Video BIOS messages not showing up on startup before BIOS POST screen

[grommit2007]

I have noticed that in PCem versions before 13, (up to PCem v12) if a graphics card had a startup message in its ROM, it would appear for a few seconds before the motherboard BIOS POST screen appears. From PCem v13 onwards, the graphics card startup messages in its ROM still appear, but they appear for seemingly less than a milisecond (or not at all depending on the selected motherboard or graphics card).

Is there a possible cause for this?

[JosepMa]

With the released version of v14, AMI winbios and Trident TVGA, I see the trident message during two seconds, so it seems there's something more in your setup that could be causing this misbehaviour. (video speed is set to default)

[ecksemmess]

I can confirm this problem (in stock v14, can't test svn at the moment). Definitely a bummer, would love to see this fixed for the v15 release!

To shed further light on how to reproduce this: only Pentium-class machines are affected. To reproduce, create a setup with any Pentium-class board (I've tested Award 430VX, Intel Premiere/PCI, Intel Advanced/EV, Intel Advanced/ZP), and any video card that displays a POST screen (I've tested Paradise Bahamas64, Number Nine 9FX, Phoenix S3 Trio64). The rest of the settings don't make any difference, the problem will happen regardless. Exactly as the OP says, the video card's POST message will flash for a vanishingly short fraction of a second before moving onto the main system POST.

While it may be tempting to write this off as a minor cosmetic problem, I would think it's definitely worth double checking the cause before finalizing v15, as it does seem like one of those things that could pay dividends in other areas if caught and fixed. The problem never occurs in 13.1, so in theory it should be easy to track down by simply identifying which commit introduced it, I guess?

[SA1988]

Well, revision 1168 caused this bug.

[ecksemmess]

Good work SA1988, thanks for checking that

[ecksemmess]

Was this ever properly fixed? If not, is there any chance of getting it nipped in the bud before the v16 release? I can't quite tell how involved the fix would be, but not having the video board's POST sure is inconvenient sometimes.

[omarsis81]

the video message length is a fix second duration? or is it depending on the processor speed? It appears the faster processor the less time the message is shown

[ecksemmess]

I don't know offhand whether most video boards' POST screens are supposed to display for shorter times on faster CPUs, but in any case that's not relevant here; this is simply a bug. The video board POST is flashing by in a minuscule fraction of a second, far too dramatic of a change from expected behavior to be explained by anything ordinary. It certainly has nothing to do with the clock speed of the emulated CPU - for example, the video POST displays normally when emulating a 100MHz 486DX4, but disappears almost instantaneously when emulating a 60MHz Pentium.

I've gone ahead and done some further testing and can report that this is NOT a recompiler bug, as it also affects WinChip and behaves identically on the recompiled and interpreted versions of the WinChip emulation. So, to recap:

- The problem isn't anything recompiler-specific
- The problem isn't specific to any particular video card
- The problem isn't clock-speed-related to any significant degree
- The problem isn't Intel-specific (WinChip behaves identically)

So, since the problem only affects Pentium class CPUs, it's likely somewhere in the CPU emulation, but it somehow still affects the recompiler and interpreter equally, AND the Intel Pentium and WinChip emulation equally. As SA1988 mentioned, rev 1168 introduced this bug, but having looked at rev 1168 I'm not really sure what to make of that.

[EluanCM]

On my real Zappa with a Pentium 75 and a Cirrus Logic GD5434 card, I can't see the video BIOS message. On PCem it blinks, but on the real hardware it disappears (because the system bios clears the output) while the monitor is syncing.

[ecksemmess]

I've got a real Zappa with a Pentium 100 and it definitely always shows video POSTs normally on every video card I've tried, I think the Cirrus Logic cards are just funny that way. Anyway, this bug is also NOT machine-specific, which I guess I should have included in the list above. Every Pentium-class mobo was showing the expected behavior until rev 1168 and then the problem abruptly started. The pre-rev-1168 behavior matches all of my real hardware, the behavior since then emphatically does not.