Tags

Following is a list of relevant tags:

0/1 Adam

• DeepSpeed Communication Compression and Hiding

1-bit Adam

• DeepSpeed Communication Compression and Hiding

1-bit LAMB

• DeepSpeed Communication Compression and Hiding

3FS

• DualPath KV Loading

4+1 View Model

• Software Design Evidence

5G

• Family Wi-Fi

64bits vs 32bits

• Arm vs X86 (unfinished)

A3

• InferenceX Data Pipeline

A5

• InferenceX Data Pipeline

ADR

• Software Design Evidence

ADXL

• Mooncake Ascend Transports

AGENTX

• Dynamo TRT-LLM and AgentX
• InferenceX Data Pipeline

AGRS

• XTuner Domino EP

AI

• AI Coding Harness
• AI Compiler
• AI Image
• Agile Governance: Balancing IPD and AI Innovation
• Code Migration And Alignment
• Codex Model Speed Benchmark
• IPD Q&A
• Introduction to AI and Machine Learning Basics
• Subagent Control Plane
• 成功的软件商业解决方案

AI AGENT

• Kimi K3 Report

AI Agent

• Evidence-Driven Goal Loop

AI CODING

• Interactive HTML Article Case

AI Chip

• AI Chip System Barriers

AI Cluster

• UB-Mesh Architecture

AI Engineering

• AI Engineering Roadmap

AI INFRA

• AI Startup Career Risk

AI Infrastructure

• AI Technology Value Stack

AI STARTUP

• AI Startup Career Risk

AI Systems

• AI Optimization Stack

AIInfra

• AI Infra Daily Radar

AISBench

• Multimodal Generation Evaluation
• VLM RL Evaluation Datasets

AIV

• AIV Direct Drive
• AIV Direct Drive Programming
• AIV MoE All-to-All
• Ascend Communication Stack
• Ascend DeepEP MoE Dispatch Optimization
• Ascend DeepEP Ref Dispatch SIMT
• DualPath KV Loading
• Mooncake Ascend Transports
• SHMEM Symmetric Memory
• TileXR-SHMEM Abstractions

ALL-TO-ALL

• AIV MoE All-to-All
• Ascend DeepEP MoE Dispatch Optimization

ALST

• AI Training Parallelism

AMAT

• Concurrent-AMAT, C-AMAT

AMD

• Bios
• GPU
• Nvidia

APR

• UB-Mesh Architecture

ASCEND

• Dynamo TRT-LLM and AgentX
• InferenceX Data Pipeline
• Mooncake Ascend Transports
• Mooncake vLLM Ascend

ASPLOS

• Conference

ASTRA-Sim

• SimAI Architecture

ASYNC I/O

• io_uring Async I/O

ATI

• GPU

ATTENTION RESIDUALS

• Kimi K3 Report

AVX

• Double 2 int8
• SIMD+SSE+AVX

Acceptance Criteria

• Requirements Analysis

Activation Offload

• XTuner Memory Optimization

Agent

• AI Engineering Roadmap
• Subagent Control Plane

Agent Harness

• AI Coding Harness

Agent Loop

• Evidence-Driven Goal Loop

Agentic RL

• Frontier Model RL

Algorithm

• Algorithm: leetcode

All-to-All

• XTuner Domino EP

AllGather

• ClusterHealthDetect A3 Performance

Alpha

• ISA & Micro-architecture

Analysis

• Analysis Software

Apt

• package manager: apt

Architecture Analysis

• Technology Insight

Architecture Governance

• Software Design Evidence

Archon

• Subagent Control Plane

Ascend

• BSND TND Operator Layout
• CloudMatrix384 LLM Serving
• ClusterHealthDetect A3 Performance
• Echo Training Simulator
• Memory-Semantic TransferQueue
• NPU Training Operators - GDN
• NPU Training Operators - GMM
• NPU Training Operators - MC2
• NPU Training Operators - RoPE MRoPE
• Next of My Ascend Career
• SimAI Architecture
• TileXR-SHMEM Abstractions

Ascend 910C

• xDeepServe on CloudMatrix384

Ascend 950

• Ascend 950 A5 Multimodal Quantization

Ascend A5

• Ascend 950 A5 Multimodal Quantization

Ascend C

• AIV Direct Drive Programming

Ascend NPU

• GLM Post-Training Beyond SFT
• GLM Post-Training Open-Source Map

AscendCL

• Ascend Data Movement Evolution

Assembly

• Assembly Arm
• Assembly X86
• Inline Assembly
• Vtune Assembly Analysis

Async

• VeRL Async
• VeRL Feature Survey

Async RL

• VeRL Async Policy

Attention

• Attention Architecture Evolution
• Attention Cache and Sequence Parallelism
• KV Cache Fundamentals

AudioVideo

• DiffusionNFT

AutoResearch

• Subagent Control Plane

AutoTP

• DeepSpeed Memory and Parallelism

Autonomous Driving

• VLA VLM + DiT

Autotuning

• DeepSpeed Observability and Autotuning

B200

• InferenceX Data Pipeline

BAM

• GPU-Initiated I/O
• NVIDIA SCADA

BATCH

• Mooncake Classic vs TENT Engine

BFS

• Parallel BFS

BHive

• BHive : An Infrastructure for Adaptive Dynamic Optimization 2003 IEEE

BLOCK DEVICE

• Mooncake File vs Block Device

BPMN

• Business Process Diagrams

BSHD

• VeRL Performance Optimization

BSND

• BSND TND Operator Layout

BT

• BitTorrent
• Video
• uTorrent

BTL

• Openmpi MPIOPT

Baka Mitai

• Deepfake - unravel/Baka Mitai

Bash

• Bash Scripting

Benchmark

• Codex Model Speed Benchmark

Benchmarking

• Technology Insight

Big-Endian

• Big-Endian & Little-Endian

Blackwell

• Inference MegaKernel

C

• Naming
• [C++ Basic] Exploring Useful Built-in Functions
• [C++ Basic] Grammar
• [C++ Basic] Types
• [C++] Destructor Order

C++

• Gperftools
• Hash map
• Naming
• [C++ Basic] Exploring Useful Built-in Functions
• [C++ Basic] Grammar
• [C++ Basic] STL Data Structure
• [C++ Basic] Types
• [C++ Basic] User-Defined Types: Class
• [C++] Destructor Order

CACHE OFFLOADING

• Mooncake Store Design

CANN

• AIV Direct Drive
• AIV Direct Drive Programming

CANN VMM

• Ascend Communication API Evolution
• Ascend Communication Runtime Evolution
• Ascend Communication Stack
• Mooncake Ascend Transports

CCD

• AMD Architecture on EPYC

CCX

• AMD Architecture on EPYC

CI

• Continuous Integration, CI

CISC

• ISA & Micro-architecture

CODE WALKTHROUGH

• Mooncake Codebase Architecture
• Mooncake TENT Request Path

COMPENSATION

• AI Startup Career Risk

CONTINUOUS BATCHING

• PagedAttention

CP

• AI Training Parallelism
• RL Data Flow

CPP

• Windows CPP Project Setup

CPU

• VeRL Local Debug on Mac

CPU Thread Socket

• AMD CPU

CUDA

• AI Chip System Barriers
• Inference MegaKernel

CUDA Allocator

• XTuner Memory Optimization

CUFILE

• Mooncake Classic NVMeoF Transport
• Mooncake File vs Block Device
• Mooncake TENT GDS
• Mooncake TENT Request Path

CV

• Introduction to AI and Machine Learning Basics

Calling Conventions

• CSAPP: Machine Programming III: Procedures

Capability Map

• Function Decomposition Methods

Career

• 1.2 Career
• 1.2 Career：1 秋招
• Career Transferable skill / Durable skills / Core capabilities
• DFX: Design for X
• Multi-Objective Decision Making
• QCC：Quality Control Circle

Career Strategy

• AI Technology Value Stack

Checkpoint

• VeRL Checkpoint

Chunk Layer

• AI Training Parallelism

Class

• [C++ Basic] Grammar
• [C++] Destructor Order

Claude Code

• AI Coding Harness

CloudMatrix384

• CloudMatrix384 LLM Serving
• xDeepServe on CloudMatrix384

Cloudflare

• Cloudflare warp proxy

ClusterHealthDetect

• ClusterHealthDetect A3 Performance

Co-design

• AI Optimization Stack

Codeforces

• Codeforces & Topcoder

Codex

• AI Coding Harness
• Codex Model Speed Benchmark

Cohousing

• Rental Checklist

Collective Communication

• UB-Mesh Architecture

Communication

• Work Reporting

Communication Logging

• DeepSpeed Observability and Autotuning

Communication-Compute Fusion

• XTuner Domino EP

Compatibility

• Modern Matchmaking System

Compiler

• AI Chip System Barriers
• AI Compiler
• AI Optimization Stack

Computer Architecture

• AI Chip System Barriers

Conference

• Conference

Context Parallelism

• Attention Cache and Sequence Parallelism

Crawler

• Crawler
• WebCrawler first try

DDR

• CPU vs GPU

DEEPEP

• Ascend DeepEP MoE Dispatch Optimization
• Ascend DeepEP Ref Dispatch SIMT

DFD

• Data Flow Analysis Methods

DFX

• DFX: Design for X
• RL DFX Metrics
• RL Data Flow

DFlash

• VeRL Feature Survey
• VeRL Speculative Decoding

DIRECT STORAGE

• Mooncake Classic NVMeoF Transport
• Mooncake NDS Integration
• Mooncake TENT GDS

DISAGGREGATION

• Long-Context KV Cache Systems

DISTRIBUTED STORAGE

• Mooncake Store Design

DLP

• ParallelIntroduction

DMA

• Ascend Communication Stack
• DMA Communication Terminology
• Scratchpad and NoC

DNS

• DNS
• PersonalWebsiteDomain

DP

• AI Training Parallelism
• DP
• HDMI

DPO

• AI Post Traning: DPO + MPO

DRAM

• CPU vs GPU
• DRAM types: Size, Latency, Bandwidth, Energy Consumption
• Processing In/near Memory

DSA

• Attention Architecture Evolution
• Attention Cache and Sequence Parallelism

DanceGRPO

• AI Post Traning: DanceGRPO
• Business Trip: 2601-2602 verl + DanceGRPO

DanceOPD

• On-Policy Distillation

Data Lineage

• Data Flow Analysis Methods

Data Placement

• Distributed KV Cache Management

DataStates

• DeepSpeed I/O, Offload, and Asynchrony

Database

• Database basics

Databases

• Sql

Dataset

• VLM RL Evaluation Datasets

Debug

• VeRL Local Debug on Mac

Decision Analysis

• Multi-Objective Decision Making
• Requirements Analysis

Decode

• KV Cache Fundamentals

Deep Learning

• Introduction to AI and Machine Learning Basics

DeepEP

• XTuner Domino EP

DeepNVMe

• DeepSpeed I/O, Offload, and Asynchrony

DeepSeek-R1

• CloudMatrix384 LLM Serving

DeepSpeed

• DeepSpeed Communication Compression and Hiding
• DeepSpeed I/O, Offload, and Asynchrony
• DeepSpeed Memory and Parallelism
• DeepSpeed MoE and Model Compression
• DeepSpeed Observability and Autotuning

Deepfake

• Deepfake - unravel/Baka Mitai

Design Patterns

• Software Design Evidence

DevLog

• [DevLog] PLAN
• [DevLog]24Q3P1 - Optimize PTA With Thread Affinity
• [DevLog]24Q4P2 - Lazy Initialize during old dispatcher way

DiT

• VLA VLM + DiT

Diffusion

• DiffusionNFT

Diffusion Language Model

• Diffusion LLM Post-Training

Diffusion Model

• VLA VLM + DiT

Digital Worker

• AI Documentation Workflow
• My Digital Worker
• My Digital Worker : AutoMoneyMaker - AutoTrader
• My Digital Worker : Model / Software Usage
• My Digital Worker : New Coding Way
• My Digital Worker : New Coding Way Part0 —— Building AI-Coding Env
• My Digital Worker : Work with AI
• Personal Advantage Workflow

Disk

• Disk Space Cleanup

Distributed Systems

• Distributed KV Cache Management

Distributed Training

• Echo Training Simulator
• SimAI Architecture

Domain Model

• Data Flow Analysis Methods

Domino

• DeepSpeed Communication Compression and Hiding

Domino EP

• XTuner Domino EP

DualPath

• DualPath KV Loading

Dynamic Programming

• Dynamic Programming

ECC

• Cuda Program Basic

EDP

• AI Training Parallelism

EP

• AI Training Parallelism

ER Model

• Data Flow Analysis Methods

ESOP

• AI Startup Career Risk

ETP

• AI Training Parallelism

EXTENSIBILITY

• Mooncake Classic vs TENT Engine

Echart

• Echart

Echo

• Echo Training Simulator

Ed2k

• Ed2k

Evals

• AI Engineering Roadmap

Evaluation

• Multimodal Generation Evaluation
• VLM RL Evaluation Datasets

Evidence

• Evidence-Driven Goal Loop

Executable file

• Linux Executable file: Structure & Running
• Linux Executable file: Structure & Running 2

Expert Parallelism

• CloudMatrix384 LLM Serving
• NPU Training Operators - GMM
• NPU Training Operators - MC2
• XTuner Domino EP

Explain

• Understand deeper ： Why xxx is useful?
• Where Do AI Ideas Come From? A Deeper Look

FABRIC MEMORY

• Mooncake Ascend Transports

FAST

• Function Decomposition Methods

FBS

• Function Decomposition Methods

FLA

• BSND TND Operator Layout

FLOPs Profiler

• DeepSpeed Observability and Autotuning

FMA

• XeonScalable

FP8

• Inference Quantization Formats

FPDT

• AI Training Parallelism

FSDP

• NPU Training Operators - GDN

FSDP2

• AI Training Parallelism
• Kimi K3 NPU Training
• VeRL Backend Parallelism
• XTuner Memory Optimization

FUSELINK

• FuseLink Multi-NIC Communication

Fabric Memory

• Ascend Communication API Evolution
• Ascend Communication Runtime Evolution
• Ascend Communication Stack

FeatureMatrix

• VeRL Feature Matrix

FlashAttention

• AI Optimization Stack

FlashInfer

• BSND TND Operator Layout

Flow-Factory

• Multimodal Generation Evaluation

FlowServe

• xDeepServe on CloudMatrix384

Flowchart

• Business Process Diagrams

Foundation Model Training

• AI Technology Value Stack

Frontier Models

• Frontier Model RL

Function Tree

• Function Decomposition Methods

Functional Decomposition

• Function Decomposition Methods

FurtherStudy

• PersonalWebsiteDomain

GCN

• GNN

GDDR6x

• CPU vs GPU

GDN

• Attention Architecture Evolution
• Attention Cache and Sequence Parallelism
• BSND TND Operator Layout
• NPU Training Operators - GDN
• Training Performance Model
• vLLM Inference Profiling

GDS

• GPU-Initiated I/O
• Mooncake Classic NVMeoF Transport
• Mooncake File vs Block Device
• Mooncake NDS Integration
• Mooncake TENT GDS
• Mooncake TENT Request Path

GIL

• Python MPI

GLIBC

• Library GLIBC

GLM-5

• GLM Post-Training Beyond SFT
• GLM Post-Training Open-Source Map

GLM-5.2

• GLM Post-Training Beyond SFT
• GLM Post-Training Open-Source Map
• Inference Quantization Formats

GMM

• NPU Training Operators - GMM

GMT

• GPU-Initiated I/O

GNN

• GNN

GNU

• GNU Assembly File

GPT

• Disk

GPT-5.6

• Codex Model Speed Benchmark

GPTQ

• Inference Quantization Formats

GPU

• CPU / GPU Performance Model based on Interval Analysis

GPU Benchmark

• InferenceX Data Pipeline

GPU COMMUNICATION

• FuseLink Multi-NIC Communication

GPU DIRECT STORAGE

• GPU-Initiated I/O
• Tutti SSD-Backed KV Cache
• io_uring Async I/O
• vLLM KV Offloading and GDS

GPU OFFLOADING

• TurboBus PCIe Bandwidth Pooling

GPU-Aware MPI

• GPU-Centric Communication

GPU-Centric Communication

• GPU-Centric Communication

GPU-INITIATED I/O

• GPU-Initiated I/O

GPU-INITIATED IO

• NVIDIA SCADA

GPUDIRECT STORAGE

• Mooncake TENT GDS
• NVIDIA SCADA

GPUDirect

• GPU-Centric Communication

GQA

• Attention Architecture Evolution

GRPO

• BSND TND Operator Layout
• Diffusion LLM Post-Training
• NPU Training Operators - GDN
• RL Algorithms: PPO-RLHF & GRPO-family

Game

• Game Streaming & Video Streaming
• UnimportantView: Game

Goal

• Evidence-Driven Goal Loop

H100

• InferenceX Data Pipeline

H200

• InferenceX Data Pipeline

H2D

• ClusterHealthDetect A3 Performance

HARDWARE-SOFTWARE CO-DESIGN

• WaferLLM Co-Design

HCCL

• AIV Direct Drive
• AIV Direct Drive Programming
• Ascend Communication API Evolution
• Ascend Communication Stack
• ClusterHealthDetect A3 Performance
• DualPath KV Loading
• Echo Training Simulator
• Mooncake Ascend Transports
• NPU Training Operators - MC2
• SHMEM Symmetric Memory
• SimAI Architecture
• TileXR-SHMEM Abstractions

HCCS

• AIV Direct Drive
• Ascend Communication Stack
• Ascend Interconnect Evolution

HCOMM

• Ascend Communication API Evolution
• Ascend Communication Runtime Evolution

HDMI

• HDMI

HIGH AVAILABILITY

• Mooncake KV Metadata

HIXL

• Mooncake Ascend Transports

HPCAI

• HPCAI

HPL

• HPL.dat file detailed explanation

HPL-PL

• VtuneOptimize

HTML

• Interactive HTML Article Case

Hardware Software Co-design

• Computing Systems Optimization

HiXL

• Ascend Communication API Evolution
• Ascend Communication Stack
• SHMEM Symmetric Memory

Hopper

• Nvidia Arch : Ampere & Hopper & Pascal

Housing

• Rental Checklist

Huawei

• Huawei Training Camp

Hugo

• AI Documentation Workflow

HybridEP

• XTuner Domino EP

HyperParallel-MoE

• XTuner Domino EP

ILP

• ParallelIntroduction

INFERENCE

• Long-Context KV Cache Systems

INFERENCEX

• Dynamo TRT-LLM and AgentX
• InferenceX Data Pipeline

IO_URING

• io_uring Async I/O

IP

• PersonalWebsiteDomain

IPC

• Ascend Communication Runtime Evolution
• Ascend Data Movement Evolution
• DMA Communication Terminology

IPCC

• Double 2 int8
• IPCC2022final

IPD

• Agile Governance: Balancing IPD and AI Innovation
• IPD Q&A

IPO

• AI Startup Career Risk
• Data Flow Analysis Methods

Image

• Image

Inference

• AI Chip System Barriers
• VeRL Rollout Inference

Inference Quantization

• Inference Quantization Formats

Inference Serving

• AI Optimization Stack

Jellyfin

• Jellyfin

Jupyter

• Jupyter AI

KDA

• Attention Architecture Evolution
• Kimi K3 NPU Training
• Kimi K3 Report

KIMI

• Kimi K3 NPU Training
• Kimi K3 Report
• Mooncake

KV CACHE

• Long-Context KV Cache Systems
• Mooncake
• Mooncake Codebase Architecture
• Mooncake KV Metadata
• Mooncake NDS Integration
• Mooncake Store Design
• PagedAttention
• Tutti SSD-Backed KV Cache
• vLLM KV Offloading and GDS

KV CONNECTOR

• Mooncake Codebase Architecture

KV Cache

• Attention Cache and Sequence Parallelism
• CloudMatrix384 LLM Serving
• Distributed KV Cache Management
• DualPath KV Loading
• KV Cache Fundamentals

KVCACHE

• Mooncake vLLM Ascend

Kavita

• Komga

Kernel

• AI Optimization Stack

Kernel Launch

• AIV Direct Drive Programming

Kimi K3

• Inference Quantization Formats

Knowledge Distillation

• On-Policy Distillation

Komga

• Komga

Kunpeng

• Huawei Kunpeng workload
• Kunpeng
• Static Code Analyzer on Kunpeng Overview

LATENTMOE

• Kimi K3 Report

LDA

• Latent Dirichlet Allocation (2003)

LINUX

• io_uring Async I/O

LLM

• Long-Context KV Cache Systems
• Scaling Law

LLM Application

• AI Engineering Roadmap

LLM INFERENCE

• WaferLLM Co-Design

LLM Inference

• Inference MegaKernel
• xDeepServe on CloudMatrix384

LLM SERVING

• Dynamo TRT-LLM and AgentX
• Mooncake
• Mooncake Store Design
• PagedAttention
• Tutti SSD-Backed KV Cache

LLM SYSTEMS

• TurboBus PCIe Bandwidth Pooling

LLM Serving

• CloudMatrix384 LLM Serving
• Distributed KV Cache Management

LLM Wiki

• AI Documentation Workflow

LLaDA

• Diffusion LLM Post-Training

LONG CONTEXT

• Long-Context KV Cache Systems
• Tutti SSD-Backed KV Cache

LeetCode

• Algorithm: leetcode

Leetcode

• Leetcode

LegacyBugs

• PersonalWebsiteDomain
• Web Server: Nginx V.S. Apache2

Linking

• C program compile&run process

Lock

• Lock & Synchronization

Long Context

• Distributed KV Cache Management

MAC

• Macbook Config

MATERIAL FOR MKDOCS

• Interactive HTML Article Case

MC2

• NPU Training Operators - MC2
• TileXR-SHMEM Abstractions
• XTuner Domino EP

MCA

• Openmpi MPIOPT

MCDA

• Multi-Objective Decision Making

MESH NOC

• WaferLLM Co-Design

METADATA

• Mooncake KV Metadata

MFU

• Training Performance Model
• VeRL Performance Optimization

MHA

• Attention Architecture Evolution
• Attention Cache and Sequence Parallelism

MILAN

• AMD Architecture on EPYC

MINDIE

• Dynamo TRT-LLM and AgentX

MIPS

• ISA & Micro-architecture

MKDOCS

• Interactive HTML Article Case

MLA

• Attention Architecture Evolution

MLP

• ParallelIntroduction

MOE

• AIV MoE All-to-All
• Ascend DeepEP MoE Dispatch Optimization
• Ascend DeepEP Ref Dispatch SIMT
• Kimi K3 Report

MOONCAKE

• Mooncake
• Mooncake Ascend Transports
• Mooncake Classic NVMeoF Transport
• Mooncake Classic vs TENT Engine
• Mooncake Codebase Architecture
• Mooncake File vs Block Device
• Mooncake KV Metadata
• Mooncake NDS Integration
• Mooncake Store Design
• Mooncake TENT GDS
• Mooncake TENT Request Path
• Mooncake vLLM Ascend
• vLLM KV Offloading and GDS

MOPD

• GLM Post-Training Beyond SFT
• On-Policy Distillation

MPI

• Dynamic pool dispatch
• IPCC Preliminary SLIC Optimization 5: MPI + OpenMP
• IPCC Preliminary SLIC Optimization 6: Non-blocking MPI
• MPI
• Memory Semantics vs RDMA
• Python MPI
• Why MPI_Init is slow

MPI option

• Openmpi MPIOPT

MPI_Init

• Why MPI_Init is slow

MPK

• Inference MegaKernel

MPS

• VeRL Local Debug on Mac

MQA

• Attention Architecture Evolution

MRoPE

• NPU Training Operators - RoPE MRoPE

MTE

• Ascend Data Movement Evolution
• DMA Communication Terminology
• Memory Semantics vs RDMA
• Memory-Semantic TransferQueue

MTP

• VeRL Speculative Decoding
• XTuner Memory Optimization

MULTI-NIC

• FuseLink Multi-NIC Communication

MULTIMODAL

• Kimi K3 Report

MXFP8

• Inference Quantization Formats

Mac

• VeRL Local Debug on Mac

Machine Learning

• Introduction to AI and Machine Learning Basics
• ML Optimizer
• Model Migrate Accuracy
• Tensor Metrics

Marriage

• Modern Matchmaking System

Matchmaking

• Modern Matchmaking System

MegaKernel

• Inference MegaKernel

MegaMoE

• Inference MegaKernel

Megatron

• Kimi K3 NPU Training
• VeRL Backend Parallelism
• VeRL Performance Optimization

Megatron Bridge

• NPU Training Operators - RoPE MRoPE

Megatron Core

• NPU Training Operators - RoPE MRoPE

MemFabric

• Memory Semantics vs RDMA
• Memory-Semantic TransferQueue

Memory Hierarchy

• Scratchpad and NoC

Memory Optimization

• XTuner Memory Optimization

Memory Semantics

• Ascend Communication Runtime Evolution
• Ascend Communication Stack
• Ascend Data Movement Evolution
• Memory Semantics vs RDMA
• Memory-Semantic TransferQueue

Mesh Interconnect Architecture

• XeonScalable

Metrics

• RL DFX Metrics

Micro

• Micro2023: Utopia

Micro-Fusion

• Microarchitecture: Micro-Fusion & Macro-Fusion

Micro-architecture

• ISA & Micro-architecture

Microarchitecture

• Microarchitecture: Micro-Fusion & Macro-Fusion
• Microarchitecture: Out-Of-Order execution(OoOE/OOE) & Register Renaming
• Microarchitecture: Pipeline of Intel Core CPUs
• Microarchitecture: Zero (one) idioms & Mov Elimination

MindSpeed

• NPU Training Operators - RoPE MRoPE

MindSpeed-MM

• AI Training Parallelism
• NPU Training Operators - GMM
• NPU Training Operators - MC2
• VeRL Backend Parallelism

MixZ++

• DeepSpeed Memory and Parallelism

MoE

• AI Training Parallelism
• AIV Direct Drive
• CloudMatrix384 LLM Serving
• DeepSpeed MoE and Model Compression
• Inference MegaKernel
• Kimi K3 NPU Training
• NPU Training Operators - GMM
• NPU Training Operators - MC2
• Training Performance Model
• VeRL Router Replay
• XTuner Domino EP
• XTuner Memory Optimization
• xDeepServe on CloudMatrix384

MoQ

• DeepSpeed MoE and Model Compression

Model Compression

• DeepSpeed MoE and Model Compression

Module

• Module Command

Money

• Contribution Allocation
• Money Maker ideas

Monitor

• DeepSpeed Observability and Autotuning

MoonEP

• XTuner Domino EP

Mooncake

• Ascend Communication API Evolution

Mount

• Mount Network Disk

Mov-Elimination

• Microarchitecture: Zero (one) idioms & Mov Elimination

Multi-Agent

• Subagent Control Plane

Multimodal

• Multimodal RL

Multimodal Generation

• Ascend 950 A5 Multimodal Quantization

Multimodal Model

• Scaling Law

Muon

• XTuner Memory Optimization

NAS

• Migrate From Synology DS220J to UGREEN DX4600
• Nas Disk Speed Test
• Synology NAS Configuration

NCCL

• Echo Training Simulator
• FuseLink Multi-NIC Communication
• GPU-Centric Communication
• SimAI Architecture

NDS

• Mooncake NDS Integration

NFT

• DiffusionNFT

NIXL

• vLLM KV Offloading and GDS

NLP

• Introduction to AI and Machine Learning Basics

NPU

• AI Chip System Barriers
• AI Infra Daily Radar
• BSND TND Operator Layout
• Kimi K3 NPU Training
• Mooncake NDS Integration
• NPU Training Operators - GDN
• NPU Training Operators - GMM
• NPU Training Operators - MC2
• NPU Training Operators - RoPE MRoPE
• VeRL Async Policy
• VeRL Performance Optimization

NUMA

• ClusterHealthDetect A3 Performance
• NUMA perf
• XeonScalable

NV

• Nvidia

NVIDIA

• NVIDIA SCADA

NVIDIA DYNAMO

• Dynamo TRT-LLM and AgentX

NVLINK

• FuseLink Multi-NIC Communication
• TurboBus PCIe Bandwidth Pooling

NVLink

• GPU-Centric Communication
• Inference MegaKernel

NVME

• GPU-Initiated I/O
• NVIDIA SCADA
• io_uring Async I/O

NVMEOF

• Mooncake Classic NVMeoF Transport
• Mooncake File vs Block Device

NVSHMEM

• GPU-Centric Communication

Nas

• Jellyfin
• Komga

NoC

• Ascend Interconnect Evolution
• Scratchpad and NoC

Nvidia

• Cuda Optimize
• GPU
• Nvidia
• Nvidia Nsight
• Nvprof

O2

• Intel Compile Options

O3

• Intel Compile Options
• Vtune Assembly Analysis

OPD

• GLM Post-Training Beyond SFT
• GLM Post-Training Open-Source Map
• On-Policy Distillation

OS

• Lock & Synchronization

OSI

• Ascend Communication Axes

OSS

• Picbed Migration: from Aliyun OSS to cloudflare R2

Observability

• Evidence-Driven Goal Loop

Omni

• Multimodal RL

OmniNFT

• DiffusionNFT

Open Source

• AI Technology Value Stack
• GLM Post-Training Open-Source Map

OpenCode

• AI Coding Harness

OpenLDAP

• OpenLDAP

OpenMP

• Dynamic pool dispatch

OpenSHMEM

• SHMEM Symmetric Memory

OpenWRT

• OpenWRT on router

Openssl

• Safe File Transport

Options

• AMD Epyc Compiler Options
• GCC Compiler Option 1 : Optimization Options
• GCC Compiler Option 2 : Preprocessor Options
• Intel Compile Options

OutOfOrder

• Microarchitecture: Out-Of-Order execution(OoOE/OOE) & Register Renaming

PAGEDATTENTION

• Mooncake Store Design
• PagedAttention

PAPER

• Tutti SSD-Backed KV Cache

PCIE

• TurboBus PCIe Bandwidth Pooling

PCIe

• Ascend Interconnect Evolution
• Motherboard & PCI-e & UPI

PD Disaggregation

• DualPath KV Loading

PGAS

• Ascend Communication API Evolution
• Ascend Communication Runtime Evolution
• SHMEM Symmetric Memory

PIM

• PIM Simulator
• Processing In/near Memory

PKI

• OpenVPN

PLMR

• WaferLLM Co-Design

PML

• Openmpi MPIOPT

PP

• AI Training Parallelism

PPO

• RL Algorithms: PPO-RLHF & GRPO-family

PPT

• Graph/diagram/visualization/PPT Examples Yet to Be Explored

PREFILL-DECODE

• Mooncake

PT

• BitTorrent
• uTorrent

PTA

• Debug/Profile/Devlop Tools of PTA
• [DevLog] PLAN
• [DevLog]24Q3P1 - Optimize PTA With Thread Affinity
• [DevLog]24Q4P2 - Lazy Initialize during old dispatcher way

PTX

• cuda Assembly:PTX & SASS

PVE

• Proxmox VE（PVE）on J4125

Pagerank

• Graph Algorithms: Pagerank

Parallel

• Parallel BFS

Parallelism

• Kimi K3 NPU Training
• VeRL Backend Parallelism

Perf

• Perf

Performance

• AI Infra Daily Radar
• Codex Model Speed Benchmark
• VeRL Performance Optimization

Performance Engineering

• AI Technology Value Stack

Performance Model

• ClusterHealthDetect A3 Performance
• Echo Training Simulator
• Training Performance Model

Performance Modeling

• Computing Systems Optimization
• Personal Advantage Workflow

Performance Simulation

• SimAI Architecture

Pi

• Subagent Control Plane

PicBed

• PicBed & OSS

PicGo

• PicBed & OSS

Pin

• Intel Pin
• Old Pintool Upgrade with newest pin

Pipeline

• Microarchitecture: Pipeline of Intel Core CPUs

Post Training

• Scaling Law

Post-Training

• On-Policy Distillation

PostTraining

• Frontier Model RL

Powershell

• Powershell

Prefill

• KV Cache Fundamentals

Presentation

• Presentation & Visualization : PPT

Priority

• Personal Advantage Workflow

Privacy

• Modern Matchmaking System

Probability Theory

• Probability Theory

Product Discovery

• Requirements Analysis

Professional Skills

• Work Reporting

Prompt Engineering

• AI Engineering Roadmap

PyG

• PyTorchGeometric

PyPTO

• PyPTO

PyTorch

• Code Migration And Alignment
• PyTorch_VS_TensorFlow
• VeRL Local Debug on Mac

PyTorch Profiler

• DeepSpeed Observability and Autotuning

Pytorch

• Pytorch 1 ：Basic Components
• Pytorch 2 ：more conceptions about training and inference
• Pytorch 2.5 ：Dataset & Dataloader
• Pytorch 3 ：Model & Training
• Pytorch 4 ：Save & Load & Pretrain
• Pytorch 5 ： Distributed Training & Parallelism (Mem Re-allocated)
• Pytorch 6 ：Visualization
• Pytorch 7 ：Memory Optimization(Freeing GPU/NPU Memory Early)
• Pytorch 8 ：Hyperparameter

QAT

• GLM Post-Training Beyond SFT

QCC

• QCC：Quality Control Circle

QuantitativeFinance

• Why Choose Quantitative Finance

Quantization

• Ascend 950 A5 Multimodal Quantization
• GLM Post-Training Beyond SFT

Qwen3-Omni

• On-Policy Distillation

Qwen3-VL

• NPU Training Operators - RoPE MRoPE

Qwen3.5

• BSND TND Operator Layout
• Inference Quantization Formats
• NPU Training Operators - GDN
• NPU Training Operators - GMM
• Training Performance Model

RAG

• AI Documentation Workflow
• AI Engineering Roadmap

RAID

• Disk

RAW

• Microarchitecture: Zero (one) idioms & Mov Elimination

RDMA

• AIV Direct Drive
• Ascend Communication Axes
• Ascend Communication Stack
• Ascend Data Movement Evolution
• Distributed KV Cache Management
• DualPath KV Loading
• FuseLink Multi-NIC Communication
• Memory Semantics vs RDMA
• Mooncake
• Mooncake Store Design
• Mooncake vLLM Ascend

RGB Lab

• IPCC Preliminary SLIC Analysis part2 : Run process

RHLF

• AI Post Traning: DPO + MPO
• AI Post Traning: DanceGRPO

RISC

• ISA & Micro-architecture
• RISC-V

RISC-V

• RISC-V

RL

• AI Post Traning: DPO + MPO
• AI Post Traning: DanceGRPO
• Agent & Agentic RL
• DiffusionNFT
• Fast Debug: VeRL example
• Frontier Model RL
• Kimi K3 Report
• Memory-Semantic TransferQueue
• Multimodal Generation Evaluation
• Multimodal RL
• RL Algorithms: PPO-RLHF & GRPO-family
• RL DFX Metrics
• RL Data Flow
• RL Infra Series
• RL Next: Meta-Learning
• RL: Training Inference Mismatch
• RL: xPU Mismatch - metrics
• The Mechanics of RL: How Inference Sampling Shapes the Probability Landscape
• Train Stages: Pretrain, Mid-Train(CT), SFT, RL
• VLM RL Evaluation Datasets
• VeRL
• VeRL Async
• VeRL Backend Parallelism
• VeRL Checkpoint
• VeRL Feature Matrix
• VeRL Feature Survey
• VeRL Performance Optimization
• VeRL Rollout Inference
• VeRL Router Replay
• VeRL Speculative Decoding
• VeRL Training Flow

RLHF

• VeRL Async Policy

RLInfra

• RL DFX Metrics
• RL Data Flow
• RL Infra Series

RMA

• AIV Direct Drive Programming

ROMA

• AMD Architecture on EPYC

RTX 3090

• Nvidia Arch : Ampere & Hopper & Pascal

Ramulator

• Zsim

Register-Renaming

• Microarchitecture: Zero (one) idioms & Mov Elimination

Reinforcement Learning

• GLM Post-Training Beyond SFT
• GLM Post-Training Open-Source Map
• Scaling Law

Relationship Design

• Modern Matchmaking System

Reliability Engineering

• Software Design Evidence

Rental

• Rental Checklist

Reporting

• Personal Advantage Workflow
• Work Reporting

Requirements Engineering

• Requirements Analysis

Ring Attention

• AI Training Parallelism

Risk Management

• Work Reporting

RoCE

• AIV Direct Drive
• Ascend Communication Axes
• Ascend Communication Stack
• Ascend Interconnect Evolution

RoPE

• NPU Training Operators - RoPE MRoPE

RouterReplay

• VeRL Router Replay

Rust

• Rust

SASS

• Cuda Optimize : Vectorized Memory Access
• cuda Assembly:PTX & SASS

SCADA

• NVIDIA SCADA

SCALE-UP

• TurboBus PCIe Bandwidth Pooling

SDMA

• Ascend Data Movement Evolution

SE

• SE

SFT

• Diffusion LLM Post-Training

SHMEM

• AIV Direct Drive Programming
• AIV MoE All-to-All
• Ascend Communication API Evolution
• Ascend Communication Runtime Evolution
• Ascend Communication Stack
• DMA Communication Terminology
• Memory Semantics vs RDMA
• Memory-Semantic TransferQueue
• SHMEM Symmetric Memory
• TileXR-SHMEM Abstractions

SIMD

• Arm - Neon
• Assembly Arm
• Double 2 int8
• Intel® Intrinsics Guide
• Manual AVX256 SIMD
• SIMD+SSE+AVX

SIMT

• Ascend DeepEP Ref Dispatch SIMT

SLIC

• Hybrid Multithreaded/OpenMP + MPI parallel Programs
• IPCC Preliminary SLIC Analysis
• IPCC Preliminary SLIC Analysis part2 : Run process
• IPCC Preliminary SLIC Analysis part3 : Hot spot analysis
• IPCC Preliminary SLIC Analysis part4 : cluster environment
• IPCC Preliminary SLIC Case1/2/3
• IPCC Preliminary SLIC Optimization 2
• IPCC Preliminary SLIC Optimization 3
• IPCC Preliminary SLIC Optimization 4: EnforceLabelConnectivity
• IPCC Preliminary SLIC Optimization 5: MPI + OpenMP
• IPCC Preliminary SLIC Optimization 6: Non-blocking MPI
• IPCC Preliminary SLIC algorithm
• IPCC Preliminary SLIC test

SMA

• VeRL Performance Optimization

SNAPSHOT

• Mooncake KV Metadata

SP

• AI Training Parallelism
• RL Data Flow

SPDK

• GPU-Initiated I/O

SRAM

• Scratchpad and NoC

SSD

• Tutti SSD-Backed KV Cache

SSD OFFLOADING

• Mooncake KV Metadata

SSE

• SIMD+SSE+AVX

STEPFUN

• AI Startup Career Risk

STORAGE-NEXT

• NVIDIA SCADA

SWAP

• Linux Executable file: Structure & Running

Scale-Out

• Ascend Interconnect Evolution

Scale-Up

• Ascend Interconnect Evolution

Scaling Law

• Scaling Law

Scratchpad

• Scratchpad and NoC

Sequence Diagram

• Business Process Diagrams

Server

• Disk

Services

• Services Systemd Systemclt

SimAI

• SimAI Architecture

Simulation

• Echo Training Simulator

Skill

• Evidence-Driven Goal Loop

Skylake

• XeonScalable

Slurm

• SlurmCommand

Software Architecture

• Software Design Evidence

SpeculativeDecoding

• VeRL Speculative Decoding

Staleness

• VeRL Async Policy

Stanford

• Other SSG

State Machine

• Business Process Diagrams

Streaming

• Codex Model Speed Benchmark

Subagent

• AI Coding Harness

SuperPixel

• IPCC Preliminary SLIC algorithm

Survey

• Multimodal RL

Swimlane Diagram

• Business Process Diagrams

System Optimization

• Computing Systems Optimization

Systemclt

• Services Systemd Systemclt

Systemd

• Services Systemd Systemclt

TCP/IP

• Ascend Communication Axes

TENSORRT-LLM

• Dynamo TRT-LLM and AgentX

TENT

• Mooncake Classic vs TENT Engine
• Mooncake Codebase Architecture
• Mooncake File vs Block Device
• Mooncake NDS Integration
• Mooncake TENT GDS
• Mooncake TENT Request Path

TILEXR

• AIV MoE All-to-All

TLB

• Context Switch

TLP

• ParallelIntroduction

TND

• BSND TND Operator Layout
• Kimi K3 NPU Training

TP

• AI Training Parallelism

TRANSFER ENGINE

• Mooncake Classic NVMeoF Transport
• Mooncake Classic vs TENT Engine
• Mooncake Codebase Architecture
• Mooncake File vs Block Device

TRANSPORT

• Mooncake Classic vs TENT Engine

TURBOBUS

• TurboBus PCIe Bandwidth Pooling

TUTTI

• Tutti SSD-Backed KV Cache
• io_uring Async I/O

Team Architecture

• Computing Systems Optimization

Technical Due Diligence

• Technology Insight

Technical Feasibility

• Requirements Analysis

Technical Leadership

• Computing Systems Optimization

Technology Radar

• Technology Insight

Technology Strategy

• Technology Insight

Terminal

• Linux Terminal

Threat Modeling

• Software Design Evidence

Thunderbolt

• USB & Thunderbolt & Type-C

TileXR

• Ascend Communication Stack
• DMA Communication Terminology
• TileXR-SHMEM Abstractions

Topcoder

• Codeforces & Topcoder

Training

• VeRL Training Flow

Training Systems

• AI Optimization Stack

TransferQueue

• Memory-Semantic TransferQueue
• VeRL Feature Survey

Transformer

• KV Cache Fundamentals

Transformerless

• xDeepServe on CloudMatrix384

Triton

• AutoFuse
• NPU Training Operators - GDN
• TileLang & Mega-kernel
• Triton & Triton Ascend

Type-C

• USB & Thunderbolt & Type-C

UB-Mesh

• UB-Mesh Architecture

UBSHMEM

• Mooncake Ascend Transports

UCX

• GPU-Centric Communication

UDMA

• Ascend Communication Stack
• Ascend Data Movement Evolution
• Ascend DeepEP MoE Dispatch Optimization
• DMA Communication Terminology
• TileXR-SHMEM Abstractions

UPI

• Motherboard & PCI-e & UPI

URMA

• Ascend DeepEP MoE Dispatch Optimization
• Ascend DeepEP Ref Dispatch SIMT

USP

• AI Training Parallelism
• Attention Cache and Sequence Parallelism

Ubuntu

• Ubuntu server reInstall
• package manager: apt

UltraEP

• XTuner Domino EP

Ulysses

• AI Training Parallelism

Ulysses-Offload

• AI Training Parallelism
• DeepSpeed I/O, Offload, and Asynchrony

UniGRPO

• Multimodal RL

UnifiedBus

• Ascend Communication Axes
• Ascend Communication Stack
• Ascend Data Movement Evolution
• Ascend Interconnect Evolution
• DMA Communication Terminology
• Memory Semantics vs RDMA
• UB-Mesh Architecture

Unreal

• EpicUnrealEngine

Useful

• Understand deeper ： Why xxx is useful?
• Where Do AI Ideas Come From? A Deeper Look

VBench

• Multimodal Generation Evaluation

VEFX-Bench

• Multimodal Generation Evaluation

VERL

• vLLM KV Offloading and GDS

VLA

• VLA VLM + DiT

VLLM

• InferenceX Data Pipeline
• Mooncake Store Design
• Mooncake vLLM Ascend
• PagedAttention
• vLLM KV Offloading and GDS

VLM

• Frontier Model RL
• Ideas around Vision-Language Models (VLMs) / Reasoning Models
• Multimodal Generation Evaluation
• VLA VLM + DiT
• VLM RL Evaluation Datasets

VNC

• VNC

VP

• AI Training Parallelism

VPN

• VPN

VR

• Virtual reality (VR)
• Waifu2x 8K Upscaling Bug

Value Capture

• AI Technology Value Stack

Value of Information

• Multi-Objective Decision Making

VeOmni

• Diffusion LLM Post-Training
• VeRL Backend Parallelism

VeRL

• AI Infra Daily Radar
• BSND TND Operator Layout
• Memory-Semantic TransferQueue
• NPU Training Operators - GDN
• Training Performance Model
• VLM RL Evaluation Datasets
• VeRL Async
• VeRL Async Policy
• VeRL Backend Parallelism
• VeRL Checkpoint
• VeRL Feature Matrix
• VeRL Feature Survey
• VeRL Performance Optimization
• VeRL Rollout Inference
• VeRL Router Replay
• VeRL Speculative Decoding
• VeRL Training Flow

VeRL-Omni

• Ascend 950 A5 Multimodal Quantization
• Multimodal Generation Evaluation

Vectorization

• IPCC Preliminary SLIC Optimization 6: Non-blocking MPI

Video

• Jellyfin

Virtual memory

• Linux Executable file: Structure & Running

Visualization

• AI Model Visualization
• Echart
• Visualization Ranking

Vscode

• Vscode

Vue

• Vue

W8A4

• Inference Quantization Formats

W8A8

• Inference Quantization Formats

WAFER-SCALE

• WaferLLM Co-Design

WAFERLLM

• WaferLLM Co-Design

WAR

• Microarchitecture: Zero (one) idioms & Mov Elimination

WAW

• Microarchitecture: Zero (one) idioms & Mov Elimination

WBS

• Function Decomposition Methods

WLAN

• Family Wi-Fi

WQE

• AIV Direct Drive

Work Management

• Personal Advantage Workflow

X86

• CSAPP: Machine Programming III: Procedures

XCCL

• xDeepServe on CloudMatrix384

XLA

• AI Chip System Barriers

XTuner

• XTuner Domino EP
• XTuner Memory Optimization

Xeon

• XeonScalable

ZeRO++

• DeepSpeed Memory and Parallelism

ZeRO-Offload

• DeepSpeed Memory and Parallelism

ZenFlow

• DeepSpeed I/O, Offload, and Asynchrony

Zero-Idiom

• Microarchitecture: Zero (one) idioms & Mov Elimination

Zsim

• Zsim

abi

• C++ ABI

agent

• My Digital Worker
• My Digital Worker : AutoMoneyMaker - AutoTrader
• My Digital Worker : Model / Software Usage
• My Digital Worker : New Coding Way
• My Digital Worker : New Coding Way Part0 —— Building AI-Coding Env
• My Digital Worker : Work with AI

ai

• AI Documentation Workflow
• AI Model Visualization
• Building Large-Scale AI Systems on Ascend: Training, Inference, and Multimodal Optimization
• Business Trip: 2601-2602 verl + DanceGRPO
• Deploy OpenLLM to one A100
• HPCAI
• My Digital Worker
• My Digital Worker : AutoMoneyMaker - AutoTrader
• My Digital Worker : Model / Software Usage
• My Digital Worker : New Coding Way
• My Digital Worker : New Coding Way Part0 —— Building AI-Coding Env
• My Digital Worker : Work with AI
• Personal Advantage Workflow
• Probability Theory
• Pytorch 1 ：Basic Components
• Pytorch 2 ：more conceptions about training and inference
• Pytorch 2.5 ：Dataset & Dataloader
• Pytorch 3 ：Model & Training
• Pytorch 4 ：Save & Load & Pretrain
• Pytorch 5 ： Distributed Training & Parallelism (Mem Re-allocated)
• Pytorch 6 ：Visualization
• Pytorch 7 ：Memory Optimization(Freeing GPU/NPU Memory Early)
• Pytorch 8 ：Hyperparameter

algorithm

• DMD & DMDR

amd

• AMD CPU
• AMD Epyc Compiler Options
• GCC Compiler Option 1 : Optimization Options

ampere

• Nvidia Arch : Ampere & Hopper & Pascal

anaconda

• Conda
• Jupyter AI

anime

• Anime Super Resolution to 4K & Interpolation to 120 fps
• Diary 230827: 上海二次元之旅
• UnimportantView: Anime Recommendation
• UnimportantView: Film & TV(Anime) Works Rating

aocc

• AOCC

apache

• Web Server: Nginx V.S. Apache2

app

• Buying and Selling on Xianyu: Tips and Insights

apple

• Arm vs X86 (unfinished)

apt

• LinuxCommand: system info
• Network Basic
• Unix Basic

architecture

• GPU
• Microarchitecture: Micro-Fusion & Macro-Fusion
• Microarchitecture: Out-Of-Order execution(OoOE/OOE) & Register Renaming
• Microarchitecture: Pipeline of Intel Core CPUs

arm

• Arm vs X86 (unfinished)
• Assembly Arm

array

• [C++ Basic] STL Data Structure

ascend

• vLLM Inference Profiling

assembly

• Disassembly file: objdump
• GNU Assembly File

auto

• Linux Auto Run : crontab
• Reboot Task

avx

• Vtune Assembly Analysis

avx256

• Manual AVX256 SIMD

bank

• State Owned Enterprise

batch

• PyTorchGeometric

benchmark

• Aarch64 CPU Benchmarks for Performance Degrade
• LLVM Mca ：with BHive (2019)

bilibili

• Piano: Transcribe Piano Sheet Music from Video using AI model

blog

• Hexo
• Web Server: Nginx V.S. Apache2

broadcast

• ServerLogin

bug

• Windows Common Bugs
• Zsim-tlb: bug

bugs

• C Program Compile Problems
• Linux Terminal

business

• 1.3 making-money business

c

• C program compile&run process
• C++ ABI
• Inline Assembly
• Python

c++

• PlantUML
• Programming Principle & Code Test
• Thread

cache

• Cache
• Memory Consistency and Cache Coherence

calloc

• Memalloc

cat

• Nas Disk Speed Test
• Synology NAS Configuration

centos

• Yum

chatgpt

• LLM Usage 1: prompt engineer

chivier

• IPCC Preliminary SLIC Optimization 1

chsh

• FTP

clang

• llvm

clash

• Clash Config 4 yourself
• Clash on LAN/linux/Dockers
• ClashX Pro and Wireguard on Macbook In School Net
• OpenWRT on router
• Wireguard

class

• PlantUML
• Python Class
• [C++ Basic] User-Defined Types: Class

cloud

• Database basics

clustering

• IPCC Preliminary SLIC algorithm

cmake

• cmake

code

• How To Read Code

color

• Colorful Commands

comic

• Calibre and its Pugins for e-hentai Books
• Komga

command

• Colorful Commands

commands

• C program compile&run process

compile

• C Program Compile Problems
• C program compile&run process
• C++ ABI
• GNU Assembly File
• llvm

compile options

• IPCC Preliminary SLIC Optimization 2

compression

• Tar Zip Rar

conda

• Conda

context switch

• Context Switch

cpp

• [C++ Basic] Grammar
• [C++] Destructor Order

cpu

• AMD CPU
• Aarch64 CPU Benchmarks for Performance Degrade
• CPU vs GPU

cpu flags

• AMD CPU

crontab

• Linux Auto Run : crontab

cs

• Code Project Reading
• Overview of Compute system

css

• Css & Scss

ctags

• Linux Terminal

cuda

• Cuda Optimize
• Cuda Optimize : Stencil
• Cuda Optimize : Vectorized Memory Access
• Cuda Program Basic
• Nvidia Nsight
• Nvprof
• The CUDA Execution Model

cuobjdump

• Cuda Optimize : Vectorized Memory Access

dLLM

• Diffusion LLM Post-Training

ddns

• OpenWRT on router

debug

• C & C++: Debug Print like icecream in Python
• Fast Debug: VeRL example
• Servers

deepseek

• [LLM]: DeekSeekV3

diagram

• Graph/diagram/visualization/PPT Examples Yet to Be Explored

disassembly

• Disassembly file: objdump

disk

• Disk
• Disk C: make room for installation
• Migrate From Synology DS220J to UGREEN DX4600
• Mount Network Disk
• Nas Disk Speed Test
• Ubuntu server reInstall

dit

• World Model/UFMs/Omni-Modal: AR vs DiT

div

• LLVM Mca ：with BHive (2019)

dnat

• Firewall

dnf

• Unix Basic

docker

• BitTorrent
• Clash on LAN/linux/Dockers
• Docker
• Docker On Win10
• OpenVPN in Docker
• Podman

docuwiki

• Deploy Dokuwiki to localhost

dokuwiki

• Dokuwiki

domain

• Miscellaneous
• PersonalWebsiteDomain
• Web Server: Nginx V.S. Apache2

domestic

• New Generation Shenwei System

dram

• CPU vs GPU
• RAM

echart

• Linux CMD

email

• Script: Email notifications when machine is free

english

• English Paper Writing

enter

• Linux Terminal

entertainment

• Komga

epic

• EpicUnrealEngine

epyc

• AMD Epyc Compiler Options

ethernet

• ServerLogin

family

• Family Wi-Fi
• Home Network

filesystem

• Disk
• File System

firewall

• Firewall

firewalld

• Linux Network Command Guide

flags

• Assembly X86

flood fill

• Flood fill Algorithm

focusk

• Presentation & Visualization : PPT

fork

• LLVM Mca ：with BHive (2019)

forward

• IP Forward
• SSHForward
• Wireguard

fpic

• C program compile&run process

fstab

• Disk

ftp

• FTP
• FTP 2 : debug
• USTC Network Information Center
• WebCrawler first try

fun

• (Research) Team/Lab Organization
• 0 Overview
• 1.1 Living Needs & Meaning
• 1: Target2chase
• 2.2 Social Part
• 2: Courage to move on
• 3 EfficientJumpingRunning
• 3.2 taskPriority
• 3.3 EfficientWorkLearning
• 6 FPS
• AI Hardware & Accelerators
• AI Infra: 10k-GPU cluster
• AI Training Optimization
• Anime Auto add Chinese Subtitle
• AntiCheat
• Audio
• Balance (Efficient) work & life time
• Benchmark
• Blind Date 1st
• Blind Date 1st(2)
• Blind Date Tips
• Blog writing
• Burnout Monitor : Healthy Body Model + hair/heart-aware exercise
• CProgramReading
• CV Model
• Chrome://tracing
• Classical AI Models
• Colorful Life (TOP)
• Cuda Driver Runtime
• Data Link
• Data Structure Summary
• Datastruture: Tree
• DeviceExpansion
• Disease And Prevention
• Disordered Ideas
• Excel
• Experiments For PIM Motivation
• FPGA
• Financial Values
• Future Plans: House & Car
• GUIAgents
• Go Templates
• HTML
• Host-Core With PIM-Core In 3D-stacked Mem
• Huawei Ascend Domain-Specific Architectures : DaVinci
• Important Date
• Inference Basic
• Inference Optimization
• Japanese
• Keyboard
• LLM Model
• LLM Model Basic
• Lab homepage Template & Website builders choice
• LinuxFolderInstall
• Mathematical Logic & Algebraic structure
• Mkdocs
• Muon Optimizer
• Muon Optimizer + FSDP
• OOTD: outfit of the day
• Open &Free Multimodel AI Tools
• OpenCL Basic
• OpenWRTNetworkManage
• Parallel_sort
• Personal Image Management
• Php
• Piano: Transcribe Piano Sheet Music from Video using AI model
• Postgraduate dormitory
• Predictor
• RL Weekly News
• Research logic
• Salary & Tax & Insurance
• Scientifically Concocting Glasses
• Search, Ads, and Recommendation
• Security
• Social Science
• Synology terminal
• TODO
• Team Cooperation / Relationship
• Tmux
• Topology
• Training Data Usage
• Turing Machine & P versus NP problem
• URLs
• UnimportantView: Film & TV(Anime) Works Rating
• UnimportantView: Game
• User Kernel Mode
• Wake On Lan(Wol)
• Weekly
• When & How 4 team presentation page & knowledge database pool
• 宛如泥潭的大型项目开发困境

function call

• CSAPP: Machine Programming III: Procedures

game

• EpicUnrealEngine
• Switch emulator on PC

gateway

• ServerLogin

gcc

• C program compile&run process
• GCC Compile Error
• GCC Compiler Option 1 : Optimization Options
• GCC Compiler Option 2 : Preprocessor Options
• Inline Assembly

gdb

• GDB
• IPCC Preliminary SLIC Optimization 1
• IPCC Preliminary SLIC test

gdbgui

• GDB

gef

• GDB

gem5

• Gem5

gif

• Image

git

• Git Lfs
• Git Push 2 Homepage
• Git Standardization
• Git Submodule: Data & Code Repository Separate
• Introduction to Git Commands
• Network Basic

github

• Git Standardization
• Github Access
• Introduction to Git Commands
• PicBed & OSS

github action

• Hexo

glibc

• Context Switch

gnu

• C program compile&run process

go

• Go Install and Command
• Go mod
• Golang Syntax
• Web Design 2 : Content Organization & Link Content using go template

golang

• Go Install and Command
• Go mod
• Golang Syntax

gprof

• IPCC Preliminary SLIC Analysis part3 : Hot spot analysis

gpt

• LLM Usage 1: prompt engineer

gpu

• CPU vs GPU
• GPU
• Nvidia Arch : Ampere & Hopper & Pascal

graph

• Graph/diagram/visualization/PPT Examples Yet to Be Explored

group

• Servers

gzip

• Tar Zip Rar

h265

• BitTorrent
• Video

hash

• Hash map
• [C++ Basic] STL Data Structure

header

• C program compile&run process

health

• 2.3: Health Quantify

heap

• Linux Executable file: Structure & Running

hexo

• Hexo

home

• Family Wi-Fi
• Home Network

homepage

• Git Push 2 Homepage
• Homepage Template Conflict
• How SSG Get Work? & Hugo theme creation
• Miscellaneous
• Web Server: Nginx V.S. Apache2

hpc

• HPCAI
• HPCthreeLaws

html

• Css & Scss
• How SSG Get Work? & Hugo theme creation

htop

• Cpu Time
• Linux Network Command Guide

http

• WebCrawler first try

huawei

• Forecasting Housing Prices Around Huawei Xicen Base: A Home Buying Plan for Qingpu District
• Huawei Kunpeng workload
• Travel and Business Trip Checklist

hugo

• Homepage Template Conflict
• How SSG Get Work? & Hugo theme creation
• Hugo

hw

• Ascend Extension for PyTorch Installation
• Debug/Profile/Devlop Tools of PTA
• Working Tips

icc

• Intel Compile Options

icecream

• C & C++: Debug Print like icecream in Python

icpc

• IPCC Preliminary SLIC Optimization 2
• Intel Compile Options

incomplete

• AOCC
• Cache
• GCC Compiler Option 1 : Optimization Options
• GCC Compiler Option 2 : Preprocessor Options
• Memalloc

inference

• SGLang
• Speculative Decoding & eagle3
• Vllm Basic
• vLLM Inference Profiling
• vllm-omni & DiT Inference Accelerate

initd

• Reboot Task

intel

• Intel Compile Options
• Intel Pin
• Intel SDM(Software Developer's Manual)
• Microarchitecture: Micro-Fusion & Macro-Fusion
• Microarchitecture: Zero (one) idioms & Mov Elimination
• Old Pintool Upgrade with newest pin

interval model

• Victima: feature extension

ip

• IP Forward
• IPV4 && IPV6
• Linux Auto Run : crontab
• Linux Network Command Guide
• ServerLogin
• USTC Network Information Center

ipcc

• Hybrid Multithreaded/OpenMP + MPI parallel Programs
• IPCC Preliminary SLIC Analysis
• IPCC Preliminary SLIC Analysis part2 : Run process
• IPCC Preliminary SLIC Analysis part3 : Hot spot analysis
• IPCC Preliminary SLIC Analysis part4 : cluster environment
• IPCC Preliminary SLIC Case1/2/3
• IPCC Preliminary SLIC Optimization 1
• IPCC Preliminary SLIC Optimization 2
• IPCC Preliminary SLIC Optimization 3
• IPCC Preliminary SLIC Optimization 4: EnforceLabelConnectivity
• IPCC Preliminary SLIC Optimization 5: MPI + OpenMP
• IPCC Preliminary SLIC Optimization 6: Non-blocking MPI
• IPCC Preliminary SLIC algorithm
• IPCC Preliminary SLIC test
• Training course - IPCC 5 Optimize common tools
• VtuneOptimize

iptable

• OpenVPN in Docker

iptables

• Firewall
• Linux Network Command Guide

ipv4

• IPV4 && IPV6

ipv6

• Home Network
• IPV4 && IPV6

ispc

• Language

ive

• Jang Wonyoung 장원영

izone

• Jang Wonyoung 장원영

j4125

• Proxmox VE（PVE）on J4125

java

• Java

jekyll

• Other SSG

job

• Interview
• Interview in english

jpg

• Image

k-fold cross validation

• Pytorch 1 ：Basic Components
• Pytorch 2 ：more conceptions about training and inference
• Pytorch 3 ：Model & Training
• Pytorch 4 ：Save & Load & Pretrain
• Pytorch 5 ： Distributed Training & Parallelism (Mem Re-allocated)
• Pytorch 6 ：Visualization

kill

• Linux Terminal

kpop

• Jang Wonyoung 장원영
• Waifu2x 8K Upscaling Bug

kunpeng 920

• Arm vs X86 (unfinished)

latex

• Latex
• Overleaf On Ubuntu

linux

• Cpu Time
• Disk Space Cleanup
• Linux CMD
• LinuxCommand: system info
• Servers

live2d

• WebCrawler first try

llm

• Deploy OpenLLM to one A100
• LLM Usage 1: prompt engineer

llvm

• Kunpeng
• LLVM Mca : huawei HiSilicon's TSV110 work
• LLVM Mca ：with BHive (2019)
• LLVM-MCA: Install&RunTests
• LLVM-MCA: docs
• llvm
• llvm Backend
• llvm Pass

llvm-mca

• LLVM-MCA: docs
• Static Code Analysis
• uops.info: Characterizing Latency, Throughput, and Port Usage of Instructions on Intel Microarchitectures (2019)

local-dev

• VeRL Local Debug on Mac

localhost

• Localhost

logic core

• AMD CPU

loop

• Disk

lscpu

• AMD CPU

lsof

• Firewall
• Linux Network Command Guide

macbook

• ClashX Pro and Wireguard on Macbook In School Net
• Macbook Config
• Wireguard

make

• Make
• cmake

malloc

• Memalloc

manga

• Calibre and its Pugins for e-hentai Books

map

• Hash map

master

• Subscribe

mca

• Kunpeng
• LLVM Mca : huawei HiSilicon's TSV110 work
• LLVM Mca ：with BHive (2019)
• LLVM-MCA: Install&RunTests
• LLVM-MCA: docs

memory

• Benchmark : GUPS
• DRAM types: Size, Latency, Bandwidth, Energy Consumption
• RAM

micro-op fusions

• Static Code Analysis

minicoda

• IPCC Preliminary SLIC Optimization 1

mkdocs

• Homepage Template Conflict
• Web Design 4 : Customize Markdown Grammar In SSG

moonlight

• Game Streaming & Video Streaming

mount

• Disk

mpi

• MPI Compilers

mpicc

• MPI Compilers

mpiicc

• MPI Compilers

mtr

• Tcpdump & wireshark

nD-FullMesh

• UB-Mesh Architecture

name

• Naming your work

nas

• Clash on LAN/linux/Dockers
• Ugreen Nas
• Webdav
• Wireguard

nasm

• LLVM Mca ：with BHive (2019)

neon

• Arm - Neon

network

• Network Basic
• OpenVPN in Docker
• TCP & UDP

networkmanager

• Tcpdump & wireshark

newline

• Linux Terminal

nginx

• Deploy Dokuwiki to localhost
• Miscellaneous
• Web Server: Nginx V.S. Apache2

npm

• npm

nsight

• Deploy Stable Diffusion to A100
• Nvidia Nsight

nvidia

• Nvidia Arch : Ampere & Hopper & Pascal

nvprof

• Nvprof

omni

• DiffSynth & ms-swift
• VeOmni

oneApi

• oneApi

oneapi

• Intel Advisor

openmp

• Disassembly file: objdump
• OpenMP
• OpenMP Reductions

openmpi

• Openmpi MPIOPT

openvpn

• OpenVPN
• OpenVPN in Docker
• Wireguard

opkg

• Migrate From Synology DS220J to UGREEN DX4600

optimization

• Optimization Outline

optimize

• Cuda Optimize : Stencil

optimizer

• ML Optimizer

overleaf

• Overleaf On Ubuntu

page table

• Address Translation

pam

• FTP

paper

• DualPath KV Loading
• FuseLink Multi-NIC Communication
• GPU-Centric Communication
• GPU-Initiated I/O
• LLVM Mca ：with BHive (2019)
• Latent Dirichlet Allocation (2003)
• Micro2023: Utopia
• NVIDIA SCADA
• TurboBus PCIe Bandwidth Pooling
• uops.info: Characterizing Latency, Throughput, and Port Usage of Instructions on Intel Microarchitectures (2019)

parallel

• ParallelIntroduction

password

• Keepass

pdf

• Komga

perf

• Gperftools
• NUMA perf

piano

• Piano Piece Playing Practice
• Piano: Transcribe Piano Sheet Music from Video using AI model

ping

• Network Basic

pip

• Pip Cache
• Pip Package
• pip

png

• Image

podman

• Podman

port

• Firewall

ppt

• Presentation & Visualization : PPT

prefetch

• Cache

price

• Forecasting Housing Prices Around Huawei Xicen Base: A Home Buying Plan for Qingpu District

process

• Linux Executable file: Structure & Running

profiling

• vLLM Inference Profiling

proxy

• Clash on LAN/linux/Dockers
• Cloudflare warp proxy
• Introduction to Git Commands
• Network Basic
• SSHForward

ps

• Cpu Time

pull request

• Introduction to Git Commands

pyc

• Python

pyo

• Python

pypy

• Python

python

• Debug/Profile/Devlop Tools of PTA
• Latent Dirichlet Allocation (2003)
• Pip Package
• Python
• Python Class
• Python Graph & Visualization
• Python MPI
• Python: DataStructure
• PythonRegex
• WebCrawler first try

pytorch

• Ascend Extension for PyTorch Installation
• Debug/Profile/Devlop Tools of PTA

qt

• Nvidia Nsight

ram

• RAM

ranking

• Visualization Ranking

rar

• Tar Zip Rar

rating

• UnimportantView: Film & TV(Anime) Works Rating

readelf

• Disassembly file: objdump

reboot

• Reboot Task
• Servers

reduce

• OpenMP Reductions

regex

• PythonRegex

register

• Assembly X86
• CSAPP: Machine Programming III: Procedures

registers

• Arm vs X86 (unfinished)

report

• Research outline

rl

• VeRL Local Debug on Mac

route

• Localhost

router

• OpenWRT on router
• Switch & Router & CLI Config

rpm

• RPM Package

safety

• Safe File Transport

scons

• Scons

scss

• Css & Scss

security

• Keepass

segmentation fault

• IPCC Preliminary SLIC test

server

• FTP
• Servers
• Ubuntu server reInstall

sh

• Bash Scripting

simd

• Vtune Assembly Analysis

skill

• AI Documentation Workflow

slime

• GLM Post-Training Beyond SFT
• GLM Post-Training Open-Source Map

smanga

• Komga

smb

• Webdav

snap

• Disk

snat

• Firewall

sniper

• Victima: feature extension

socket

• AMD CPU
• Network Basic

speaking

• Phonology

sql

• Sql

sram

• RAM

ssh

• Introduction to Git Commands
• Network Basic
• SSHForward
• ssh config & X11 & jump-machine

ssl

• SSL

stack

• CSAPP: Machine Programming III: Procedures
• Linux Executable file: Structure & Running
• [C++ Basic] STL Data Structure

stencil

• Cuda Optimize : Stencil

step-video

• AI Post Traning: DPO + MPO
• AI Traning System

stream

• Game Streaming & Video Streaming
• Streaming Protocol & Streaming Coding

switch

• Switch & Router & CLI Config
• Switch emulator on PC

synology

• Nas Disk Speed Test
• Synology NAS Configuration

systemctl

• Reboot Task

systemd

• Reboot Task

t2v

• 250217 Step-Video-T2V Reading & Porting
• 260117 Step-3-VL 10B
• AI Model Memory
• Ideas around T2I2V models
• Ideas around Vision-Language Models (VLMs) / Reasoning Models
• World Model/UFMs/Omni-Modal: AR vs DiT

tar

• Tar Zip Rar

tcp

• TCP & UDP

tcpdump

• Tcpdump & wireshark

terminal

• Colorful Commands

thread

• Linux Executable file: Structure & Running
• Thread

tick

• Gem5

tips

• Script: Email notifications when machine is free

tlb

• Address Translation
• TLB: real pagewalk overhead

tmm

• TinyMediaManager

tmpfs

• Disk

tools

• Colorful Commands
• Windows Useful Tools

top

• Linux Executable file: Structure & Running
• Linux Network Command Guide

topology

• Network InterConnect

torch

• Ascend Extension for PyTorch Installation
• Torch Code

torch.dist

• torchrun

torch_npu

• NPU Training Operators - RoPE MRoPE

torchrun

• torchrun

traceroute

• Tcpdump & wireshark

uTorrent

• uTorrent

udp

• TCP & UDP

ufw

• Linux Network Command Guide

ugreen

• Ugreen Nas

unix

• Unix Basic

unravel

• Deepfake - unravel/Baka Mitai

usb

• USB & Thunderbolt & Type-C

user

• Servers

useradd

• FTP

usermod

• FTP

vLLM

• AI Infra Daily Radar
• VeRL Backend Parallelism
• VeRL Feature Survey
• VeRL Rollout Inference

vLLM-Ascend

• Ascend 950 A5 Multimodal Quantization

vLLM-Omni

• AI Infra Daily Radar
• Ascend 950 A5 Multimodal Quantization

vector

• Cuda Optimize : Vectorized Memory Access
• [C++ Basic] STL Data Structure

vectorization

• Arm - Neon
• IPCC Preliminary SLIC Analysis part3 : Hot spot analysis

verl

• Business Trip: 2601-2602 verl + DanceGRPO
• On-Policy Distillation
• VeRL Local Debug on Mac

video

• Game Streaming & Video Streaming
• Streaming Protocol & Streaming Coding
• Video

vim

• VIM

virtual machine

• VirtualBox

visualization

• Graph/diagram/visualization/PPT Examples Yet to Be Explored
• Python Graph & Visualization

vllm

• Speculative Decoding & eagle3
• Vllm Basic
• vLLM Inference Profiling
• vllm-omni & DiT Inference Accelerate

vlm

• World Model/UFMs/Omni-Modal: AR vs DiT

vpn

• Clash on LAN/linux/Dockers
• OpenWRT on router

vscode

• Windows CPP Project Setup

vscode debug

• IPCC Preliminary SLIC Analysis
• IPCC Preliminary SLIC Analysis part2 : Run process

vtune

• IPCC Preliminary SLIC Analysis part3 : Hot spot analysis
• Vtune Assembly Analysis
• VtuneOptimize

wake

• Wake Up Process

warp

• Cloudflare warp proxy
• Github Access
• The CUDA Execution Model

webdav

• Webdav

website

• Web Design 1 : Layout Overview
• Web Design 2 : Content Organization & Link Content using go template
• Web Design 3 : Future Features
• Web Design 4 : Customize Markdown Grammar In SSG

wget

• Network Basic

wifi

• Family Wi-Fi

windows

• Disk C: make room for installation
• Docker On Win10
• Windows CPP Project Setup

wireguard

• ClashX Pro and Wireguard on Macbook In School Net
• Github Access
• OpenWRT on router
• Wireguard
• Wireguard Server 2 Server in OpenWRT

wireshark

• Tcpdump & wireshark

wlt

• USTC Network Information Center

wonyoung

• Waifu2x 8K Upscaling Bug

workflow

• Git Standardization

x11

• Nvidia Nsight

x86

• Arm vs X86 (unfinished)
• Assembly X86
• Vtune Assembly Analysis

xDMA

• DMA Communication Terminology

xDeepServe

• xDeepServe on CloudMatrix384

youtube

• Speech

yum

• Unix Basic

yun

• Yum

zen2

• AMD Epyc Compiler Options
• ZEN2

zero-idioms

• Static Code Analysis

zip

• Tar Zip Rar

zsh

• Install Zsh from source

zsim

• Wake Up Process
• Zsim-tlb: bug