bugzilla-daemon at freedesktop.org
2017-Aug-23 21:18 UTC
[Nouveau] [Bug 102382] New: nv50ir: isImmediate doesn't properly work for integers if Value is typed as float
https://bugs.freedesktop.org/show_bug.cgi?id=102382
Bug ID: 102382
Summary: nv50ir: isImmediate doesn't properly work for integers
if Value is typed as float
Product: xorg
Version: git
Hardware: Other
OS: All
Status: NEW
Severity: normal
Priority: medium
Component: Driver/nouveau
Assignee: nouveau at lists.freedesktop.org
Reporter: karolherbst at gmail.com
QA Contact: xorg-team at lists.x.org
Created attachment 133728
--> https://bugs.freedesktop.org/attachment.cgi?id=133728&action=edit
TGSI
codegen should generate a texfetch instruction with a lod according to
IMM[1].xxxx, which is of value 2, but in
https://cgit.freedesktop.org/mesa/mesa/tree/src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp#n140
the isInteger(0) call assumes the Value being a F32, which leads to a is
2.8E-45 == 0 check returning true resulting into the lod argument being removed
from texfetch.
This causes a fail in the KHR-GL44.robust_buffer_access_behavior.texel_fetch
CTS test.
Other isInteger calls might be broken as well where an immediate of type float
is interpreted as int by the instruction.
--
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<https://lists.freedesktop.org/archives/nouveau/attachments/20170823/c91679cd/attachment.html>
bugzilla-daemon at freedesktop.org
2017-Aug-24 22:14 UTC
[Nouveau] [Bug 102382] nv50ir: isImmediate doesn't properly work for integers if Value is typed as float
https://bugs.freedesktop.org/show_bug.cgi?id=102382
Karol Herbst <karolherbst at gmail.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |RESOLVED
Resolution|--- |FIXED
--- Comment #1 from Karol Herbst <karolherbst at gmail.com> ---
nothing was wrong with isImmediate, but the tetexfetch was fixed in
96be442b7795a6eb3d50f4061f2b98dddc39aa4d
--
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<https://lists.freedesktop.org/archives/nouveau/attachments/20170824/56380bf7/attachment.html>
Reasonably Related Threads
- texelFetch sampler1/2DArray on nv50 gallium
- [PATCH] nvc0/ir: move sample id to second source arg to fix sampler2DMS
- [LLVMdev] Being able to know the jitted code-size before emitting
- [LLVMdev] Being able to know the jitted code-size before emitting
- [Bug 90348] New: Spilling failure of b96 merged value