Bug 83468 - [UBO] Using bool from UBO as if-statement condition asserts
Summary: [UBO] Using bool from UBO as if-statement condition asserts
Status: RESOLVED FIXED
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965 (show other bugs)
Version: git
Hardware: Other All
: medium normal
Assignee: Kenneth Graunke
QA Contact: Intel 3D Bugs Mailing List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-09-04 01:33 UTC by Ian Romanick
Modified: 2014-09-11 01:04 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
Use bool from UBO as if-statement condition (533 bytes, text/plain)
2014-09-04 01:33 UTC, Ian Romanick
Details
Use bool from UBO as if-statement condition (in FS) (401 bytes, text/plain)
2014-09-04 04:34 UTC, Kenneth Graunke
Details
Use bool for UBO as if-statement condition (in FS, v2) (423 bytes, text/plain)
2014-09-04 04:45 UTC, Kenneth Graunke
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Ian Romanick 2014-09-04 01:33:15 UTC
Created attachment 105711 [details]
Use bool from UBO as if-statement condition

The attached shader causes the following assertion failure:

shader_runner: brw_vec4_visitor.cpp:856: void brw::vec4_visitor::emit_bool_to_cond_code(ir_rvalue*, brw_predicate*): Assertion `!"not reached"' failed.

I have only tested on a fairly stale version of my gles3conform-v5 branch, so this issue may not exist in 10.3 or master branches.  If I don't submit a bug, I'll forget about it...
Comment 1 Ian Romanick 2014-09-04 01:34:20 UTC
Changing the line

    if (b1)

to

    if (b1 == true)

causes it to not assert.
Comment 2 Kenneth Graunke 2014-09-04 04:34:39 UTC
Created attachment 105714 [details]
Use bool from UBO as if-statement condition (in FS)

Yeah, we need to handle ir_unop_ubo_load in that switch statement.  In both backends.
Comment 3 Kenneth Graunke 2014-09-04 04:45:08 UTC
Created attachment 105715 [details]
Use bool for UBO as if-statement condition (in FS, v2)

Whoops, that FS test had the wrong result.  Here's a better one.
Comment 4 Kenneth Graunke 2014-09-04 07:46:52 UTC
Patches on the mailing list:
http://lists.freedesktop.org/archives/mesa-dev/2014-September/067204.html
http://lists.freedesktop.org/archives/mesa-dev/2014-September/067206.html
http://lists.freedesktop.org/archives/mesa-dev/2014-September/067205.html

The last one should fix it.  It may or may not depend on the second.  The first is just a bug fix while I was in the area :)
Comment 5 Ian Romanick 2014-09-11 01:04:14 UTC
Fixed by:

commit a20cc2796f5d55e49956ac0bc5d61ca027eec7f9
Author: Kenneth Graunke <kenneth@whitecape.org>
Date:   Thu Sep 4 00:18:45 2014 -0700

    i965: Handle ir_binop_ubo_load in boolean expression code.
    
    UBO loads can be boolean-valued expressions, too, so we need to handle
    them in emit_bool_to_cond_code() and emit_if_gen6().
    
    However, unlike most expressions, it doesn't make sense to evaluate
    their operands, then do something with the results.  We just want to
    evaluate the UBO load as a whole---which performs the read from
    memory---then load the boolean result into the flag register.
    
    Instead of adding code to handle it, we can simply bypass the
    ir_expression handling, and fall through to the default code, which will
    do exactly that.
    
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83468
    Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
    Reviewed-by: Matt Turner <mattst88@gmail.com>
    Cc: mesa-stable@lists.freedesktop.org

Commit a318e2f3 on 10.3 branch.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct.