文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > An Efficient Constraint-Based Test Selection Method for Embedded Processor Cores
侵权投诉

An Efficient Constraint-Based Test Selection Method for Embedded Processor Cores

Abstract. Software-based self-test has been proposed as a low-cost strategy for on-line periodic testing of embedded processors. However, for complex processors, the number of possible combinations of self-test routines may be huge, and the selection of a

presented later in Section 4, has three testable components (c=3), and the self-test library includes four routines (s=4) for each component. A test program for this processor is composed of three routines, one for each component. Thus, there are 64 (sc) possible test programs for the Java processor with the defined self-test library. However, only a subset of the possible test programs can actually be performed by a real-time embedded processor. In fact, as detailed in [7], some test programs are too large and do no fit the available memories. In other cases, the test program has a long execution time, leading to deadlines misses in the application. Therefore, a method for the definition of a suitable test program is necessary.

We have proposed such a method in [7], considering small processors. First, the method evaluates all possible combinations of self-test routines (64 possible programs, in the case of the Java processor mentioned above). Test programs that do not meet real-time and memory requirements are eliminated. Then, from the remaining suitable programs, the one that minimizes a single cost factor defined by the designer (data or program memory size, or test period) is chosen. In addition to the selected set of self-test routines, the method provides the minimum test period (time between two executions of the self-test program) for which real-time tasks meet their deadlines as defined by the application.

Although the solution proposed in [7] ensures the periodic execution of the test program and the fulfillment of the target platform and application limits, an exhaustive search is performed to find the best test program. Therefore, the method is inappropriate when the solutions space is very large. This may happen when a complex embedded processor is used, as we show next.

Let us consider, for instance, the test of a Very Long Instruction Word (VLIW) processor, like the C6000TM DSP core from Texas Instruments, with eight functional units and two register banks [8] (c=10). If the self-test library with four routines per component is used as before (s = 4), there will be 410 = 1,048,576 possible test programs. If we include other self-test routines with test response compaction code, and the test of other components, following the test priority presented in [4]), the search for a suitable solution will become unfeasible with an exhaustive method.

Thus, it is clear that an efficient heuristic to select a suitable set of self-test routines is necessary. Such a heuristic is proposed in the next section, using a constraint-based approach, where the system-imposed requirements guide the selection process. With this approach, one ensures that a feasible test program is generated. At the same time, the search method tries to find a solution that minimizes a cost function defined by the designer. The cost function is a combination of weighted cost variables, such as memories sizes, power consumption, test execution time, etc, where different weights define the optimization priority of each cost factor.

3. Test Program Search Method

The selection process receives four inputs: a self-testing library, the application characteristics, the system constraints, and the cost function to be optimized.

The self-test library contains several self-test routines for each testable component of the processor. Different SBST approaches can be used to test each component [3-

上一页第3页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email