Diverse: Solace v3.2 :: AEP Emulation Page - Emulation News

Release 3.2: Sep 26, 2005

New Features

Previously, Solace´s speed regulation routine would kill time until the next time slice by polling the message queue. This caused Solace to appear to use 100% of the CPU cycles (which it was), but in fact if any other application needed CPU attention, the constant polling of the message queue caused Solace to yield. Nevertheless, it did cause more than one person to think Solace was a CPU hog, so I´ve changed the speed regulation logic to use the Sleep() function, even though it has very poor granularity.
Related to this, I changed the code that updates the status bar with emulation performance statistics to update about once per second whether or not the emulator was in unregulated speed mode. It is easier to read now, and the stats shown are the average over that previous second.
Internal change (same flush, different plumbing): Code now dynamically attaches handlers to I/O port addresses instead of the old static scheme.

Bug Fixes

The emulator rate control was broken for AMD-based systems, at least for the Athlon system that I recently upgraded to. The high-resolution performance counter runs 1000x faster than the same counter on my Celeron, which resulted in some overflows using 32 bit math. I rearranged the calculations to prevent that overflow.
I dropped a line of code somehow, resulting in overlays not working properly when more than one was enabled simultaneously.
The debugger SAVE and SAVEH commands were both saving in Intel HEX format.
There was a bug where audio, once disabled, couldn´t be re-enabled after starting up again. Now you can enable/disable it at will without problems.
If the user was viewing the tail part of the log in the debugger window and then resized the window to make it larger, the region after the end of the log would appear as garbage; now it gets cleared properly.
When a program does an IN from an undriven I/O address, the emulator used to return 0x00. It now returns 0xFF because the real hardware has pullups on the data bus so undriven addresses return 0xFF. This was caught because the program STARW.ECB was counting on this to detect if there was a joystick at a given I/O port.

Known Bugs

If audio is enabled and the CPU is in unregulated speed mode, and disk accesses are made sporadically (say from doing CP/M DIR commands), the peak CPU speed achieved will settle down to different values. For instance, on my machine, the emulated speed will be 117 MHz, but sometimes after performing a DIR, the speed will become 77 MHz, and a subsequent DIR may nudge it back to 117 or not. This will be investigated for the next release.
The code relating to the virtual disk access is in a horrible state. A number of times I´ve started putting in the plumbing to support both virtual northstar (.svn) and virtual helios (.svh) disk files along with the emulation of the helios disk controller. Distractions have kept me from that task, and rather than delay this bug fix release, the code is going out as-is. Don´t look at it or you may go blind.