🚀 go-pugleaf

In article <davis-0108951934350001@pppsb1.itg.ti.com>,
Tom L. Davis <davis@adrs1.dseg.ti.com> wrote:
>In article <3vlv3n$8o0@nnrp3.primenet.com>, english@primenet.com (Lawson
>English) wrote:
>
>:Chad Irby (cirby@news.gate.net) wrote:
>:: Jeff Medcalf (214) 280-5970 IBM SDO (medcalf@ppc5.svo.dfw.ibm.com) wrote:
>:
>:: : That's pretty funny, considering that Win95 won't *run* on PPC systems.
>:
>:: Strangely enough, Windows 95 *will* run on a PPC Mac with SoftWindows.  
>:: With a reasonably fast Mac, you can get it up to fast 386/slow 486 speeds.
>:
>:ON a 9500/132, the speed of SoftWIndows 2.0 beta running Windows 95 Gold is 
>:said to rival that of an 80486DX2 33 [sic] system.
>:Since SoftWIndows was *optimized* so that the emulator would be bypassed 
>:for critical calls like GDI stuff, and since the beta system of 2.0 probably 
>:isn't optimized for all WIndows 95 calls, this is quite an interesting 
>:datum, IMHO.
>
>If Apple's new emulator (on the 9500) can outperform all 040 Macs, why
>can't Softwindows do the same relative to a 66 MHz 486DX2 (which I assume
>is equivalent to a 33 MHz 040)?

Apple's emulator doesn't worry about emulating hardware. Either the 
68K code in an app is using API calls to get the job done, or the 
hardware being accessed by 68K code (an emulated driver, say) actually 
exists in the same form. Not true the 486 emulator running on the 
PowerPC system with no VGA chip, no 16[45]50 UART chip, etc etc etc to
let the emulated entity manipulate directly.

A great example of this would be a an emulated game, say something a couple
years old like Maelstrom or Hellcats. They write directly to the screen
and run great in emulation - the screen's still there the same way they
expect it to be. Not true for DOOM-PC trying to run in VGA ModeX under 
SoftWindows 2.0 :(

Apple's emulator has the advantage of inheriting only code that is more
or less well-behaved (or to put it another way, the emulators opt to 
provide undefined behavior in the case of code that violates the guidelines).

Self-modifying code without cache-flushing calls. SoftWindows probably
has to deal with that, Apple's emulator(s) don't [choose not to]. That's
probably the biggest issue alone.

Apple's emulator is a pretty pure "CPU only" emulator. SoftWindows has
to deal with CPU and more [although their new technology is supposed
to be radically different from 1.0].

I have edited the followup line.

Rob Barris
Quicksilver Software Inc.
rbarris@quicksilver.com