akkartik
/
mu
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

.

This commit is contained in:
Kartik K. Agaram 2021-09-13 04:53:38 -07:00
parent 35c8c8edba
commit 77207192ea
1 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
linux/x86_approx.md
View File

@ -110,6 +110,8 @@ the fourth decimal place.
Among the x86 instructions Mu supports, two are described in the Intel manual Among the x86 instructions Mu supports, two are described in the Intel manual
as "approximate": `reciprocal` (`rcpss`) and `inverse-square-root` (`rsqrtss`). as "approximate": `reciprocal` (`rcpss`) and `inverse-square-root` (`rsqrtss`).
Intel introduced these instructions as part of its SSE expansion in 1999. When Intel introduced these instructions as part of its SSE expansion in 1999. When
it upgraded SSE to SSE2 (in 2000), most of its single-precision floating-point it upgraded SSE to SSE2 (in 2000), most of its scalar[1] single-precision
instructions got upgraded to double-precision — but not these two. So floating-point instructions got upgraded to double-precision — but not
they seem to be an evolutionary dead-end. these two. So they seem to be an evolutionary dead-end.
[1] Thanks boulos for feedback: https://news.ycombinator.com/item?id=28501429#28507118