From mboxrd@z Thu Jan 1 00:00:00 1970 From: bugzilla-daemon@freedesktop.org Subject: [Bug 95285] [radeonsi] Tomb Raider: enabling TressFX makes hair disappear Date: Thu, 05 May 2016 17:07:14 +0000 Message-ID: References: Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============1462130973==" Return-path: Received: from culpepper.freedesktop.org (culpepper.freedesktop.org [131.252.210.165]) by gabe.freedesktop.org (Postfix) with ESMTP id 226ED6E9FF for ; Thu, 5 May 2016 17:07:14 +0000 (UTC) In-Reply-To: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" To: dri-devel@lists.freedesktop.org List-Id: dri-devel@lists.freedesktop.org --===============1462130973== Content-Type: multipart/alternative; boundary="14624680341.2FfaE3bc.28759"; charset="UTF-8" --14624680341.2FfaE3bc.28759 Date: Thu, 5 May 2016 17:07:14 +0000 MIME-Version: 1.0 Content-Type: text/plain https://bugs.freedesktop.org/show_bug.cgi?id=95285 Kai changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |kenneth@whitecape.org --- Comment #4 from Kai --- (In reply to Ilia Mirkin from comment #2) > Try some of Ken's patches at > > https://cgit.freedesktop.org/~kwg/mesa/log/?h=tombraider-2 Thanks for the pointer, Ilia! I've applied the following patches: fb4f8f9 glsl: Make opt_constant_variable() bail in useless cases. a6bebce glsl: Remove bonus tree walking in opt_constant_folding(). b01114f glsl: Consolidate duplicate copies of constant folding. 2040a80 glsl: Avoid excess tree walking when folding ir_dereference_arrays. 285e966 glsl: Don't do constant propagation in opt_constant_folding. 862fb32 glsl: Make lower_const_arrays_to_uniforms work directly on constants. 341ee51 glsl: Don't constant propagate arrays Additionally, I think I should mention, that I've applied locally, since I'm using a LLVM > 268259. Just in case that could have an impact. Anyway, with those patches applied the game freezes the system (ie. the system stops responding to any kind of input) while entering the main menu. The only output I could capture was "[0505/185411:WARNING:x11_util.cc(1490)] X error received: serial 1001, error_code 3 (BadWindow (invalid Window parameter)), request_code 4, minor_code 0 (X_DestroyWindow)" CCing Kenneth, maybe he has an idea which of his patches could cause this. > There are some issues which tressfx happens to hit. (And some other option > too...) Do you mean bug 95190 or something else? I'm seeing bug 95190 as well. Therefore I've kept post processing disabled so far. > However at least on nouveau, tressfx is still a fail even with > those. [Sorry, don't have the game myself, so don't have further details. > Just an FYI with the current state of affairs.] No need to apologize! Any piece of information is appreciated. As far as access to TR goes: did you try contacting Feral? They were willing to give access to XCOM 2 to some Mesa developers IIRC, maybe they're willing to do the same here? (In reply to Nicolai Hähnle from comment #3) > If the patches mentioned by Ilia don't help, an apitrace showing the problem > would be useful. Ok, I'll do that. -- You are receiving this mail because: You are the assignee for the bug. --14624680341.2FfaE3bc.28759 Date: Thu, 5 May 2016 17:07:14 +0000 MIME-Version: 1.0 Content-Type: text/html changed bug 95285
What Removed Added
CC   kenneth@whitecape.org

Comment # 4 on bug 95285 from
(In reply to Ilia Mirkin from comment #2)
> Try some of Ken's patches at
> 
> https://cgit.freedesktop.org/~kwg/mesa/log/?h=tombraider-2

Thanks for the pointer, Ilia! I've applied the following patches:
fb4f8f9 glsl: Make opt_constant_variable() bail in useless cases.
a6bebce glsl: Remove bonus tree walking in opt_constant_folding().
b01114f glsl: Consolidate duplicate copies of constant folding.
2040a80 glsl: Avoid excess tree walking when folding ir_dereference_arrays.
285e966 glsl: Don't do constant propagation in opt_constant_folding.
862fb32 glsl: Make lower_const_arrays_to_uniforms work directly on constants.
341ee51 glsl: Don't constant propagate arrays

Additionally, I think I should mention, that I've applied
<https://patchwork.freedesktop.org/patch/85101/> locally, since I'm using a
LLVM > 268259. Just in case that could have an impact.

Anyway, with those patches applied the game freezes the system (ie. the system
stops responding to any kind of input) while entering the main menu. The only
output I could capture was "[0505/185411:WARNING:x11_util.cc(1490)] X error
received: serial 1001, error_code 3 (BadWindow (invalid Window parameter)),
request_code 4, minor_code 0 (X_DestroyWindow)"
CCing Kenneth, maybe he has an idea which of his patches could cause this.

> There are some issues which tressfx happens to hit. (And some other option
> too...)

Do you mean bug 95190 or something else? I'm seeing bug 95190 as well.
Therefore I've kept post processing disabled so far.

> However at least on nouveau, tressfx is still a fail even with
> those. [Sorry, don't have the game myself, so don't have further details.
> Just an FYI with the current state of affairs.]

No need to apologize! Any piece of information is appreciated. As far as access
to TR goes: did you try contacting Feral? They were willing to give access to
XCOM 2 to some Mesa developers IIRC, maybe they're willing to do the same here?


(In reply to Nicolai Hähnle from comment #3)
> If the patches mentioned by Ilia don't help, an apitrace showing the problem
> would be useful.
Ok, I'll do that.


You are receiving this mail because:
  • You are the assignee for the bug.
--14624680341.2FfaE3bc.28759-- --===============1462130973== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: base64 Content-Disposition: inline X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KZHJpLWRldmVs IG1haWxpbmcgbGlzdApkcmktZGV2ZWxAbGlzdHMuZnJlZWRlc2t0b3Aub3JnCmh0dHBzOi8vbGlz dHMuZnJlZWRlc2t0b3Aub3JnL21haWxtYW4vbGlzdGluZm8vZHJpLWRldmVsCg== --===============1462130973==--