This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| """Matmul K-block size vs numerical error, standalone. | |
| C = A @ B, tiled over BLOCK_M x BLOCK_N x BLOCK_K. Sweeps BLOCK_K -- the | |
| number of serial tl.dot steps K is split into (K/BLOCK_K) -- and measures | |
| error against an fp64 reference. tl.dot is opaque from Triton source, so | |
| --dump-ptx exists to count the mma.sync/fma per BLOCK_K directly. | |
| python matmul_error_vs_reduction_depth.py --K 1024 4096 16384 --dtype fp16 fp32 | |
| python matmul_error_vs_reduction_depth.py --dump-ptx --dtype fp16 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| """Reduction depth vs numerical error, standalone. | |
| Row-wise sum in Triton. BLOCK_SIZE trades serial chunk depth (N/BLOCK_SIZE) | |
| against in-chunk tree depth (log2(BLOCK_SIZE)); error is measured against an | |
| fp64 reference and plotted against the sum of the two. For fp16 storage both | |
| accumulator dtypes are swept, so that pair differs in exactly one variable. | |
| python accumulation_error_vs_reduction_depth.py --N 1024 4096 16384 --dtype fp16 fp32 | |
| python accumulation_error_vs_reduction_depth.py --dump-ptx --dtype fp16 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| #!/usr/bin/env python3 | |
| """Simple HTTP Server With Upload. | |
| This module builds on BaseHTTPServer by implementing the standard GET | |
| and HEAD requests in a fairly straightforward manner. | |
| see: https://gist.github.com/UniIsland/3346170 | |
| """ | |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| #!/usr/bin/env python3 | |
| """ | |
| Very simple HTTP server in python for logging requests | |
| Usage:: | |
| ./server.py [<port>] | |
| """ | |
| from http.server import BaseHTTPRequestHandler, HTTPServer | |
| import logging | |
| class S(BaseHTTPRequestHandler): |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| #!/usr/bin/env python | |
| """Simple HTTP Server With Upload. | |
| This module builds on BaseHTTPServer by implementing the standard GET | |
| and HEAD requests in a fairly straightforward manner. | |
| """ |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| #!/bin/sh | |
| # | |
| # Read-only Root-FS for Raspian | |
| # | |
| # Modified 2018 by Zhong Sheng to support using filesystem other than tmpfs as upper filesystem | |
| # | |
| # Modified 2015 by Pascal Rosin to work on raspian-ua-netinst with | |
| # overlayfs integrated in Linux Kernel >= 3.18. | |
| # | |
| # Originally written by Axel Heider (Copyright 2012) for Ubuntu 11.10. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| %YAML 1.2 | |
| --- | |
| # See http://www.sublimetext.com/docs/3/syntax.html | |
| name: ARM Map | |
| file_extensions: | |
| - map | |
| scope: text.map | |
| # first_line_match: |