[Libclc-dev] [PATCH 4/4] Implement sin builtin for float types

[Matt Arsenault]

On Jul 2, 2014, at 1:55 PM, Jeroen Ketema <j.ketema at imperial.ac.uk> wrote:

> 
> Hi Tom,
> 
> atan and atan2 clearly need implementations, as there are no intrinsics for them. However, why are implementations of cos and sin for float needed? The intrinsics are typed so they could just be expanded appropriately in the backend.
> 
> On another note: It’s difficult to check the code in the patches. Is the intention to implement some standard textbook algorithm?
> 
> Jeroen
> 

The intrinsics do not have the precision or well defined edge case behavior required by the standard.The LLVM math intrinsics are sufficient for native_* math functions, and maybe in some of the fast math modes, but not as the standard function behavior.


> On 02 Jul 2014, at 21:11, Tom Stellard <thomas.stellard at amd.com> wrote:
> 
>> This double version still uses @llvm.sin.
>> ---
>> generic/include/clc/math/sin.h |  9 ++----
>> generic/lib/SOURCES            |  1 +
>> generic/lib/math/sin.cl        | 69 ++++++++++++++++++++++++++++++++++++++++++
>> 3 files changed, 73 insertions(+), 6 deletions(-)
>> create mode 100644 generic/lib/math/sin.cl
>> 
>> diff --git a/generic/include/clc/math/sin.h b/generic/include/clc/math/sin.h
>> index e50b71a..6d4cf5a 100644
>> --- a/generic/include/clc/math/sin.h
>> +++ b/generic/include/clc/math/sin.h
>> @@ -1,6 +1,3 @@
>> -#undef sin
>> -#define sin __clc_sin
>> -
>> -#define __CLC_FUNCTION __clc_sin
>> -#define __CLC_INTRINSIC "llvm.sin"
>> -#include <clc/math/unary_intrin.inc>
>> +#define __CLC_BODY <clc/math/sin.inc>
>> +#include <clc/math/gentype.inc>
>> +#undef __CLC_BODY
>> diff --git a/generic/lib/SOURCES b/generic/lib/SOURCES
>> index 4d30669..73a47d2 100644
>> --- a/generic/lib/SOURCES
>> +++ b/generic/lib/SOURCES
>> @@ -38,6 +38,7 @@ math/mad.cl
>> math/mix.cl
>> math/clc_nextafter.cl
>> math/nextafter.cl
>> +math/sin.cl
>> math/sincos.cl
>> math/sincos_helpers.cl
>> relational/all.cl
>> diff --git a/generic/lib/math/sin.cl b/generic/lib/math/sin.cl
>> new file mode 100644
>> index 0000000..b32ee4e
>> --- /dev/null
>> +++ b/generic/lib/math/sin.cl
>> @@ -0,0 +1,69 @@
>> +/*
>> + * Copyright (c) 2014 Advanced Micro Devices, Inc.
>> + *
>> + * Permission is hereby granted, free of charge, to any person obtaining a copy
>> + * of this software and associated documentation files (the "Software"), to deal
>> + * in the Software without restriction, including without limitation the rights
>> + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
>> + * copies of the Software, and to permit persons to whom the Software is
>> + * furnished to do so, subject to the following conditions:
>> + *
>> + * The above copyright notice and this permission notice shall be included in
>> + * all copies or substantial portions of the Software.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
>> + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
>> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
>> + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
>> + * THE SOFTWARE.
>> + */
>> +
>> +#include <clc/clc.h>
>> +
>> +#include "math.h"
>> +#include "sincos_helpers.h"
>> +
>> +_CLC_OVERLOAD _CLC_DEF float sin(float x)
>> +{
>> +    int ix = as_int(x);
>> +    int ax = ix & 0x7fffffff;
>> +    float dx = as_float(ax);
>> +
>> +    float r0, r1;
>> +    int regn = argReductionS(&r0, &r1, dx);
>> +
>> +    float ss = sinf_piby4(r0, r1);
>> +    float cc = cosf_piby4(r0, r1);
>> +
>> +    float s = (regn & 1) != 0 ? cc : ss;
>> +    s = as_float(as_int(s) ^ ((regn > 1) << 31) ^ (ix ^ ax));
>> +
>> +    s = ax >= PINFBITPATT_SP32 ? as_float(QNANBITPATT_SP32) : s;
>> +
>> +    //Subnormals
>> +    s = x == 0.0f ? x : s;
>> +
>> +    return s;
>> +}
>> +
>> +_CLC_UNARY_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, float, sin, float);
>> +
>> +#ifdef cl_khr_fp64
>> +
>> +#pragma OPENCL EXTENSION cl_khr_fp64 : enable
>> +
>> +#define __CLC_FUNCTION __clc_sin_intrinsic
>> +#define __CLC_INTRINSIC "llvm.sin"
>> +#include <clc/math/unary_intrin.inc>
>> +#undef __CLC_FUNCTION
>> +#undef __CLC_INTRINSIC
>> +
>> +_CLC_OVERLOAD _CLC_DEF double sin(double x) {
>> +    return __clc_sin_intrinsic(x);
>> +}
>> +
>> +_CLC_UNARY_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, double, sin, double);
>> +
>> +#endif
>> -- 
>> 1.8.1.5
>> 
>> 
>> _______________________________________________
>> Libclc-dev mailing list
>> Libclc-dev at pcc.me.uk
>> http://www.pcc.me.uk/cgi-bin/mailman/listinfo/libclc-dev
> 
> 
> _______________________________________________
> Libclc-dev mailing list
> Libclc-dev at pcc.me.uk
> http://www.pcc.me.uk/cgi-bin/mailman/listinfo/libclc-dev