Mail Archives: djgpp/2000/01/05/11:25:12
On 4 Jan 00, at 19:55, Stephen Howe wrote:
>
> Kalum Somaratna aka Grendel wrote in message
<snip>.
>
> >>So one way would be to run a small computationally intensive
> >>bechmark program (does anyone have any ideas?) compiled under
> >>WatcomC++ 11(or whatever is the latest version) and DJGPP/GCC
> >>2.952 with full optimizations and see which performs better. And
> >>maybe we can publish the results on this forum for everyone to see.
>
> Your right. But I would be looking separately at
>
> (i) integer optimisation
> (ii) fp optimisation
> (iii) space and time saving optimisations
> (iv) loop optimisations
> (v) stack lifetime optimisations
> (vi) tail optimisations
> etc.
>
> You will find that it wins on some tests and does badly on others. I
> am aware of it strengths and its weaknesses from having used it for 13
> years. My ideal C & C++ compiler and tools would be a combination from
> all the PC compilers that is on the market.
I indeed wish that there was such a compiler, Sigh.
I too like Watcom (it was my first compiler for developing 32bit
extended dos apps) and IMHO the code produced by Watcom
version 10.x was better than GCC 2.81.
However I haven't checked about the 2.952 versions of GCC and
there optimisations.
If you have or know of any testing/benchmark program please
consider posting it. I'm sure many of the members of this group
would be interested in the results from compiling the programs with
full optimisations with watcom and djgpp.
>
> Speaking honestly, there are a number of areas where the compiler &
> libraries could be improved. For instance, the 32-bit maths libraries
> are built assuming that your executables (32-bit DOS, Win32, OS/2 etc)
> could be running on a 386+287 combination which implies fwait opcodes.
> Well how many people are running on that combination? Not many, I bet.
I might be wrong but I don't think there would be any on this
newsgroup using a 386+287 for serious work in 2000. My 386 is
been put to good use as a doorstopper ;-)
> For everyone running 486 or better, these can be removed. So that
> means numerical maths tests could be faster on Pentiums if the
> libraries wer built differently.
It is indeed suprising how much software (which cannot practically
be run on a 386) for example that trash windoze, are compiled
using only 386 specific instructions. IMHO they really shoud ditch
386 and consider optimizing for at least the 486 and using 486
specific intructions (the ideal would be pentium specific
instructions). The good thing about free software like linux is that
you can recompile the kernel for whatever processor you want.
> At this point are you talking about 32-bit DOS? If so, would you not
> mostly be measuring the DOS extenders performance (Tenberry, Causeway,
> PMODE/W, DOS32 etc) rather than the compiled code? That would be
> tricky to factor out. DOS4GW 2.01 _is_ faster than DOS4GW 1.97 (this
> is the version that comes with Watcom), I have measured it.
Could you please tell me where you got DOS4GW 2.01. I myself
have been on the look out for a newer version than 1.97.
Take Care,
Kalum
- Raw text -