home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!sun-barr!cs.utexas.edu!zaphod.mps.ohio-state.edu!sdd.hp.com!news.cs.indiana.edu!news.orst.edu!pmontgom
- From: pmontgom@math.orst.edu (Peter Montgomery)
- Subject: Re: trapping speculative ops (LONG)
- Message-ID: <Btvou1.6F9@news.orst.edu>
- Sender: usenet@news.orst.edu (Usenet News admin)
- Nntp-Posting-Host: ilanga.math.orst.edu
- Organization: Oregon State University Math Department
- References: <CLIFFC.92Aug28085924@antigone.rice.edu> <1992Aug31.224611.5196@odin.diku.dk>
- Date: Tue, 1 Sep 1992 02:42:00 GMT
- Lines: 31
-
- In article <1992Aug31.224611.5196@odin.diku.dk> thorinn@diku.dk
- (Lars Henrik Mathiesen) writes:
- >To summarize: trap bits are proposed to enable a compiler to move a
- >potentially trapping operation outside of a condition. As a special
- >case, this would allow many, if not all, of the same optimizations
- >that can be done if a load through a NULL pointer yields zeroes.
-
- On the early models in the CDC Cyber series, such
- as the CDC 6400 and 6600, floating point errors
- (e.g., overflow, division by zero) generated special
- results (e.g., indefinite quantity). They did not generate
- a trap until those NaN-like results were used as operands
- to other floating point instructions.
-
- Sometimes, as in A = B + C/D where D = 0,
- the error would be detected a few instructions after
- the NaN-like result was generated and used as another operand.
- Often, however, there would be a long gap.
- Certain operations like converting an indefinite quantity to integer
- generated no trap. Since integer division used floating point
- instruction, the code I/J where J = 0 could proceed without a trap.
-
- The behavior changed on some later models, like the CDC 7600.
- Those models gave a trap whenever the bad result was generated
- (or a few instructions later). It was often possible, when looking
- at core dumps, to see the bad operands still in the registers.
- --
- Peter L. Montgomery Internet: pmontgom@math.orst.edu
- Dept. of Mathematics, Oregon State Univ, Corvallis, OR 97331-4605 USA
-
- Energy conservation principle: It doesn't matter.
-