[B! python][kernel] slay-tã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

slay-t id:slay-t

pythonã¨kernelã«é–¢ã™ã‚‹slay-tã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (5)

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

${{author_name}}$
{{author_name}}{{created}}
{{ #comment }}{{ comment }}{{ /comment }}
- {{ label }}

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

ç¬¬692å›žã€€sysfsã‚„bpftoolã‚’ç”¨ã„ãŸeBPFã®æ´»ç”¨ | gihyo.jp
ç¬¬688å›žã¨ç¬¬690å›žã§ã¯ã€ã‚«ãƒ¼ãƒãƒ«ã®ãƒˆãƒ¬ãƒ¼ã‚·ãƒ³ã‚°ãƒ„ãƒ¼ãƒ«ã¨ã—ã¦æ³¨ç›®ã•ã‚Œã¦ã„ã‚‹eBPFã‚’æ´»ç”¨ã™ã‚‹ãŸã‚ã®ãƒ„ãƒ¼ãƒ«ã¨ã—ã¦BCCã‚’ç´¹ä»‹ã—ã¾ã—ãŸã€‚ã—ã‹ã—ãªãŒã‚‰ã€BCCã ã‘ãŒeBPFã‚’æ‰±ãˆã‚‹ãƒ„ãƒ¼ãƒ«ã¨ã„ã†ã‚ã‘ã§ã¯ã‚ã‚Šã¾ã›ã‚“ã€‚ä»Šå›žã¯ãƒ„ãƒ¼ãƒ«ãªã—ã«åˆ©ç”¨ã§ãã‚‹sysfsã‚„ã€ã‚ˆã‚Šãƒ¦ãƒ¼ã‚¶ãƒ¼ãƒ•ãƒ¬ãƒ³ãƒ‰ãƒªãƒ¼ãªãƒˆãƒ¬ãƒ¼ã‚·ãƒ³ã‚°ãƒ„ãƒ¼ãƒ«ã§ã‚ã‚‹bpftoolã‚’ç´¹ä»‹ã—ã¾ã™ã€‚ Pythonç‰ˆBCCã®å•é¡Œç‚¹ ã“ã‚Œã¾ã§ç´¹ä»‹ã—ã¦ã„ãŸBPF Compiler Collectionï¼ˆBCCï¼‰ã®ãƒ„ãƒ¼ãƒ«ã¯ã„ãšã‚Œã‚‚ãƒ•ãƒãƒ³ãƒˆã‚¨ãƒ³ãƒ‰ã¨ã—ã¦Pythonã‚’ä½¿ã£ã¦ã„ã¾ã—ãŸã€‚ã¤ã¾ã‚Šåˆ©ç”¨è€…ã¯ã¾ãšPythonã‚¹ã‚¯ãƒªãƒ—ãƒˆã‚’èµ·å‹•ã—ã€ãã®ä¸ã§eBPFã®ã‚ªãƒ–ã‚¸ã‚§ã‚¯ãƒˆã‚’ã‚³ãƒ³ãƒ‘ã‚¤ãƒ«ã—ã€ãƒãƒ¼ãƒ‰ã™ã‚‹ã“ã¨ã§ã‚ˆã†ã‚„ããƒˆãƒ¬ãƒ¼ã‚¹ãŒå§‹ã¾ã£ã¦ã„ãŸã®ã§ã™ã€‚ å®Ÿè¡Œç’°å¢ƒã§BPFã‚ªãƒ–ã‚¸ã‚§ã‚¯ãƒˆã‚’ãƒ“ãƒ«ãƒ‰ã™ã‚‹å¿…è¦ãŒã‚ã‚‹ã“ã®æ–¹æ³•ã«ã¯ã„ãã¤ã‹ã®å•é¡Œç‚¹ãŒå˜åœ¨ã—ã¾ã™ã€‚ å®Ÿè¡Œç’°å¢ƒã«ã‚³ãƒ³ãƒ‘ã‚¤ãƒ©ã‚’ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã™ã‚‹å¿…è¦ãŒã‚ã‚‹
slay-t 2021/11/18
ã‚«ãƒ¼ãƒãƒ«

ãƒ„ãƒ¼ãƒ«

data

event

co

python

kernel

ubuntu

software
ãƒªãƒ³ã‚¯
PyTorchã§GPUã®è¨ˆç®—æ™‚é–“ã‚’æ£ã—ãè¨ˆæ¸¬ã™ã‚‹ - ã¾ã£ãŸã‚Šå‹‰å¼·ãƒŽãƒ¼ãƒˆ
æ¸¬å®šçµæžœ çµæžœã¨ã—ã¦ã€é–“é•ã£ãŸæ¸¬å®šæ–¹æ³•ã ã¨CPUã¨GPUã‚’æ¯”è¼ƒã™ã‚‹ã¨ã€Œ1883å€é€Ÿããªã‚Šã¾ã—ãŸï¼ã€ã¨ã„ã†ä¸»å¼µã‚’ã—ã¦ã—ã¾ã†ã“ã¨ã«ãªã‚Šã¾ã™ã€‚ã¡ãªã¿ã«GPUã§1000å€ãªã‚“ã¦æ•°å—ãŒå‡ºã¦ããŸã‚‰ç¢ºå®Ÿã«ã©ã“ã‹é–“é•ãˆã¦ã„ã¾ã™ã€‚å®Ÿéš›ã€ä»Šå›žã®ã‚±ãƒ¼ã‚¹ã§ã¯æœ¬å½“ã¯ã€Œç´„59å€é€Ÿããªã‚Šã¾ã—ãŸï¼ã€ã¨ã„ã†ã®ãŒæ£ã—ã„çµæžœã«ãªã‚Šã¾ã™ã€‚ torch.cuda.synchronize()ã¨torch.cuda.Eventã‚’ä½¿ã£ãŸå ´åˆã®é•ã„ ä»Šå›žtorch.cuda.synchronize()ã¨torch.cuda.Event ã®2ç¨®é¡žã‚’ç´¹ä»‹ã—ã¾ã—ãŸã€‚å ´åˆã«ã‚ˆã£ã¦ã¯ä½¿ã„åˆ†ã‘ã‚’ã—ãŸã»ã†ãŒã„ã„ã®ã§ã“ã®äºŒã¤ã®é•ã„ã‚’èª¬æ˜Žã—ã¦ã„ãã¾ã™ã€‚ torch.cuda.synchronize() ã‚’åˆ©ç”¨ã—ãŸå ´åˆã€ç°¡å˜ãªã®ã§æ¸¬å®šã—ã‚„ã™ã„ã®ã§ã„ã„ã®ã§ã€ã±ã£ã¨æ¸¬å®šã—ãŸã„å ´åˆã¯ã“ã¡ã‚‰ã®æ–¹æ³•ãŒæ¥½ã§ã‚ˆã„ã‹ã¨æ€ã„ã¾ã™ã€‚ãŸã ã€ã“ã¡ã‚‰ã®æ–¹æ³•ã¯kernelé–¢æ•°ã®ç™ºè¡Œã¨æ¸¬
slay-t 2021/03/22
GPU

kernel

event

CPU

å‹‰å¼·

æ©Ÿæ¢°å¦ç¿’

python

ã‚³ãƒ³ãƒ”ãƒ¥ãƒ¼ã‚¿
ãƒªãƒ³ã‚¯
A visual debugger for Jupyter
Most of the progress made in software projects comes from incrementalism. The ability to quickly see the outcome of an execution and iterate has been one of the main reasons for the success of Jupyter, especially in scientific exploratory workflows. Jupyter users like to experiment in the notebook, and to use the notebook as an interactive communication tool. However, for more classical software d
slay-t 2020/04/02
kernel

blog

debug

channel

code

python

Jupyter

jupyterlab

JupyterNotebook

debugger
ãƒªãƒ³ã‚¯
CuPyã®ElementwiseKernelã§æ¥½ã«GPUã®æ©æµã‚’å—ã‘ã‚‹ - Qiita
Deleted articles cannot be recovered. Draft of this article would be also deleted. Are you sure you want to delete this article? Powered by DLHacks ã¯ã˜ã‚ã¾ã—ã¦ã€ãƒžãƒ³ãƒœã‚¦ã§ã™ã€‚æ™®æ®µã¯ä¿¡å·å‡¦ç†ã‚„ç”»åƒè§£æžã‚’ã‚„ã£ã¦ã„ã¾ã™ã€‚ å”çªã§ã™ãŒã€Pythonã‹ã‚‰ãƒ”ãƒ³ãƒã‚¤ãƒ³ãƒˆã«CUDAã®æ©Ÿèƒ½ã‚’åˆ©ç”¨ã§ãã‚‹CuPyã®ElementwiseKernelã‚’ç´¹ä»‹ã—ã¾ã™ã€‚ CUDAã¨è¨€ã£ã¦ã‚‚æã‚Œã‚‹ã“ã¨ã¯ãªãã€ã€ŒC++ã®è¨˜æ³•ã‚’ãªã‚“ã¨ãªãç†è§£ã—ã¦ã„ã‚‹ã€ãƒ¬ãƒ™ãƒ«ã®äººã§ã‚‚ç°¡å˜ã«åˆ©ç”¨ã§ãã¾ã™ã€‚ ã¯ã˜ã‚ã« Pythonã¯æ›¸ãã‚„ã™ãèªã¿ã‚„ã™ã„ã®ã§ç§ã¯å¥½ãã§ã™ã€‚ ã—ã‹ã—ã€è†¨å¤§ãªä¿¡å·ã‚„å¤§é‡ã®ç”»åƒã‚’Pythonã§æ‰±ã£ã¦ã„ã‚‹ã¨ã€ã©ã†ã—ã¦ã‚‚é€Ÿåº¦ãŒæ°—ã«ãªã‚Šã ã—ã¾ã™ã€‚ ã‹ã¨ã„ã£ã¦ã€å…¨ã¦ã‚’C++ã‚„CUDA
slay-t 2018/08/31
kernel

C

python

CUDA

GPU
ãƒªãƒ³ã‚¯
Linux kernel 3.9 ã®æ–°æ©Ÿèƒ½ SO_REUSEPORT ã‚’è©¦ã—ã¦ã¿ã‚‹ - nigakyã®ãƒ–ãƒã‚°
Linux kernel 3.9 ã®ãƒžãƒ¼ã‚¸ã‚¦ã‚£ãƒ³ãƒ‰ã‚¦ã§ã¯ SO_REUSEPORT ã¨ã„ã†ã‚½ã‚±ãƒƒãƒˆã‚ªãƒ—ã‚·ãƒ§ãƒ³ãŒãƒžãƒ¼ã‚¸ã•ã‚Œã¾ã—ãŸã€‚ Merge branch 'soreuseport' Â· c617f39 Â· torvalds/linux Â· GitHub ã“ã‚Œã¯åŒä¸€ãƒãƒ¼ãƒˆã«è¤‡æ•°ã®ãƒªã‚¹ãƒŠãƒ¼ï¼ˆlisten ã‚½ã‚±ãƒƒãƒˆï¼‰ãŒ bind ã§ãã‚‹ã‚ˆã†ã«ãªã‚‹ã¨ã„ã†ã‚‚ã®ã§ã€Webã‚µãƒ¼ãƒãªã©å˜ä¸€ã®ãƒãƒ¼ãƒˆã«å¤šãã®ã‚³ãƒã‚¯ã‚·ãƒ§ãƒ³ãŒæ¥ã‚‹ã‚ˆã†ãªãƒ¯ãƒ¼ã‚¯ãƒãƒ¼ãƒ‰ã§ã€è¤‡æ•°ãƒ—ãƒã‚»ã‚¹ã§ã†ã¾ãè² è·åˆ†æ•£ãŒã§ãã‚‹ã‚ˆã†ã«ãªã‚‹ãã†ã§ã™ã€‚ ï¼ˆã“ã‚Œã¾ã§ã ã¨ä¸€äººãŒaccept()ã—ã¦ãã‚Œãžã‚Œã® worker ã«æ¸¡ã™ã¨ã„ã†ã‚ˆã†ãªãƒ¢ãƒ‡ãƒ«ãŒã‚ã‚Šã¾ã—ãŸãŒã€ã“ã‚Œã§ã¯ accept() ã™ã‚‹éƒ¨åˆ†ãŒãƒœãƒˆãƒ«ãƒãƒƒã‚¯ã«ãªã£ã¦ã—ã¾ã„ãŒã¡ã§ã—ãŸã€‚ï¼‰ BSDã§ã¯å…ƒã€… SO_REUSEPORT ã‚ªãƒ—ã‚·ãƒ§ãƒ³ã¯ã‚ã£ãŸã‚ˆã†ã§ã™ãŒã€ãƒžãƒ«ãƒã‚ãƒ£ã‚¹ãƒˆé€šä¿¡ã§ä½¿ã†ã‚‚ã®ã®ã‚ˆã†ã§ã™ã€‚Linux
slay-t 2013/05/16
python

kernel

linux

network

bind

port

socket
ãƒªãƒ³ã‚¯
1