stackprof VS speedscope

stackprof can be used alone/by itself to generate flamegraphs for arbitrary Ruby code.

I’d use something like stackprof ( https://github.com/tmm1/stackprof ) to see where the time is going. If you already have suspicions you can use it to get information about a specific method / few lines of Ruby but there’s also a rack middleware.

stackprof

Other profilers, such as stackprof, trace everything that’s happening by line. These types of profilers usually need some instrumentation to be configured, as shown below:

Either https://github.com/tmm1/stackprof for cpu or https://github.com/SamSaffron/memory_profiler for memory. In practice profiling and removing allocations also gives a large perf boost.

At first, I tried to benchmark the method using the benchmark library. (The profiler uses stackprof)

Glad to hear you like it! Those flame graph screenshots are taken from https://www.speedscope.app/ .

Looks like speedscope. https://www.speedscope.app/

Flamegraphs simply visualize this process by placing each of these recorded stacks side by side. The resulting visualization looks like "flames", hence a "flame graph". If you do this visualization where the "parent" of all the stack frames is on the top, rather than the bottom, you get a "waterfall graph", because it looks like a waterfall. It's the same thing though. Speedscope and DevTools visualize using the waterfall format, but I still call them flamegraphs anyway.

While doing research on Ruby profiling I found Shopify's blog post on "How to Fix Slow Code in Ruby". Though the entire post was extremely insightful, it lead me to Shopify's app_profiler library, which can be used to automatically profile code and redirect the output to a local instance of speedscope. Having worked previously with Flame Graphs of CPU stack traces collected using perf.