQUACS) has chosen a suite of Sinectonalysis libraries 
including this library for its development purposes. The NEUROQUACS team cited Sinectonalysis for 
its excellent performance, fast technical support, and functions which cover the widest task 
spectrum.
•	Also Spectrum Signal Processing has tested these libraries for ISO 9000 compliance.

Availability

•	Immediate

Company Background and Contact Information

Sinectonalysis, Inc. develops high-performance software for digital signal processors and 
parallel-processing environments for applications in signal processing, advanced algebra, image 
processing, and other scientific and engineering functions. The company specializes in optimizing 
algorithms and assembly code to take advantage of the TMS320C40 and TMS320C30 architectures. 
Sinectonalysis also offers consulting and code development services.
Contact:	Alexander Orenshteyn
		3702 S. Virginia St.
		Suite G-12 #164
		Reno, Nevada 89502-6034 
		(702) 345-0148
		Fax: (702) 345-0149
		e-mail: sinecto@clark.net