Thu, 9 Feb 2023 00:37:09 -0800 (PST)

From: | gah4 <gah4@u.washington.edu> |

Newsgroups: | comp.compilers |

Date: | Thu, 9 Feb 2023 00:37:09 -0800 (PST) |

Organization: | Compilers Central |

References: | 23-01-092 23-02-003 23-02-019 23-02-025 23-02-026 23-02-029 23-02-034 |

Injection-Info: | gal.iecc.com; posting-host="news.iecc.com:2001:470:1f07:1126:0:676f:7373:6970"; logging-data="44137"; mail-complaints-to="abuse@iecc.com" |

Keywords: | arithmetic, history, comment |

Posted-Date: | 10 Feb 2023 13:16:36 EST |

In-Reply-To: | 23-02-034 |

On Wednesday, February 8, 2023 at 8:50:19 AM UTC-8, Hans-Peter Diettrich wrote:

(snip)

*> That's correct, the inprecise representation of FP numbers allows for*

*> such tricks. The hidden bit trick can be used again with the FP*

*> exponents, as I outlined in my Dynamic Floating Point Exponents proposal*

To continue discussion about OoO and the 360/91, S/360 specifies

floating point divide with a truncated quotient. The 91 uses a

Newton-Raphson divide algorithm, using its high-speed multiplier,

to generate a rounded quotient.

Along with imprecise interrupts, that is on the list of incompatibilities

with other S/360 models.

As for hidden bit, S/360 uses base 16 floating point, so no hidden bit.

[I actually programmed a /91 in Fortran, so, yeah. -John]

Post a followup to this message

Return to the
comp.compilers page.

Search the
comp.compilers archives again.