Ming tslmy

The NEON rasterizer in neon_overlay.cpp is already the most taxing single pass in the pipeline. It runs only over the screen-space footprint of the dice — roughly 60,000–80,000 pixels per frame when several dice are in flight. Per batch of four pixels, the NEON path handles perspective-correct UV interpolation with two vector instructions, then falls back to scalar code for the texture fetch, tangent-space lighting, and framebuffer blend. Roughly ~20 floating-point operations per pixel, processed four at a time.

A separable Gaussian blur — the standard cheap blur — cannot skip empty pixels. It must iterate across the full framebuffer (750 × 560 = 420,000 pixels) in two passes (horizontal + vertical), reading $(2r+1)$ neighbor values per pixel per pass. For a radius-3 kernel (barely visible softening), that is 420,000 × 7 × 2 = 5.88 million memory reads plus the same number of multiply-accumulates. Expressed as "equivalent work" against the rasterizer:

$$\frac{420{,}000 \times 2(2r+1)}{80{,}000 \times 5}

The framework I had to temper with is IOKit, for it didn't include header files by default:

darwin/Platform.c:30:10: fatal error: 'IOKit/IOKitLib.h' file not found
   30 | #include <IOKit/IOKitLib.h>
      |          ^~~~~~~~~~~~~~~~~~
darwin/Platform.c:30:10: note: did not find header 'IOKitLib.h' in framework 'IOKit' (loaded from '/Volumes/Xcode/Xcode.app/Contents/Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS6.1.sdk/System/Library/Frameworks')
9 errors generated.

Discovered this unexpected behavior while using Together AI as an OpenAI-compatible API with LlamaIndex:

        Settings.llm = OpenAILike(
            model="meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo",
            api_base="https://api.together.xyz/v1",
            temperature=0.1,
            api_key=api_key,
            is_function_calling_model=True,

	#!/usr/bin/env python
	import random

	from llama_index.agent.openai import OpenAIAgent
	from llama_index.core import Settings
	from llama_index.core.tools import FunctionTool
	from llama_index.llms.openai_like import OpenAILike
	from pydantic import Field

	from llama_index.core import (
	Settings,
	SimpleDirectoryReader,
	StorageContext,
	VectorStoreIndex,
	)
	from llama_index.legacy.vector_stores import Neo4jVectorStore
	from llama_index.llms.ollama import Ollama

	Settings.llm = Ollama(

	import logging
	from datetime import datetime
	from github import GitHubIntegration


	class GitHubCheckHandler(logging.handler):
	"""
	A logging handler that sends messages to GitHub as a Check.
	"""

	openapi: 3.0.0
	paths:
	/account/approve_registers:
	post:
	responses:
	"200":
	description: approve accounts successful
	"400":
	description: approve accounts have some fails
	summary: get account infomation from database

	<settings>
	<servers>
	<server>
	<id>repo-for-snapshots</id>
	<username>ming</username>
	<password>{ABCDEFGHIJKLMNOP=}</password>
	</server>
	<server>
	<id>repo-for-releases</id>
	<username>ming</username>

	plugins {
	// The gradle-release plugin is designed to work similar to the Maven release plugin. (https://github.com/researchgate/gradle-release)
	id 'net.researchgate.release' version '3.0.2'
	}

	plugins {
	// Applies security defined for a repository in Maven's settings.xml file to MavenRepository elements in a Gradle build - both upload and download. (https://plugins.gradle.org/plugin/org.hibernate.build.maven-repo-auth)
	id "org.hibernate.build.maven-repo-auth" version "3.0.4"
	}