Valgrind - Overview

Recently, I explored Valgrind's capabilities. In this post, I will provide a high level overview of this tool.

Valgrind is a GPLv2 tool that can be used to find functional and performance issues in linux user space C/C++ applications. It is architected as an extendable/modular framework to which a custom sub-tool can be hooked without touching core part of Valgrind. In this modular framework, Valgrind offers each of it's standard capabilities as sub-tools. Valgrind today has 6 production quality sub-tools & 3 experimental sub-tools.

The standard sub-tools help to:

• detect run-time memory errors - leaks, corruptions, buffer over tuns etc (memcheck)
• detect locking issues in multi-thread programming (helgrind, DRD)
• debug performance issues
• Instruction & Data (L1/L2) cache hits/misses at per-function, per-module whole program level & also detail the code which is causing the misses (cachegrind, callgrind)
• code profiling (use callgrind & then feed output of callgrind to visualization tool Kcachegrind)
• detect memory hogs using memory snapshot profiling (massif)

The experimental sub-tools help to:

• detect stack and global array corruptions using behavioral heuristics (SGCheck)
• heap usage statistics that help optimize improve data structure field grouping, reduce unnecessary allocations, etc.. (DHAT)
• cpu cycle level profiler that helps in architecture research (BBV)

Among these use cases, Valgrind is popular for detecting run time memory errors (memcheck sub-tool). So, I will cover Valgrind's memcheck sub-tool in my next posts - Valgrind's Memcheck: Detect user space run time memory errors, Valgrind's Memcheck + Valkyrie: Detect user space run time memory errors.

Reference:
http://valgrind.org/