stuff.cpp

#include <benchmark/benchmark.h>
#include <stan/math.hpp>
#include <utility>
#include "toss_me.hpp"
#include "callback_bench_impl.hpp"

static void tcrossprod(benchmark::State& state) {
  using stan::math::var;
  using stan::math::promote_scalar;
  using stan::math::tcrossprod;
  using stan::math::sum;
  Eigen::MatrixXd x_val = Eigen::MatrixXd::Random(state.range(0), state.range(0));

  for (auto _ : state) {
    var lp = 0;
    auto start = std::chrono::high_resolution_clock::now();
    {
    auto x = promote_scalar<var>(x_val);
    auto x_cross = tcrossprod(x);
    lp += sum(x_cross);
    }
    lp.grad();
    auto end = std::chrono::high_resolution_clock::now();
    auto elapsed_seconds =
      std::chrono::duration_cast<std::chrono::duration<double>>(end - start);

    state.SetIterationTime(elapsed_seconds.count());
    stan::math::recover_memory();
    benchmark::ClobberMemory();
  }
}

// The start and ending sizes for the benchmark
int start_val = 2;
constexpr int end_val = 4096;
constexpr int extra_alloc = 4096 * 1.2;
BENCHMARK_TEMPLATE(toss_me, extra_alloc);
BENCHMARK(tcrossprod)->RangeMultiplier(2)->Range(start_val, end_val)->UseManualTime();
BENCHMARK_MAIN();