Intel® Software Network

Return to main Intel® Trace Analyzer and Collector page

Intel® Trace Analyzer and Collector 7.1 for Linux* or Windows* Compute Cluster Server 2003 allows you to analyze MPI performance, speed up parallel application runs, locate hotspots and bottlenecks, and compare trace files with graphics providing extensively detailed analysis and aligned timelines. Below are screenshots of some of the many views:


Timeline Views and Parallelism Display

• Displays concurrent behavior of parallel applications
• Displays application activities, event source code locations, and message-passing along time axis
• (see Figure 1)

Advanced GUI

• Manages displays and statistics with fast interface
• Uses object point-and-zoom for enhanced detail, context-sensitive sub-menus
• Couples displays to allow focused analysis
• Supplies automatic updates of recomputed statistics
• Provides timeline displays, call-graph, performance profile for function groups and communication in a specific phase of parallel execution

Display Scalability

• Navigates through trace data levels of abstraction: cluster, node, process, thread, and function hierachies (classes)

Detailed and Aggregate Views

• Examines aspects of application runtime behavior, grouped by functions or processes
• Easily identifies the amount of time spent e.g. in MPI communication
• Easily see the performance differences between two program runs (see Figure 2)

Ease of Use

• Offers user-friendly application programming interface (API) to control or record user events
• Adds versatile recording and analysis of counter data (see Figure 3)

Metrics Tracking

Communication Statistics

• Displays communication patterns of parallel applications
• Displays metrics for an arbitrary time interval
• Through comparison it keeps track of the performance increase of an algorithm change
• Figure 4 shows the same algorithm with synchronous communication (left) vs. asynchronous communication (right) with communication overhead (red)

Execution Statistics

• Provides subroutine execution metrics down to the level of call-tree characteristics
• Though binary, compiler-driven and source code instrumentation it offers very detailed analysis beyond MPI

Profiling Library

• Records distributed, event-based trace data

Statistics Readability

• Logs information for function calls, sent messages, and collective operations

Return to main Intel® Trace Analyzer and Collector page