xref: /aosp_15_r20/external/deqp/doc/testspecs/GLES31/functional.compute.indirect_dispatch.txt (revision 35238bce31c2a825756842865a792f8cf7f89930)

-------------------------------------------------------------------------
drawElements Quality Program Test Specification
-----------------------------------------------

Copyright 2014 The Android Open Source Project

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

     http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-------------------------------------------------------------------------
    Indirect Compute Dispatch

Tests:
 + dEQP-GLES31.functional.compute.indirect_dispatch.*

Includes:
 + glDispatchComputeIndirect()
 + Single and multiple dispatch calls from a single buffer
 + One or several work groups
 + One or several items per work group
 + Offsets within buffer
 + Empty dispatch commands (0 work groups)
 + Buffers uploaded from application side
 + Buffers generated using compute shaders
   - Synchronization with GL_COMMAND_BARRIER_BIT

Excludes:
 + Negative tests
 + Robustness tests

Description:

Indirect dispatch tests allocate a buffer and fill it with one or more
dispatch commands. In upload_buffer cases commands are written using
glBufferData(). In gen_in_compute cases a compute shader is first dispatches,
that writes out the commands into the buffer by accessing the command buffer
as a SSBO. Memory barrier is issued after the first compute shader in
gen_in_compute cases.

Actual indirect compute tasks use a simple compute shader, that has access to
a SSBO with two fields. One is reference work group count (uvec3) and another
is output uint field, that is incremented atomically for each invocation,
where comparison between gl_NumWorkGroups and the reference count passed.

Once the command buffer has been generated, one or more indirect dispatch
calls are made. SSBO binding offset is adjusted between each call to allocate
a different portion of the input/output buffer for each call. After calls have
been issued, result buffer is mapped for reading and pass counts are verified.
No explicit synchronization call is made prior to mapping the buffer so GL is
responsible of synchronizing the tasks.