[Piglit] [PATCH] glsl-1.10: Add test for bug in i965 gl_FrontFacing optimization
Ian Romanick
idr at freedesktop.org
Wed Mar 11 15:01:17 PDT 2015
On 03/11/2015 02:32 PM, Emil Velikov wrote:
> On 11 March 2015 at 21:19, Ian Romanick <idr at freedesktop.org> wrote:
>> From: Ian Romanick <ian.d.romanick at intel.com>
>>
>> These patterns
>>
>> gl_FrontFacing ? -1.0 : 1.0
>> gl_FrontFacing ? 1.0 : -1.0
>> gl_FrontFacing ? 0.0 : -1.0
>>
>> are recognized and optimized by the i965 driver. However, it
>> incorrectly optimizes
>>
>> int x = gl_FrontFacing ? 1 : -1;
>>
>> as
>>
>> int x = gl_FrontFacing ? floatBitsToInt(1.0) : floatBitsToInt(-1.0);
>>
>> NOTE: This test currently fails on Mesa master and 10.5 branches.
>>
>> Signed-off-by: Ian Romanick <ian.d.romanick at intel.com>
>> ---
>> .../fs-frontfacing-ternary-1-neg-1.shader_test | 32 ++++++++++++++++++++++
>> 1 file changed, 32 insertions(+)
>> create mode 100644 tests/spec/glsl-1.10/execution/fs-frontfacing-ternary-1-neg-1.shader_test
>>
>> diff --git a/tests/spec/glsl-1.10/execution/fs-frontfacing-ternary-1-neg-1.shader_test b/tests/spec/glsl-1.10/execution/fs-frontfacing-ternary-1-neg-1.shader_test
>> new file mode 100644
>> index 0000000..8d73bcd
>> --- /dev/null
>> +++ b/tests/spec/glsl-1.10/execution/fs-frontfacing-ternary-1-neg-1.shader_test
>> @@ -0,0 +1,32 @@
>> +[require]
>> +GLSL >= 1.10
>> +
>> +[vertex shader passthrough
> ^
> Missing closing bracket ] ?
Yes... it's also missing in the test that I copied,
tests/spec/glsl-1.10/execution/fs-frontfacing-ternary-1.0-neg-1.0.shader_test.
Weird.
> -Emil
More information about the Piglit
mailing list