Update WebGPU CTS (2024-07)
Categories
(Core :: Graphics: WebGPU, task, P1)
Tracking
()
Tracking | Status | |
---|---|---|
firefox131 | --- | fixed |
People
(Reporter: ErichDonGubler, Assigned: ErichDonGubler)
References
(Blocks 1 open bug)
Details
Attachments
(23 files)
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review | |
Bug 1909753, part 2.17 promote tests whose structure didn't change, but now pass r=#webgpu-reviewers
48 bytes,
text/x-phabricator-request
|
Details | Review | |
48 bytes,
text/x-phabricator-request
|
Details | Review |
Assignee | ||
Updated•7 months ago
|
Updated•7 months ago
|
Comment hidden (obsolete) |
Comment hidden (obsolete) |
Updated•7 months ago
|
Assignee | ||
Comment 3•6 months ago
|
||
Assignee | ||
Comment 4•6 months ago
|
||
Assignee | ||
Comment 5•6 months ago
|
||
Not sure how we hadn't caught these before. 😅
Assignee | ||
Comment 6•6 months ago
|
||
Updated•6 months ago
|
Updated•6 months ago
|
Assignee | ||
Comment 7•6 months ago
|
||
Assignee | ||
Comment 8•6 months ago
|
||
Assignee | ||
Comment 9•6 months ago
|
||
Assignee | ||
Comment 10•6 months ago
|
||
webgpu:shader,execution,expression,call,builtin,textureNumLevels:depth:*
, implemented by gpuweb/cts#3855.webgpu:shader,execution,expression,call,builtin,textureLoad:storage_texel_formats:*
, introduced by gpuweb/cts#3849.webgpu:shader,execution,expression,unary,address_of_and_indirection:deref_swizzle:*
, introduced by gpuweb/cts#3869.webgpu:shader,validation,expression,call,builtin,pack2x16float:*
, introduced by TODO.webgpu:shader,validation,expression,call,builtin,texture{Dimensions,Num{Layers,Levels,Samples}}:*
, introduced by gpuweb/cts#3689.webgpu:shader,validation,statement,continue:*
, introduced by gpuweb/cts#3688.webgpu:shader,validation,statement,{if,loop,return,switch,while}:*
, introduced by gpuweb/cts#3645 and moved around by gpuweb/cts#3742.webgpu:shader,validation,types,pointer:missing_type:*
, introduced bygpuweb/cts
#3696.
This list is likely incomplete, because I really needed to make this tractable. Sorry! I upkept the above for my own sanity, and think it's interesting for record-keeping, but not necessary useful for a reviewer.
Assignee | ||
Comment 11•6 months ago
|
||
This appears to be because of new trailing comma subcases (which aren't represented in the WPT test path) from gpuweb/cts#3696, rather than an actual regression.
Assignee | ||
Comment 12•6 months ago
|
||
webpgpu:api,validation,buffer,mapping:getMappedRange,disjoinRanges_many
was renamed todisjointRanges_many
(see also @jimb's PR making the move at gpuweb/cts#3838)webgpu:shader,validation,decl,ptr_spelling:*
was merged withwebgpu:shader,validation,types,pointer:*
ingpuweb/cts
#3714.webgpu:shader,validation,{decl,var_access_mode,parse,var_and_let}:*
→webgpu:shader,validation,decl,var:*
, merged in [gpuweb/cts#3713](https:// github.com/gpuweb/cts/pull/3713).…:initializer_type:*
had its structure changed to accommodate the merge.
webgpu:shader,validation,parse,align:*
→webgpu:shader,validation,shader_io,align:*
, moved by gpuweb/cts#3712,webgpu:shader,validation,parse,{break,break_if,compound,const_assert,continuing,discard,statement_behavior}
, moved by gpuweb/cts#3742webgpu:shader,validation,parse,break_if:non_bool_param:*
→webgpu:shader,validation,statement,break_if:condition_type:*
, moved by gpuweb/cts#3671, affected bygpuweb/cts
#3690.
webgpu:shader,validation,parse_binary_ops:*
→webgpu:shader,validation,expression,binary,parse:*
This list is likely incomplete, because I really needed to make this tractable. Sorry! I upkept the above for my own sanity, and think it's interesting for record-keeping, but not necessary useful for a reviewer.
Assignee | ||
Comment 13•6 months ago
|
||
This (non-exhaustively) adds or implements the following test paths:
-
webgpu:api,validation,render_pipeline,depth_stencil_state:depth_bias:*
, introduced by gpuweb/cts#3863. Tracked with bug 1911312. -
webgpu:shader,execution,expression,access,array,index:matrix:*
, introduced by gpuweb/cts#3738. -
webgpu:shader,execution,expression,call,builtin,atomics,atomicSub:sub_i32_min:*
, introduced by gpuweb/cts#3728. -
webgpu:shader,execution,expression,call,builtin,fwidth{,Coarse,Fine}:f32:*
, implemented by gpuweb/cts#3686. -
webgpu:shader,execution,expression,call,builtin,textureNumLayers:arrayed:*
, implemented by gpuweb/cts#3853. -
webgpu:shader,execution,expression,call,builtin,textureNumLevels:sampled:*
, implemented by gpuweb/cts#3855. -
webgpu:shader,execution,expression,call,builtin,textureNumSamples:sampled:*
, implemented by gpuweb/cts#3856. -
webgpu:shader,execution,expression,call,builtin,textureLoad:*
, with:{arrayed,sampled_1d,sampled_3d}
implemented by gpuweb/cts#3852sampled_2d
implemented by gpuweb/cts#3861.
-
webgpu:shader,execution,expression,call,builtin,textureSample:*
, apparently changed by gpuweb/cts#3877. -
webgpu:shader,execution,expression,call,builtin,textureStore:*
had the following tests added:- By gpuweb/cts#3781:
out_of_bounds
out_of_bounds_array
- By gpuweb/cts#3841:
bgra8unorm_swizzle
texel_formats
It's unclear why these are failing, but I suspect part 2.7.
- By gpuweb/cts#3781:
-
webgpu:shader,execution,expression,constructor,zero_value:vector_prefix:*
, introduced by gpuweb/cts#3734. -
webgpu:shader,validation,decl,var:address_space_access_mode:*
, introduced bygpuweb/cts
#3695. -
webgpu:shader,validation,decl,var:shader_stage:*
, introduced by gpuweb/cts#3736. -
webgpu:shader,validation,expression,access,vector:*
was modified in gpuweb/cts#3708:- renamed
…:vector:*
toconcrete
- added
abstract
- renamed
-
webgpu:shader,validation,expression,binary,*
:…,bitwise_shift:*
:…:partial_eval_errors:*
, introduced bygpuweb/cts
#3796…:{shift_left_concrete,shift_right_concrete}:*
had new cases introduced in gpuweb/cts#3726.…:{shift_left_abstract,shift_right_abstract}:*
, introduced bygpuweb/cts
#3833.
…,{add_sub_mul:*,and_or_xor:*,bitwise_shift:scalar_vector:*,div_rem:scalar_vector:*}
modified bygpuweb/cts
#3648.
-
webgpu:shader,validation,expression,call,builtin,*
:…,dot:*
, introduced in gpuweb/cts#3579.…,fma:*
, introduced bygpuweb/cts
#3577.…,ldexp:*
:…:{args,must_use,values}:*
introduced by gpuweb/cts#3615.…:partial_values:*
introduced by gpuweb/cts#3798.
…,mix:*
, introduced by TODO.…,modf:*
, introduced by TODO.…,pow:*
, introduced by TODO.…,refract:*
, introduced by TODO.…,saturate:*
, introduced by TODO.…,texture{Dimensions,Num{Layers,Levels,Samples}}:*
, introduced by gpuweb/cts#3689.
-
webgpu:shader,validation,functions,alias_analysis:swizzles:*
, introduced by gpuweb/cts#3869. -
webgpu:shader,validation,shader_io,align:parsing:duplicate
, introduced by gpuweb/cts#3692. -
webgpu:shader,validation,statement,{for,loop,return}:*
cases introduced by gpuweb/cts#3645 and moved around by gpuweb/cts#3742. -
webgpu:shader,validation,types,pointer:{access_mode,address_space,type}:*
, introduced bygpuweb/cts
#3649.
This list is incomplete, because I really needed to make this tractable. Sorry! I upkept the above for my own sanity, and think it's interesting for record-keeping, but not necessary useful for a reviewer.
Assignee | ||
Comment 14•6 months ago
|
||
webgpu:api,operation,uncapturederror:constructor:*
was unimplemented, and then removed.webgpu:api,operation,uncapturederror:iff_uncaptured:*
was finally implemented (which triviallyPASS
ed before), and fails in theout-of-memory
case.
Assignee | ||
Comment 15•6 months ago
|
||
These failures (non-exhaustively) include:
webgpu:api,operation,uncapturederror:*
webgpu:api,rendering,depth_clip_clamp:*
webgpu:api,validation,render_pipeline,inter_stage:interpolation_sampling:*
This list is incomplete, because I really needed to make this tractable. Sorry! I upkept the above for my own sanity, and think it's interesting for record-keeping, but not necessary useful for a reviewer.
These started breaking in gpuweb/cts
#3814. Neither we nor Chrome stable (127.0.6533.89) pass this test ATOW, so 🤷🏻♂️ whatevs!
Assignee | ||
Comment 16•6 months ago
|
||
Quite a few tests regressed, with the only recent change being gpuweb/cts#3821. These failures (non-exhaustively) include:
webgpu:api,validation,buffer,create:{size,usage}:*
webgpu:compat,api,validation,encoding,programmable,pipeline_bind_group_compat:twoDifferentTextureViews,{render_pass,compute_pass},{used,unused}:*
This list is incomplete, because I really needed to make this tractable. Sorry! I upkept the above for my own sanity, and think it's interesting for record-keeping, but not necessary useful for a reviewer
Assignee | ||
Comment 17•6 months ago
|
||
Assignee | ||
Comment 18•6 months ago
|
||
I found the following:
webgpu:compat,api,validation,encoding,programmable,pipeline_bind_group_compat:twoDifferentTextureViews,{render_pass,compute_pass},used:*
- These started actually failing in part 2.9, but are also affected here.
webgpu:compat,api,validation,render_pipeline,fragment_state:colorState:*
…and there may be more, but I really needed to make this tractable. Sorry! I upkept the above for my own sanity, and think it's interesting for record-keeping, but not necessary useful for a reviewer.
Assignee | ||
Comment 19•6 months ago
|
||
This movement was introduced in gpuweb/cts#3769.
These would have been included in part 2.7, except that a breakage via @interpolate(…, either)
was introduced (see also part 2.7).
Assignee | ||
Comment 20•6 months ago
|
||
With so many tests starting to enter the webgpu-long
track, the tasks are starting to take > 70 minutes on Windows. Sounds like a good time to split 'em up!
There's enough of these coming into webgpu-long
that I expect we'll be making it worth CI time on all platforms shortly.
Assignee | ||
Comment 21•6 months ago
|
||
Assignee | ||
Comment 22•6 months ago
|
||
If I saw a test that took longer than 1 min. to execute, than I marked it as webgpu-long
. Works pretty well; one can see that non-*-long
WebGPU CI task chunks are executing very consistently now. 😀
Assignee | ||
Comment 23•6 months ago
|
||
Assignee | ||
Comment 24•6 months ago
|
||
Assignee | ||
Comment 25•6 months ago
|
||
Assignee | ||
Updated•6 months ago
|
Comment 26•6 months ago
|
||
Comment 27•6 months ago
|
||
bugherder |
Comment 28•6 months ago
|
||
Assignee | ||
Updated•6 months ago
|
Comment 29•6 months ago
|
||
bugherder |
https://hg.mozilla.org/mozilla-central/rev/f296491b5136
https://hg.mozilla.org/mozilla-central/rev/2868446de6df
https://hg.mozilla.org/mozilla-central/rev/353c8cbdfa7a
https://hg.mozilla.org/mozilla-central/rev/860d96bdcd69
https://hg.mozilla.org/mozilla-central/rev/4ea84dd05c32
https://hg.mozilla.org/mozilla-central/rev/e7b536d026a2
https://hg.mozilla.org/mozilla-central/rev/cc8e4c0d5276
https://hg.mozilla.org/mozilla-central/rev/5424ce25eebe
https://hg.mozilla.org/mozilla-central/rev/46842e355b83
https://hg.mozilla.org/mozilla-central/rev/8df332be3ac3
https://hg.mozilla.org/mozilla-central/rev/0f0a6f299665
https://hg.mozilla.org/mozilla-central/rev/c9348783d3bf
https://hg.mozilla.org/mozilla-central/rev/e8f44cadcc6a
https://hg.mozilla.org/mozilla-central/rev/3d77ae1a68ea
https://hg.mozilla.org/mozilla-central/rev/5e668e6fa9e1
https://hg.mozilla.org/mozilla-central/rev/160efb2294aa
https://hg.mozilla.org/mozilla-central/rev/545b3601f6ff
https://hg.mozilla.org/mozilla-central/rev/aeff4370e885
https://hg.mozilla.org/mozilla-central/rev/590598212025
Description
•