#wine-staging

any idea how to fix this? it would remove 3 bugs at once https://bugs.winehq.org/show_bug.cgi?id=38266

04Bug 38266: normal, P2, ---, wine-bugs, UNCONFIRMED , Precision of math functions needs to be improved

there is an explanation of the issue

jarkko: is probably easy to fix, either requires changes to the FPU flags (easy to test), or alternatively an assembler implementation for those math functions - would you mind opening a wine-staging bug for this?

slackner: i will

fpu and assembler? ugh

DarkPlayer: its not that hard, i even implemented fpu stuff into my proof of concept compiler back then ^^

i don't even know which registers you need to preserve in cdecl or stdcall when using the fpu

DarkPlayer: not sure to which api function exp(...) maps, but some of the wine functions have a wrapper - the arguments are passed directly in the FPU, and wine then stores them in double variables

which could also explain the loss of precision

fpu = 80bit, double=64 bit

slackner: gcc supports special float datatypes to store all 80bit ;-)

DarkPlayer: well, then its probably even easier to fix ^^

(if it also has math functions for such types)

__float80 and __float128

04Bug 186: normal, ---, ---, bugs, UNCONFIRMED , Precision of math functions needs to be improved

jarkko: thx, the bot will also announce it shortly i assume ;)

ah "long double" is also 80 bit

it's usually faster than this

can this impact real wolrd applications?

13 digits is a lot

slackner: from glibc:

Function: double exp (double x)

Function: float expf (float x)

Function: long double expl (long double x)

jarkko: it could indeed affect real world apps, but only those which are programmed very badly ;)

expl

10New Wine Staging bug 186 filed by jarkko_korpi@hotmail.com.

04Bug http://bugs.wine-staging.com/show_bug.cgi?id=186 normal, ---, ---, bugs, UNCONFIRMED , Precision of math functions needs to be improved

10New Wine Staging bug 187 filed by michael@fds-team.de.

04Bug http://bugs.wine-staging.com/show_bug.cgi?id=187 normal, ---, ---, bugs, CONFIRMED , kernel32: Return correct device type for cd devices without medium

DarkPlayer: are you sure you dont need DRIVE_DVD too?

case DEVICE_CDROM: type = DRIVE_CDROM; break;

case DEVICE_DVD: type = DRIVE_CDROM; break;

this code is part of the mount manager

really simple code addition anyway

10New Wine Staging bug 188 filed by michael@fds-team.de.

04Bug http://bugs.wine-staging.com/show_bug.cgi?id=188 normal, ---, ---, bugs, CONFIRMED , mountmgr.sys: Write usable device paths into HKLM\SYSTEM\MountedDevices

[06staging/master][0541c3e67](00http://is.gd/2So65b) 02Sebastian Lackner: Added patch to return correct device type for cd devices without medium.

[06staging/master][05006db23](00http://is.gd/VVznqQ) 02Sebastian Lackner: Added patch to fix device paths in HKLM\SYSTEM\MountedDevices.

trying to make a hack that lets 3dmark vantage see dx10 gpu....meanwhile i have had few crashes and the crashing point seems to change

10sebastian@fds-team.de changed the Status on bug 188 from CONFIRMED --- to RESOLVED FIXED.

04Bug http://bugs.wine-staging.com/show_bug.cgi?id=188 normal, ---, ---, bugs, RESOLVED FIXED, mountmgr.sys: Write usable device paths into HKLM\SYSTEM\MountedDevices

10sebastian@fds-team.de changed the Status on bug 187 from CONFIRMED --- to RESOLVED FIXED.

04Bug http://bugs.wine-staging.com/show_bug.cgi?id=187 normal, ---, ---, bugs, RESOLVED FIXED, kernel32: Return correct device type for cd devices without medium

DarkPlayer: you wanna keep me busy, right? ^^

:-)

slackner: sure. wanna port the bugzilla plugin to py3? XD

10New Wine Staging bug 189 filed by michael@fds-team.de.

04Bug http://bugs.wine-staging.com/show_bug.cgi?id=189 normal, ---, ---, bugs, CONFIRMED , winecfg: Show unmounted devices and allow changing the device value.

btw: if anyone answered me that question from a few hours ago

i accidently rm -rf ~/ and thus had to leave very quickly :D

Keridos: oops, lol, i hope you didn't remove anything important ^^

slackner: not really, just a few scripts

have an almost identical setup at my second pc

just copied the home folder from there and patched 2 or three files

works like a charm

just have to rewrite all my wine starting scripts

for soime games

some

ah ^^

to your question:

Keridos: i am not aware of any easy method to do performance measurements. for most of the stuff we wrote separate test applications

Keridos: however, DarkPlayer was working on some patches to add a lot of measurement stuff

not sure if they are finished yet though

yeah some sort of profiling might work too?

Keridos: sure, you can also use regular linux tools

I would like to help but I am not experiences much regarding that

done some java profiling for minecraft only

but what really would help you, if i could somehow name the functions eating a lot of cpu time, correct?

DarkPlayer: care to duplicate the code?

DarkPlayer: https://bugs.winehq.org/show_bug.cgi?id=23382

04Bug 23382: normal, P2, ---, wine-bugs, UNCONFIRMED , NVIDIA Demo: Cascades needs unimplemented function d3dx10_43.dll.D3DX10CreateFontA

Comment 2

jarkko: dx10 ... not sure if its worth fixing bugs there before the wined3d people have finished upstreaming at least a minimal set of base functions ^^

i looked and changed some dx10 return codes so i could forward with 3dmark and i was suprized that there is quite a lot code

i am not the right person to say but i would assume dx10 isnt that far away

jarkko: DX10 is mostly about shaders and all that stuff and correctly Wine does not support those shaders

what's so hard about the shader stuff?

is there anyone who could even implement them?

that wined3d is undocumented and this stuff is really complex

first you need to parse the opcodes of the directx shaders add them into some complex structures and then they are converted into an GLSL shader again

depending on what kind of hardware you have, there are even different shader generators

jarkko: i am not sure, but it could indeed be possible that some of the wined3d devs already have working prototypes of working dx10 stuff - nevertheless, upstreaming takes some time, and every second patch introduces another regression, so ...

i dont believe we see csmt this year upstreamed

dont know if we even see stable release

jarkko: i am not sure either

wined3d is undocumented?

wow

jarkko: [06wine](00http://is.gd/ZSTqeU) 02John Edmonds: wined3d: Turn off message filtering temporarily in the Reset() method to allow certain messages (e.g. WM_ACTIVATEAPP) through.

any idea why 1 spec is different?

jarkko: d3dx10_43 is where the implementation will be later (currently a stub), the rest just redirects to d3dx10_43

jarkko: thats the usual method to avoid reimplementing the function for all versions of the dll

i tried the demo just now and i dont know if the font is even the issue now

=>0 0x005f0e3e in cascades (+0x1f0e3e) (0x0033e2dc) and similar lines

fixme:dxgi:dxgi_device_init Ignoring adapter type. same as 3dmark vantage. i quess it wants to know if we have dx10