RaVeN-05 wrote on 2025-12-02, 18:55:

yes old build is 26 fps max at starting position of Scorched planet vs 33 fps max at last build.
DMA still faster

Thanks Raven! It's better but not great.

RaVeN-05 wrote on 2025-12-02, 18:55:

yes old build is 26 fps max at starting position of Scorched planet vs 33 fps max at last build.
DMA still faster

Try adding this to the joystick section. On dual core machines this needs to be added. I don't know what the situation will be on your machine.
[joystick]
multithreaded=false

[EDIT] Nope does not work[/EDIT]

For VRITB170.CGL Chat GPT suggested this, function id and name pointer table:

1cglInitScreen_        ordinal  4  entry_offset 0x01EA  name_offset 0x01ED
2cglQueryScreen_       ordinal  5  entry_offset 0x01FB  name_offset 0x01FE
3cglClearScreen_       ordinal  6  entry_offset 0x020C  name_offset 0x020F
4cglCloseScreen_       ordinal  7  entry_offset 0x021D  name_offset 0x0220
5cglSetWindowClip_     ordinal  8  entry_offset 0x022E  name_offset 0x0231
6cglInitPan_           ordinal  9  entry_offset 0x0245  name_offset 0x0248
7cglPanPos_            ordinal 10  entry_offset 0x0253  name_offset 0x0256
8cglSetLineStyle_      ordinal 11  entry_offset 0x0260  name_offset 0x0263
9cglSetFillPattern_    ordinal 12  entry_offset 0x0272  name_offset 0x0275
10cglSetStipplePattern_ ordinal 13  entry_offset 0x0287  name_offset 0x028A
11cglSetDepthMode_      ordinal 14  entry_offset 0x029C  name_offset 0x029F
12cglSetStencilMode_    ordinal 15  entry_offset 0x02AE  name_offset 0x02B1
13cglGetDepthResult_    ordinal 16  entry_offset 0x02C1  name_offset 0x02C4
14cglWaitVerticalBlank_ ordinal 17  entry_offset 0x02D5  name_offset 0x02D8
15cglPlot_              ordinal 18  entry_offset 0x02EC  name_offset 0x02EF
16cglGetPixel_          ordinal 19  entry_offset 0x02FA  name_offset 0x02FD
17cglDrawLine_          ordinal 20  entry_offset 0x030B  name_offset 0x030E
18cglDrawRect_          ordinal 21  entry_offset 0x031B  name_offset 0x031E
19cglFillRect_          ordinal 22  entry_offset 0x0329  name_offset 0x032C
20
21cglClearBuffer_       ordinal 23  entry_offset 0x0362  name_offset 0x0365
22cglSetConstant_       ordinal 24  entry_offset 0x0374  name_offset 0x0377
23cglSwapBuffer_        ordinal 25  entry_offset 0x0386  name_offset 0x0389
24cglUpLoad_            ordinal 26  entry_offset 0x0397  name_offset 0x039A
25cglDownLoad_          ordinal 27  entry_offset 0x03A4  name_offset 0x03A7
26cglSetPalette_        ordinal 28  entry_offset 0x03B3  name_offset 0x03B6
27cglGetPalette_        ordinal 29  entry_offset 0x03C4  name_offset 0x03C7
28cglSetPickRegion_     ordinal 30  entry_offset 0x03D5  name_offset 0x03D8
29cglGetPickResult_     ordinal 31  entry_offset 0x03E9  name_offset 0x03EC
30cglSetExtent_         ordinal 32  entry_offset 0x03FD  name_offset 0x0400
31cglGetExtentResult_   ordinal 33  entry_offset 0x040D  name_offset 0x0410
32cglGetConfig_         ordinal 34  entry_offset 0x0423  name_offset 0x0426
33cglSetCursor_         ordinal 35  entry_offset 0x0433  name_offset 0x0436
34cglSetCursorPos_      ordinal 36  entry_offset 0x0443  name_offset 0x0446
35cglSetupInterrupt_    ordinal 37  entry_offset 0x0456  name_offset 0x0459
36cglReleaseInterrupt_  ordinal 38  entry_offset 0x046B  name_offset 0x046E
37cglBoardPresent_      ordinal 39  entry_offset 0x0482  name_offset 0x0485
38cglQueryFeature_      ordinal 40  entry_offset 0x0495  name_offset 0x0498
39cglEventRegister_     ordinal 41  entry_offset 0x04A8  name_offset 0x04AB
40cglEventDeregister_   ordinal 42  entry_offset 0x04BC  name_offset 0x04BF
41cglBoardConfigure_    ordinal 43  entry_offset 0x04D2  name_offset 0x04D5

Here are the exports and their corresponding file offsets (hex, from the start of the .CGL file):

1cglVersionControl – offset 0x02350
2cglSendStream – offset 0x02434
3cglGetPixel – offset 0x03444
4cglPutPixel – offset 0x03390
5cglInitScreen – offset 0x01878
6cglQueryScreen – offset 0x01674
7cglClearScreen – offset 0x038C8
8cglCloseScreen – offset 0x01F34
9cglSetWindowClip – offset 0x03960
10cglInitPan – offset 0x02040
11cglPanPos – offset 0x021CC
12cglSetLineStyle – offset 0x0314C
13cglSetFillPattern – offset 0x0315C
14cglSetStipplePattern – offset 0x032B8
15cglSetDepthMode – offset 0x030CC
16cglDrawHorizLine – offset 0x0275C
17cglDrawVertLine – offset 0x02884
18cglDrawLine – offset 0x02558
19cglDrawScanLine – offset 0x029A0
20cglDrawRectangle – offset 0x02638
21cglDrawTriangle – offset 0x02D28
22cglDrawTriangleList – offset 0x02E44
23cglDrawPolygon – offset 0x02B38
24cglDrawPolygonList – offset 0x02C3C
25cglSetPalette – offset 0x03B50
26cglSetPaletteEntry – offset 0x03A44
27cglInterpPalette – offset 0x03C14
28cglSetViperPageMode – offset 0x022C4
29cglSwapViperPages – offset 0x03D00
30cglAllocateTmiBuffer – offset 0x003D8
31cglFreeTmiBuffer – offset 0x004B8
32cglDrawBitmap – offset 0x00564
33cglSetScreenTransform – offset 0x00A24
34cglSetScreenTransformFix – offset 0x009D8
35cglSetScreenTransform32 – offset 0x00A6C
36cglSetObjectTransform – offset 0x00B5C
37cglSetObjectTransformFix – offset 0x00B18
38cglSetObjectTransform32 – offset 0x00BA4
39cglEnableRescale – offset 0x00C00
40cglSetRescale – offset 0x00C30
41cglGetRescale – offset 0x00DF0
42cglLoadFrame – offset 0x010D4
43cglUnloadFrame – offset 0x011F0
44cglBoardConfigure – offset 0x000EC

I'm going to try patching VRITB170.CGL function by function to see if I can get it to go.

A bit of progress.

VRITB170.CGL offset 0x00010551 cglQueryScreen.

Need to decode the other functions and figure out where the function entry points are set.

Early Attempts to test Formula 1:
The game can be installed trough otvdm , thanks.
And it seems could not detect Randition, and doesnot installs f1red.exe file.
So i extracted it by UniExtract Tool a archiver which supports way nearly all archive types.
placed f1red.exe into newly installed formula1 game and ready to go.

in Game:

Performance variety goes from 19 fps, average 35 fps, max 60 fps(vsync).
The game seems rendered perfectly.

Only suspicious 2 things:
Wheels at far distance on foggy weather , not fogged , stays dark .
Cars very rarely 'jumps' off of their positions a little bit up.

RaVeN-05 wrote on 2025-12-06, 17:04:

Early Attempts to test Formula 1: The game can be installed trough otvdm , thanks. And it seems could not detect Randition, and […]
Show full quote

Early Attempts to test Formula 1:
The game can be installed trough otvdm , thanks.
And it seems could not detect Randition, and doesnot installs f1red.exe file.
So i extracted it by UniExtract Tool a archiver which supports way nearly all archive types.
placed f1red.exe into newly installed formula1 game and ready to go.

in Game:

Performance variety goes from 19 fps, average 35 fps, max 60 fps(vsync).
The game seems rendered perfectly.

Only suspicious 2 things:
Wheels at far distance on foggy weather , not fogged , stays dark .
Cars very rarely 'jumps' off of their positions a little bit up.

Ok, I'll test this on the v2200 tomorrow. The v1000 still has issues rendering file explorer, randomly locks up, with rredline drivers. I saw a post here where someone said v1000 struggles to render the desktop with the full acceleration slider set to max on an Athlon xp, but works fine on older, slower cpus.

RaVeN-05 wrote on 2025-12-06, 17:04:

Early Attempts to test Formula 1: The game can be installed trough otvdm , thanks. And it seems could not detect Randition, and […]
Show full quote

Early Attempts to test Formula 1:
The game can be installed trough otvdm , thanks.
And it seems could not detect Randition, and doesnot installs f1red.exe file.
So i extracted it by UniExtract Tool a archiver which supports way nearly all archive types.
placed f1red.exe into newly installed formula1 game and ready to go.

in Game:

Performance variety goes from 19 fps, average 35 fps, max 60 fps(vsync).
The game seems rendered perfectly.

Only suspicious 2 things:
Wheels at far distance on foggy weather , not fogged , stays dark .
Cars very rarely 'jumps' off of their positions a little bit up.

There's also a problem with the water spray texture from the rear of cars under wet weather not being sized correctly sometimes. I'm going to try to fix that today.

The cars jumping *might* just be LOD for the model changing. On real hardware it changes, but the framerate is low so it's less dramatic. I'll also see what's up with the fogging on the wheels.

RaVeN-05 wrote on 2025-12-06, 17:04:

Early Attempts to test Formula 1:
The game can be installed trough otvdm , thanks.
And it seems could not detect Randition, and doesnot installs f1red.exe file.

You have to tick the checkbox manually for all the versions you need. I reinstalled this earlier today, on my W98 machine, because the F1 wrapper, (RReady.V2k), wouldn't work with the reference drivers. I had to revert to Diamond drivers to get this to work. I should probably update the thread F1 V2k.RReady thread.

If you've got time to kill, I've got a couple of clips for comparison, they're not the same, but might help with identifying issues:
RReady: https://youtu.be/lMIdBnGKQBs
V2100: https://youtu.be/HFhbocqwHoc

Does anyone know what this means (from here:https://www.felixcloutier.com/x86/intn:into:int3:int1):

1Each of the INT n, INTO, and INT3 instructions generates a general-protection exception (#GP) if the CPL is greater than the DPL value in the selected gate descriptor in the IDT. In contrast, the INT1 instruction can deliver a #DB

This seems to be preventing me from patching bits of code out of ICR2.

It also seems to block port IO, this also throws an exception without every calling my code. There doesn't seem to be any bits of empty space to add code.

compared two videos:
Yes everything rendered correctly.

The game rarely CTD.
Do not use Alt+tabbing. Hangs app.

Rarely tire trails on asphalt gets corrupted texture.
Rarely shadow under cars gets corrupted texture.

The performance is variable fps.

Sometimes tire trails seen trough road.
Sometimes water splash effects has seen trough other polygons. Those two probably not fixable due to nature of how Rredline API works.

RaVeN-05 wrote on 2025-12-10, 18:30:

The game rarely CTD.
Do not use Alt+tabbing. Hangs app.

Thanks Raven for taking a look.

Having task manager running does the same thing, such as when starting a race.

On newer faster CPUs, F1 runs at a solid 60 fps. I think there was a user here who has a 14the gen Intel CPU who gets a solid 60 fps everywhere. It's mostly CPU bound.

There is a problem with the water spray texture having the wrong size. It's intermittent and I have trouble reproducing it. I also can't tweak the performance of this game. Profiler's crash because of the mixed calling convention.

Still chugging along with widescreen. Progress is a bit slow. The player car can be moved to *almost* match the camera. This mirroring of position will have to be matched exactly:

https://youtu.be/_wn1jfKtNhI

Die by the Sword:
Seems everything perfect.
One thing is suspect - game prints messages on top of screen , when text message visible game drops their performance twice.

on your videos this is not happens, it have solid fps, could be drops by VL_LockBuffer and VL_UnLockBuffer?

Also, i have access to my hardware.

RaVeN-05 wrote on Today, 12:11:

Die by the Sword:
Seems everything perfect.
One thing is suspect - game prints messages on top of screen , when text message visible game drops their performance twice.

Thanks for testing this!

What's the frame rate like? The problem with the text is that it repeatedly calls V_LockBuffer/V_UnlockBuffer to render parts of each character (maybe entire characters at a time). This forces the CPU to stall until the GPU finishes what's it's doing and then the GPU stalls until the CPU calls V_UnlockBuffer. It's used to directly read from or write to GPU RAM. Unlike with DOS where games directly write to video RAM (Descent 2 Speedy3D has a problem with this at the moment) , with RRedline you need to lock the buffer (and stall the GPU and then stall both). Usually this is used with the front buffer and not a texture surface. To make matters worse, DBTS calls this repeatedly per frame. I looked at making this synchronous, but that runs into all sorts of issues with synchronisation. I'll see if I can speed it up a bit tomorrow. Oddly enough this runs without much of an issue on real hardware.

From the RRedline SDK docs (page 52 of the Word doc):

1Use of this call can severely hurt performance.  Performance penalties of 50% have been observed due to inappropriate use.  Never use this call to do blits or drawing with the CPU in a scene that the Vérité also draws to.  Either draw everything with the CPU, or draw everything with the Vérité, but absolutely do not attempt to draw with both at the same time.
2The problem is that when V_LockBuffer is called, the CPU stalls until the Vérité has finished processing all issued commands.  The Vérité stalls while the CPU performs operations on the locked surface.  Typically, the Vérité continues to stall while the CPU transforms and lights the next batch of polygons.  The Vérité starts working again when the polygon commands are issued. Regardless of the optimization of the assembly code for the CPU, the resulting overall throughput of the graphics system is severely reduced.

ah this is very frequently used in Quake engine games.
Well game capped at 18 fps and when text prints it halfes fps and then 9 fps.

RaVeN-05 wrote on Today, 12:45:

ah this is very frequently used in Quake engine games.
Well game capped at 18 fps and when text prints it halfes fps and then 9 fps.

You can also turn on frame interpolation for this game. On my old sandybridge i7 with a gtx660 it runs at 10 fps all the time.

On a Pentium Gxxx (Haswell) with hd4000 it runs at 4 fps.

On my dev machine, old Ryzen it runs at 18 fps constantly.

RaVeN-05 wrote on Today, 12:45:

ah this is very frequently used in Quake engine games.
Well game capped at 18 fps and when text prints it halfes fps and then 9 fps.

VQuake 2 and vHexen 2 call this infrequently. Vq2 uses it for the screenshot function. Vh2's screenshot doesn't work with rready. I think it uses some other win98 way of capturing the screen buffer, which I can't trap. Vq1 has no equivalent of this but it might be querying registers to simulate it. That might be why it's slow.

The v_lockbuffer call doesn't come back until rendering has finished. It returns a pointer in system memory to a copy of the GPU space buffer. Verité has stopped rendering at this point. The app can read or modify what's in the buffer and then call v_unlockbuffer. Verité then updates the buffer in GPU memory with the system buffer and then v_unlockbuffer comes back.

Monster Truck Madness 2 uses it for HUD and hud edge antialiasing (rready cheats here because the game expects a 640x480 16 bit buffer) but calls it only once per frame, so doesn't slowdown. Yes, the game does rendered frame and HUD edge intersection antialiasing using the cpu.