Skip to content

Instantly share code, notes, and snippets.

View tsg-'s full-sized avatar

Tushar Gohad tsg-

  • Intel
  • 01:24 (UTC -07:00)
View GitHub Profile
@tsg-
tsg- / Dockerfile.cpu.amx
Last active March 5, 2026 07:11
AMX folder from ofiplugin
# =============================================================================
# vLLM CPU — AMX ENABLED (Intel Sapphire Rapids / Emerald Rapids)
# Ubuntu 24.04 LTS, Python 3.12, gcc-13
#
# Build:
# git clone https://github.com/vllm-project/vllm.git
# cd vllm
# docker build \
# -f Dockerfile.cpu.amx \
# --build-arg VLLM_CPU_AMXBF16=1 \

GPU Networking Reading List


1. FOUNDATIONAL SURVEYS ON GPU INTERCONNECTS

Primary GPU Interconnect Evaluation

  • “Evaluating Modern GPU Interconnect: PCIe, NVLink, NV-SLI, NVSwitch and GPUDirect” (2019)
    • Authors: Ang Li, Shuaiwen Leon Song, Jieyang Chen, Jiajia Li, Xu Liu, Nathan Tallent, Kevin Barker
  • arXiv: 1903.04611
@tsg-
tsg- / ceph_connector_v1.py
Created August 8, 2025 02:15
Simple CephKVConnector (vLLM v1)
import io
import gzip
import asyncio
import rados
import torch
import logging
from typing import TYPE_CHECKING, Any, Optional, Dict, Set
from vllm.distributed.kv_transfer.kv_connector.v1.base import (
KVConnectorBase_V1, KVConnectorMetadata, KVConnectorRole
)
@tsg-
tsg- / ceph_connector.py
Last active April 16, 2025 16:36
Simple CephKVConnector (vLLM v0)
import os
import io
import gzip
import asyncio
import rados
import torch
import logging
from typing import List, Tuple, Optional, Dict
from vllm.distributed.kv_transfer.kv_connector.base import KVConnectorBase
@tsg-
tsg- / nat.sh
Created February 28, 2024 17:58
#!/bin/sh
#
IPTABLES=/sbin/iptables
DEPMOD=/sbin/depmod
MODPROBE=/sbin/modprobe
EXTIF="eno1"
INTIF1="ens785f0"
echo "Using EXTIF=$EXTIF, INTIF=$INTIF2, INTIF=$INTIF1"
-- example script that adds a query string
local threadcounter = 1
local threads = {}
function setup(thread)
thread:set("id", threadcounter)
table.insert(threads, thread)
threadcounter = threadcounter + 1
@tsg-
tsg- / rbd.cc
Created November 29, 2017 04:59
/*
* rbd engine
*
* IO engine using Ceph's librbd to test RADOS Block Devices.
*
*/
#include <rbd/librbd.h>
#include "global/global_init.h"
#include "global/global_context.h"
# Thanks to Sudarshan Muralidhar for this test program
from pyeclib.ec_iface import ECDriver
from struct import pack
from random import randint
def random_mutate(share, mutations):
share = list(share)
rands = [randint(0, 255) for i in xrange(mutations)]
@tsg-
tsg- / gist:d59d34a664ec93d5833b
Created December 8, 2015 21:58
Swift unit test failue: test_drop_privileges
======================================================================
FAIL: test_drop_privileges (test.unit.common.test_utils.TestUtils)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/root/swift/test/unit/common/test_utils.py", line 1820, in test_drop_privileges
self.assertEqual(set(groups), set(os.getgroups()))
AssertionError: Items in the first set but not the second:
0
"""Fail immediately, with the given message."""
>> raise self.failureException('Items in the first set but not the second:\n0')
import os
import platform
import sys
from ctypes.util import find_library
from distutils.sysconfig import EXEC_PREFIX as _exec_prefix
from distutils.sysconfig import get_python_lib
from distutils.sysconfig import get_python_inc
platform_arch = platform.architecture()
default_python_libdir = get_python_lib()