GPU-GEMM generator
Project description
GPU-GEMM generator for the Discontinuous Galerkin method
Installation
For users
pip install gemmforge
For developers
git clone https://github.com/ravil-mobile/gemmforge.git gemmforge
cd gemmforge
pip install -e .
Usage
from gemmforge import DenseMatrix, GemmGenerator, GenerationError
from gemmforge import arch
arch = arch.produce("nvidia", "sm_60")
mat_a = DenseMatrix(num_rows=56,
num_cols=9,
addressing="strided",
bbox=[0, 0, 55, 8],
transpose=False)
mat_b = DenseMatrix(num_rows=9,
num_cols=9,
addressing="strided",
bbox=[0, 0, 8, 8],
transpose=False)
mat_c = DenseMatrix(num_rows=56,
num_cols=9,
bbox=[0, 0, 55, 8],
addressing="strided",
transpose=False)
try:
gen = GemmGenerator(arch, "float")
gen.generate(mat_a, mat_b, mat_c, alpha=1.1, beta=1.1)
print(gen.get_kernel())
print(gen.get_launcher())
print(gen.get_launcher_header())
except GenerationError as err:
print("ERROR: {}".format(err))
raise err
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
gemmforge-0.0.201.tar.gz
(19.4 kB
view hashes)
Built Distribution
Close
Hashes for gemmforge-0.0.201-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 5759428902528ed4e395c5640642fbb8e3c9aa047c44c758fc3720776a41968f |
|
MD5 | 57a5b929a121e4b7a125dff1dfc58fe5 |
|
BLAKE2b-256 | 7912bba5b1e1f54b1ffe609b9595dbd4330e8016391e14c2ba80decd6a1ce519 |