sgemm_gpu_kernel_performance (Q6037794)
From MaRDI portal
OpenML dataset with id 44961
Language | Label | Description | Also known as |
---|---|---|---|
English | sgemm_gpu_kernel_performance |
OpenML dataset with id 44961 |
Statements
11
0 references
**Data Description**\N\NThis data set measures the running time of a matrix-matrix product A\*B = C, where all matrices have size 2048 x 2048, using a parameterizable SGEMM GPU kernel with 241600 possible parameter combinations. For each tested combination, 4 runs were performed and their results are reported as the 4 last columns. All times are measured in milliseconds.\N\NThere are 14 parameter, the first 10 are ordinal and can only take up to 4 different powers of two values, and the 4 last variables are binary. Out of 1327104 total parameter combinations, only 241600 are feasible (due to various kernel constraints). This data set contains the results for all these feasible combinations.\N\NThe experiment was run on a desktop workstation running Ubuntu 16.04 Linux with an Intel Core i5 (3.5GHz), 16GB RAM, and a NVidia Geforce GTX 680 4GB GF580 GTX-1.5GB GPU. We use the 'gemm_fast' kernel from the automatic OpenCL kernel tuning library 'CLTune'.\N\N\N**Attribute Description**\N\N1-2. *MWG*, *NWG* - per-matrix 2D tiling at workgroup level: {16, 32, 64, 128} (integer)\N3. *KWG* - inner dimension of 2D tiling at workgroup level: {16, 32} (integer)\N4-5. *MDIMC*, *NDIMC* - local workgroup size: {8, 16, 32} (integer)\N6-7. *MDIMA*, *NDIMB* - local memory shape: {8, 16, 32} (integer)\N8. *KWI* - kernel loop unrolling factor: {2, 8} (integer)\N9-10. *VWM*, *VWN* - per-matrix vector widths for loading and storing: {1, 2, 4, 8} (integer)\N11-12. *STRM*, *STRN* - enable stride for accessing off-chip memory within a single thread: {0, 1} (categorical)\N13-14. *SA*, *SB* - per-matrix manual caching of the 2D workgroup tile: {0, 1} (categorical)\N15-18. *Run1*, *Run2*, *Run3*, *Run4* - performance times in milliseconds for 4 independent runs using the same parameters, ranging between 13.25 and 3397.08.
0 references
22 December 2022
0 references
Run1
0 references
https://doi.org/10.1109/MCSoC.2015.10
0 references
0
0 references
0
0 references
15
0 references
241,600
0 references
0
0 references
0
0 references
15
0 references
0
0 references