VOGONS


SBEMU: Sound Blaster emulation on AC97

Topic actions

Reply 2000 of 2017, by RayeR

User metadata
Rank Oldbie
Rank
Oldbie

I use R.Loew's RAMDISK 2.0 package (RAMDSK64.COM of size 2673B).
I had set it to load
DEVICE=C:\WINDOWS\HIMEMEX.SYS /S /V
so using sysenter mode and it works without VDPMI so I didn't try other methods. Now I see there's prothook method involving a VXD. But in my case the RAMDSK64.COM cause immediate restart when loaded from autoexec, so before Windows start and any VXD gets loaded. But if you thing this sysenter method cause crash I can try with porthook...

Last edited by RayeR on 2026-11-17, 07:07. Edited 2 times in total.

Gigabyte GA-P67-DS3-B3, Core i7-2600K @4,5GHz, 8GB DDR3, 128GB SSD, GTX970(GF7900GT), SB Audigy + YMF724F + DreamBlaster combo + LPC2ISA

Reply 2001 of 2017, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
RayeR wrote on 2026-11-17, 03:49:
I use R.Loew's RAMDISK 2.0 package (RAMDSK64.COM of size 2673B). I had set it to load DEVICE=C:\WINDOWS\HIMEMEX.SYS /S /V so us […]
Show full quote

I use R.Loew's RAMDISK 2.0 package (RAMDSK64.COM of size 2673B).
I had set it to load
DEVICE=C:\WINDOWS\HIMEMEX.SYS /S /V
so using sysenter mode and it works without VDPMI so I didn't try other methods. Now I see there's prothook method involving a VXD. But in my case the RAMDSK64.COM cause immediate restart when loaded from autoexec, so before Windows start and any VXD gets loaded. But if you thing this sysenter method cause crash I can try with porthook...

Thanks for the info, I will test with /S /V after current problems solved.
You can try prothook with Windows but I'm afraid it won't work with VDPMI.

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 2002 of 2017, by RayeR

User metadata
Rank Oldbie
Rank
Oldbie

OK, the method with prothook works, no reset, ramdisk is available under windows...

Gigabyte GA-P67-DS3-B3, Core i7-2600K @4,5GHz, 8GB DDR3, 128GB SSD, GTX970(GF7900GT), SB Audigy + YMF724F + DreamBlaster combo + LPC2ISA

Reply 2003 of 2017, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
RayeR wrote on 2026-11-17, 07:08:

OK, the method with prothook works, no reset, ramdisk is available under windows...

Cool, does windows starts with VDPMI?
Does it crash/reset before windows startup? I guess it will do in pure dos 🤣.

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 2004 of 2017, by DoZator

User metadata
Rank Member
Rank
Member
crazii wrote on 2026-11-16, 06:09:
I hadn't test SBEMU with windows but I didn't some changes and tests, now Windows can start with SBEMU loaded. but it DOES NOT m […]
Show full quote

I hadn't test SBEMU with windows but I didn't some changes and tests, now Windows can start with SBEMU loaded. but it DOES NOT mean sbemu can always work after windows exit.
IF windows does NOT have audio driver and after exiting to dos, sbemu is likely to work,
otherwise the audio driver may breaks thing and make SBEMU not working even when fresh-installed after windows exiting.

I've test with win3.x and win95 without sound driver and after exiting SBEMU works, on win98se with sound driver, it does work after Windows exits.

Hi. Thank you for paying attention to this. I’ve also encountered exactly this situation before. What helped was saving the PCI registers of the sound card before launching Windows (with a cold start) and then restoring them to their original values when entering MS-DOS mode. After that, everything started working. I added all of this to DOSSTART.BAT so that the registers would always return to their original values automatically every time MS-DOS was entered. And this fully resolved the problem. The same, by the way, applied to USB - initially, it also didn’t work at all after exit from Windows (the mouse and keyboard didn’t respond) when the USB driver was installed in Windows. However, restoring the PCI registers of the USB controller (in a similar way) also resolves this problem (naturally, all of this was also added to DOSSTART.BAT). It’s good to see that VDPMI now takes all these features into account.

By the way, what is the current status of the main SBEMU branch based on HDPMI32I? Will it no longer be developed, or are you still inclined to support it, at least at the most minimal level? I just wanted to suggest adding a "/U" parameter so that we could finally have the ability to unload SBEMU from memory (when needed) without having to hardware reboot the PC. This feature is very much needed (I still use this version of SBEMU from time to time).

Reply 2005 of 2017, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
DoZator wrote on 2026-11-17, 08:23:

Hi. Thank you for paying attention to this. I’ve also encountered exactly this situation before. What helped was saving the PCI registers of the sound card before launching Windows (with a cold start) and then restoring them to their original values when entering MS-DOS mode. After that, everything started working. I added all of this to DOSSTART.BAT so that the registers would always return to their original values automatically every time MS-DOS was entered. And this fully resolved the problem. The same, by the way, applied to USB - initially, it also didn’t work at all after exit from Windows (the mouse and keyboard didn’t respond) when the USB driver was installed in Windows. However, restoring the PCI registers of the USB controller (in a similar way) also resolves this problem (naturally, all of this was also added to DOSSTART.BAT). It’s good to see that VDPMI now takes all these features into account.

Thank you for the information. I was guessing it's a ACPI power issue but it turns out a PCI configuration register change. I will check to fix that.

DoZator wrote on 2026-11-17, 08:23:

By the way, what is the current status of the main SBEMU branch based on HDPMI32I? Will it no longer be developed, or are you still inclined to support it, at least at the most minimal level? I just wanted to suggest adding a "/U" parameter so that we could finally have the ability to unload SBEMU from memory (when needed) without having to hardware reboot the PC. This feature is very much needed (I still use this version of SBEMU from time to time).

It's currently have new drivers merged, and bug fixed but compatibility game problem won't fixed, because I am not skilled enough to maintain or merge the HDPMI32i ASM source code. Some compatibility problem is fixed in vdpmi, e.g. the monkey island divided by 0 & space quest v sb detection problems, those real mode program compatibility issues cannot be fixed in SBEMU not even in a DPMI server.

Is there any reason that you prefer the HDPMI32i version? I can add /U but it may not be fully unloaded because it hooks 2 IVT entries, one may be skipped, but the other, for installation detection, if hooked by other programs later, still needs some bytes to work.

With newly added VDPMI PM Hook for RM, the VDPMI build of SBEMU now uses zero DOS memory (no conventional, UMB and such), no IVT hooks, fully resident in himem with full function as before. so it can be unloaded more easily and cleanly. And it doesn't feed silent audio data to the PCI sound card anymore. the change is in the last release linked to the github issue, not uploaded here yet. before that even if it's idle in dos commandline, SBEMU is still busy feed silent data to sound card, which is dumb but simple enough to work. this is a known issue for a long time, but I haven't fixed it til now because there's too much to do.

Personally I tend to add /U for the VDPMI build (but not done it yet because I failed to see the needs, as now it is totally idle in the background, and VDPMI can start windows without reset the PC. unless there's not enough RAM installed, e.g. less than 64MB). if you have specific reasons, I can also try to add an unclean implementation to the HDPMI32i build.

EDIT:
It comes to me a cleaner but hacky way to do unloading with HDPMI32i, maybe I can test my idea soon.

Last edited by crazii on 2026-11-17, 11:36. Edited 1 time in total.

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 2006 of 2017, by DoZator

User metadata
Rank Member
Rank
Member
crazii wrote on 2026-11-17, 09:25:

Is there any reason that you prefer the HDPMI32i version? I can add /U but it may not be fully unloaded because it hooks 2 IVT entries, one may be skipped, but the other, for installation detection, if hooked by other programs later, still needs some bytes to work.

Yes, I was just about to write about all of this in more detail. During the discussion of the new universal multi-core emulator Voodoo, it suddenly turned out that, for some reason, VDPMI, unfortunately, currently does not provide the option to enable APIC after exiting Windows, so that multi-core functionality can be used again in MS-DOS mode. Without VDPMI, this works fine. HDPMI32I also provides this option.

Yes, I think it would probably be worth trying to implement at least some way to dump SBEMU at the present moment, as it turns out (provided it’s not too difficult and won’t take up too much of your time).

It would be ideal, of course, if SBEMU in the future combined all the approaches that exist today and was able to work under any conditions, providing opportunities depending on those very conditions. A universal SBEMU )

Reply 2007 of 2017, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
DoZator wrote on 2026-11-17, 11:35:
Yes, I was just about to write about all of this in more detail. During the discussion of the new universal multi-core emulator […]
Show full quote
crazii wrote on 2026-11-17, 09:25:

Is there any reason that you prefer the HDPMI32i version? I can add /U but it may not be fully unloaded because it hooks 2 IVT entries, one may be skipped, but the other, for installation detection, if hooked by other programs later, still needs some bytes to work.

Yes, I was just about to write about all of this in more detail. During the discussion of the new universal multi-core emulator Voodoo, it suddenly turned out that, for some reason, VDPMI, unfortunately, currently does not provide the option to enable APIC after exiting Windows, so that multi-core functionality can be used again in MS-DOS mode. Without VDPMI, this works fine. HDPMI32I also provides this option.

Yes, I think it would probably be worth trying to implement at least some way to dump SBEMU at the present moment, as it turns out (provided it’s not too difficult and won’t take up too much of your time).

It would be ideal, of course, if SBEMU in the future combined all the approaches that exist today and was able to work under any conditions, providing opportunities depending on those very conditions. A universal SBEMU )

That's very interesting, VDPMI doesn't touch APIC at all, so I assume APIC is disabled by default. so how does the voodoo emulator enables APIC and then unable to do that after windows exit? Is my assumption is incorrect that APIC is enabled by BIOS by default, used for legacy PIC emulation?

I will surely investigate more about this, thanks.

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 2008 of 2017, by DoZator

User metadata
Rank Member
Rank
Member
crazii wrote on 2026-11-17, 09:25:

EDIT:
It comes to me a cleaner but hacky way to do unloading with HDPMI32i, maybe I can test my idea soon.

I already like this idea! It definitely won’t make things worse.

Reply 2009 of 2017, by DoZator

User metadata
Rank Member
Rank
Member
crazii wrote on 2026-11-17, 11:42:

That's very interesting, VDPMI doesn't touch APIC at all, so I assume APIC is disabled by default. so how does the voodoo emulator enables APIC and then unable to do that after windows exit? Is my assumption is incorrect that APIC is enabled by BIOS by default, used for legacy PIC emulation?

I will surely investigate more about this, thanks.

Before Windows, everything works fine (including VDPMI). Windows disables APIC, but the emulator can temporarily enable it while the game is running, so that multicore support works in Win9x (but after the game ends, APIC is disabled again). And after switching to MS-DOS mode, you can enable APIC either without anything at all (just bare HIMEM), or if you additionally load HDPMI32I on top of it. And if VDPMI is used, multicore support cannot be enabled after Windows. It’s not yet very clear what this is related to. Furthermore, there are other situations in which HDPMI32I is preferable (at least at this stage). Therefore, I still sometimes use this version in places where it works well.

Reply 2010 of 2017, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
DoZator wrote on 2026-11-17, 12:03:
crazii wrote on 2026-11-17, 11:42:

That's very interesting, VDPMI doesn't touch APIC at all, so I assume APIC is disabled by default. so how does the voodoo emulator enables APIC and then unable to do that after windows exit? Is my assumption is incorrect that APIC is enabled by BIOS by default, used for legacy PIC emulation?

I will surely investigate more about this, thanks.

Before Windows, everything works fine (including VDPMI). Windows disables APIC, but the emulator can temporarily enable it while the game is running, so that multicore support works in Win9x (but after the game ends, APIC is disabled again). And after switching to MS-DOS mode, you can enable APIC either without anything at all (just bare HIMEM), or if you additionally load HDPMI32I on top of it. And if VDPMI is used, multicore support cannot be enabled after Windows. It’s not yet very clear what this is related to. Furthermore, there are other situations in which HDPMI32I is preferable (at least at this stage). Therefore, I still sometimes use this version in places where it works well.

Very interesting. Is the voodoo emulator installed before VDPMI? does it work when it's installed AFTER VDPMI? I know when VDPMI map UMA for UMB/EMS, it ignores the ACPI table (only checks&skips ROM signature), the ACPI table may be inaccessible after memory map, its region will be used for UMB/EMS, so that APIC may not work, because APIC's data pointer is located in ACPI interface. that doesn't affect Windows because windows uses its own page table.
I can check the ACPI signature to avoid map its UMA region. It may work If you can confirm whether the voodoo emulator works before/after loading VDPMI.

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 2011 of 2017, by RayeR

User metadata
Rank Oldbie
Rank
Oldbie
crazii wrote on 2026-11-17, 07:54:

Cool, does windows starts with VDPMI?
Does it crash/reset before windows startup? I guess it will do in pure dos 🤣.

Yes, since vdpmi 0.33 Windows 98 started with VDPMI. Problem was just with loading the himemex+ramdisk for swap file in config+autoexec. Without VDPMI the himemex works with Sysenter method with windows but cause reset when ramdsk64.com loaded with VDPMI. As I delete the /s switch it use default prothook method that is compatible with VDPMI. I also had to add and load that prothook.vxd in system.ini (without it the ramdisk present drive letter but is not accessible under windows).

Strange thing is that when I boot to dos only I got warning message from ramdsk64.com that prot hook must be installed but when booting windows path this message don't appear. I have to check it how ramdsk64.com detect it.
Btw can you recompile ramdsk64.asm in working state? I tried to do it with jwasm, compiled without error but binary didn't work and i can see the binary is cca 300B larger than original R.Loew's binary. There's no batch/makefile so I even don't know what assembler he used but probably some MASM that should be fully replacable with jwasm.

Maybe you can try to check why sysenter method cause reset with vdpmi. To test it's needed just dos environment, load himemex.sys /S then load VDPMI and last load ramdsk64.com in autoexec...

Gigabyte GA-P67-DS3-B3, Core i7-2600K @4,5GHz, 8GB DDR3, 128GB SSD, GTX970(GF7900GT), SB Audigy + YMF724F + DreamBlaster combo + LPC2ISA

Reply 2012 of 2017, by DoZator

User metadata
Rank Member
Rank
Member
crazii wrote on 2026-11-17, 13:43:
DoZator wrote on 2026-11-17, 12:03:
crazii wrote on 2026-11-17, 11:42:

That's very interesting, VDPMI doesn't touch APIC at all, so I assume APIC is disabled by default. so how does the voodoo emulator enables APIC and then unable to do that after windows exit? Is my assumption is incorrect that APIC is enabled by BIOS by default, used for legacy PIC emulation?

I will surely investigate more about this, thanks.

Before Windows, everything works fine (including VDPMI). Windows disables APIC, but the emulator can temporarily enable it while the game is running, so that multicore support works in Win9x (but after the game ends, APIC is disabled again). And after switching to MS-DOS mode, you can enable APIC either without anything at all (just bare HIMEM), or if you additionally load HDPMI32I on top of it. And if VDPMI is used, multicore support cannot be enabled after Windows. It’s not yet very clear what this is related to. Furthermore, there are other situations in which HDPMI32I is preferable (at least at this stage). Therefore, I still sometimes use this version in places where it works well.

Very interesting. Is the voodoo emulator installed before VDPMI? does it work when it's installed AFTER VDPMI? I know when VDPMI map UMA for UMB/EMS, it ignores the ACPI table (only checks&skips ROM signature), the ACPI table may be inaccessible after memory map, its region will be used for UMB/EMS, so that APIC may not work, because APIC's data pointer is located in ACPI interface. that doesn't affect Windows because windows uses its own page table.
I can check the ACPI signature to avoid map its UMA region. It may work If you can confirm whether the voodoo emulator works before/after loading VDPMI.

Yes, the emulator works in both cases. It generally supports VDPMI (with the exceptions described above). When the emulator cannot enable multi‑core support, it continues to run but switches to single‑core mode.

Reply 2013 of 2017, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
RayeR wrote on 2026-11-17, 15:29:

Btw can you recompile ramdsk64.asm in working state? I tried to do it with jwasm, compiled without error but binary didn't work and i can see the binary is cca 300B larger than original R.Loew's binary. There's no batch/makefile so I even don't know what assembler he used but probably some MASM that should be fully replacable with jwasm.

Sure, I'll see what I can do.

RayeR wrote on 2026-11-17, 15:29:

Maybe you can try to check why sysenter method cause reset with vdpmi. To test it's needed just dos environment, load himemex.sys /S then load VDPMI and last load ramdsk64.com in autoexec...

vdpmi need shandle the privilege instruction for it. current it doesn't. I think may be able to fix that.

DoZator wrote on 2026-11-17, 16:09:

Yes, the emulator works in both cases. It generally supports VDPMI (with the exceptions described above). When the emulator cannot enable multi‑core support, it continues to run but switches to single‑core mode.

Thanks for the info, I will check it.

As the github issues now reports fine, I'm putting the last release here:

update for VDPMI:
1.separate loader (VDPMI.EXE and core (VDPMI.LX) file.
Previously they're stubbed together, but I found in CONFIG.SYS dos always loads the whole file no matter what. It ignores the MZ size and extended NE/PE/LE/LX signature. but the loader will load itself (through argv[0]) again, so there are two full file reads.
separating the loader and core file solves that. it also potentially fix that dos cannot load it if the file size is too large in CONFIG.SYS, exceeding ~500K.
2.add buffer for reading VDPMI.CFG. skip reading VDPMI.CFG if VDPMI is already installed.
3.improve memory copy on loading
4.workaround a wcc bug that write extra word to a int pointer.
5.remove unused ~60K stack from core image file. now it does not have stack segment.
6.VDS: fix function 0x8105: scatter/gather lock region returned with incorrect physical address and size. this fixes IO for disk that have PCI IDE DMA enabled.
7.VDMA: fix physical DMA controller passthrough with incorrect IO ports and not even un-masked, which never starts DMA transfer. now floppy disk can be accessed properly.
8.fix keyboard driver bugs. now LED works. may fix lockups for some PCs too. also fixed CTMOUSE freeze a while and fail to detect mouse when random keys pressed during the time window AFTER vdpmi is loadded and BEFORE CTMOUSE is loaded.
9.fix "VIntFlag=0" in VDPMI.CFG
10.bug fxies for vdpmi is loaded from command line, and workaround MEM.EXE crash for it.
11.fix ROTT random freeze on exit (seen on some PCs but may be not for others)
12.bugfix on DOS driver list processing for WinME/DOS8.0
13.add support for Joe & Mac: Caveman Ninja. the game is incompatible with vdpmi. I guess it won't work in windows. have to add special handlings.
also fix divided by 0 error for it (patched/modded edition), now it can run on modern fast CPU. game speed is fine but maybe not accurate, music still plays slow.
14.DPMI memory pool now have 2 instances, one for normal clients, the other for "client drivers" like sbemu.
new entries in [DPMI] section in VDPMI.CFG
ClientDriverMem=11 ;MB
ClientDrivers=SBEMU.EXE
the dedicated driver memory pool keeps SBEMU alive during windows sessions. normal memory pool will be freed for windows.
15.vdpmi core now resident in high address of XMS. along with the client driver memory.
I also tried to move physical memory used for EMS/UMB to high address, but win95 refuse to start.

SBEMU changes:
1. now resident fully in himem, no DOS memory (no conventional/UMB) usage.
2.auto stops sound card playback if no SB/OPL/MPU data. start playing if SB/OPL/MPU plays.
previously SBEMU always plays silent sound samples in the background, even if it is idle in dos command line.
3.fix ADPCM crash, seen in Duke Nukem 2.
4.MMIO base now is at high address of XMS.

Don't forget to copy the new file: VDPMI.LX to your disk.

next TODOs:
merge PC speaker emu to VDPMI build, now it have some problems.
with /DPMI=0 to disable DPMI, it is possible to load HDPMI32i and its SBEMU edition, but currently not stable and have problems.

The attachment vdpmi0.36.zip is no longer available

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 2014 of 2017, by RayeR

User metadata
Rank Oldbie
Rank
Oldbie

I have special interest in points 6,7 if it could fix my broken vdma/vds on disk connected via sata3 that I have to fix with xhdd.sys (but this forces windows to use 16bit HDD mode - int13h). I don't exctly remember what was the problem and how xhdd fix it maybe I can find in old emails with Jack Ellis.

Also 2 memory pools idea sound interesting. Some games crashes with too much XMS/DPMI mem so it would be good to limit memory for them while for whitelisted programs that needs/can use more mem then give it to them...

Gigabyte GA-P67-DS3-B3, Core i7-2600K @4,5GHz, 8GB DDR3, 128GB SSD, GTX970(GF7900GT), SB Audigy + YMF724F + DreamBlaster combo + LPC2ISA

Reply 2015 of 2017, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
RayeR wrote on Yesterday, 13:43:

I have special interest in points 6,7 if it could fix my broken vdma/vds on disk connected via sata3 that I have to fix with xhdd.sys (but this forces windows to use 16bit HDD mode - int13h). I don't exctly remember what was the problem and how xhdd fix it maybe I can find in old emails with Jack Ellis.

I only tested with VMWare Player, MSDOS7 will use VDS for its PCI bus master DMA enabled IDE controller, but it seems working fine without xhdd.sys. the memory mapped UMB are still handed by vdpmi's int13h hook, but I think it's not necessary with VDS present. the int13h hook only exist for buggy BIOS legacy USB emulation that perhaps xhdd.sys doesn't check, and MSDOS7 also failed to detect, it's not a real IDE controller anyways.

Perhaps you can test with vdpmi but be careful with your disk data, when debugging in VMWare player, the disk was wholely ***ked up and cannot boot. 🤣

RayeR wrote on Yesterday, 13:43:

Also 2 memory pools idea sound interesting. Some games crashes with too much XMS/DPMI mem so it would be good to limit memory for them while for whitelisted programs that needs/can use more mem then give it to them...

I haven't thought about that, it's just needed for TSRs when DPMI should terminate its memory for windows.
Without that you can still adjust the DPMI memory amount to be larger for TSRs, so that the left are suitable for legacy programs.
It's just a little bit of convenience that you may not need to adjust the memory size if you preserve big pool for TSRs.

BTW I've added a hack to hook pm interrupt vector for the HDPMI build, everything seems working fine except after unloading SBEMU, the pm int hook are restored, then the foreground instance exists and the IDT is restored to an earlier snapshot (right before the second instance starts) that the pm int vectors are still hooked. 🤣 IDT backup/restore are safe approaches and vdpmi does the same.

So I have to go the normal way - using the rm hook and with other hacky methods. now the unloading with /U works for some simple tests. The rm IRQ handler is disabled again and can be tested more.
@DoZator you can test the latest release: https://github.com/crazii/SBEMU/releases/tag/ … 1.0.0-beta.7rc2

EDIT: You can check the notes in readme.txt for /U, generally you should unload PM TSRs with reversed order of installation otherwise the IDT maybe restored to an invalid hooked state.

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 2016 of 2017, by DoZator

User metadata
Rank Member
Rank
Member
crazii wrote on Today, 13:41:
BTW I've added a hack to hook pm interrupt vector for the HDPMI build, everything seems working fine except after unloading SBEM […]
Show full quote

BTW I've added a hack to hook pm interrupt vector for the HDPMI build, everything seems working fine except after unloading SBEMU, the pm int hook are restored, then the foreground instance exists and the IDT is restored to an earlier snapshot (right before the second instance starts) that the pm int vectors are still hooked. 🤣 IDT backup/restore are safe approaches and vdpmi does the same.

So I have to go the normal way - using the rm hook and with other hacky methods. now the unloading with /U works for some simple tests. The rm IRQ handler is disabled again and can be tested more.
@DoZator you can test the latest release: https://github.com/crazii/SBEMU/releases/tag/ … 1.0.0-beta.7rc2

EDIT: You can check the notes in readme.txt for /U, generally you should unload PM TSRs with reversed order of installation otherwise the IDT maybe restored to an invalid hooked state.

Thank you! Now everything works correctly. I went back to DOS, launched HDPMI32I+SBEMU, and checked how the sound works in SETMAIN (Everything is fine — there is sound). Then I unloaded SBEMU using the "/U" parameter, and it unloaded correctly and completely, along with HDPMI32I (I didn’t use the "-R" parameter for it now, so this is a sign that the behavior is correct).

The only additional action that needed to be performed to get back to Windows after this was to run "setupds /s" so that the PCI registers of the sound card would be cleared and returned to their normal state (This is also quite an expected behavior, as this was also required to be done earlier when I was still actively testing the preliminary alpha version of VDPMI+SBEMU, and it was the same in this regard). If this is not done, Windows will not boot.

The new VDPMI should, in theory, already take this point into account and maintain all registers within the normal range on its own (without the need for external tools). I’ll try to figure this out using version 36.

Reply 2017 of 2017, by Plombo

User metadata
Rank Newbie
Rank
Newbie

Tried Windows 3.11 again with the new VDPMI 0.36.

  • I can now launch Windows while SBEMU is running, and nothing crashes! Hooray!
  • The SB16 driver installer (sbbasic.exe from here) detects "sbemu /T6" as a Sound Blaster 16 (good), but crashes with a VDPMI error during installation. The installer does work with SBEMU+QEMM+HDPMI, so this is a VDPMI issue.
  • Going back to VDPMI and running Windows after this, it's unable to load the SB16 driver, so I still don't get any sound. Windows says to make sure the file exists, but I'm guessing it's actually the driver failing somehow?
  • The SBEMU version included with VDPMI 0.36 doesn't include my recent YMF7x4 noise fix, so it makes a horrible cacophony while SBEMU is running 🙁