[Libclc-dev] [PATCH v2 DON'T MERGE 14/14] half_powr: Implement using powr

[Jeroen Ketema via Libclc-dev]

> On 29 Jan 2018, at 04:02, Jan Vesely <jan.vesely at rutgers.edu> wrote:
> 
> On Thu, 2018-01-18 at 16:19 -0500, Jan Vesely wrote:
>> On Thu, 2018-01-18 at 20:43 +0100, Jeroen Ketema wrote:
>>> This whole patch set LGTM.
>> 
>> thanks for quick review
>> 
>>> 
>>>> On 18 Jan 2018, at 00:41, Jan Vesely via Libclc-dev <libclc-dev at lists.llvm.org> wrote:
>>>> 
>>>> v2: Use full precision implementation
>>>> 
>>>> Signed-off-by: Jan Vesely <jan.vesely at rutgers.edu>
>>>> ---
>>>> Fails CTS as well. It should probably be left out for now.
>>> 
>>> I don’t have any issues with committing this, as the real failure is not in this code.
> 
> Hi Jeroen,
> 
> would you prefer this version(v2) for consistency, or v3?

Personally, I would just go for v2.

Jeroen

> 
> thanks,
> Jan
> 
>> 
>> I've posted a workaround that passes the CTS. I'm not sure which
>> version is preferable. since the workaround clearly hides an issue in
>> the original powr implementation.
>> 
>> Jan
>> 
>>> 
>>> Jeroen
>>> 
>>>> 
>>>> generic/include/clc/clc.h            | 1 +
>>>> generic/include/clc/math/half_powr.h | 7 +++++++
>>>> generic/lib/SOURCES                  | 1 +
>>>> generic/lib/math/half_powr.cl        | 6 ++++++
>>>> 4 files changed, 15 insertions(+)
>>>> create mode 100644 generic/include/clc/math/half_powr.h
>>>> create mode 100644 generic/lib/math/half_powr.cl
>>>> 
>>>> diff --git a/generic/include/clc/clc.h b/generic/include/clc/clc.h
>>>> index 930bbfc..fa48639 100644
>>>> --- a/generic/include/clc/clc.h
>>>> +++ b/generic/include/clc/clc.h
>>>> @@ -78,6 +78,7 @@
>>>> #include <clc/math/half_log.h>
>>>> #include <clc/math/half_log10.h>
>>>> #include <clc/math/half_log2.h>
>>>> +#include <clc/math/half_powr.h>
>>>> #include <clc/math/half_recip.h>
>>>> #include <clc/math/half_rsqrt.h>
>>>> #include <clc/math/half_sin.h>
>>>> diff --git a/generic/include/clc/math/half_powr.h b/generic/include/clc/math/half_powr.h
>>>> new file mode 100644
>>>> index 0000000..c4e5242
>>>> --- /dev/null
>>>> +++ b/generic/include/clc/math/half_powr.h
>>>> @@ -0,0 +1,7 @@
>>>> +#define __CLC_BODY <clc/math/binary_decl_tt.inc>
>>>> +#define __CLC_FUNCTION half_powr
>>>> +
>>>> +#include <clc/math/gentype.inc>
>>>> +
>>>> +#undef __CLC_BODY
>>>> +#undef __CLC_FUNCTION
>>>> diff --git a/generic/lib/SOURCES b/generic/lib/SOURCES
>>>> index 9dbd35e..5aa2ba6 100644
>>>> --- a/generic/lib/SOURCES
>>>> +++ b/generic/lib/SOURCES
>>>> @@ -112,6 +112,7 @@ math/half_exp2.cl
>>>> math/half_log.cl
>>>> math/half_log10.cl
>>>> math/half_log2.cl
>>>> +math/half_powr.cl
>>>> math/half_recip.cl
>>>> math/half_rsqrt.cl
>>>> math/half_sin.cl
>>>> diff --git a/generic/lib/math/half_powr.cl b/generic/lib/math/half_powr.cl
>>>> new file mode 100644
>>>> index 0000000..53174d4
>>>> --- /dev/null
>>>> +++ b/generic/lib/math/half_powr.cl
>>>> @@ -0,0 +1,6 @@
>>>> +#include <clc/clc.h>
>>>> +
>>>> +#define __CLC_FUNC powr
>>>> +#define __CLC_BODY <half_binary.inc>
>>>> +#define __FLOAT_ONLY
>>>> +#include <clc/math/gentype.inc>
>>>> -- 
>>>> 2.14.3
>>>> 
>>>> _______________________________________________
>>>> Libclc-dev mailing list
>>>> Libclc-dev at lists.llvm.org
>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/libclc-dev
>>> 
>>> 
> 
> -- 
> Jan Vesely <jan.vesely at rutgers.edu <mailto:jan.vesely at rutgers.edu>>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/libclc-dev/attachments/20180131/aac96ea7/attachment-0001.html>