Stream User’s Guide
Download 0.95 Mb.
|
- Navigate this page:
- 8.8Performance
- 9Stream Program Development
- 9.1Invoke spide
8.7TimersThe Timers section of the Component API chapter above describes built-in timers. spm_demo includes calls to print timer data from built-in timers and from a timer defined in gsr_pipeline. Run the app version of spm_demo on stream processor hardware to see some built-in timer data: $ ./app_sys sample.bmp result.bmp ...
Startup: 8501397 ns DSP MIPS load: 40701437 ns Total: 863282282 ns Done!
$ ./app_sys sample.bmp result.bmp ...
file_in cmd handler: 223015 ns file_in execute: 204558486 ns gsr cmd handler: 0 ns gsr execute: 16611011 ns file_out cmd handler: 257153 ns file_out execute: 680860124 ns Startup: 26330698 ns DSP MIPS load: 56076100 ns Total: 993550214 ns Done!
gsr_pipeline also defines a gsr timer, used to measure GSR performance. It prints timer data to the gsr log if the log’s enable mask is set accordingly. To see this GSR performance data, add option --spi_log_mask=gsr,2 to the program’s argument list. 8.8PerformanceYou can use spperf to evaluate the performance of a Stream program simulated by spsim. The Stream Reference Manual describes how to run spperf. To generate performance information using spperf:
Source gsr_pipeline.sc already contains spi_trace_start and spi_trace_stop calls in gsr_pipeline. Type: $ mkdir -p ../build/sp16_profile/bin $ spc -o ../build/sp16_profile/bin/testbench -p -m testbench \ file_io.c gsr_pipeline.sc testbench/spimain.c # compile $ spsim ../build/sp16_profile/bin/testbench --spi_trace_file=spm_demo.sbt \ data/sample.bmp data/result.bmp # simulate $ spperf ../build/sp16_profile/bin/testbench \ spm_demo.sbt spm_demo_tcs.sbt \ -o spm_demo.html # analyze You can open the generated HTML file in any browser. The Optimization chapter gives much more detailed information about performance, including a description of the tables in the HTML file. 9Stream Program DevelopmentThe preceding chapters introduced basic notions of Stream programming and showed how to use command line tools to compile and run a Stream program. This chapter describes Stream program development using the Stream Processors integrated development environment (IDE), called spide. You can use spide to edit, build, execute, and debug Stream programs. spide is based on the Eclipse extensible open development platform. www.eclipse.org gives general information about Eclipse. Stream program development under spide typically steps through a series of development modes:
The programmer uses functional mode to create a functionally correct version of an application, uses profile mode to evaluate its performance, modifies the program based on the performance data, and then repeats this cycle as needed. Once the programmer is satisfied with the result, release mode produces a final version of the program. spide provides two additional modes that are used less frequently:
This chapter uses the spm_demo demo program in directory demos/spm_demo of the Stream distribution as a concrete example. It describes the use of spide to build and run spm_demo in each of the modes described above. 9.1Invoke spideMake sure your PATH includes the Stream distribution bin directory before you invoke spide. Then invoke spide: $ spide & spide keeps information in a directory called a workspace. spide uses the workspace as the default location for any project files you create. You can have multiple workspaces and switch workspaces within the IDE. Each workspace contains one or more projects, where a project is a group of related files (sources, binaries, and data). Each project defines one or more modules; a module represents an executable or library built by spide from the source files in a project. When you build a project with multiple modules, the IDE builds each module in the project. The first time you invoke spide, it displays a banner and then asks you to select a location for your workspace. After you specify a workspace directory, spide displays an empty IDE window (Figure 1): Figure 1: Stream IDE Window 
This window shows the Stream perspective, as indicated near the upper right corner. A perspective is an editor and a group of views that together provide a development environment. You can use the Stream perspective to edit, build, and run Stream applications. A view is a pane within a perspective. The Stream perspective above contains a Stream Projects view on the left, an Outline view on the right, and several additional views at the bottom. You can right-click on the header of a view to maximize, minimize, or detach it. The editor is the area of a window not occupied by a view. Hovering over any icon on the IDE toolbar produces pop-up help information. The IDE toolbar contains pull-down selections for architecture (sp16 or sp8) and mode (functional, functional_fast, debug, profile, or release), as well as pull-down icons for build (the hammer  ), debug (the bug ), and run (the green right arrow ). Other toolbar icons allow you to create a new project, to save the current project, and to print.
The Stream Projects view in Figure 1 above is initially empty, as the workspace does not yet contain any projects. After you import or create a Stream project, the Stream Projects view shows its structure. The project’s structure generally corresponds to the spide workspace directory structure, but it may contain additional “virtual” folders that do not exist in the workspace. A project contains the following top-level folders:
If you delete a project, be sure to delete the project directory from your IDE workspace. Deleting a project from the Stream Projects view does not delete the associated project files. Download 0.95 Mb. Share with your friends:
|