We don’t know how many real uses cases it has affected because any problems would have been found and fixed during testing.
I’m sure that issues were rare but I doubt I was the only person in the history of the world who wasted time trying to understand why sin(pi) was giving such inaccurate results when the documentation said it should not. Forgive me for using the documentation to understand how fsin works.
Your claim that the “typical” use of sin is with an argument in cycles (?) or degrees is odd and I don’t believe it. I think most software would use radians and only convert to degrees for human consumption.
Anyway, the documentation was wrong. Now it’s right.

“Only if you start doing funny things with fsin” – I’m not sure why invoking sin(pi) is a “funny thing”. Regardless, the problem is that Intel’s documentation was wrong. That’s it. If you want developers to know that sin(pi) will not have a lot of precision then you just need to make sure your documentation reflects that. Now it does, so I’m calling that a victory.

They missed that because you get less than 5 decimals digits only if you start doing funny things with fsin, and if you use fsin for what it was intended for – you get accurate results. If someone wrote software which depends on what funny result sin(pi) returned, it’s bad software to start with and fixing sin wouldn’t help it.

It has affected precisely 0 (zero) real use cases. Nobody puts put pi-sin(pi) in production code, and even if they do, it’s entirely their fault, not Intel’s. It lead to 0 errors in flight control software and CAD. Because rounding errors in the input of sin dominates. Moreover, the typical use of sin is with argument in cycles or degrees, which get converted to radians with low, 53-bit precision, not your fancy 106+ precision.

Mike! It’s so good to hear from you. Yes, I remember that vividly. It was an amazing demonstration of generosity and community. Thanks again for welcoming me into your home and letting me borrow your computer.

Hope you are doing well.

Not sure if you remember or not but during your travels in Australia you found yourself in need of an Amiga system to do a little work on CygnusEd.

I believe you attended a local user group and asked if anyone had a reasonable development system you could use.

I offered mine and used that as an excuse to buy a higher resolution monitor.

I still have that amiga 2000 and it still boots.

it was good to see that machine being used for real work ~35 years ago.

Best regards,

Mike Cruse

]]> https://randomascii.wordpress.com/2025/10/22/reflections-on-my-tech-career-part-1/#comment-105112 Sat, 07 Feb 2026 03:23:27 +0000 http://randomascii.wordpress.com/?p=4184#comment-105112 HI Bruce,

Hope you are doing well.

Not sure if you remember or not but during your travels in Australia you found yourself in need of an Amiga system to do a little work on CygnusEd.

I believe you attended a local user group and asked if anyone had a reasonable development system you could use.

I offered mine and used that as an excuse to buy a higher resolution monitor.

I still have that amiga 2000 and it still boots.

it was good to see that machine being used for real work ~35 years ago.

Best regards,

Mike Cruse

Hope you are doing well.

Not sure if you remember or not but during your travels in Australia you found yourself in need of an Amiga system to do a little work on CygnusEd.

I believe you attended a local user group and asked if anyone had a reasonable development system you could use.

I offered mine and used that as an excuse to buy a higher resolution monitor.

I still have that amiga 2000 and it still boots.

it was good to see that machine being used for real work ~35 years ago.

Best regards,

Mike Cruse

Thank you for the kind words Patrick. I enjoyed teaching at DigiPen and I think that the public speaking practice which it gave me was very helpful in both my professional and personal life.

Enjoy your retirement, Bruce!

Patrick Hackett