barsh@stsci.EDU (John Barshinger) (11/24/87)
The following is my experience with the Lattice 4.0 compiler and
utilizing the 68010 instead of the 68000 in my AMIGA...
Lattice C 4.0 Performance Report
--------------------------------
1. Make of VT100 source from scratch.
2. Compile/Link of the drystone program.
3. Compile/Link of the drystone program using base reg. addressing/no stack
check/smalldata/smallcode
4. Executable size of #1.
5. Executable size of #2.
6. Executable size of #3.
7. Results of execution of #2.
8. Results of execution of #3.
Lattice 3.1 | Lattice 4.0 | %improvement
============================================================================
1. 17 min, 18 sec | 15 min, 52 sec | 9%
2. 1 min, 16 sec | 1 min, 08 sec | 11%
3. 1 min, 18 sec | 1 min, 08 sec | 13%
4. 71892 bytes | 52228 bytes | 27%
5. 15600 bytes | 11668 bytes | 25%
6. 14068 bytes | 10980 bytes | 22%
7. 689 drys/sec | 826 drys/sec | 20%
8. 740 drys/sec | 900 drys/sec | 22%
I would say that the Lattice upgrade has been well worth it and the
continued support from both Lattice Customer Support and Lattice
Technical Support is much appreciated.
68000 vs. 68010 Comparisons
---------------------------
These are timings that I've done of the above utilizing the 68010
processor instead of the 68000. The 68010 is drop-in compatible
with the Amiga 68000 system. However, there are some programs out
there that will cause a GURU since they access status registers directly
instead of calling the AmigaDOS routine to do it. There is a "handler"
in the public domain that will fix this problem for such programs, I
believe it's called decigel. I can get it if somebody wants really
wants it.
note: Lattice 4.0 compiler used
68000 | 68010 | %improvement
============================================================================
1. 15 min, 52 sec | 15 min, 10 sec | 4.4%
2. 1 min, 08 sec | 1 min, 04 sec | 5.9%
3. 1 min, 08 sec | 1 min, 04 sec | 5.9%
7. 826 drys/sec | 869 drys/sec | 5.2%
8. 900 drys/sec | 944 drys/sec | 4.9%
So, it looks like in general doing the things I do, the 68010 buys
me about 5%, which isn't real significant, but I have nothing better
to do with the chip...
Disclaimer: I am in no way affiliated with Lattice or Motorola. I am
just passing along the info as I see it for whomever is interested.
jrbii
~~~~~
--
John Barshinger
Space Telescope Science Institute
Baltimore, MD 21218
plink:jrbii | bix:jrbii | barsh@stsci.edubarsh@stsci.EDU (John Barshinger) (11/25/87)
in article <111@obi-wan>, barsh@stsci.EDU (John Barshinger) says: > > The following is my experience with the Lattice 4.0 compiler and > utilizing the 68010 instead of the 68000 in my AMIGA... > This is in response to the mail messages I have been getting regarding the above named performance analysis: Numbers 7 & 8 (the drystone program executions) are not time values. They are number of drystones executed per second average. The moral of the story is DRYer is indeed better... jrbii ~~~~~ -- John Barshinger Space Telescope Science Institute Baltimore, MD 21218 plink:jrbii | bix:jrbii | barsh@stsci.edu