mirror of
git://sourceware.org/git/glibc.git
synced 2024-12-03 04:01:43 +08:00
3c0265394d
This patch implements modf/modff optimization for POWER by focus on FP operations instead of relying in integer ones.
181 lines
4.9 KiB
Makefile
181 lines
4.9 KiB
Makefile
# 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/>.
|
||
|
||
|
||
# Makefile for benchmark tests. The only useful target here is `bench`.
|
||
|
||
# Adding a new function `foo`:
|
||
# ---------------------------
|
||
|
||
# - Append the function name to the bench variable
|
||
|
||
# - Define foo-ITER with the number of iterations you want to run. Keep it
|
||
# high enough that the overhead of clock_gettime is only a small fraction of
|
||
# the total run time of the test. A good idea would be to keep the run time
|
||
# of each test at around 10 seconds for x86_64. That is just a guideline,
|
||
# since some scenarios may require higher run times.
|
||
|
||
# - Define foo-ARGLIST as a colon separated list of types of the input
|
||
# arguments. Use `void` if function does not take any inputs. Put in quotes
|
||
# if the input argument is a pointer, e.g.:
|
||
|
||
# malloc-ARGLIST: "void *"
|
||
|
||
# - Define foo-RET as the type the function returns. Skip if the function
|
||
# returns void. One could even skip foo-ARGLIST if the function does not
|
||
# take any inputs AND the function returns void.
|
||
|
||
|
||
# - Make a file called `foo-inputs` with one input value per line, an input
|
||
# being a comma separated list of arguments to be passed into the function.
|
||
# See pow-inputs for an example.
|
||
|
||
subdir := benchtests
|
||
bench := exp pow rint sin cos tan atan modf \
|
||
slowexp slowpow slowsin slowcos slowtan slowatan
|
||
|
||
# exp function fast path
|
||
exp-ITER = 5e8
|
||
exp-ARGLIST = double
|
||
exp-RET = double
|
||
LDFLAGS-bench-exp = -lm
|
||
|
||
# pow function fast path
|
||
pow-ITER = 2e8
|
||
pow-ARGLIST = double:double
|
||
pow-RET = double
|
||
LDFLAGS-bench-pow = -lm
|
||
|
||
rint-ITER = 250000000
|
||
rint-ARGLIST = double
|
||
rint-RET = double
|
||
LDFLAGS-bench-rint = -lm
|
||
|
||
# exp function slowest path
|
||
slowexp-ITER = 3e5
|
||
slowexp-ARGLIST = double
|
||
slowexp-RET = double
|
||
slowexp-INCLUDE = slowexp.c
|
||
LDFLAGS-bench-slowexp = -lm
|
||
|
||
# sin function fast path
|
||
sin-ITER = 3e9
|
||
sin-ARGLIST = double
|
||
sin-RET = double
|
||
LDFLAGS-bench-sin = -lm
|
||
|
||
# cos function fast path
|
||
cos-ITER = 3e9
|
||
cos-ARGLIST = double
|
||
cos-RET = double
|
||
LDFLAGS-bench-cos = -lm
|
||
|
||
# tan function fast path
|
||
tan-ITER = 3e9
|
||
tan-ARGLIST = double
|
||
tan-RET = double
|
||
LDFLAGS-bench-tan = -lm
|
||
|
||
# atan function fast path
|
||
atan-ITER = 6e9
|
||
atan-ARGLIST = double
|
||
atan-RET = double
|
||
LDFLAGS-bench-atan = -lm
|
||
|
||
# pow function slowest path
|
||
slowpow-ITER = 1e5
|
||
slowpow-ARGLIST = double:double
|
||
slowpow-RET = double
|
||
slowpow-INCLUDE = slowpow.c
|
||
LDFLAGS-bench-slowpow = -lm
|
||
|
||
# sin function slowest path
|
||
slowsin-ITER = 3e7
|
||
slowsin-ARGLIST = double
|
||
slowsin-RET = double
|
||
slowsin-INCLUDE = slowsin.c
|
||
LDFLAGS-bench-slowsin = -lm
|
||
|
||
# cos function slowest path
|
||
slowcos-ITER = 3e7
|
||
slowcos-ARGLIST = double
|
||
slowcos-RET = double
|
||
slowcos-INCLUDE = slowcos.c
|
||
LDFLAGS-bench-slowcos = -lm
|
||
|
||
# tan function slowest path
|
||
slowtan-ITER = 3e7
|
||
slowtan-ARGLIST = double
|
||
slowtan-RET = double
|
||
slowtan-INCLUDE = slowtan.c
|
||
LDFLAGS-bench-slowtan = -lm
|
||
|
||
# atan function slowest path
|
||
slowatan-ITER = 3e8
|
||
slowatan-ARGLIST = double
|
||
slowatan-RET = double
|
||
slowatan-INCLUDE = slowatan.c
|
||
LDFLAGS-bench-slowatan = -lm
|
||
|
||
|
||
|
||
# Rules to build and execute the benchmarks. Do not put any benchmark
|
||
# parameters beyond this point.
|
||
|
||
include ../Makeconfig
|
||
include ../Rules
|
||
|
||
binaries-bench := $(addprefix $(objpfx)bench-,$(bench))
|
||
|
||
# This makes sure CPPFLAGS-nonlib and CFLAGS-nonlib are passed
|
||
# for all these modules.
|
||
cpp-srcs-left := $(binaries-bench:=.c)
|
||
lib := nonlib
|
||
include $(patsubst %,$(..)cppflags-iterator.mk,$(cpp-srcs-left))
|
||
|
||
bench-deps := bench-skeleton.c Makefile
|
||
|
||
run-bench = $(test-wrapper-env) \
|
||
GCONV_PATH=$(common-objpfx)iconvdata LC_ALL=C \
|
||
$($*-ENV) $(rtld-prefix) $${run}
|
||
|
||
bench-clean:
|
||
rm -f $(binaries-bench) $(addsuffix .o,$(binaries-bench))
|
||
|
||
bench: $(binaries-bench)
|
||
{ for run in $^; do \
|
||
echo "Running $${run}" >&2; \
|
||
$(run-bench); \
|
||
done; } > $(objpfx)bench.out-tmp; \
|
||
if [ -f $(objpfx)bench.out ]; then \
|
||
mv -f $(objpfx)bench.out $(objpfx)bench.out.old; \
|
||
fi; \
|
||
mv -f $(objpfx)bench.out-tmp $(objpfx)bench.out
|
||
|
||
$(binaries-bench): %: %.o \
|
||
$(sort $(filter $(common-objpfx)lib%,$(link-libc))) \
|
||
$(addprefix $(csu-objpfx),start.o) $(+preinit) $(+postinit)
|
||
$(+link)
|
||
|
||
$(objpfx)bench-%.c: %-inputs $(bench-deps)
|
||
{ if [ -n "$($*-INCLUDE)" ]; then \
|
||
cat $($*-INCLUDE); \
|
||
fi; \
|
||
$(..)scripts/bench.pl $(patsubst %-inputs,%,$<) \
|
||
$($*-ITER) $($*-ARGLIST) $($*-RET); } > $@-tmp
|
||
mv -f $@-tmp $@
|