stu
Build automation (by kunegis)
gpgpu-loadbalancerx
Simple load-balancing library for balancing GPGPU workloads between a GPU and a CPU or any number of devices in a computer or multiple computers. (by tugrul512bit)
stu | gpgpu-loadbalancerx | |
---|---|---|
1 | 4 | |
37 | 1 | |
- | - | |
6.7 | 2.6 | |
2 months ago | over 2 years ago | |
C++ | C++ | |
GNU General Public License v3.0 only | GNU General Public License v3.0 only |
The number of mentions indicates the total number of mentions that we've tracked plus the number of user suggested alternatives.
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
stu
Posts with mentions or reviews of stu.
We have used some of these posts to build our list of alternatives
and similar projects. The last one was on 2022-02-14.
-
C++ Show and Tell - Experiment
Stu – Build tool written in C++14, intended for large data science projects (rather than for compilation). Can be compared to Make, but with special features that are hard/impossible to recreate, e.g. output/plot.[languages.txt].eps will build all files output/plot.$lang.eps, for $lang taken from the file languages.txt. It all sounds very simple but has turned out to be extremely useful for generating the website http://konect.cc/ ; for years I had used and researched other tools, and none was really adequate.
gpgpu-loadbalancerx
Posts with mentions or reviews of gpgpu-loadbalancerx.
We have used some of these posts to build our list of alternatives
and similar projects. The last one was on 2022-02-14.
-
vectorAdd.cu sample load-balanced on 3 GPUs
/** * Copyright 1993-2015 NVIDIA Corporation. All rights reserved. * * Please refer to the NVIDIA end user license agreement (EULA) associated * with this source code for terms and conditions that govern your use of * this software. Any use, reproduction, disclosure, or distribution of * this software and related documentation outside the terms of the EULA * is strictly prohibited. * */ /** * Vector addition: C = A + B. * * This sample is a very basic sample that implements element by element * vector addition. It is the same as the sample illustrating Chapter 2 * of the programming guide with some additions like error checking. */ #include // For the CUDA runtime routines (prefixed with "cuda_") #include #include // for load balancing between 3 different GPUs // https://github.com/tugrul512bit/gpgpu-loadbalancerx/blob/main/LoadBalancerX.h #include "LoadBalancerX.h" /** * CUDA Kernel Device code * * Computes the vector addition of A and B into C. The 3 vectors have the same * number of elements numElements. */ __global__ void vectorAdd(const float *A, const float *B, float *C, int numElements) { int i = blockDim.x * blockIdx.x + threadIdx.x; if (i < numElements) { C[i] = A[i] + B[i]; } } #include #include int main(void) { int numElements = 15000000; int numElementsPerGrain = 500000; size_t size = numElements * sizeof(float); float *h_A = (float *)malloc(size); float *h_B = (float *)malloc(size); float *h_C = (float *)malloc(size); for (int i = 0; i < numElements; ++i) { h_A[i] = rand()/(float)RAND_MAX; h_B[i] = rand()/(float)RAND_MAX; } /* * default tutorial vecAdd logic cudaMemcpy(d_A, h_A, size, cudaMemcpyHostToDevice); cudaMemcpy(d_B, h_B, size, cudaMemcpyHostToDevice); int threadsPerBlock = 256; int blocksPerGrid =(numElements + threadsPerBlock - 1) / threadsPerBlock; vectorAdd<<>>(d_A, d_B, d_C, numElements); cudaGetLastError(); cudaMemcpy(h_C, d_C, size, cudaMemcpyDeviceToHost); */ /* load-balanced 3-GPU version setup */ class GrainState { public: int offset; int range; std::map d_A; std::map d_B; std::map d_C; ~GrainState(){ for(auto a:d_A) cudaFree(a.second); for(auto b:d_B) cudaFree(b.second); for(auto c:d_C) cudaFree(c.second); } }; class DeviceState { public: int gpuId; int amIgpu; }; LoadBalanceLib::LoadBalancerX lb; lb.addDevice(LoadBalanceLib::ComputeDevice({0,1})); // 1st cuda gpu in computer lb.addDevice(LoadBalanceLib::ComputeDevice({1,1})); // 2nd cuda gpu in computer lb.addDevice(LoadBalanceLib::ComputeDevice({2,1})); // 3rd cuda gpu in computer // lb.addDevice(LoadBalanceLib::ComputeDevice({3,0})); // CPU single core for(int i=0;i( [&,i](DeviceState gpu, GrainState& grain){ if(gpu.amIgpu) { cudaSetDevice(gpu.gpuId); cudaMalloc((void **)&grain.d_A[gpu.gpuId], numElementsPerGrain*sizeof(float)); cudaMalloc((void **)&grain.d_B[gpu.gpuId], numElementsPerGrain*sizeof(float)); cudaMalloc((void **)&grain.d_C[gpu.gpuId], numElementsPerGrain*sizeof(float)); } }, [&,i](DeviceState gpu, GrainState& grain){ if(gpu.amIgpu) { cudaSetDevice(gpu.gpuId); cudaMemcpyAsync(grain.d_A[gpu.gpuId], h_A+i, numElementsPerGrain*sizeof(float), cudaMemcpyHostToDevice); cudaMemcpyAsync(grain.d_B[gpu.gpuId], h_B+i, numElementsPerGrain*sizeof(float), cudaMemcpyHostToDevice); } }, [&,i](DeviceState gpu, GrainState& grain){ if(gpu.amIgpu) { int threadsPerBlock = 1000; int blocksPerGrid =numElementsPerGrain/1000; vectorAdd<<>>(grain.d_A[gpu.gpuId], grain.d_B[gpu.gpuId], grain.d_C[gpu.gpuId], numElements-i); } else { for(int j=0;j de(3); for(int i=0;i<100;i++) { nanoseconds += lb.run(); } for(auto v:de) std::cout<
- I created a load-balancer for multi-gpu projects.
-
C++ Show and Tell - Experiment
Here is Nvidia's vectorAdd example modified for 3-GPU load balancing.
What are some alternatives?
When comparing stu and gpgpu-loadbalancerx you can also consider the following projects:
libletlib - C++ framework for the impatient.
Blackjack_V1.02 - Extension of my old Blackjack game with Qt for C++
osmanip - A cross-platform library for output stream manipulation using ANSI escape sequences.
SHA256-Implementation - A program that implements the SHA256 algorithm and generates the binary+hexdigest of a string input.
TensorComprehensions - A domain specific language to express machine learning workloads.
ftl - Freestanding template library
dmpower - Interactive terminal D&D helper toolbox program for Dungeon Masters, players, and worldbuilders.
stu vs libletlib
gpgpu-loadbalancerx vs libletlib
stu vs Blackjack_V1.02
gpgpu-loadbalancerx vs osmanip
stu vs osmanip
gpgpu-loadbalancerx vs SHA256-Implementation
stu vs TensorComprehensions
gpgpu-loadbalancerx vs Blackjack_V1.02
stu vs ftl
gpgpu-loadbalancerx vs dmpower
stu vs SHA256-Implementation
stu vs dmpower