C64 BASIC Compilers Speed Test using the Quicksort algorithm

I use the Abacus Basic64 and Compute!s Gazette Sprint compilers in a speed test using a visual quicksort demo

Links

Sprint Compiler

Lightning Sort

00:00 Introduction
00:36 The Compilers
00:58 BASIC Quicksort
02:56 Abacus BASIC64 compiler
03:42 BASIC64 Optimizer 1 (default)
05:38 BASIC64 Optimizer 2, PCODE|Integer
05:45 BASIC64 Compiler directives
06:57 BASIC64 Optimizer 2, ML|Integer
07:41 Sprint compiler
09:11 Modifying QSORT.B
10:24 Sprint compile
11:19 Speed Test
13:42 Comments
14:08 Lightning Sort

[andrewdunbar]

When I read "compilers speed tests" I nearly spat my coffee on my keysboard!

[stephanan]

Mit dem Basic Boss, unter Nutzung der Direktiven, habe ich eine TIME: 1.26 secs. erreicht.
Schneller geht nur noch selbstgeschriebener, optimierter Assemblercode. :-)

[KellyMurphy]

It would be interesting to see a comparison of hand written assembly qsort routines.

[vcv]

I bought this BITD too, never bothered with the type-in from Gazette. I never got more than a 3x increase using the Abacus product on the Rectan 3-D surface plotting program by Tim Colvin, from Compute May 84. A lot of SIN, COS and float->integer conversions it clearly taxed the capability of the compiler which still had to call the transcendental and floating point math functions from the ROM.

[kangarht]

wow I only knew austro speed basic compiler, never would have expected such speeds on 256 numbers not even from compiled basic

[stephanan]

I'm sorry I wrote in German. I only noticed a little later that this was an English-language YouTube channel.


With the Basic Boss, using the directives, I have a TIME: 1.26 secs. reached.
The only thing that can be done faster is self-written, optimized assembly code. :-)

[jeffyp]

"Thanks for watching"😉 "and be careful out there'😈

ok i hit like!

😂

[trueKENTUCKY]

Oh God I'm not sober enough for this

[PhilipvanderMatten]

Wonder how quick the C64 would be with this sort programmed entirely in 6510 code.

[mikechappell]

You should have used the Sprint version for the benchmark. It didn't need to do any expression parsing, while the original did.

[henkeboy]

How would the result co pare to a pure machinecode program?

[larswadefalk]

Some things come to my mind about your ML example. Did that use direct access to the screen buffer, or did it go via the kernel print? Maybe the basic code did the same. Because I suppose the printing itself was included in the benchmark. Then I also wonder if it wouldn't be possible to create a compiler that actually translate everything into pure assembly, I mean even using an optimized library for everything that otherwise is called in the basic rom. Because I suppose every basic command is still being processed in the basic rom, only the translation is left out?

[drphilxr]

there's no option in advanced options to return to the main menu lol- return is it. trial and error

[BillAnt]

Well quick sort is great for not completely random lists, but for completely random bubble sort tends to be faster.

[TheStuffMade]

Quick Sort will always be O(n log n) no matter how much you optimize it.

[jeffyp]

those first two file names @9:05 are bs