Intel® VTune™ Amplifier
This topic provides a list of new features for all existing versions of the Intel® VTune™ Amplifier. Typically, each new version of the product includes all features released in the previous product version unless a feature is specific to a particular market segment. For example, support for Android* application analysis is specific to the mobile and embedded platform segment addressed by the Intel System Studio, so this type of update is covered in the what's new list for the VTune Amplifier for Systems only.
Optimized project configuration that provides a straightforward workflow for specifying an analysis target (formerly, Project Properties dialog box) and analysis type within the same Choose Target and Analysis Type window
Optimized configuration for the Microarchitecture Analysis that now includes platform-independent analysis types
New Memory Access analysis that helps identify memory-related issues, including bandwidth issues formerly covered by the Bandwidth analysis (now deprecated). For Linux targets, Memory Access analysis can be configured to attribute performance events to memory objects (data structures).
Hardware Events viewpoint, displaying an estimated count and/or the number of collected samples, replaced Hardware Event Counts and Hardware Event Sample Counts viewpoints
Support for GPU hardware metrics and GPU Architecture Diagram available for systems with Intel HD Graphics
Analysis of applications in a virtualization environment (KVM) for Linux kernels (version 3.2 and higher) and QEMU (version 1.4 and higher)
Task analysis for Android* framework and Linux* Ftrace events extended with an option to set up the threshold for slow/fast task duration in the Summary window
Remote event-based sampling analysis on SoFIA leveraging an existing sampling driver on the target
Support for the ? CLI argument for filter, group-by and column options to get a list of option values available for the specified report
CLI reports synchronized with the data provided in the Top-down and Bottom-up views of the GUI viewpoints
Intel Processor Event Reference integrated into the VTune Amplifier help system
Energy analysis support for Linux targets via Intel SoC Watch collector
New VTune Amplifier for Systems Installation Guide providing detailed instructions on installing the product, building/loading drivers, and so on
Optimized project configuration that provides a straightforward workflow for specifying an analysis target (formerly, Project Properties dialog box) and analysis type within the same Choose Target and Analysis Type window
Optimized configuration for the Microarchitecture Analysis that now includes platform-independent analysis types
New Memory Access analysis that helps identify memory-related issues, including bandwidth issues formerly covered by the Bandwidth analysis (now deprecated). For Linux targets, Memory Access analysis can be configured to attribute performance events to memory objects (data structures).
OpenMP* analysis extended to display atomic operations cost and iteration counts (min/max/average) per parallel loop and identifying the iterations insufficient to saturate working threads
Hardware Events viewpoint, displaying an estimated count and/or the number of collected samples, replaced Hardware Event Counts and Hardware Event Sample Counts viewpoints
Support for GPU hardware metrics and GPU Architecture Diagram available for systems with Intel HD Graphics
New trace-mpi knob CLI option and Trace MPI GUI target configuration option for enabling collectors to determine MPI rank IDs in case of a non-Intel MPI library implementation.
Support for the ? CLI argument for filter, group-by and column options to get a list of option values available for the specified report
CLI reports synchronized with the data provided in the Top-down and Bottom-up views of the GUI viewpoints
Intel Processor Event Reference integrated into the VTune Amplifier help system
New VTune Amplifier XE Installation Guide providing detailed instructions on installing the product, building/loading drivers, and so on
Information on packet submission available on the Timeline for Intel Media SDK applications running on Linux systems with Intel HD Graphics
Support for the __itt_detach API to detach collection from all processes
Attach to Process target type supported for user-mode sampling and tracing analysis of Java* applications on Linux OS
MPI rank ID embedded to an MPI process name to better distinguish multiple ranks for MPI analysis
Basic Hotspots, Locks and Waits and hardware event-based stack sampling collection supported for RT kernel and RT applications for Linux targets
Hardware event-based stack sampling collection supported for kernel-mode threads
KVM guest OS profiling based on the Linux Perf* tool (for Linux targets)
Support for Intel® Atom™ x7 Z8700 and x5 Z8500/X8400 processor series (Cherry Trail) including GPU analysis
OpenMP* analysis enhancements with:
Potential Gain expansion by parallelization inefficiencies representing their wall time cost
Precise trace-based imbalance calculation that is especially useful for profiling of small region instances
Detailed analysis by barrier-to-barrier region segments to explore performance of OpenMP work-sharing constructs and barrier cost inside a region
OpenMP and MPI multi-rank analysis on a compute node with:
Per-rank Intel MPI communication busy wait time detection and displaying the metric in the Summary, grid and Timeline views
Selective rank profiling for MPI applications, including PMU analysis for multiple ranks on a node by using Intel MPI library 5.0.3 or higher
VTune Amplifier XE command line generation for selective rank profiling through Intel Trace Analyzer and Collector user interface
Super Tiny display mode added for the Timeline pane to easily identify problem areas for results with multiple processes/threads
Platform window replacing Tasks and Frames window and providing CPU, GPU, and Bandwidth metrics data distributed over time
General Exploration analysis views extended to display confidence indication (greyed out font) for non-reliable metrics data resulted, for example, from the low number of collected samples
GPU usage analysis for OpenCL™ applications extended to display compute-originated batch buffers on the GPU software queue in the Timeline pane (Linux* target only)
New filtering mode for command line reports to display data for the specified column names only
Intel® Transactional Synchronization Extensions (Intel TSX) Hotspots analysis providing Precise Clockticks data for the Intel microarchitecture code name Haswell
Bandwidth analysis improvements including:
Support for the 5th Generation Intel® Core™ processors (code name: Broadwell) and Intel microarchitecture code name Silvermont
Intel® QuickPath Interconnect (Intel® QPI) Bandwidth data analysis for server platforms
Total, Read and Write Bandwidth timeline areas merged into single area
Grouping by package for the CPU Time timeline area
GPU analysis improvements including:
Support for the OpenCL application analysis (for Intel HD Graphics) and GPU usage analysis (except for the GPU hardware metrics) for Linux* targets
Intel Media SDK program analysis for Linux systems with Intel HD Graphics
Compute extended counter set support for GPU analysis on the 5th Generation Intel® Core™ processors (code name: Broadwell)
Compute basic counter set replacing the Global/local accesses set for GPU analysis
Loop trip count collection level supported for the Advanced Hotspots and custom hardware-event-based sampling analysis types
Support for a collection of system-wide context switches that enable an analysis of detailed scheduling layout for all threads on the system and identify the nature of context switches for a thread (preemption or synchronization)
Configuration for Android framework and Linux* Ftrace events collection via Events Library in the Custom Analysis
Support for analysis of ITT API applications running in the Attach to Process mode
Average calculation for bandwidth and power metrics added for Energy Analysis data
Support for Perf* based driverless hardware event-based sampling collection with stacks
New Analyzing Energy Usage on an Android* Platform tutorial, providing instructions for analyzing energy usage on an Android target system.
Spin and Overhead Time metrics classified by reasons for OpenMP* analysis
Direct access to the source analysis for an OpenMP parallel region when using the /OpenMP Region/.. granularity in the grid
Support for Intel microarchitecture code name Haswell-EP and Broadwell
Intel Energy Profiler analysis support for remote Windows* target systems via data collection with the Intel SoC Watch command line tool and importing the collected data to the VTune Amplifier project on the host
CPU Sleep States and CPU Frequency analysis appear under Energy Analysis, instead of the previous Power Analysis
GPU analysis on remote Android* target systems (with Intel HD Graphics) from supported host systems.
CPU/GPU Concurrency analysis using a predefined configuration to explore GPU usage per basic hardware metrics and correlate this data with the CPU usage at the same time frames
Support for custom data collection of Android framework events and processor chipset events
Lightweight integration of the VTune Amplifier to the Eclipse* IDE with the toolbar access to the standalone version of the VTune Amplifier
Java analysis support on Android systems for the code generated with the ART* compiler
OS X* host support for performing automated remote collection on other platforms and viewing collected results
Support for driverless hardware event-based sampling (EBS) collection for Advanced Hotspots, General Exploration and Custom EBS analysis types on Linux and Android systems
New Finding Hotspots on an Android Platform tutorial accompanied with the C++ code example and providing instructions for the Basic Hotspots analysis on a remote Android system
New SIMD Width metric added to the list of supported OpenCL™ kernel analysis metrics
Extended OpenMP analysis providing a CPU utilization histogram and metrics on serial time and potential tuning gain for parallel regions
Stack collection enabled for Intel Xeon Phi™ coprocessor analysis types
Improved Intel Xeon Phi coprocessor analysis workflow using the Target system configuration options (GUI and CLI mode)
An option to create custom grouping levels and apply them to grid views
Support for importing *.perf files with the event-based sampling data collected by the Linux Perf tool
An option to limit the call stack size and minimize collection overhead for custom hardware event-based sampling analysis results
An option to display verbose collection and finalization messages in the Collection Log window
Support for importing csv files with instantaneous data collected out of the VTune Amplifier with the external collector
Improved OpenMP region analysis with the OpenMP Region grouping level in the Bottom-up window and the Region Instance Duration histogram in the Summary window
Auto-rebuild of sampling and power drivers at system boot time after Linux* kernel update
OS X* host support for performing automated remote collection on other platforms and viewing collected results
New TSX Exploration analysis type for 4th generation Intel® Core™ processors based on the Intel microarchitecture code name Haswell
Summary window extended with hyperlinks for Top Hotspots and performance metrics navigating to the Bottom-up grid view
Call Stack Mode default setting changed from Only user functions to User functions + 1 for better understanding of library usage