Return to the
comp.compilers page.
Search the
comp.compilers archives again.
| Related articles |
|---|
| Re: Why is Cobol ignored in compiler textbooks? preston@dawn.cs.rice.edu (1992-04-22) |
| Cobol picture data ericj@csn.org (Eric Jacobsen) (1992-04-22) |
| Re: Cobol picture data reid@csgrad.cs.vt.edu (1992-04-23) |
| Re: Cobol picture data md@sco.COM (1992-04-27) |
| Newsgroups: | comp.compilers |
| From: | reid@csgrad.cs.vt.edu (Thomas F. Reid) |
| Keywords: | Cobol |
| Organization: | VPI&SU Computer Science Department, Blacksburg, VA |
| References: | 92-04-099 92-04-109 |
| Date: | Thu, 23 Apr 1992 16:29:41 GMT |
>(Say what you will about efficiency, but it's a language for data base
>operations. I/O is the bottleneck. Although for non-I/O-bound
>operations, it's reeaalllyyy ssllooww.)
Don't be so quick to assume that. In the mid 1980s, a project I worked on
analyzed a well-known-DBMS in search of performance improvements in very
large databases. Initial investigations showed that the DBMS was CPU
bound - 87-93% on a standalone VAX 11/780. Caching the indices in a 32 MB
ram disk brought very little improvement. So we put a monitor on it to
discover in what subroutines it was spending most of its time. A
significant number of them were conversion routines. It seemed that in
order to have a null value, they used binary zeroes in their BCD numbers
and non-null numbers had their digits increased by one. Thus, to perform
A + B, they had to subtract one from each digit of both A and B, perform
the addition, and then add one to each digit to store it. Microcoding the
worst 50 or so routines brought over a 50% impovement.
Tom Reid
--
Return to the
comp.compilers page.
Search the
comp.compilers archives again.