CompuServe test bed

CompuServe Thread

Forum unknown · Talk to the Trade

#Lattice C 3.10

7 messages in this thread
#62875From: Ed SilvaApr 8, 1987 10:38 PM
Hello: I am trying to get an update of the newer Lattice 3.10 version. what are the pluses and minuses of getting, also does it really suupport FFP and/or IEEE math routines directly without having to kludge? how big is the linked code? how about compiler/nker speed any improvements. also on the brochure that CBM send to future software developers Lattice is suppose to have a 3D library anybody knows and/or uses it if so please tell me what it is? Ed.
#62883From: Ariel ButlerApr 8, 1987 11:00 PM
Ed: If you are suffering along with Lattice 3.03, then by all means please DO upgrade to 3.10. While the code size reduction with standard compile options is modest, there IS a new merged hunk mode for small programs which will reduce their size somewhat more. As far as the math support goes, it does appear to work with the new Lattice IEEE routines (which are allegedly 5-10 times faster than the old ones), the Motorola FFP routines, and the runtime Math Library on the Workbench disk. I say "allegedly" because we haven't used the math rout- ines extensively. I can say lots of stuff about the docs though. They're GREAT. At least, great relative to the old stuff! Then, of course, there's BLINK which you might already be using. Plus, an assembler, some Unix tools (Grep,Extract,Splat,etc..) that I've used and are handy. If it comes across that I'm happy with the product, then I'm conveying the correct impression. <wink> ab
#63026From: Ed SilvaApr 9, 1987 9:05 PM
What does the assembler does…? will it generate clean machine code or is it for compiling SEPARATE 68000 source and the Link it (or BLink it) to your run time modules. what do you mean that it does support FFP,IEEE ,do you mean mean that when i load #include <mathh.FFP> or whatever name is for the FFP,it will use them when i use sines,sqrt() and such? Ed.
#63149From: Ariel ButlerApr 10, 1987 3:03 PM
Ed: The assembler has at least one purpose, and that is to reassemble the "c.a" module they provide (if you make changes to it). They claim that it is a reasonably powerful Macro assembler, but not being experienced on 68000 assemblers, I can't confirm it. I have used it to redo "c.a", though, and it is relatively quick. And yes, it does produce BLINK'able code. No, the "math.h" include file is used for ALL the libraries. Essentially, you have the choice of linking with either of two compile-time libraries (lcm.lib or the Motorola FFP lib). There is also a math library that you can link to dynamically at run-time using the OpenLibrary call and MathBase (see the RKM about this). Which one to use will of course depend on your applicat- ion. The IEEE code is 80 bit, and produces more accurate results than the Motorola FFP, which uses 32 bit representation. Ariel P.S. I might have the libraries slightly mixed up, I only use lcm.lib which I know is IEEE. ;;C_AMIG: – Job 114: BTW, I read your AC column all the time!
#63149From: Ariel ButlerApr 10, 1987 3:03 PM
Ed: The assembler has at least one purpose, and that is to reassemble the "c.a" module they provide (if you make changes to it). They claim that it is a reasonably powerful Macro assembler, but not being experienced on 68000 assemblers, I can't confirm it. I have used it to redo "c.a", though, and it is relatively quick. And yes, it does produce BLINK'able code. No, the "math.h" include file is used for ALL the libraries. Essentially, you have the choice of linking with either of two compile-time libraries (lcm.lib or the Motorola FFP lib). There is also a math library that you can link to dynamically at run-time using the OpenLibrary call and MathBase (see the RKM about this). Which one to use will of course depend on your applicat- ion. The IEEE code is 80 bit, and produces more accurate results than the Motorola FFP, which uses 32 bit representation. Ariel P.S. I might have the libraries slightly mixed up, I only use lcm.lib which I know is IEEE. ;;C_AMIG: – Job 114: BTW, I read your AC column all the time!
#63026From: Ed SilvaApr 9, 1987 9:05 PM
What does the assembler does…? will it generate clean machine code or is it for compiling SEPARATE 68000 source and the Link it (or BLink it) to your run time modules. what do you mean that it does support FFP,IEEE ,do you mean mean that when i load #include <mathh.FFP> or whatever name is for the FFP,it will use them when i use sines,sqrt() and such? Ed.
#62883From: Ariel ButlerApr 8, 1987 11:00 PM
Ed: If you are suffering along with Lattice 3.03, then by all means please DO upgrade to 3.10. While the code size reduction with standard compile options is modest, there IS a new merged hunk mode for small programs which will reduce their size somewhat more. As far as the math support goes, it does appear to work with the new Lattice IEEE routines (which are allegedly 5-10 times faster than the old ones), the Motorola FFP routines, and the runtime Math Library on the Workbench disk. I say "allegedly" because we haven't used the math rout- ines extensively. I can say lots of stuff about the docs though. They're GREAT. At least, great relative to the old stuff! Then, of course, there's BLINK which you might already be using. Plus, an assembler, some Unix tools (Grep,Extract,Splat,etc..) that I've used and are handy. If it comes across that I'm happy with the product, then I'm conveying the correct impression. <wink> ab