[Piglit] [PATCH 1/2] glsl-1.10 / glsl-1.20: Don't use ortho in variable indexing tests.

[Matt Turner]

On Tue, Jun 12, 2018 at 11:38 AM, Ian Romanick <idr at freedesktop.org> wrote:
> On 06/11/2018 11:22 PM, Jordan Justen wrote:
>> From: Matt Turner <mattst88 at gmail.com>
>>
>> Using ortho makes shrinking the window size much more difficult, which
>> makes running tests in simulation take much longer.
>>
>> Ref: https://lists.freedesktop.org/archives/piglit/2013-November/008298.html
>> [jordan.l.justen at intel.com: convert from shell script to mako]
>> Signed-off-by: Jordan Justen <jordan.l.justen at intel.com>
>> Cc: Kenneth Graunke <kenneth at whitecape.org>
>> ---
>>  .../templates/gen_variable_index_read_tests/helpers.mako   | 7 ++-----
>>  .../gen_variable_index_write_tests/fs.shader_test.mako     | 3 ---
>>  .../templates/gen_variable_index_write_tests/helpers.mako  | 4 ++--
>>  .../gen_variable_index_write_tests/vs.shader_test.mako     | 3 ---
>>  4 files changed, 4 insertions(+), 13 deletions(-)
>>
>> diff --git a/generated_tests/templates/gen_variable_index_read_tests/helpers.mako b/generated_tests/templates/gen_variable_index_read_tests/helpers.mako
>> index 155eb8c4e..3ca306daf 100644
>> --- a/generated_tests/templates/gen_variable_index_read_tests/helpers.mako
>> +++ b/generated_tests/templates/gen_variable_index_read_tests/helpers.mako
>> @@ -179,9 +179,6 @@ void main()
>>  <%def name="emit_test_vectors(params)" filter="dedent">
>>    <%block filter="newlines">
>>    [test]
>> -  clear color 0.5 0.5 0.5 0.5
>> -  clear
>> -  ortho
>>    % if params.mode == 'uniform' and params.glsl_version == 110 and params.test_array_dim == 0:
>>      uniform ${params.cxr_type} m ${matrix_data(1, params.matrix_dim, delim=' ')}
>>    % endif
>> @@ -220,8 +217,8 @@ void main()
>>            x = x_base + 15 * column - 10
>>            y = 15 * row - 10
>>          %>
>> -        draw rect ${x} ${y} 10 10
>> -        probe rgb ${x + 5} ${y + 5} 0.0 1.0 0.0
>> +        draw rect -1 -1 2 2
>> +        probe all rgb 0.0 1.0 0.0
>
> It's a little surprising that this helps performance.  If I'm reading
> this correctly, this test tries to draw a bunch of 10x10 blocks for each
> test.  This is getting replaced with redrawing the whole window.  This
> also weakens the test.  Since there's no clear between draws, if every
> draw after the first does nothing, the test will incorrectly pass.

As I said 5 years ago, this patch is not intended to improve
performance. It's to make shrinking window sizes simpler, which is
itself intended to improve performance.