This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] benchtests: Add strtod benchmark
- From: Will Newton <will dot newton at linaro dot org>
- To: Adhemerval Zanella <azanella at linux dot vnet dot ibm dot com>
- Cc: "GNU C. Library" <libc-alpha at sourceware dot org>
- Date: Tue, 5 Nov 2013 00:33:32 -0800
- Subject: Re: [PATCH] benchtests: Add strtod benchmark
- Authentication-results: sourceware.org; auth=none
- References: <5277E7CF dot 30801 at linux dot vnet dot ibm dot com>
On 4 November 2013 10:30, Adhemerval Zanella
<azanella@linux.vnet.ibm.com> wrote:
> Hi all,
>
> This patch add a benchtest for 'strtod'. I used as possible input the testing
> values from both 'stdlib/tst-strtod.c' and 'stdlib/tst-atof2.c'. I'm aware it
> might not be the best inputs that might cover all the strtod cases or even the
> user realcases, however adjustment can me made in the future.
>
> ---
>
> 2013-11-04 Adhemerval Zanella <azanella@linux.vnet.ibm.com>
>
> * benchtests/Makefile: Add bench-strtod.
> * benchtests/bench-strtod.c: New file: strtod benchtest.
>
> --
>
> diff --git a/benchtests/Makefile b/benchtests/Makefile
> index aec395a..5eb65e5 100644
> --- a/benchtests/Makefile
> +++ b/benchtests/Makefile
> @@ -32,7 +32,9 @@ string-bench := bcopy bzero memccpy memchr memcmp memcpy memmem memmove \
> string-bench-ifunc := $(addsuffix -ifunc, $(string-bench))
> string-bench-all := $(string-bench) $(string-bench-ifunc)
>
> -benchset := $(string-bench-all)
> +stdlib-bench := strtod
> +
> +benchset := $(string-bench-all) $(stdlib-bench)
>
> LDLIBS-bench-acos = -lm
> LDLIBS-bench-acosh = -lm
> diff --git a/benchtests/bench-strtod.c b/benchtests/bench-strtod.c
> new file mode 100644
> index 0000000..72151f9
> --- /dev/null
> +++ b/benchtests/bench-strtod.c
> @@ -0,0 +1,119 @@
> +/* Measure strtod implementation.
> + Copyright (C) 2013 Free Software Foundation, Inc.
> + This file is part of the GNU C Library.
> +
> + The GNU C Library is free software; you can redistribute it and/or
> + modify it under the terms of the GNU Lesser General Public
> + License as published by the Free Software Foundation; either
> + version 2.1 of the License, or (at your option) any later version.
> +
> + The GNU C Library is distributed in the hope that it will be useful,
> + but WITHOUT ANY WARRANTY; without even the implied warranty of
> + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + Lesser General Public License for more details.
> +
> + You should have received a copy of the GNU Lesser General Public
> + License along with the GNU C Library; if not, see
> + <http://www.gnu.org/licenses/>. */
> +
> +#define TEST_MAIN
> +#define TEST_NAME "strtod"
> +
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include "bench-timing.h"
> +
> +#define INNER_LOOP_ITERS 65536
> +
> +struct ltest
> + {
> + const char *str;
> + };
Is there a reason why this struct is needed rather than using a const
char * directly?
> +static const struct ltest inputs[] =
> + {
> + { "1e308" },
> + { "100000000e300" },
> + { "0x1p1023" },
> + { "0x1000p1011" },
> + { "0x1p1020" },
> + { "0x0.00001p1040" },
> + { "1e-307" },
> + { "0.000001e-301" },
> + { "0.0000001e-300" },
> + { "0.00000001e-299" },
> + { "1000000e-313" },
> + { "10000000e-314" },
> + { "100000000e-315" },
> + { "0x1p-1021" },
> + { "0x1000p-1033" },
> + { "0x10000p-1037" },
> + { "0x0.001p-1009" },
> + { "0x0.0001p-1005" },
> + { "12.345" },
> + { "12.345e19" },
> + { "-.1e+9" },
> + { ".125" },
> + { "1e20" },
> + { "0e-19" },
> + { "4\00012" },
> + { "5.9e-76" },
> + { "0x1.4p+3" },
> + { "0xAp0" },
> + { "0x0Ap0" },
> + { "0x0A" },
> + { "0xA0" },
> + { "0x0.A0p8" },
> + { "0x0.50p9" },
> + { "0x0.28p10" },
> + { "0x0.14p11" },
> + { "0x0.0A0p12" },
> + { "0x0.050p13" },
> + { "0x0.028p14" },
> + { "0x0.014p15" },
> + { "0x00.00A0p16" },
> + { "0x00.0050p17" },
> + { "0x00.0028p18" },
> + { "0x00.0014p19" },
> + { "0x1p-1023" },
> + { "0x0.8p-1022" },
> + { "Inf" },
> + { "-Inf" },
> + { "+InFiNiTy" },
> + { "0x80000Ap-23" },
> + { "1e-324" },
> + { "0x100000000000008p0" },
> + { "0x100000000000008.p0" },
> + { "0x100000000000008.00p0" },
> + { "0x10000000000000800p0" },
> + { "0x10000000000000801p0" },
> + { NULL }
> +};
> +
> +int
> +do_bench (void)
> +{
> + const size_t iters = INNER_LOOP_ITERS;
> + const struct ltest *lt;
> +
I guess there should be a call to TIMING_INIT here. Not sure it is
essential but it would be consistent at least.
> + for (lt = inputs; lt->str != NULL; ++lt)
> + {
> + char *ep;
> + timing_t start, stop, cur;
> +
> + printf ("Input %-24s:", lt->str);
> + TIMING_NOW (start);
> + for (size_t i = 0; i < iters; ++i)
> + strtod(lt->str, &ep);
> + TIMING_NOW (stop);
> +
> + TIMING_DIFF (cur, start, stop);
> + TIMING_PRINT_MEAN ((double) cur, (double) iters);
> + putchar ('\n');
> + }
> +
> + return 0;
> +}
> +
> +#define TEST_FUNCTION do_bench ()
> +
> +#include "../test-skeleton.c"
Otherwise looks good to me.
--
Will Newton
Toolchain Working Group, Linaro