Posted on 05/04/16 0
Intel® VTune™ Amplifier provides an option to view a hierarchy of the loops in your application call tree and identify code sections for optimization. We used option "-knob collection-detail=stack-call-and-tripcount" to collect data. # amplxe-cl -c advanced-hotspots -knob collection-detail=stack...
Posted on 04/07/16 0
Python is an interpreting language and it doesn’t use compiler to generate binary execution code. Cython is an interpreting language too (but C-extension), it can be built to native code. VTune™ Amplifier XE 2017 beta can fully support of reporting hot functions both in Python code and Cython cod...
Posted on 03/18/16 0
The user might be aware of two kinds of problem when using VTune(TM) Amplifier XE to profile an application which uses signal or signalstack. Case 1. Limited signal stack sizeSymptom: Application runs smoothly without VTune(TM) Amplifier XE, but if use VTune Amplifier to profile, will get error...
Posted on 02/14/16 0
VTune Amplifier XE 2016 版部分值得尝试的新特性： 1. Driveless hardware event-based sampling 由于某种原因，VTune的驱动在某些环境下不能安装，如没有root权限或系统是基于虚拟的环境，等等。但是Perf工具是可以运行的，可惜没有图形界面，不便于分析。VTune™ Amplifier XE 2016 可以实现与Perf的无缝对接，也即，直接使用VTune Amplifier的命令，可以采集Perf的性能数据，其结果可以在VTune Amplifier的图形界面浏览。具体可以看这篇博客 2. Hardwa...
Posted on 01/21/16 1
Former VTune(TM) Amplifier XE can be installed in guest Linux* operation system based on Virtual Machine, however only user mode data collectors (they are basic hotspots, concurrency, locksandwaits analysis) can be supported, event-based cannot be supported because of "invisible" hardware PMU. S...
Posted on 12/10/15 0
Background VTune™ Amplifier XE provides data collector, command line, Graphic User Interface components. Some systems have no GUI supported (no GNU* X11 libraries installed, may act as target device with customized operation system, etc), so only data collector and command line of VTune Amplifie...
Posted on 11/16/15 0
We know that user can check document of VTune(TM) Amplifier XE to know supported events for any specific Intel(R) processor. Another way, on GUI if user modifies any existing event-based analysis to add/remove event(s), the combo-list box displays all supported events for current running system,....
New Memory Access Analysis in VTune™ Amplifier XE 2016 Supports of Tracing Memory Allocation/De-allocation Used in Your Code
Posted on 09/10/15 0
New memory access analysis in XE 2016 version, combines functions from old Memory Access and Bandwidth. And there is a new feature to trace system memory allocation / de-allocation in your code. Note that this feature uses of instrumenting system memory allocation / de-allocation APIs, it will increase additional runtime overhead. The limitations are: 1. It is only available in Linux* operation system 2. Only support Sandy Bridge processors or later
Posted on 08/12/15 0
If you compile and run your OpenMP* code with Intel Compiler 13.1 Update 2 or later, use advanced-hotspots from VTune(TM) Amplifier XE 2015 Update 4 to get important metrics, they can be categorized into "Serial Time" and "Parallel Region Time". Also “OpenMP Potential Gain” is provided to let you know if you have more works to optimize code. Meanwhile, VTune Amplifier highlight on:
Use "column" option to display data on selective columns in the report of Intel® VTune™ Amplifier XE
Posted on 07/15/15 0
Tutorial showing how to use the column option in Intel® VTune™ Amplifier XE to display data on selective columns in report output.