[Piglit] [PATCH 5/5] cl: Add signbit(float) relational tests
Aaron Watry
awatry at gmail.com
Tue Jun 17 06:50:00 PDT 2014
On Thu, Jun 12, 2014 at 6:40 PM, Tom Stellard <tom at stellard.net> wrote:
> On Wed, Jun 11, 2014 at 02:14:15PM -0500, Aaron Watry wrote:
>> Signed-off-by: Aaron Watry <awatry at gmail.com>
>> ---
>> generated_tests/generate-cl-relational-builtins.py | 9 +++++++++
>> 1 file changed, 9 insertions(+)
>>
>> diff --git a/generated_tests/generate-cl-relational-builtins.py b/generated_tests/generate-cl-relational-builtins.py
>> index f22f5d1..833a836 100644
>> --- a/generated_tests/generate-cl-relational-builtins.py
>> +++ b/generated_tests/generate-cl-relational-builtins.py
>> @@ -34,6 +34,7 @@ CLC_VERSION_MIN = {
>> 'isgreater' : 10,
>> 'isgreaterequal' : 10,
>> 'isnotequal' : 10,
>> + 'signbit' : 10
>> }
>>
>> DATA_TYPES = ['float']
>> @@ -81,6 +82,14 @@ tests = {
>> [0.0, float("nan"), 1.0, 1, float("nan"), float("inf"), 1.0, 0.5 ], # Arg0
>> [0.0, float("nan"), 1.0, float("nan"), 1, float("inf"), 0.5, 1.0 ] # Arg1
>> ]
>> + },
>> + 'signbit' : {
>> + 'arg_types': [I, F],
>> + 'function_type': 'ttt',
>> + 'values': [
>> + [0, TRUE, 0, 0, TRUE, 0, TRUE ], # Result
>
> According to the man page for signbit, -inf, inf, and nan all have a
> signbit. I'm assuming OpenCL C is the same since it is based on C99,
> so the inf case should be TRUE and I think it would be a good idea to
> add a case for nan.
I'll do that (add [+-]NAN and -INF tests) and send a v2 of this patch
separate from the rest of the series, which I think it ready to go.
>
> Also, have you verified that float(0)*-1 generates -0.0 ?
Yes... after quite a bit of searching, that was the only way I found
to get python to generate a negative 0. Maybe there's others, but
float(-0.0) and several other variants all generate a positive 0
instead.
--Aaron
>
> -Tom
>
>> + [0.0, float(0)*-1, 1.0, 1, float("-inf"), float("inf"), -1.0 ] # Arg0
>> + ]
>> }
>> }
>>
>> --
>> 1.9.1
>>
More information about the Piglit
mailing list