uber-archive/go-torchArchived
Go Torch
go-torch is a profiling tool for capturing the execution state of Go programs and transforming raw binary data into visual representations of program performance. It functions as a flame graph profiler and performance visualization utility that identifies expensive code paths through the collection of CPU and memory stack traces.
The tool features a network-capable remote process profiler that connects to endpoints to capture and export execution profiles from Go binaries. It utilizes stochastic profiling to synthesize execution data into call graphs, allowing for the identification of bottlenecks in running applications.
The project covers the full pipeline of profiling data collection, from stack trace sampling and remote profile acquisition to call graph synthesis. These workflows culminate in the generation of SVG-based flame graphs that visualize function depth and execution frequency.
Features
- Flame Graph Visualizers - Generates SVG-based flame graphs from processed call graph data to visualize function depth and execution frequency.
- CPU Profilers - Captures CPU and memory usage data from remote endpoints or local binaries to generate raw call graphs.
- Remote Profiling Streams - Retrieves binary execution data over a network connection to analyze processes running on separate host machines.
- Synthesis Workflows - Transforms raw stack trace samples into a hierarchical tree structure representing program execution flow.
- Call Stack Samplers - Captures periodic snapshots of the call stack to statistically estimate resource usage without recording every function call.
- Remote Profiling Interfaces - Connects to remote endpoints to capture and export execution profiles from Go binaries over a network.
- Application Stack Trace Capturers - Captures the execution state of running programs to identify where time is spent and locate bottlenecks.
- Go Runtime Profilers - Provides a profiler that reads Go runtime profiling data to generate call graphs and flame graphs.
- Stochastic Profilers - Utilizes stochastic sampling to synthesize execution data into call graphs for bottleneck identification.
- Application Performance Profiling - Captures CPU and memory usage from running Go processes over a network to debug production performance.
