lava

A Highlevel Python Wrapper for Vulkan's Compute API

https://github.com/osanj/lava

Science Score: 13.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • CITATION.cff file
  • codemeta.json file
    Found codemeta.json file
  • .zenodo.json file
  • DOI references
  • Academic publication links
  • Committers with academic emails
  • Institutional organization owner
  • JOSS paper metadata
  • Scientific vocabulary similarity
    Low similarity (11.1%) to scientific vocabulary

Keywords

gpu numpy parallel-computing shaders vulkan vulkan-compute
Last synced: 6 months ago · JSON representation

Repository

A Highlevel Python Wrapper for Vulkan's Compute API

Basic Info
  • Host: GitHub
  • Owner: osanj
  • Language: Python
  • Default Branch: master
  • Homepage:
  • Size: 207 KB
Statistics
  • Stars: 17
  • Watchers: 3
  • Forks: 4
  • Open Issues: 7
  • Releases: 7
Topics
gpu numpy parallel-computing shaders vulkan vulkan-compute
Created about 7 years ago · Last pushed 6 months ago
Metadata Files
Readme

README.md

alert_status coverage

Lava

Lava allows leveraging Vulkan compute shaders from Python in an easy manner without writing any C++. Like CUDA kernels, Vulkan compute shaders can be used to move heavy computation into the GPU. For most use cases Lava will expect and return numpy arrays. Overall, Lava is similar to PyCuda or PyOpenCL.

Installation

pip install lava

The Vulkan SDK needs to be installed and its environment variables need to be set. See the LunarG guide for Linux and Windows.

See release 0.4.0 for some older Vulkan SDK versions which are no longer available on the LunarG download page.

Support

Sofar lava was tested on: * Nvidia Tesla V100, Ubuntu 18.04 + Nvidia 415 (✓ works) * Nvidia 1080 GTX, Ubuntu 18.04 + Nvidia 415 (✓ works) * Nvidia 1080 GTX, Ubuntu 18.04 + Nvidia 396 (✗ does not work, driver issue) * Nvidia 1080 GTX, Windows 10 (✓ works) * Intel HD, Ubuntu 18.04 + Mesa (✓ works)

Features

  • Automatic memory alignment and parsing of arbitrary complex block interfaces
    • Supported: scalars (int, uint, float, double), vectors, matrices, multidimensional arrays and structs
    • Partially supported: bool
    • Not supported: dynamic arrays
    • Further supported: ssbo's and ubo's, std140 and std430 layouts
    • Multidimensional arrays of scalars, vectors or matrices are expected and parsed as respective numpy arrays
  • CPU, GPU and staged buffers
  • Intuitive shader execution
    • Block definitions are parsed from the compiled shader bytecode
    • Buffers are bound with a single line of code
    • Buffers and shaders are checked for compatibility, other sanity checks are performed

Example

Below is the lava version of Erkaman's vulkan minimal compute:

```python import cv2 as cv import lava as lv import numpy as np

start session on first device

session = lv.Session(lv.devices()[0])

compile glsl code

shaderpath = "shader.comp" shader = lv.Shader(session, lv.compileglsl(shader_path))

allocate buffer and take its definition from the shader (see glsl: float[HEIGHT][WIDTH][3] imageData)

bufferout = lv.StagedBuffer.fromshader(session, shader, binding=0)

compute minimum amount of work groups

workgroupsize = 32 x = int(3200 / float(workgroupsize) + 0.5) y = int(2400 / float(workgroupsize) + 0.5) z = 1

bind buffer to binding 0 (see glsl: layout(std430, binding = 0) buffer buf)

stage = lv.Stage(shader, {0: bufferout}) stage.record(x, y, z) stage.runand_wait()

retrieve output as numpy array

im = buffer_out["imageData"]

cv.imwrite("mandelbrot.png", np.around(255 * im).astype(np.uint8)) ```

```glsl

version 450

extension GLARBseparateshaderobjects : enable

define WIDTH 3200

define HEIGHT 2400

define WORKGROUP_SIZE 32

layout ( localsizex = WORKGROUPSIZE, localsizey = WORKGROUPSIZE, localsizez = 1 ) in;

layout(std430, binding = 0) buffer buf { float[HEIGHT][WIDTH][3] imageData; };

void main() { if(glGlobalInvocationID.x >= WIDTH || glGlobalInvocationID.y >= HEIGHT) return;

float x = float(glGlobalInvocationID.x) / float(WIDTH); float y = float(glGlobalInvocationID.y) / float(HEIGHT);

vec2 uv = vec2(x, y); float n = 0.0; vec2 c = vec2(-.445, 0.0) + (uv - 0.5) * (2.0 + 1.7 * 0.2), z = vec2(0.0); const int M = 128; for (int i = 0; i < M; i++) { z = vec2(z.x * z.x - z.y * z.y, 2. * z.x * z.y) + c; if (dot(z, z) > 2) break; n++; }

float t = float(n) / float(M); vec3 d = vec3(0.3, 0.3 ,0.5); vec3 e = vec3(-0.2, -0.3 ,-0.5); vec3 f = vec3(2.1, 2.0, 3.0); vec3 g = vec3(0.0, 0.1, 0.0); vec3 color = d + e * cos( 6.28318 * (f * t + g) );

int xIndex = int(glGlobalInvocationID.x); int yIndex = int(glGlobalInvocationID.y); imageData[yIndex][xIndex][0] = color.r; imageData[yIndex][xIndex][1] = color.g; imageData[yIndex][xIndex][2] = color.b; } ```

Owner

  • Name: Jonas Schuepfer
  • Login: osanj
  • Kind: user
  • Location: Munich

GitHub Events

Total
  • Watch event: 2
  • Pull request event: 1
Last Year
  • Watch event: 2
  • Pull request event: 1

Committers

Last synced: over 1 year ago

All Time
  • Total Commits: 94
  • Total Committers: 2
  • Avg Commits per committer: 47.0
  • Development Distribution Score (DDS): 0.043
Past Year
  • Commits: 0
  • Committers: 0
  • Avg Commits per committer: 0.0
  • Development Distribution Score (DDS): 0.0
Top Committers
Name Email Commits
Jonas Schuepfer j****r@g****m 90
Jonas Schuepfer 7****j 4

Issues and Pull Requests

Last synced: 6 months ago

All Time
  • Total issues: 13
  • Total pull requests: 13
  • Average time to close issues: 14 days
  • Average time to close pull requests: about 5 hours
  • Total issue authors: 2
  • Total pull request authors: 3
  • Average comments per issue: 1.46
  • Average comments per pull request: 0.0
  • Merged pull requests: 9
  • Bot issues: 0
  • Bot pull requests: 3
Past Year
  • Issues: 0
  • Pull requests: 2
  • Average time to close issues: N/A
  • Average time to close pull requests: N/A
  • Issue authors: 0
  • Pull request authors: 1
  • Average comments per issue: 0
  • Average comments per pull request: 0.0
  • Merged pull requests: 0
  • Bot issues: 0
  • Bot pull requests: 2
Top Authors
Issue Authors
  • osanj (12)
  • gregc-91 (1)
Pull Request Authors
  • osanj (9)
  • dependabot[bot] (3)
  • KurniawanZATY (1)
Top Labels
Issue Labels
enhancement (7) bug (2) spike (2)
Pull Request Labels
dependencies (3) python (2)

Packages

  • Total packages: 1
  • Total downloads:
    • pypi 177 last-month
  • Total dependent packages: 0
  • Total dependent repositories: 4
  • Total versions: 7
  • Total maintainers: 1
pypi.org: lava

Highlevel Wrapper for Vulkan's Compute API

  • Versions: 7
  • Dependent Packages: 0
  • Dependent Repositories: 4
  • Downloads: 177 Last month
Rankings
Dependent repos count: 7.5%
Dependent packages count: 10.1%
Average: 13.3%
Stargazers count: 15.3%
Downloads: 16.6%
Forks count: 16.9%
Maintainers (1)
Last synced: 6 months ago

Dependencies

requirements.txt pypi
  • cffi ==1.11.5
  • future ==0.17.1
  • numpy ==1.16.3
  • pycparser ==2.19
  • pytest ==5.1.0
  • pytest-cov ==2.7.1
  • vulkan ==1.1.99.1
setup.py pypi
  • vulkan *