https://bugs.winehq.org/show_bug.cgi?id=50093
Bug ID: 50093 Summary: Lara Croft and the guardian of light have very low performance without neckloop even at lowest resolution Product: Wine Version: unspecified Hardware: x86-64 OS: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: -unknown Assignee: wine-bugs@winehq.org Reporter: andy86@fastwebnet.it Distribution: ---
Created attachment 68566 --> https://bugs.winehq.org/attachment.cgi?id=68566 plain wine console output at 640x480 with lowest settings and +fps,+tid,+timestamp
Lara Croft and the guardian of light runs well at 60 fps (and at 100 and more fps if vsync is disabled) on the menu, but during the game it has a performance drop of about 14 fps even at the lowest resolution and with the settings as low as possible.
CPU usage is around 20-35% for the game for a maximium total of approx 60% consider steam and wineserver, and 5-10% max GPU load measured at 640x480 resolution and lowest possible graphics settings and approx 300mb of RAM so it's do not appear as an hardware bottleneck.
I tried some old versions (about 3.x) of wine to exclude other regressions (like bug 46942) and the performance on all tested versions are identical.
I also tried wine-staging with pba patches and all mentioned wine version with unsupported dxvk and dgvoodo2 and performance are identical also on it.
So it seems that the problem is not in d3d* and something others goes wrong in Wine, but I am not able to define exactly what.
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |performance CC| |andy86@fastwebnet.it
https://bugs.winehq.org/show_bug.cgi?id=50093
--- Comment #1 from andy andy86@fastwebnet.it --- Created attachment 68580 --> https://bugs.winehq.org/attachment.cgi?id=68580 last 100000 lines of wine output with relay, wined3d, timestamp and tid killing wineserver during gameplay/issue
Trying several WINEDEBUG options I found these wined3d error:
15982.773:01ec:warn:d3d_perf:texture_resource_sub_resource_map Mapping a dynamic texture without WINED3D_MAP_DISCARD. 15982.773:01ec:warn:d3d:wined3d_context_gl_bind_shader_resources No resource view bound at index 0, X.
With the last of two repeated for 7 times with various X values.
The reference to "d3d_perf", also looking the comments for spawning lines in the code, would seem quite clearly the cause of the performance degradation.
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- Distribution|--- |ArchLinux Summary|Lara Croft and the guardian |Lara Croft and the guardian |of light have very low |of light have very low |performance without |performance without |neckloop even at lowest |neckloop even at lowest |resolution |resolution due to Mapping a | |dynamic texture without | |WINED3D_MAP_DISCARD Component|-unknown |directx-d3d
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- Version|unspecified |5.21
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |regression
--- Comment #2 from andy andy86@fastwebnet.it --- Hi, reading the other tests on appdb this bug seems to be existing at least from wine-1.6 but game is report as run smootly and flawlessly about wine 1.3.x, this might suggest a regression, but the involved versions of wine are too old to be run or compiled on updated systems... is there another way I don't know to detect such early-regression?
Also this bug seems reporting only by nvidia users, but running game in nouveau drivers bug still existing, so can it be directly related to nvidia cards regardless of the driver used?
I do not have a non-nvidia configuration suitable to test the game to determine it.
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- Summary|Lara Croft and the guardian |Lara Croft and the guardian |of light have very low |of light have very low |performance without |performance without |neckloop even at lowest |neckloop even at lowest |resolution due to Mapping a |resolution due to Mapping a |dynamic texture without |dynamic texture without |WINED3D_MAP_DISCARD |WINED3D_MAP_DISCARD | |(Suspected regression)
https://bugs.winehq.org/show_bug.cgi?id=50093
--- Comment #3 from Austin English austinenglish@gmail.com --- (In reply to andy from comment #2)
Hi, reading the other tests on appdb this bug seems to be existing at least from wine-1.6 but game is report as run smootly and flawlessly about wine 1.3.x, this might suggest a regression, but the involved versions of wine are too old to be run or compiled on updated systems... is there another way I don't know to detect such early-regression?
No _good_options...
It's time consuming, but if you're adventurous, see https://wiki.winehq.org/Reverse_Regression_Testing
Basically, you need to find what patches made it so that wine does build on modern systems, then you need to cherry pick those commits when building the older versions. The further back you go, however, the more cherry picking that may be required.
Alternatively, again, if you're adventurous and have lots of time, you could install an older distro (i.e., one released around the time of wine-1.6) and build the older versions on that (assuming you have spare disk space/etc.).
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |austinenglish@gmail.com
--- Comment #4 from andy andy86@fastwebnet.it --- Thank you, this is about what I was afraid of.
I had also already tried to build wine inside a very old live (2013) but without much success.
So before "being even more adventurous" I'd like to know what are the chances that a developer might be able (sooner or later) to fix this bug even if I don't find the bad commmit?
And what are the chances that a developer could (sooner or later) fix this bug if I find the bad commit?
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- Version|5.21 |6.0-rc1
--- Comment #5 from andy andy86@fastwebnet.it --- Still present in 6.0-rc1
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- Version|6.0-rc1 |6.0-rc2 Summary|Lara Croft and the guardian |Lara Croft and the guardian |of light have very low |of light have very low |performance without |performance without |neckloop even at lowest |neckloop even at lowest |resolution due to Mapping a |resolution (Suspected |dynamic texture without |regression) and crashes |WINED3D_MAP_DISCARD |using new builtin vulkan |(Suspected regression) |renderer
https://bugs.winehq.org/show_bug.cgi?id=50093
--- Comment #6 from andy andy86@fastwebnet.it --- Created attachment 68918 --> https://bugs.winehq.org/attachment.cgi?id=68918 Crash backtrace with builtin vulkan renderer
Still present in wine 6.0-rc2.
Game crashes using builtin vulkan renderer.
Attacched crash backtrace is with wine 5.20 because with 6.0rc2 crash is handled by application and winedbg do not output anything.
https://bugs.winehq.org/show_bug.cgi?id=50093
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Version|6.0-rc2 |5.21
--- Comment #7 from Austin English austinenglish@gmail.com --- Please leave the version at the earliest known affected version.
https://bugs.winehq.org/show_bug.cgi?id=50093
Zebediah Figura z.figura12@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |z.figura12@gmail.com
--- Comment #8 from Zebediah Figura z.figura12@gmail.com --- (In reply to andy from comment #4)
So before "being even more adventurous" I'd like to know what are the chances that a developer might be able (sooner or later) to fix this bug even if I don't find the bad commmit?
And what are the chances that a developer could (sooner or later) fix this bug if I find the bad commit?
I can't give particularly specific answers, as it's always hard to guess how much time it'll take to solve a bug, but generally it's much easier to fix a regression from a known patch than to debug one from scratch, especially when the fundamental problem is something like poor performance. Even if nothing else, it generally allows the offending patch to be reverted.
Also, please file a separate bug report for the problem with the Vulkan renderer, rather than polluting this one.
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- See Also| |https://bugs.winehq.org/sho | |w_bug.cgi?id=50332
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- Summary|Lara Croft and the guardian |Lara Croft and the guardian |of light have very low |of light have very low |performance without |performance without |neckloop even at lowest |neckloop even at lowest |resolution (Suspected |resolution (Suspected |regression) and crashes |regression) |using new builtin vulkan | |renderer |
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- Attachment #68918|0 |1 is obsolete| |
--- Comment #9 from andy andy86@fastwebnet.it --- Comment on attachment 68918 --> https://bugs.winehq.org/attachment.cgi?id=68918 Crash backtrace with builtin vulkan renderer
(In reply to Zebediah Figura from comment #8)
I can't give particularly specific answers, as it's always hard to guess how much time it'll take to solve a bug, but generally it's much easier to fix a regression from a known patch than to debug one from scratch, especially when the fundamental problem is something like poor performance. Even if nothing else, it generally allows the offending patch to be reverted.
Thank you. Thank you. All in all I'm starting to think that finding a regression here might really take too long to be worth it.
Also, please file a separate bug report for the problem with the Vulkan renderer, rather than polluting this one.
Here: https://bugs.winehq.org/show_bug.cgi?id=50332
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- See Also| |https://bugs.winehq.org/sho | |w_bug.cgi?id=50311
--- Comment #10 from andy andy86@fastwebnet.it --- Seems to be still present in wine-6.0-rc4-6-gff09f14867e.
Consider that I can no longer test properly due to bug 50311, I can only run the game in a specific dirty prefix with dxvk enabled (in a clean prefix bug 50311 occurs even with dxvk enabled).
https://bugs.winehq.org/show_bug.cgi?id=50093
--- Comment #11 from andy andy86@fastwebnet.it --- Still present in wine-6.1-233-gbb009426719
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- See Also|https://bugs.winehq.org/sho | |w_bug.cgi?id=50311 |
--- Comment #12 from andy andy86@fastwebnet.it --- Still present in wine-6.2-271-g4981785f0fb
https://bugs.winehq.org/show_bug.cgi?id=50093
--- Comment #13 from andy andy86@fastwebnet.it --- Still present in wine-6.6
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords|regression | Summary|Lara Croft and the guardian |Lara Croft and the guardian |of light have very low |of light have very low |performance without |performance due to Steam |neckloop even at lowest |DRM/CEG |resolution (Suspected | |regression) | Component|directx-d3d |-unknown See Also| |https://bugs.winehq.org/sho | |w_bug.cgi?id=24505 CC| |hverbeet@gmail.com
--- Comment #14 from andy andy86@fastwebnet.it --- Still present in wine 6.7.xyz, but...
After reading bug 24505 I also try a "patched" executable and I got the exact same result: fps increased up to 59-60 (with vsync active and medium-high settings).
So appdb tests performance differences are not between wine's versions but effectively beetwen Steam and Retail version...
https://bugs.winehq.org/show_bug.cgi?id=50093
andy andy86@fastwebnet.it changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |oguilherme@protonmail.com