[Libclc-dev] [PATCH 03/15] Implement fast_normalize builtin v3
Jan Vesely
jan.vesely at rutgers.edu
Fri Apr 10 16:23:26 PDT 2015
On Tue, 2015-04-07 at 18:05 +0000, Tom Stellard wrote:
> This implementation was ported from the AMD builtin library
> and has been tested with piglit, OpenCV, and the ocl conformance tests.
>
> v2:
> - Remove f suffix from constant in double implementations.
> - Consolidate implementations using the .cl/.inc approach.
>
> v3:
> - Use __CLC_FPSIZE instead of __CLC_FP{32,64}
> ---
> generic/include/clc/clc.h | 1 +
> generic/include/clc/geometric/fast_normalize.h | 24 ++++++++++++++
> generic/include/clc/geometric/fast_normalize.inc | 24 ++++++++++++++
> generic/include/clc/geometric/floatn.inc | 8 +++++
> generic/lib/SOURCES | 1 +
> generic/lib/geometric/fast_normalize.cl | 40 ++++++++++++++++++++++++
> generic/lib/geometric/fast_normalize.inc | 39 +++++++++++++++++++++++
> 7 files changed, 137 insertions(+)
> create mode 100644 generic/include/clc/geometric/fast_normalize.h
> create mode 100644 generic/include/clc/geometric/fast_normalize.inc
> create mode 100644 generic/lib/geometric/fast_normalize.cl
> create mode 100644 generic/lib/geometric/fast_normalize.inc
>
> diff --git a/generic/include/clc/clc.h b/generic/include/clc/clc.h
> index b5ecda3..f3eaea2 100644
> --- a/generic/include/clc/clc.h
> +++ b/generic/include/clc/clc.h
> @@ -129,6 +129,7 @@
> #include <clc/geometric/dot.h>
> #include <clc/geometric/fast_distance.h>
> #include <clc/geometric/fast_length.h>
> +#include <clc/geometric/fast_normalize.h>
> #include <clc/geometric/length.h>
> #include <clc/geometric/normalize.h>
>
> diff --git a/generic/include/clc/geometric/fast_normalize.h b/generic/include/clc/geometric/fast_normalize.h
> new file mode 100644
> index 0000000..cba69c2
> --- /dev/null
> +++ b/generic/include/clc/geometric/fast_normalize.h
> @@ -0,0 +1,24 @@
> +/*
> + * Copyright (c) 2014,2015 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.
> + */
> +
> +#define __CLC_BODY <clc/geometric/fast_normalize.inc>
> +#include <clc/geometric/floatn.inc>
> diff --git a/generic/include/clc/geometric/fast_normalize.inc b/generic/include/clc/geometric/fast_normalize.inc
> new file mode 100644
> index 0000000..3ef8f86
> --- /dev/null
> +++ b/generic/include/clc/geometric/fast_normalize.inc
> @@ -0,0 +1,24 @@
> +/*
> + * Copyright (c) 2014,2015 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.
> + */
> +
> +
> +_CLC_OVERLOAD _CLC_DECL __CLC_FLOATN fast_normalize(__CLC_FLOATN p);
> diff --git a/generic/include/clc/geometric/floatn.inc b/generic/include/clc/geometric/floatn.inc
> index fb7a9ae..2c8e2d8 100644
> --- a/generic/include/clc/geometric/floatn.inc
> +++ b/generic/include/clc/geometric/floatn.inc
> @@ -1,8 +1,11 @@
> #define __CLC_FLOAT float
> +#define __CLC_FPSIZE 32
>
> #define __CLC_FLOATN float
> +#define __CLC_SCALAR
> #include __CLC_BODY
> #undef __CLC_FLOATN
> +#undef __CLC_SCALAR
>
> #define __CLC_FLOATN float2
> #include __CLC_BODY
> @@ -17,14 +20,18 @@
> #undef __CLC_FLOATN
>
> #undef __CLC_FLOAT
> +#undef __CLC_FPSIZE
>
> #ifdef cl_khr_fp64
>
> #define __CLC_FLOAT double
> +#define __CLC_FPSIZE 64
>
> #define __CLC_FLOATN double
> +#define __CLC_SCALAR
> #include __CLC_BODY
> #undef __CLC_FLOATN
> +#undef __CLC_SCALAR
>
> #define __CLC_FLOATN double2
> #include __CLC_BODY
> @@ -39,6 +46,7 @@
> #undef __CLC_FLOATN
>
> #undef __CLC_FLOAT
> +#undef __CLC_FPSIZE
>
> #endif
>
> diff --git a/generic/lib/SOURCES b/generic/lib/SOURCES
> index 92067ea..a4a017a 100644
> --- a/generic/lib/SOURCES
> +++ b/generic/lib/SOURCES
> @@ -38,6 +38,7 @@ geometric/distance.cl
> geometric/dot.cl
> geometric/fast_distance.cl
> geometric/fast_length.cl
> +geometric/fast_normalize.cl
> geometric/length.cl
> geometric/normalize.cl
> integer/abs.cl
> diff --git a/generic/lib/geometric/fast_normalize.cl b/generic/lib/geometric/fast_normalize.cl
> new file mode 100644
> index 0000000..d3d7846
> --- /dev/null
> +++ b/generic/lib/geometric/fast_normalize.cl
> @@ -0,0 +1,40 @@
> +/*
> + * Copyright (c) 2014,2015 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>
> +
> +_CLC_OVERLOAD _CLC_DEF float fast_normalize(float p) {
> + return normalize(p);
> +}
> +
> +#ifdef cl_khr_fp64
> +
> +#pragma OPENCL EXTENSION cl_khr_fp64 : enable
> +
> +_CLC_OVERLOAD _CLC_DEF double fast_normalize(double p) {
> + return normalize(p);
> +}
similar here. the specs only mention fast_normalize for floatN, not
double.
jan
> +
> +#endif
> +
> +#define __CLC_BODY <fast_normalize.inc>
> +#include <clc/geometric/floatn.inc>
> diff --git a/generic/lib/geometric/fast_normalize.inc b/generic/lib/geometric/fast_normalize.inc
> new file mode 100644
> index 0000000..fa3467c
> --- /dev/null
> +++ b/generic/lib/geometric/fast_normalize.inc
> @@ -0,0 +1,39 @@
> +/*
> + * Copyright (c) 2014,2015 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.
> + */
> +
> +#if __CLC_FPSIZE == 32
> +#define ZERO 0.0f
> +#else
> +#define ZERO 0.0
> +#endif
> +
> +#ifndef __CLC_SCALAR
> +
> +// Only handle vector implementations
> +_CLC_OVERLOAD _CLC_DEF __CLC_FLOATN fast_normalize(__CLC_FLOATN p) {
> + __CLC_FLOAT l2 = dot(p, p);
> + return l2 == ZERO ? p : p * half_rsqrt(l2);
> +}
> +
> +#endif
> +
> +#undef ZERO
--
Jan Vesely <jan.vesely at rutgers.edu>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: This is a digitally signed message part
URL: <http://lists.llvm.org/pipermail/libclc-dev/attachments/20150410/12b7e3f7/attachment.sig>
More information about the Libclc-dev
mailing list