michaeljclark / CMakeLists.txt

Last active December 9, 2024 21:47

five ways to invoke the LLVM disassembler

	cmake_minimum_required (VERSION 3.12)

	project(llvmdis)

	find_package(LLVM REQUIRED CONFIG)
	include_directories(${LLVM_INCLUDE_DIRS})
	add_definitions(${LLVM_DEFINITIONS_LIST})
	add_definitions(-Wall)

	add_executable(llvmdisc llvmdisc.c)

michaeljclark / glyph.md

Last active March 9, 2025 00:18

a super regular RISC that encodes constants in immediate blocks

glyph

a super regular RISC that encodes constants in immediate blocks.

introduction

glyph is a super regular RISC architecture that encodes constants in a secondary constant stream accessed via an immediate base register. the immediate base register branches like the program counter and the call instructions set (pc,ib) together for procedure calls and returns.

michaeljclark / slurphash.c

Created June 12, 2024 08:47

SipHash derivative using carryless multiplies instead of rotates

	/*
	SlurpHash 0-pre-alpha AVX implementation

	(SipHash derivative using carryless multiplies instead of rotates)
	------------------------------------------------------------------

	Copyright (c) 2012-2016 Jean-Philippe Aumasson <[email protected]>
	Copyright (c) 2012-2014 Daniel J. Bernstein <[email protected]>
	Copyright (c) 2017 Salvatore Sanfilippo <[email protected]>
	Copyright (c) 2024 Michael Clark <[email protected]>

michaeljclark / pythag.c

Created March 4, 2024 08:26

find largest 32-bit Pythagorean integer triple with a 64-bit square

	#include <stdio.h>

	/*
	* cc -O2 pythag.c -o pythag
	*
	* find largest 32-bit Pythagorean integer triple with a 64-bit square
	*/

	typedef unsigned long long ulong;

michaeljclark / pattern-synth.py

Last active September 9, 2024 11:32

combinatorial bit pattern synthesizer in python

	#!/usr/bin/env python3

	# MIT license
	#
	# combinatorial bit pattern synthesizer in python
	#
	# Copyright (c) 2024 Michael Clark <[email protected]>
	#
	# Permission to use, copy, modify, and distribute this software for any
	# purpose with or without fee is hereby granted, provided that the above

michaeljclark / refcounting.md

Last active October 18, 2023 19:40

Work in progress draft on the design of a reference counting system.

References

Work in progress draft on the design of a reference counting system.

Primitive operations on reference counters

Summary of atomic operations required to implement reference counting. All reference counter operations take a pointer to a reference counter object. Reference to relaxed, acquire and release memory ordering on the primitive operations primarily refer to the barrier that protects the

michaeljclark / cpuident.c

Last active July 30, 2022 08:17

cmake configure time CPU feature detection for x86 processors

	#include <stdio.h>
	#include <stdlib.h>

	#if defined __GNUC__ && (defined __i386__ \|\| defined __x86_64__)
	#define HAS_X86_CPUID 1
	#include <cpuid.h>
	static inline void __x86_cpuidex(int reg[], int level, int count)
	{ __cpuid_count(level, count, reg[0], reg[1], reg[2], reg[3]); }
	#elif defined _MSC_VER && (defined _M_IX86 \|\| defined _M_X64)
	#define HAS_X86_CPUID 1

michaeljclark / vf128.txt

Last active November 9, 2024 12:50

table of single-byte encodings for the vf128 variable-length floating-point number format

	0 ▄▄▄▄ 0.0000 1 ▄▄▄▟ 0.0625 2 ▄▄▄▙ 0.1250 3 ▄▄▄█ 0.1875 4 ▄▄▟▄ 0.2500 5 ▄▄▟▟ 0.3125 6 ▄▄▟▙ 0.3750 7 ▄▄▟█ 0.4375
	8 ▄▄▙▄ 0.5000 9 ▄▄▙▟ 0.5625 10 ▄▄▙▙ 0.6250 11 ▄▄▙█ 0.6875 12 ▄▄█▄ 0.7500 13 ▄▄█▟ 0.8125 14 ▄▄█▙ 0.8750 15 ▄▄██ 0.9375
	16 ▄▟▄▄ 1.0000 17 ▄▟▄▟ 1.0625 18 ▄▟▄▙ 1.1250 19 ▄▟▄█ 1.1875 20 ▄▟▟▄ 1.2500 21 ▄▟▟▟ 1.3125 22 ▄▟▟▙ 1.3750 23 ▄▟▟█ 1.4375
	24 ▄▟▙▄ 1.5000 25 ▄▟▙▟ 1.5625 26 ▄▟▙▙ 1.6250 27 ▄▟▙█ 1.6875 28 ▄▟█▄ 1.7500 29 ▄▟█▟ 1.8125 30 ▄▟█▙ 1.8750 31 ▄▟██ 1.9375
	32 ▄▙▄▄ 2.0000 33 ▄▙▄▟ 2.1250 34 ▄▙▄▙ 2.2500 35 ▄▙▄█ 2.3750 36 ▄▙▟▄ 2.5000 37 ▄▙▟▟ 2.6250 38 ▄▙▟▙ 2.7500 39 ▄▙▟█ 2.8750
	40 ▄▙▙▄ 3.0000 41 ▄▙▙▟ 3.1250 42 ▄▙▙▙ 3.2500 43 ▄▙▙█ 3.3750 44 ▄▙█▄ 3.5000 45 ▄▙█▟ 3.6250 46 ▄▙█

michaeljclark / bench_asn1.txt

Created February 25, 2022 23:41

benchmark results from crefl bench_asn1 benchmark for strtod, snprintf, vlu, leb, vf128, asn1 float and integer encoding

	# crefl serialization benchmarks run on Intel Core i9-7980XE CPU

	benchmark count time(s) op(ns) ops/s MiB/s
	------------------------ ------- ------- ------- ------------- ---------
	[ 0] f64-read-strtod-dec 1M 0.11 109.65 9,119,975 69.580
	[ 1] f32-read-strtof-dec 1M 0.09 86.43 11,570,658 44.139
	[ 2] f32-write-snprintf-dec 1M 0.14 143.62 6,962,688 26.561
	[ 3] f64-write-snprintf-dec 1M 0.20 203.81 4,906,604 37.434
	[ 4] f64-asn.1-read-byptr 1M 0.03 30.61 32,667,961 249.237
	[ 5] f64-asn.1-read-byval 1M 0.03 28.96 34,532,112 263.459

michaeljclark / phydate128.py

Last active November 20, 2022 20:53

phydate128 — a 128-bit date format with 4-bit prefix and 124-bit counter

	#!/usr/bin/env python3

	# phydate128 — a 128-bit variable width datetime format
	#
	# struct { int<4> timebase; uint<4> length; bytes<length> counter; }
	#
	# prefix byte is composed of a 4-bit signed frequency base and 4-bit length
	# frequency base is signed two's complement where 0b0000 is femtoseconds
	# length indicates the size of the variable length counter field in bytes.
	#

Michael Clark michaeljclark

glyph

introduction

References

Primitive operations on reference counters