LibJacket | Documentation | LibJacket is designed for engineers, scientists, and analysts who want maximum performance and maximum leverage of GPU resources, without hassling with messy low-level programming details. It runs on NVIDIA GPUs in any system, including servers, workstations, and laptops.
LibJacket is Comprehensive- It supports single- and double-percision floating point values, complex numbers, and booleans.
- It supports manipulating vectors, matrices, and N-dimensional arrays.
- It includes routines for arithmetic, linear algebra, statistics, imaging, signal processing, and related algorithms.
| |  Learn More  |
LibJacket is Fast- It outperforms CPU libraries
- It is optimized for any CUDA-enabled GPU. The same code will run on everything: laptops, desktops, servers.
- It includes thousands of lines of optimized device code.
- It defers computation so it can analyze and batch instructions to increase arithmetic intensity and memory throughput while avoiding unnecessary temporary allocations.
- It combines and enhances all the best CUDA libraries available, including the fastest FFT, BLAS, and LAPACK implementations.
Example UsageThe library can be used by itself in C/C++ applications or integrated with your existing CUDA code. Here's a stripped down example of Monte-Carlo estimation of Pi: #include <stdio.h>
#include <jacket.h>
using namespace jkt;
int main() {
int n = 20e6; // 20 million random samples
f32 x = f32::rand(n,1), y = f32::rand(n,1);
// how many fell inside unit circle?
float pi = 4.0 * sum_vector(sqrt(x*x + y*y) < 1) / n;
printf("pi = %g\n", pi);
return 0;
}
Base LicenseThe base LibJacket license enables you to access all LibJacket functions and core runtime system for use on a single GPU. Addon Products- LibJacket MGL enables Jacketized code to run on multiple GPUs simultaneously in a workstation, supporting from 2 to 8 total GPUs.
- LibJacket HPC enables Jacketized code to run on multiple GPUs simultaneously in a cluster.
- LibJacket DLA adds many advanced double-precision linear algebra functions.
- LibJacket SLA adds sparse matrix support and linear algebra functions.
- LibJacket PGI Compatibility adds support for PGI compilers. This is included with LibJacket Base License.
| LibJacket | Documentation | LibJacket是專為希望能達到最高的性能和最大限度地利用GPU資源，而不想被低級編程細節被麻煩到的工程師，科學家和分析師們，。
它運行在任何NVIDIA GPU系統，包括服務器，工作站和筆記本電腦的。
LibJacket is Comprehensive它支持單一精度和雙精度浮點值，複數，和邏輯數據（booleans）。 支持操縱向量，矩陣和N維數組。它包括算術，線性代數，統計，成像，信號處理，及相關演算法的常式。 | | Learn More |
LibJacket is Fast· 它優於 CPU 資料庫 · 它為啟用CUDA的GPU優化。相同的代碼將一切上運行：筆記本 電腦，臺 式機，服務器上運行。 · 它包括數千行優化代碼設備。 · 它推遲計算為了分析和處理指令，同時增加算術強度和內存輸送量，避免不必要的臨時配分。 · 它結合並提高了所有CUDA最好的庫，包括最快的FFT，BLAS和LAPACK的設備。
Example Usage資料庫可以被用在C/C++的應用程式或結合你現有的CUDA code: 範例: Monte-Carlo estimation of Pi: #include <stdio.h>
#include <jacket.h>
using namespace jkt;
int main() {
int n = 20e6; // 20 million random samples
f32 x = f32::rand(n,1), y = f32::rand(n,1);
// how many fell inside unit circle?
float pi = 4.0 * sum_vector(sqrt(x*x + y*y) < 1) / n;
printf("pi = %g\n", pi);
return 0;
}
Addon Products- LibJacket MGL enables Jacketized code to run on multiple GPUs simultaneously in a workstation, supporting from 2 to 8 total GPUs.
- LibJacket HPC enables Jacketized code to run on multiple GPUs simultaneously in a cluster.
- LibJacket DLA adds many advanced double-precision linear algebra functions.
- LibJacket SLA adds sparse matrix support and linear algebra functions.
- LibJacket PGI Compatibility adds support for PGI compilers. This is included with LibJacket Base License.
|
|
