Keren Zhou Jokeren

We couldn’t find that file to show.

Multiplexer

How many operations are still pending?
How does a thread ensure its activities are transferred before it ends? Because metrics are stored in thread local maps, it needs to get all activities and attribute them before it dies.
```
 opencl-api.c

 device_finalizer_register vs thread finalizer
```

pending_operations

Install spack

git clone https://github.com/spack/spack.git
export SPACK_ROOT=/path/to/spack
export PATH=${SPACK_ROOT}/bin:${PATH}
source ${SPACK_ROOT}/share/spack/setup-env.sh

Clone HPCToolkit

Prepare environment

module swap PrgEnv-intel PrgEnv-gnu
module unload darshan

Install spack

spack configs

	#include <iostream>
	#include <chrono>

	void __attribute__ ((noinline)) init(int *arr, size_t length) {
	for (auto i = 0; i < length; ++i) {
	arr[i] = 1;
	}
	}

	class NumMatrix {
	private:
	std::vector<std::vector<int> > tree;

	void columnInit(std::vector<int> &tree_column, std::vector<int> &matrix_column) {
	size_t columns = matrix_column.size();
	for (size_t i = columns; i < tree_column.size(); ++i) {
	tree_column[i] = matrix_column[i - columns];
	}
	for (size_t i = columns - 1; i > 0; --i) {

	template<type T>
	void build(T[] tree, T[] vals, size_t n) {
	if (n == 0) {
	return;
	}
	for (auto i = n; i < 2 * n; ++i) {
	tree[i] = vals[i];
	}
	for (auto i = n - 1; i > 0; --i) {
	tree[i] = tree[i * 2] + tree[i * 2 + 1];

	I left files that I am not familar with.

	spack/package.py \| 131 ++
	spack/packages.yaml \| 16 +-

	Add new packages for NVIDIA GPUs

	src/lib/analysis/Args.hpp \| 3 +

	Initialize instruction mix analysis files. All the features are implemented in master-gpu-instruction-analyzer branch.