DIVIDE(3clc)

💖 NAME 💖

okay bestie, so this is the half_divide and native_divide page — the division girlies of the OpenCL manual, from the OpenCL Manual. it's giving "math but make it split things in two" 💅

the vibes, formally:

gentype half_divide(gentype x, gentype y);

gentype native_divide(gentype x, gentype y);

---

✨ DESCRIPTION ✨

so half_divide computes x / y. that's it, that's the girl. but bestie, the accuracy?? it's implemented with a minimum of 10-bits of accuracy, i.e. an ULP value less than or equal to 8192 ulp. so like… she's fast, she's not a perfectionist, we respect her anyway 🫶

okay and now for something completely different, and yes this is literally what the original says: native_cos computes cosine over an implementation-defined range. the maximum error is implementation-defined. (not me telling you about cosine on a division page 💀 the docs really said "while we're here".)

a few ground rules, no cap:

• the vector versions of the math functions operate component-wise. the description is percomponent. so like, each lane does its own little thing, independently, iconic behavior ✨
• the built-in math functions are not affected by the prevailing rounding mode in the calling environment, and always return the same value as they would if called with the round to nearest even rounding mode. so don't come for them with your environment variables, they're unbothered 💅
• for any specific use of a function, the actual type has to be the same for all arguments and the return type, unless otherwise specified. match your types, bestie, that's the vibe check.

🌸 about that native_ prefix 🌸

the functions with the native_ prefix may map to one or more native device instructions and will typically have better performance compared to the corresponding functions (without the native__ prefix). speed, queen 🔥

but — and this is the important part — the accuracy (and in some cases the input range(s)) of these functions is implementation-defined. so it depends on your device. she's fast but she's mysterious, iykyk.

🦋 the gentype thing 🦋

gentype indicates that the functions can take float, float2, float3, float4, float8 or float16 as the type for the arguments. pick your size, they all slay 💖

---

📜 SPECIFICATION 📜

OpenCL Specification[1]

---

👀 SEE ALSO 👀

mathFunctions(3clc)

---

👑 AUTHORS 👑

The Khronos Group

---

⚖️ COPYRIGHT ⚖️

Copyright © 2007-2011 The Khronos Group Inc.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and/or associated documentation files (the "Materials"), to deal in the Materials without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Materials, and to permit persons to whom the Materials are furnished to do so, subject to the condition that this copyright notice and permission notice shall be included in all copies or substantial portions of the Materials.

legal bestie said "include the notice in all copies or substantial portions" and we do NOT argue with legal bestie, periodt.

---

🎀 NOTES 🎀

1. OpenCL Specification page 244, section 6.12.2 - Math Functions

---

okay that's the whole page, you're basically a division expert now, no cap 🌈💜

and for the sign-off, this lil gem came from The Khronos Group, dated 01/14/2021, and it's filed under the manual page name HALF_DIVIDE, NATIVE_(3clc) — go forth and divide, bestie 💅✨