diff options
| -rw-r--r-- | thesis/parts/implementation.tex | 43 |
1 files changed, 30 insertions, 13 deletions
diff --git a/thesis/parts/implementation.tex b/thesis/parts/implementation.tex index 43ed2ba..d175d63 100644 --- a/thesis/parts/implementation.tex +++ b/thesis/parts/implementation.tex @@ -4,24 +4,41 @@ \section{Tooling integration} -As well as a standalone compiler, the Rust toolchain provides an official package manager, which is widely adopted. -We interface with cargo through the \code{cargo metadata} command, which allows us to work seamlessly with virtually any Rust project. +get project metadata from cargo +available benchmarks and source directories +works with most projects +could be expanded to run as cargo command -From cargo, we retrieve information about the user's project source directories, the available benchmarks and tests, and the location of the build cache. +parse minimal amount of information from criterion benchmark +most common benchmarking tool, closest there is to a standard +should be easy to adapt if/when cargo ships proper benchmarking support -\section{Cost model generation} +benchmarking of container implementations also outputs this format, but without using criterion harness +allows more flexibility, ie pre-processing results -Generation of cost models is implemented in three locations: +each trait has its own set of benchmarks, which run different workloads +benchmarker trait doesn't have Ns +example benchmarks for hashmap and vec -\begin{itemize} -\item The \code{candelabra_benchmarker} crate, which provides code for benchmarking anything that implements the primrose library's traits -\item The \code{candelabra::cost::benchmark} module, which generates, runs, and parses benchmarks for container types in the primrose library -\item The \code{candelabra::cost::fit} module, which fits a linear regression model to a set of observations and allows for -\end{itemize} +\code{candelabra::cost::benchmark} generates code which just calls candelabra_benchmarker methods +Ns are set there, and vary from [...] -\todo{Fitting of cost models} -\todo{Profiling wrapper} -\todo{Selection and comparison code} +fitting done with least squares in \code{candelabra::cost::fit} +list other methods tried +simple, which helps 'smooth out' noisy benchmark results + +profiler type in \code{primrose_library::profiler}} +wraps an 'inner' implementation and implements whatever operations it does, keeping track of number of calls +on drop, creates new file in folder specified by env variable + +primrose results generated in \code{primrose::codegen}, which is called in \code{candelabra::profiler} +picks the first valid candidate - performance doesn't really matter for this case +each drop generates a file, so we get details of every individual collection allocated +\todo{immediately aggregate these into summary statistics, for speed} +\todo{mention benchmark repetition} + +estimate a cost for each candidate: op(avg_n) * op_times for each op +pick the smallest one \todo{Caching} \todo{Other implementation details?} |