From: sinner@cafe.net (The Sinner) Newsgroups: comp.emulators.apple2 Subject: Emulator Info Date: 12 Apr 1995 05:02:30 GMT Organization: Cafe.Net Internet Access System Lines: 89 Message-ID: <3mfmt6$3gf@scipio.cyberstore.ca> NNTP-Posting-Host: espresso.cafe.net X-Newsreader: TIN [version 1.2 PL2] Please reply to wnight@cafe.net At the suggestion of a few people, I have performed actual speed tests as opposed to just guessing how fst the emulators are. But first some corrections to what I said earlier... Applewin does have keyboard equivalents for all of it's commands. I just couldn't find a reference to them in the docs. It also is not slow, in fact is just a bit slower than the dos-based emulators which aren't quite at well implemented. (And significantly faster than Sim2e) I performed a series of 5 tests (which I will upload in the following message) on my GS with a 8mhz Zip Chip, the GS running at slow speed (which I used for the //e times in the testing program), and then Applewin, Apl2em-2, Sim2e, and Applemu on my 486. In case you care to know exactly what I was running to get the times that I did, my machine is a 486DX4/100 with 48megs of ram, a Cirrus Logic 5426 1MB VLB video card, and QEMM 7.5. The tests were a simple basic for-next loop, a text-scrolling test, basic HGR routines, ASM counting (no output), and an ASM HGR routine. For-Next |Scrolling|Basic HGR|ASM Loops|ASM HGR |Average ---------|---------|---------|---------|---------|--------- ZIP GS 11.2|5.35|8.4 |4.94|9.2 |6.08|24 |5.58|4.7 |5.74| |5.54 ---------|---------|---------|---------|---------|--------- AplWin1 61.5|.975|40.5|1.02|58.1|.963|125 |1.07|27.8|.971| |.999 ----------------------------------------------------------- Apl2em2 7.1 |8.45|5.1 |8.13|8.2 |11.4|11.7|11.4|12 |2.25| |7.41 ---------|---------|---------|---------|---------|--------- AppleMu 7 |8.57|4.9 |8.46|8.6 |6.51|11.7|11.4|12.1|2.23| |7.43 ---------|---------|---------|---------|---------|--------- Sim2e 19.1|3.14|42.1|.985|35.2|1.59|28.5|4.7 |118 |.228| |2.13 ---------|---------|---------|---------|---------|--------- AplWin2 8.8 |5.35|9.5 |4.36|18.3|3.06|17.5|7.65|17.9|1.5 | |4.38 ---------|---------|---------|---------|---------|--------- Average 25.6|6.37|15.4|5.48|17.6|5.56|8.94|8.79|40 |1.55| Parts of this may not be clear so I'll explain what the number on the chart mean. The first number under each test is the number of seconds it took it to execute on my machine. The second is the number of times faster than the //e this was. Fractional numbers mean that it was slower! The score in the last column is the average multiple by which the speed of the emulator on my machine exceeded a //e. The average score in the bottom row is how well the compilers performed on everage. This show what things (graphics) are the hardest to emulate quickly. AplWin1 & 2 are not different program, but rather different settings in the program Applewin. Number 1 is run with the control speed option on. This forces it to try to run as slowly as an apple//e for games and timing sensitive programs. You can see that it is almost exactly as fast as a //e and that any difference is probably due to a slow finger on the stopwatch. Number 2 is without the speed limiting on. This lets it run at full speed. The //GS is also excluded from the compilation of averages, along with AplWin1, because Aplwin1 is trying to run slowly and the //gs does different things at different speeds that the IBM so the figures wouldn't correspond with the emulators. The bottom row of averages is for the last four emulators. A note on Sim2e. I performed the testing with Debugging turned off, Page Flipping turned on, and Flashing Text turned off. Sim2e scored remarkably low in the ASM HGR text but seems to play games at quite close to proper speed. I wonder if it has some speed limiting option I didn't notice or if the graphics speed is simply compensated for by the speed of the rest of it. Comments on the program SPEED.TEST... it's just a simple little thing I whipped up in an hour or two, but I tried to cover a range of functions. Due to hardware limitations in some cases (machines and emulators without clocks) you will need to time the tests with a stopwatch. But... considering none of this is really important, I didn't think that little mistakes mattered much. Just remember to take this into account. Any comments on this, answers to any of my questions, problems with my testing methods etc... would be appreciated. P.S. If you want to use any of this or the program in the next message in the faq or any other document you are welcome to do so. Here are some benchmark results, using SpeedTest 2.2.2 (by Clayten Hamacher) on my 60 MHz Pentium. AppleWin Apl2Em Applemu Time Ratio Time Ratio Time Ratio Memory Read Test 9.39 12.69 14.58 8.18 9.33 12.77 Memory Write Test 7.21 16.53 17.14 6.95 7.66 15.56 Zero Page Read Test 11.55 10.35 14.73 8.12 11.67 10.24 Zero Page Write Test 9.31 12.84 11.39 10.50 8.92 13.40 Indirect References 9.97 12.00 13.73 8.72 8.80 13.60 Branch Calculation 10.72 11.12 21.96 5.43 11.21 10.64 Register Variables 7.15 11.65 9.02 9.24 8.36 9.96 Instruction Fetch 6.64 18.05 9.47 12.66 8.42 14.23 CPU Instruction Mix 11.35 10.23 18.96 6.12 11.02 10.54 CPU INDEX 12.83 8.43 12.53 Direct Text Writes #1 13.22 9.08 49.73 2.42 49.78 2.41 Direct Text Writes #2 18.19 6.62 50.93 2.36 50.99 2.36 BIOS Text Writes 27.35 4.36 37.90 3.15 35.20 3.39 TEXT INDEX 6.69 2.64 2.72 Direct Low-Res Writes 16.12 7.26 28.74 4.07 28.33 4.13 Direct High-Res Writes 16.30 7.02 161.77 0.71 162.70 0.70 VIDEO INDEX 7.14 2.39 2.42 RWTS Full Disk Read 10.58 9.23 16.73 5.84 112.11 0.87 DISK INDEX 9.23 5.84 0.87 Some notes: The version of AppleWin that I used is the current code, which is somewhat faster than the March Beta. Also, I switched it to 64k Apple ][+ mode before executing a couple of the benchmarks which would be totally unfair otherwise. Most especially, I think that the "BIOS Text Writes" benchmark is not a good benchmark, because what it really compares is the speeds of the BIOS's rather than the speeds of the emulators. The Enhanced Apple //e's BIOS was slower at text writes than the ][+ BIOS. I didn't include SimIIE in the benchmarks because, to be totally honest, I got tired of waiting for it to complete benchmarks. On my machine it is about 3-4 times slower than the other emulators at CPU emulation, and about 100 times slower at video and disk emulation. As you can see, AppleWin is the fastest overall emulator. There were a few CPU benchmarks where Applemu was slightly faster; this is because it emulates a 48k machine and therefore does not have to provide any memory banking support at all. Maybe I should add a 48k emulation mode to AppleWin. Anyway, we now how SpeedTest results for four emulators; Kevin Lund posted results for STM a while ago. Alex, I noticed that you added a section on SpeedTest to the FAQ; perhaps you can add benchmark results as well. For anyone running a UNIX box, I would like to see benchmark results for some of the UNIX emulators. Especially ZaniWOK, because the folks at ARDI have made some very interesting (suspect) claims about its speed. Mike ------------------------------------------------------------------- Hi! I ran The Sinner's speedtest on my emulator to see, how it performs. The machine was a SparcStation 20/612 under Slowlaris and my X-server ran on a Sparc Classic. For-Next |Scrolling|Basic HGR|ASM Loops|ASM HGR | ---------|---------|---------|---------|---------| Apple][ 60.0| |41.3| |56.0| |133 | |27.0| | ---------|---------|---------|---------|---------| My Emu 74.0|0.81|44.0|0.94|78.0|0.72|128 |1.04|34.0|0.79| The ASM HGR test ran without a HGR page being visible, so there was actually no traffic with the X-server. With a visible HGR page, the test ran some minutes, until i pressed the reset button :-) The scrolling test is not very good, because it fills the screen with only one character. My emulator will detect this and optimize greatly. A simple "LIST" would give more comparable results. Peter Koch (koch@dfki.uni-kl.de) --------------------------------------------------------------------