# Copyright (C) 2013-2023 Free Software Foundation, Inc.
# Copyright The GNU Toolchain Authors.
# 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
# <https://www.gnu.org/licenses/>.

# Makefile for benchmark tests.  The only useful target here is `bench`.
# Add benchmark functions in alphabetical order.

subdir := benchtests

include ../Makeconfig

bench-math := \
  acos \
  acosh \
  asin \
  asinh \
  atan \
  atan2 \
  atanh \
  cbrt \
  cos \
  cosf \
  cosh \
  erf \
  erfc \
  exp \
  exp10 \
  exp10f \
  exp2 \
  exp2f \
  expf \
  expm1 \
  fmax \
  fmaxf \
  fmin \
  fminf \
  fmod \
  fmodf \
  hypot \
  hypotf \
  ilogb \
  ilogbf \
  isfinite \
  isinf \
  isnan \
  j0 \
  j1 \
  lgamma \
  log \
  log10 \
  log1p \
  log2 \
  log2f \
  logb \
  logbf \
  logf \
  modf \
  pow \
  powf \
  rint \
  roundeven \
  roundevenf \
  sin \
  sincos \
  sincosf \
  sinf \
  sinh \
  sqrt \
  tan \
  tanh \
  tgamma \
  trunc \
  truncf \
  y0 \
  y1 \
  # bench-math

ifneq (,$(filter yes,$(float96-fcts)))
bench-math += \
  cbrtl \
  # bench-math
endif

ifneq (,$(filter yes,$(float128-fcts) $(float128-alias-fcts)))
bench-math += \
  expf128 \
  ilogbf128 \
  powf128 \
  sinf128 \
  # bench-math
endif

bench-pthread := \
  pthread-locks \
  pthread-mutex-lock \
  pthread-mutex-trylock \
  pthread-spin-lock \
  pthread-spin-trylock \
  pthread_once \
  thread_create \
  # bench-pthread

LDLIBS-bench-pthread-mutex-lock += -lm
LDLIBS-bench-pthread-mutex-trylock += -lm
LDLIBS-bench-pthread-spin-lock += -lm
LDLIBS-bench-pthread-spin-trylock += -lm

bench-string := \
  ffs \
  ffsll \
  # bench-string

# String function benchmarks.
string-benchset := \
  bzero \
  bzero-large \
  bzero-walk \
  memccpy \
  memchr \
  memcmp \
  memcmpeq \
  memcpy \
  memcpy-large \
  memcpy-random \
  memcpy-walk \
  memmem \
  memmove \
  memmove-large \
  memmove-walk \
  mempcpy \
  memrchr \
  memset \
  memset-large \
  memset-walk \
  memset-zero \
  memset-zero-large \
  memset-zero-walk \
  rawmemchr \
  stpcpy \
  stpcpy_chk \
  stpncpy \
  strcasecmp \
  strcasestr \
  strcat \
  strchr \
  strchrnul \
  strcmp \
  strcoll \
  strcpy \
  strcpy_chk \
  strcspn \
  strlen \
  strncasecmp \
  strncat \
  strncmp \
  strncpy \
  strnlen \
  strpbrk \
  strrchr \
  strsep \
  strspn \
  strstr \
  strtok \
  # string-benchset

# Build and run locale-dependent benchmarks only if we're building natively.
ifeq (no,$(cross-compiling))
wcsmbs-benchset := \
  wcpcpy \
  wcpncpy \
  wcrtomb \
  wcscat \
  wcschr \
  wcschrnul \
  wcscmp \
  wcscpy \
  wcscspn \
  wcslen \
  wcsncat \
  wcsncmp \
  wcsncpy \
  wcsnlen \
  wcspbrk \
  wcsrchr \
  wcsspn \
  wmemchr \
  wmemcmp \
  wmemset \
  # wcsmbs-benchset
else
wcsmbs-benchset :=
endif

string-benchset-all := $(string-benchset) ${wcsmbs-benchset}

ifeq (no,$(cross-compiling))
# We have to generate locales
LOCALES := \
  ar_SA.UTF-8 \
  cs_CZ.UTF-8 \
  da_DK.UTF-8 \
  el_GR.UTF-8 \
  en_GB.UTF-8 \
  en_US.UTF-8 \
  es_ES.UTF-8 \
  fa_IR.UTF-8 \
  fr_FR.UTF-8 \
  he_IL.UTF-8 \
  hi_IN.UTF-8 \
  hu_HU.UTF-8 \
  is_IS.UTF-8 \
  it_IT.UTF-8 \
  ja_JP.UTF-8 \
  pl_PL.UTF-8 \
  pt_PT.UTF-8 \
  ru_RU.UTF-8 \
  si_LK.UTF-8 \
  sr_RS.UTF-8 \
  sv_SE.UTF-8 \
  tr_TR.UTF-8 \
  vi_VN.UTF-8 \
  zh_CN.UTF-8 \
  # LOCALES
include ../gen-locales.mk
endif

hash-benchset := \
  dl-elf-hash \
  dl-new-hash \
  nss-hash \
  # hash-benchset

stdlib-benchset := \
  arc4random \
  strtod \
  # stdlib-benchset

stdio-common-benchset := sprintf

math-benchset := math-inlines

ifeq (${BENCHSET},)
benchset := \
  $(hash-benchset) \
  $(math-benchset) \
  $(stdio-common-benchset) \
  $(stdlib-benchset) \
  $(string-benchset-all) \
  # benchset
else
benchset := $(foreach B,$(filter %-benchset,${BENCHSET}), ${${B}})
endif

CFLAGS-bench-ffs.c += -fno-builtin
CFLAGS-bench-ffsll.c += -fno-builtin
CFLAGS-bench-sqrt.c += -fno-builtin
CFLAGS-bench-fmin.c += -fno-builtin
CFLAGS-bench-fminf.c += -fno-builtin
CFLAGS-bench-fmax.c += -fno-builtin
CFLAGS-bench-fmaxf.c += -fno-builtin
CFLAGS-bench-trunc.c += -fno-builtin
CFLAGS-bench-truncf.c += -fno-builtin
CFLAGS-bench-roundeven.c += -fno-builtin
CFLAGS-bench-roundevenf.c += -fno-builtin
CFLAGS-bench-isnan.c += $(config-cflags-signaling-nans)
CFLAGS-bench-isinf.c += $(config-cflags-signaling-nans)
CFLAGS-bench-isfinite.c += $(config-cflags-signaling-nans)

ifeq (${BENCHSET},)
bench-malloc := \
  malloc-simple \
  malloc-thread \
  # bench-malloc
else
bench-malloc := $(filter malloc-%,${BENCHSET})
endif

ifeq (${STATIC-BENCHTESTS},yes)
+link-benchtests = $(+link-static-tests)
link-libc-benchtests = $(link-libc-static)
libm-benchtests = $(common-objpfx)math/libm.a
thread-library-benchtests = $(static-thread-library)
else
link-libc-benchtests = $(link-libc)
+link-benchtests = $(+link-tests)
thread-library-benchtests = $(shared-thread-library)
libm-benchtests = $(libm)
endif

$(addprefix $(objpfx)bench-,$(bench-math)): $(libm-benchtests)
$(addprefix $(objpfx)bench-,$(math-benchset)): $(libm-benchtests)
$(addprefix $(objpfx)bench-,$(bench-pthread)): $(thread-library-benchtests)
$(addprefix $(objpfx)bench-,$(bench-malloc)): $(thread-library-benchtests)
$(addprefix $(objpfx)bench-,pthread-locks): $(libm-benchtests)
$(addprefix $(objpfx)bench-,pthread-mutex-locks): $(libm-benchtests)



# Rules to build and execute the benchmarks.  Do not put any benchmark
# parameters beyond this point.

# We don't want the benchmark programs to run in parallel since that could
# affect their performance.
.NOTPARALLEL:

bench-extra-objs = json-lib.o

extra-objs += $(bench-extra-objs)
others-extras = $(bench-extra-objs)

# The default duration: 1 seconds.
ifndef BENCH_DURATION
BENCH_DURATION := 1
endif

CPPFLAGS-nonlib += -DDURATION=$(BENCH_DURATION) -D_ISOMAC

# Use clock_gettime to measure performance of functions.  The default is
# to use the architecture-specific high precision timing instructions.
ifdef USE_CLOCK_GETTIME
CPPFLAGS-nonlib += -DUSE_CLOCK_GETTIME
else
# On x86 processors, use RDTSCP, instead of RDTSC, to measure performance
# of functions.  All x86 processors since 2010 support RDTSCP instruction.
ifdef USE_RDTSCP
CPPFLAGS-nonlib += -DUSE_RDTSCP
endif
endif

DETAILED_OPT :=

ifdef DETAILED
DETAILED_OPT := -d
endif

bench-deps := bench-skeleton.c bench-timing.h Makefile

run-bench = $(test-wrapper-env) \
	    $(run-program-env) \
	    $($*-ENV) $(test-via-rtld-prefix) $${run}

timing-type := $(objpfx)bench-timing-type
extra-objs += bench-timing-type.o

include ../Rules

bench-math += $(bench-libmvec)

ifeq (${BENCHSET},)
bench := \
  $(bench-math) \
  $(bench-pthread) \
  $(bench-string) \
  # bench
else
bench := $(foreach B,$(filter bench-%,${BENCHSET}), ${${B}})
endif

# NB: Use "=" instead of ":=" since sysdeps Makefiles may add more
# benches.
binaries-bench = $(addprefix $(objpfx)bench-,$(bench))
extra-objs += $(addsuffix .o,$(addprefix bench-,$(bench)))
binaries-benchset = $(addprefix $(objpfx)bench-,$(benchset))
extra-objs += $(addsuffix .o,$(addprefix bench-,$(benchset)))
binaries-bench-malloc := $(addprefix $(objpfx)bench-,$(bench-malloc))
extra-objs += $(addsuffix .o,$(addprefix bench-,$(bench-malloc)))

# This makes sure CPPFLAGS-nonlib and CFLAGS-nonlib are passed
# for all these modules.
cpp-srcs-left := \
  $(binaries-bench-malloc:=.c) \
  $(binaries-bench:=.c) \
  $(binaries-benchset:=.c) \
  $(timing-type:=.c) \
  # cpp-srcs-left
lib := nonlib
include $(patsubst %,$(..)libof-iterator.mk,$(cpp-srcs-left))

bench-clean:
	rm -f $(binaries-bench) $(addsuffix .o,$(binaries-bench))
	rm -f $(binaries-benchset) $(addsuffix .o,$(binaries-benchset))
	rm -f $(binaries-bench-malloc) $(addsuffix .o,$(binaries-bench-malloc))
	rm -f $(timing-type) $(addsuffix .o,$(timing-type))
	rm -f $(addprefix $(objpfx),$(bench-extra-objs))

# Validate the passed in BENCHSET
ifneq ($(strip ${BENCHSET}),)
VALIDBENCHSETNAMES := \
  bench-math \
  bench-pthread \
  bench-string \
  hash-benchset \
  malloc-simple \
  malloc-thread \
  math-benchset \
  stdio-common-benchset \
  stdlib-benchset \
  string-benchset \
  wcsmbs-benchset \
  # VALIDBENCHSETNAMES

INVALIDBENCHSETNAMES := $(filter-out ${VALIDBENCHSETNAMES},${BENCHSET})
ifneq (${INVALIDBENCHSETNAMES},)
$(info The following values in BENCHSET are invalid: ${INVALIDBENCHSETNAMES})
$(info The valid ones are: ${VALIDBENCHSETNAMES})
$(error Invalid BENCHSET value)
endif
endif

bench: bench-build bench-set bench-func bench-malloc

# Target to only build the benchmark without running it.  We generate locales
# only if we're building natively.
ifeq (no,$(cross-compiling))
bench-build: $(gen-locales) $(timing-type) $(binaries-bench) \
	$(binaries-benchset) $(binaries-bench-malloc)
else
bench-build: $(timing-type) $(binaries-bench) $(binaries-benchset) \
	$(binaries-bench-malloc)
endif

bench-set: $(binaries-benchset)
	for run in $^; do \
	  echo "Running $${run}"; \
	  $(run-bench) > $${run}.out; \
	done

bench-malloc: $(binaries-bench-malloc)
	for run in $^; do \
	  echo "$${run}"; \
	  if [ `basename $${run}` = "bench-malloc-thread" ]; then \
		for thr in 1 8 16 32; do \
			echo "Running $${run} $${thr}"; \
			$(run-bench) $${thr} > $${run}-$${thr}.out; \
		done;\
	  else \
		for thr in 8 16 32 64 128 256 512 1024 2048 4096; do \
		  echo "Running $${run} $${thr}"; \
		  $(run-bench) $${thr} > $${run}-$${thr}.out; \
		done;\
	  fi;\
	done

# Build and execute the benchmark functions.  This target generates JSON
# formatted bench.out.  Each of the programs produce independent JSON output,
# so one could even execute them individually and process it using any JSON
# capable language or tool.
bench-func: $(binaries-bench)
	if [ -n '$^' ] ; then \
	{ timing_type=$$($(test-wrapper-env) \
			 $(run-program-env) \
			 $(test-via-rtld-prefix) \
			 $(timing-type)); \
	  echo "{\"timing_type\": \"$${timing_type}\","; \
	  echo " \"functions\": {"; \
	  for run in $^; do \
	    op=$$($(run-bench) $(DETAILED_OPT)); \
	    ret=$$?; \
	    case "$${ret}" in \
	      77) \
	      echo "UNSUPPORTED $${run}: $${op}" >&2; \
		;; \
	      0) \
		echo "Running $${run}" >&2; \
		if [ "$${run}" != "$<" ]; then \
		  echo ","; \
		fi; \
		echo "$${op}"; \
		;; \
	      *) \
		echo "FAILED $${run}" >&2; \
		;; \
	    esac; \
	  done; \
	  echo; \
	  echo " }"; \
	  echo "}"; \
	  } > $(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; \
	  $(PYTHON) scripts/validate_benchout.py $(objpfx)bench.out \
	  scripts/benchout.schema.json; \
	fi

ifeq ($(bind-now),yes)
link-bench-bind-now = -Wl,-z,now
endif

bench-link-targets = $(timing-type) $(binaries-bench) $(binaries-benchset) \
	$(binaries-bench-malloc)

$(bench-link-targets): %: %.o $(objpfx)json-lib.o \
	$(link-extra-libs-tests) \
  $(sort $(filter $(common-objpfx)lib%,$(link-libc-benchtests))) \
  $(addprefix $(csu-objpfx),start.o) $(+preinit) $(+postinit)
	$(+link-benchtests)

$(bench-link-targets): LDFLAGS += $(link-bench-bind-now)

$(objpfx)bench-%.c: %-inputs $(bench-deps)
	{ if [ -n "$($*-INCLUDE)" ]; then \
	  cat $($*-INCLUDE); \
	fi; \
	$(PYTHON) scripts/bench.py $(patsubst %-inputs,%,$<); } > $@-tmp
	mv -f $@-tmp $@