Chiebot-Mirror
/
grpc
mirror of https://github.com/grpc/grpc.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity

Prune and update xxhash (#26879)

Browse Source 
* Add xxhash update section

* Update xxhash
pull/26898/head
Tamir Duberstein 3 years ago committed by GitHub
parent 86b08f4fcf
commit c6cb5af87d
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
22 changed files with 89 additions and 3078 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 12
      third_party/README.md
  2. 10
      third_party/xxhash/.gitattributes
  3. 47
      third_party/xxhash/.gitignore
  4. 138
      third_party/xxhash/.travis.yml
  5. 52
      third_party/xxhash/CHANGELOG
  6. 58
      third_party/xxhash/Doxyfile
  7. 506
      third_party/xxhash/Makefile
  8. 235
      third_party/xxhash/README.md
  9. 111
      third_party/xxhash/appveyor.yml
  10. 12
      third_party/xxhash/cmake_unofficial/.gitignore
  11. 173
      third_party/xxhash/cmake_unofficial/CMakeLists.txt
  12. 36
      third_party/xxhash/cmake_unofficial/README.md
  13. 4
      third_party/xxhash/cmake_unofficial/xxHashConfig.cmake.in
  14. 9
      third_party/xxhash/doc/README.md
  15. 206
      third_party/xxhash/doc/xxhash.cry
  16. 317
      third_party/xxhash/doc/xxhash_spec.md
  17. 15
      third_party/xxhash/libxxhash.pc.in
  18. 55
      third_party/xxhash/xxh3.h
  19. 770
      third_party/xxhash/xxh_x86dispatch.c
  20. 86
      third_party/xxhash/xxh_x86dispatch.h
  21. 43
      third_party/xxhash/xxhash.c
  22. 252
      third_party/xxhash/xxhash.h

12
third_party/README.md vendored
Unescape View File

@ -125,3 +125,15 @@ Since upb is vendored in the gRPC repo, you cannot use submodule to update it. P
5. Run `tools/buildgen/generate_projects.sh` to regenerate the generated files
6. Run `tools/codegen/core/gen_upb_api.sh` to regenerate upb files.
If you see breaking changes here, you may want to import upb into Google3 along with gRPC.
### Updating third_party/xxhash
The upstream xxhash repository contains a bunch of files that we don't want, so
we employ a rather manual update flow to pick up just the bits we care about:
```
git remote add xxhash https://github.com/Cyan4973/xxHash.git
git fetch xxhash
for file in LICENSE xxhash.h; do git show xxhash/dev:$file > third_party/xxhash/$file; done
git commit
```

10
third_party/xxhash/.gitattributes vendored
Unescape View File

@ -1,10 +0,0 @@
# Set the default behavior
* text eol=lf
# Explicitly declare source files
*.c text eol=lf
*.h text eol=lf
# Denote files that should not be modified.
*.odt binary

47
third_party/xxhash/.gitignore vendored
Unescape View File

@ -1,47 +0,0 @@
# objects
*.o
*.obj
*.s
# libraries
libxxhash.*
!libxxhash.pc.in
# Executables
*.exe
xxh32sum
xxh64sum
xxh128sum
xxhsum
xxhsum32
xxhsum_privateXXH
xxhsum_inlinedXXH
dispatch
tests/generate_unicode_test
# compilation chain
.clang_complete
# Mac OS-X artefacts
*.dSYM
.DS_Store
# Wasm / emcc / emscripten artefacts
*.html
*.wasm
*.js
# CMake build directories
build*/
# project managers artifacts
.projectile
# analyzer artifacts
infer-out
# test artifacts
.test*
tmp*
tests/*.unicode
tests/unicode_test*

138
third_party/xxhash/.travis.yml vendored
Unescape View File

@ -1,138 +0,0 @@
language: c
# Dump CPU info before start
before_install:
- cat /proc/cpuinfo || echo /proc/cpuinfo is not present
matrix:
fast_finish: true
include:
- name: General linux x64 tests
arch: amd64
addons:
apt:
packages:
- g++-multilib
- gcc-multilib
- cppcheck
script:
- make -B test-all
- make clean
- CFLAGS="-Werror" MOREFLAGS="-Wno-sign-conversion" make dispatch # removing sign conversion warnings due to a bug in gcc-5's definition of some AVX512 intrinsics
- make clean
- CC=g++ CFLAGS="-O1 -mavx512f -Werror" make
- make clean
- CC=g++ CFLAGS="-Wall -Wextra -Werror" make DISPATCH=1
- name: Check results consistency on x64
arch: amd64
script:
- CPPFLAGS=-DXXH_VECTOR=XXH_SCALAR make check # Scalar code path
- make clean
- CPPFLAGS=-DXXH_VECTOR=XXH_SSE2 make check # SSE2 code path
- make clean
- CPPFLAGS="-mavx2 -DXXH_VECTOR=XXH_AVX2" make check # AVX2 code path
- make clean
- CPPFLAGS="-mavx512f -DXXH_VECTOR=XXH_AVX512" make check # AVX512 code path
- make clean
- CPPFLAGS=-DXXH_REROLL=1 make check # reroll code path (#240)
- make -C tests/bench
- name: macOS General Test
os: osx
compiler: clang
script:
- CFLAGS="-Werror" make # test library build
- make clean
- make test MOREFLAGS='-Werror' | tee # test scenario where `stdout` is not the console
- name: ARM compilation and consistency checks (Qemu)
dist: xenial
arch: amd64
addons:
apt:
packages:
- qemu-system-arm
- qemu-user-static
- gcc-arm-linux-gnueabi
- libc6-dev-armel-cross
script:
# arm (32-bit)
- CC=arm-linux-gnueabi-gcc CPPFLAGS=-DXXH_VECTOR=XXH_SCALAR LDFLAGS=-static RUN_ENV=qemu-arm-static make check # Scalar code path
- make clean
# NEON (32-bit)
- CC=arm-linux-gnueabi-gcc CPPFLAGS=-DXXH_VECTOR=XXH_NEON CFLAGS="-O3 -march=armv7-a -fPIC -mfloat-abi=softfp -mfpu=neon-vfpv4" LDFLAGS=-static RUN_ENV=qemu-arm-static make check # NEON code path
- name: aarch64 compilation and consistency checks
dist: xenial
arch: arm64
script:
# aarch64
- CPPFLAGS=-DXXH_VECTOR=XXH_SCALAR make check # Scalar code path
# NEON (64-bit)
- make clean
- CPPFLAGS=-DXXH_VECTOR=XXH_NEON make check # NEON code path
# clang
- make clean
- CC=clang CPPFLAGS=-DXXH_VECTOR=XXH_SCALAR make check # Scalar code path
# clang + NEON
- make clean
- CC=clang CPPFLAGS=-DXXH_VECTOR=XXH_NEON make check # NEON code path
# We need Bionic here because the QEMU versions shipped in the older repos
# do not support POWER8 emulation, and compiling QEMU from source is a pain.
- name: PowerPC + PPC64 compilation and consistency checks (Qemu on Bionic)
dist: bionic
arch: amd64
addons:
apt:
packages:
- qemu-system-ppc
- qemu-user-static
- gcc-powerpc-linux-gnu
- gcc-powerpc64-linux-gnu
- libc6-dev-powerpc-cross
- libc6-dev-ppc64-cross
script:
- CC=powerpc-linux-gnu-gcc RUN_ENV=qemu-ppc-static LDFLAGS=-static make check # Scalar code path
- make clean
- CC=powerpc64-linux-gnu-gcc RUN_ENV=qemu-ppc64-static CPPFLAGS=-DXXH_VECTOR=XXH_SCALAR CFLAGS="-O3" LDFLAGS="-static -m64" make check # Scalar code path
# VSX code
- make clean
- CC=powerpc64-linux-gnu-gcc RUN_ENV="qemu-ppc64-static -cpu power8" CPPFLAGS=-DXXH_VECTOR=XXH_VSX CFLAGS="-O3 -maltivec -mvsx -mcpu=power8 -mpower8-vector" LDFLAGS="-static -m64" make check # VSX code path
# altivec.h redefinition issue #426
- make clean
- CC=powerpc64-linux-gnu-gcc CPPFLAGS=-DXXH_VECTOR=XXH_VSX CFLAGS="-maltivec -mvsx -mcpu=power8 -mpower8-vector" make -C tests test_ppc_redefine
- name: PPC64LE compilation and consistency checks
dist: xenial
arch: ppc64le
script:
# Scalar (universal) code path
- CPPFLAGS=-DXXH_VECTOR=XXH_SCALAR LDFLAGS=-static make check
# VSX code path (64-bit)
- make clean
- CPPFLAGS=-DXXH_VECTOR=XXH_VSX CFLAGS="-O3 -maltivec -mvsx -mpower8-vector -mcpu=power8" LDFLAGS="-static" make check
# altivec.h redefinition issue #426
- make clean
- CPPFLAGS=-DXXH_VECTOR=XXH_VSX CFLAGS="-maltivec -mvsx -mcpu=power8 -mpower8-vector" make -C tests test_ppc_redefine
- name: IBM s390x compilation and consistency checks
dist: bionic
arch: s390x
script:
# Scalar (universal) code path
- CPPFLAGS=-DXXH_VECTOR=XXH_SCALAR LDFLAGS=-static make check
# s390x code path (64-bit)
- make clean
- CPPFLAGS=-DXXH_VECTOR=XXH_VSX CFLAGS="-O3 -march=arch11 -mzvector" LDFLAGS="-static" make check
- name: cmake build test
script:
- cd cmake_unofficial
- mkdir build
- cd build
- cmake ..
- CFLAGS=-Werror make

52
third_party/xxhash/CHANGELOG vendored
Unescape View File

@ -1,52 +0,0 @@
v0.8.0
- api : stabilize XXH3
- cli : xxhsum can parse BSD-style --check lines, by @WayneD
- cli : `xxhsum -` accepts console input, requested by @jaki
- cli : xxhsum accepts -- separator, by @jaki
- cli : fix : print correct default algo for symlinked helpers, by @martinetd
- install: improved pkgconfig script, allowing custom install locations, requested by @ellert
v0.7.4
- perf: automatic vector detection and selection at runtime (`xxh_x86dispatch.h`), initiated by @easyaspi314
- perf: added AVX512 support, by @gzm55
- api : new: secret generator `XXH_generateSecret()`, suggested by @koraa
- api : fix: XXH3_state_t is movable, identified by @koraa
- api : fix: state is correctly aligned in AVX mode (unlike `malloc()`), by @easyaspi314
- api : fix: streaming generated wrong values in some combination of random ingestion lengths, reported by @WayneD
- cli : fix unicode print on Windows, by @easyaspi314
- cli : can `-c` check file generated by sfv
- build: `make DISPATCH=1` generates `xxhsum` and `libxxhash` with runtime vector detection (x86/x64 only)
- install: cygwin installation support
- doc : Cryptol specification of XXH32 and XXH64, by @weaversa
v0.7.3
- perf: improved speed for large inputs (~+20%)
- perf: improved latency for small inputs (~10%)
- perf: s390x Vectorial code, by @easyaspi314
- cli: improved support for Unicode filenames on Windows, thanks to @easyaspi314 and @t-mat
- api: `xxhash.h` can now be included in any order, with and without `XXH_STATIC_LINKING_ONLY` and `XXH_INLINE_ALL`
- build: xxHash's implementation transferred into `xxhash.h`. No more need to have `xxhash.c` in the `/include` directory for `XXH_INLINE_ALL` to work
- install: created pkg-config file, by @bket
- install: VCpkg installation instructions, by @LilyWangL
- doc: Highly improved code documentation, by @easyaspi314
- misc: New test tool in `/tests/collisions`: brute force collision tester for 64-bit hashes
v0.7.2
- Fixed collision ratio of `XXH128` for some specific input lengths, reported by @svpv
- Improved `VSX` and `NEON` variants, by @easyaspi314
- Improved performance of scalar code path (`XXH_VECTOR=0`), by @easyaspi314
- `xxhsum`: can generate 128-bit hashes with the `-H2` option (note: for experimental purposes only! `XXH128` is not yet frozen)
- `xxhsum`: option `-q` removes status notifications
v0.7.1
- Secret first: the algorithm computation can be altered by providing a "secret", which is any blob of bytes, of size >= `XXH3_SECRET_SIZE_MIN`.
- `seed` is still available, and acts as a secret generator
- updated `ARM NEON` variant by @easyaspi314
- Streaming implementation is available
- Improve compatibility and performance with Visual Studio, with help from @aras-p
- Better integration when using `XXH_INLINE_ALL`: do not pollute host namespace, use its own macros, such as `XXH_ASSERT()`, `XXH_ALIGN`, etc.
- 128-bit variant provides helper functions for comparison of hashes.
- Better `clang` generation of `rotl` instruction, thanks to @easyaspi314
- `XXH_REROLL` build macro to reduce binary size, by @easyaspi314
- Improved `cmake` script, by @Mezozoysky
- Full benchmark program provided in `/tests/bench`

58
third_party/xxhash/Doxyfile vendored
Unescape View File

@ -1,58 +0,0 @@
# Doxygen config for xxHash
DOXYFILE_ENCODING = UTF-8
PROJECT_NAME = "xxHash"
PROJECT_NUMBER = "0.8.0"
PROJECT_BRIEF = "Extremely fast non-cryptographic hash function"
OUTPUT_DIRECTORY = doxygen
OUTPUT_LANGUAGE = English
# We already separate the internal docs.
INTERNAL_DOCS = YES
# Consistency
SORT_MEMBER_DOCS = NO
BRIEF_MEMBER_DESC = YES
REPEAT_BRIEF = YES
# Warnings
QUIET = YES
# Until we document everything
WARN_IF_UNDOCUMENTED = NO
# TODO: Add the other files. It is just xxhash.h for now.
FILE_PATTERNS = xxhash.h xxh_x86dispatch.c
# Note: xxHash's source files are technically ASCII only.
INPUT_ENCODING = UTF-8
TAB_SIZE = 4
MARKDOWN_SUPPORT = YES
# xxHash is a C library
OPTIMIZE_OUTPUT_FOR_C = YES
# So we can document the internals
EXTRACT_STATIC = YES
# Document the macros
MACRO_EXPANSION = YES
EXPAND_ONLY_PREDEF = YES
# Predefine some macros to clean up the output.
PREDEFINED = "XXH_DOXYGEN=" \
"XXH_PUBLIC_API=" \
"XXH_FORCE_INLINE=static inline" \
"XXH_NO_INLINE=static" \
"XXH_RESTRICT=restrict" \
"XSUM_API=" \
"XXH_STATIC_LINKING_ONLY" \
"XXH_IMPLEMENTATION" \
"XXH_ALIGN(N)=alignas(N)" \
"XXH_ALIGN_MEMBER(align,type)=alignas(align) type"
# We want HTML docs
GENERATE_HTML = YES
HTML_OUTPUT = html
HTML_FILE_EXTENSION = .html
# Tweak the colors a bit
HTML_COLORSTYLE_HUE = 220
HTML_COLORSTYLE_GAMMA = 100
HTML_COLORSTYLE_SAT = 100
# We don't want LaTeX.
GENERATE_LATEX = NO

506
third_party/xxhash/Makefile vendored
Unescape View File

@ -1,506 +0,0 @@
# ################################################################
# xxHash Makefile
# Copyright (C) 2012-2020 Yann Collet
#
# GPL v2 License
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program 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 General Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
#
# You can contact the author at:
# - xxHash homepage: https://www.xxhash.com
# - xxHash source repository: https://github.com/Cyan4973/xxHash
# ################################################################
# xxhsum: provides 32/64 bits hash of one or multiple files, or stdin
# ################################################################
Q = $(if $(filter 1,$(V) $(VERBOSE)),,@)
# Version numbers
SED ?= sed
SED_ERE_OPT ?= -E
LIBVER_MAJOR_SCRIPT:=`$(SED) -n '/define XXH_VERSION_MAJOR/s/.*[[:blank:]]\([0-9][0-9]*\).*/\1/p' < xxhash.h`
LIBVER_MINOR_SCRIPT:=`$(SED) -n '/define XXH_VERSION_MINOR/s/.*[[:blank:]]\([0-9][0-9]*\).*/\1/p' < xxhash.h`
LIBVER_PATCH_SCRIPT:=`$(SED) -n '/define XXH_VERSION_RELEASE/s/.*[[:blank:]]\([0-9][0-9]*\).*/\1/p' < xxhash.h`
LIBVER_MAJOR := $(shell echo $(LIBVER_MAJOR_SCRIPT))
LIBVER_MINOR := $(shell echo $(LIBVER_MINOR_SCRIPT))
LIBVER_PATCH := $(shell echo $(LIBVER_PATCH_SCRIPT))
LIBVER := $(LIBVER_MAJOR).$(LIBVER_MINOR).$(LIBVER_PATCH)
CFLAGS ?= -O3
DEBUGFLAGS+=-Wall -Wextra -Wconversion -Wcast-qual -Wcast-align -Wshadow \
-Wstrict-aliasing=1 -Wswitch-enum -Wdeclaration-after-statement \
-Wstrict-prototypes -Wundef -Wpointer-arith -Wformat-security \
-Wvla -Wformat=2 -Winit-self -Wfloat-equal -Wwrite-strings \
-Wredundant-decls -Wstrict-overflow=2
CFLAGS += $(DEBUGFLAGS) $(MOREFLAGS)
FLAGS = $(CFLAGS) $(CPPFLAGS)
XXHSUM_VERSION = $(LIBVER)
UNAME := $(shell uname)
# Define *.exe as extension for Windows systems
ifneq (,$(filter Windows%,$(OS)))
EXT =.exe
else
EXT =
endif
# OS X linker doesn't support -soname, and use different extension
# see: https://developer.apple.com/library/mac/documentation/DeveloperTools/Conceptual/DynamicLibraries/100-Articles/DynamicLibraryDesignGuidelines.html
ifeq ($(UNAME), Darwin)
SHARED_EXT = dylib
SHARED_EXT_MAJOR = $(LIBVER_MAJOR).$(SHARED_EXT)
SHARED_EXT_VER = $(LIBVER).$(SHARED_EXT)
SONAME_FLAGS = -install_name $(LIBDIR)/libxxhash.$(SHARED_EXT_MAJOR) -compatibility_version $(LIBVER_MAJOR) -current_version $(LIBVER)
else
SONAME_FLAGS = -Wl,-soname=libxxhash.$(SHARED_EXT).$(LIBVER_MAJOR)
SHARED_EXT = so
SHARED_EXT_MAJOR = $(SHARED_EXT).$(LIBVER_MAJOR)
SHARED_EXT_VER = $(SHARED_EXT).$(LIBVER)
endif
LIBXXH = libxxhash.$(SHARED_EXT_VER)
XXHSUM_SRC_DIR = cli
XXHSUM_SPLIT_SRCS = $(XXHSUM_SRC_DIR)/xsum_os_specific.c \
$(XXHSUM_SRC_DIR)/xsum_output.c \
$(XXHSUM_SRC_DIR)/xsum_sanity_check.c
XXHSUM_SPLIT_OBJS = $(XXHSUM_SPLIT_SRCS:.c=.o)
XXHSUM_HEADERS = $(XXHSUM_SRC_DIR)/xsum_config.h \
$(XXHSUM_SRC_DIR)/xsum_arch.h \
$(XXHSUM_SRC_DIR)/xsum_os_specific.h \
$(XXHSUM_SRC_DIR)/xsum_output.h \
$(XXHSUM_SRC_DIR)/xsum_sanity_check.h
## generate CLI and libraries in release mode (default for `make`)
.PHONY: default
default: DEBUGFLAGS=
default: lib xxhsum_and_links
.PHONY: all
all: lib xxhsum xxhsum_inlinedXXH
## xxhsum is the command line interface (CLI)
ifeq ($(DISPATCH),1)
xxhsum: CPPFLAGS += -DXXHSUM_DISPATCH=1
xxhsum: xxh_x86dispatch.o
endif
xxhsum: xxhash.o xxhsum.o $(XXHSUM_SPLIT_OBJS)
$(CC) $(FLAGS) $^ $(LDFLAGS) -o $@$(EXT)
xxhsum32: CFLAGS += -m32 ## generate CLI in 32-bits mode
xxhsum32: xxhash.c xxhsum.c $(XXHSUM_SPLIT_SRCS) ## do not generate object (avoid mixing different ABI)
$(CC) $(FLAGS) $^ $(LDFLAGS) -o $@$(EXT)
## dispatch only works for x86/x64 systems
dispatch: CPPFLAGS += -DXXHSUM_DISPATCH=1
dispatch: xxhash.o xxh_x86dispatch.o xxhsum.c $(XXHSUM_SPLIT_SRCS)
$(CC) $(FLAGS) $^ $(LDFLAGS) -o $@$(EXT)
xxhash.o: xxhash.c xxhash.h
xxhsum.o: xxhsum.c $(XXHSUM_HEADERS) \
xxhash.h xxh_x86dispatch.h
xxh_x86dispatch.o: xxh_x86dispatch.c xxh_x86dispatch.h xxhash.h
.PHONY: xxhsum_and_links
xxhsum_and_links: xxhsum xxh32sum xxh64sum xxh128sum
xxh32sum xxh64sum xxh128sum: xxhsum
ln -sf $<$(EXT) $@$(EXT)
xxhsum_inlinedXXH: CPPFLAGS += -DXXH_INLINE_ALL
xxhsum_inlinedXXH: xxhsum.c $(XXHSUM_SPLIT_SRCS)
$(CC) $(FLAGS) $< -o $@$(EXT)
# library
libxxhash.a: ARFLAGS = rcs
libxxhash.a: xxhash.o
$(AR) $(ARFLAGS) $@ $^
$(LIBXXH): LDFLAGS += -shared
ifeq (,$(filter Windows%,$(OS)))
$(LIBXXH): CFLAGS += -fPIC
endif
ifeq ($(DISPATCH),1)
$(LIBXXH): xxh_x86dispatch.c
endif
$(LIBXXH): xxhash.c
$(CC) $(FLAGS) $^ $(LDFLAGS) $(SONAME_FLAGS) -o $@
ln -sf $@ libxxhash.$(SHARED_EXT_MAJOR)
ln -sf $@ libxxhash.$(SHARED_EXT)
.PHONY: libxxhash
libxxhash: ## generate dynamic xxhash library
libxxhash: $(LIBXXH)
.PHONY: lib
lib: ## generate static and dynamic xxhash libraries
lib: libxxhash.a libxxhash
# helper targets
AWK = awk
GREP = grep
SORT = sort
.PHONY: list
list: ## list all Makefile targets
$(Q)$(MAKE) -pRrq -f $(lastword $(MAKEFILE_LIST)) : 2>/dev/null | $(AWK) -v RS= -F: '/^# File/,/^# Finished Make data base/ {if ($$1 !~ "^[#.]") {print $$1}}' | $(SORT) | egrep -v -e '^[^[:alnum:]]' -e '^$@$$' | xargs
.PHONY: help
help: ## list documented targets
$(Q)$(GREP) -E '^[0-9a-zA-Z_-]+:.*?## .*$$' $(MAKEFILE_LIST) | \
$(SORT) | \
$(AWK) 'BEGIN {FS = ":.*?## "}; {printf "\033[36m%-30s\033[0m %s\n", $$1, $$2}'
.PHONY: clean
clean: ## remove all build artifacts
$(Q)$(RM) -r *.dSYM # Mac OS-X specific
$(Q)$(RM) core *.o *.obj *.$(SHARED_EXT) *.$(SHARED_EXT).* *.a libxxhash.pc
$(Q)$(RM) xxhsum$(EXT) xxhsum32$(EXT) xxhsum_inlinedXXH$(EXT) dispatch$(EXT)
$(Q)$(RM) xxh32sum$(EXT) xxh64sum$(EXT) xxh128sum$(EXT)
$(Q)$(RM) $(XXHSUM_SRC_DIR)/*.o $(XXHSUM_SRC_DIR)/*.obj
@echo cleaning completed
# =================================================
# tests
# =================================================
# make check can be run with cross-compiled binaries on emulated environments (qemu user mode)
# by setting $(RUN_ENV) to the target emulation environment
.PHONY: check
check: xxhsum ## basic tests for xxhsum CLI, set RUN_ENV for emulated environments
# stdin
$(RUN_ENV) ./xxhsum$(EXT) < xxhash.c
# multiple files
$(RUN_ENV) ./xxhsum$(EXT) xxhash.* xxhsum.*
# internal bench
$(RUN_ENV) ./xxhsum$(EXT) -bi0
# long bench command
$(RUN_ENV) ./xxhsum$(EXT) --benchmark-all -i0
# bench multiple variants
$(RUN_ENV) ./xxhsum$(EXT) -b1,2,3 -i0
# file bench
$(RUN_ENV) ./xxhsum$(EXT) -bi0 xxhash.c
# 32-bit
$(RUN_ENV) ./xxhsum$(EXT) -H0 xxhash.c
# 128-bit
$(RUN_ENV) ./xxhsum$(EXT) -H2 xxhash.c
# request incorrect variant
$(RUN_ENV) ./xxhsum$(EXT) -H9 xxhash.c ; test $$? -eq 1
@printf "\n ....... checks completed successfully ....... \n"
.PHONY: test-unicode
test-unicode:
$(MAKE) -C tests test_unicode
.PHONY: test-mem
VALGRIND = valgrind --leak-check=yes --error-exitcode=1
test-mem: RUN_ENV = $(VALGRIND)
test-mem: xxhsum check
.PHONY: test32
test32: clean xxhsum32
@echo ---- test 32-bit ----
./xxhsum32 -bi1 xxhash.c
.PHONY: test-xxhsum-c
test-xxhsum-c: xxhsum
# xxhsum to/from pipe
./xxhsum xxh* | ./xxhsum -c -
./xxhsum -H0 xxh* | ./xxhsum -c -
# xxhsum -q does not display "Loading" message into stderr (#251)
! ./xxhsum -q xxh* 2>&1 | grep Loading
# xxhsum does not display "Loading" message into stderr either
! ./xxhsum xxh* 2>&1 | grep Loading
# Check that xxhsum do display filename that it failed to open.
LC_ALL=C ./xxhsum nonexistent 2>&1 | grep "Error: Could not open 'nonexistent'"
# xxhsum to/from file, shell redirection
./xxhsum xxh* > .test.xxh64
./xxhsum --tag xxh* > .test.xxh64_tag
./xxhsum --little-endian xxh* > .test.le_xxh64
./xxhsum --tag --little-endian xxh* > .test.le_xxh64_tag
./xxhsum -H0 xxh* > .test.xxh32
./xxhsum -H0 --tag xxh* > .test.xxh32_tag
./xxhsum -H0 --little-endian xxh* > .test.le_xxh32
./xxhsum -H0 --tag --little-endian xxh* > .test.le_xxh32_tag
./xxhsum -H2 xxh* > .test.xxh128
./xxhsum -H2 --tag xxh* > .test.xxh128_tag
./xxhsum -H2 --little-endian xxh* > .test.le_xxh128
./xxhsum -H2 --tag --little-endian xxh* > .test.le_xxh128_tag
./xxhsum -c .test.xxh*
./xxhsum -c --little-endian .test.le_xxh*
./xxhsum -c .test.*_tag
# read list of files from stdin
./xxhsum -c < .test.xxh64
./xxhsum -c < .test.xxh32
cat .test.xxh* | ./xxhsum -c -
# check variant with '*' marker as second separator
$(SED) 's/ / \*/' .test.xxh32 | ./xxhsum -c
# bsd-style output
./xxhsum --tag xxhsum* | $(GREP) XXH64
./xxhsum --tag -H0 xxhsum* | $(GREP) XXH32
./xxhsum --tag -H1 xxhsum* | $(GREP) XXH64
./xxhsum --tag -H2 xxhsum* | $(GREP) XXH128
./xxhsum --tag -H32 xxhsum* | $(GREP) XXH32
./xxhsum --tag -H64 xxhsum* | $(GREP) XXH64
./xxhsum --tag -H128 xxhsum* | $(GREP) XXH128
./xxhsum --tag -H0 --little-endian xxhsum* | $(GREP) XXH32_LE
./xxhsum --tag -H1 --little-endian xxhsum* | $(GREP) XXH64_LE
./xxhsum --tag -H2 --little-endian xxhsum* | $(GREP) XXH128_LE
./xxhsum --tag -H32 --little-endian xxhsum* | $(GREP) XXH32_LE
./xxhsum --tag -H64 --little-endian xxhsum* | $(GREP) XXH64_LE
./xxhsum --tag -H128 --little-endian xxhsum* | $(GREP) XXH128_LE
# check bsd-style
./xxhsum --tag xxhsum* | ./xxhsum -c
./xxhsum --tag -H32 --little-endian xxhsum* | ./xxhsum -c
# xxhsum -c warns improperly format lines.
echo '12345678 ' >>.test.xxh32
./xxhsum -c .test.xxh32 | $(GREP) improperly
echo '123456789 file' >>.test.xxh64
./xxhsum -c .test.xxh64 | $(GREP) improperly
# Expects "FAILED"
echo "0000000000000000 LICENSE" | ./xxhsum -c -; test $$? -eq 1
echo "00000000 LICENSE" | ./xxhsum -c -; test $$? -eq 1
# Expects "FAILED open or read"
echo "0000000000000000 test-expects-file-not-found" | ./xxhsum -c -; test $$? -eq 1
echo "00000000 test-expects-file-not-found" | ./xxhsum -c -; test $$? -eq 1
@$(RM) .test.*
.PHONY: armtest
armtest: clean
@echo ---- test ARM compilation ----
CC=arm-linux-gnueabi-gcc MOREFLAGS="-Werror -static" $(MAKE) xxhsum
.PHONY: clangtest
clangtest: clean
@echo ---- test clang compilation ----
CC=clang MOREFLAGS="-Werror -Wconversion -Wno-sign-conversion" $(MAKE) all
.PHONY: cxxtest
cxxtest: clean
@echo ---- test C++ compilation ----
CC="$(CXX) -Wno-deprecated" $(MAKE) all CFLAGS="-O3 -Wall -Wextra -Wundef -Wshadow -Wcast-align -Werror -fPIC"
.PHONY: c90test
ifeq ($(NO_C90_TEST),true)
c90test:
@echo no c90 compatibility test
else
c90test: CPPFLAGS += -DXXH_NO_LONG_LONG
c90test: CFLAGS += -std=c90 -Werror -pedantic
c90test: xxhash.c
@echo ---- test strict C90 compilation [xxh32 only] ----
$(RM) xxhash.o
$(CC) $(FLAGS) $^ $(LDFLAGS) -c
$(RM) xxhash.o
endif
.PHONY: usan
usan: CC=clang
usan: CXX=clang++
usan: ## check CLI runtime for undefined behavior, using clang's sanitizer
@echo ---- check undefined behavior - sanitize ----
$(MAKE) clean
$(MAKE) test CC=$(CC) CXX=$(CXX) MOREFLAGS="-g -fsanitize=undefined -fno-sanitize-recover=all"
.PHONY: staticAnalyze
SCANBUILD ?= scan-build
staticAnalyze: clean ## check C source files using $(SCANBUILD) static analyzer
@echo ---- static analyzer - $(SCANBUILD) ----
CFLAGS="-g -Werror" $(SCANBUILD) --status-bugs -v $(MAKE) all
CPPCHECK ?= cppcheck
.PHONY: cppcheck
cppcheck: ## check C source files using $(CPPCHECK) static analyzer
@echo ---- static analyzer - $(CPPCHECK) ----
$(CPPCHECK) . --force --enable=warning,portability,performance,style --error-exitcode=1 > /dev/null
.PHONY: namespaceTest
namespaceTest: ## ensure XXH_NAMESPACE redefines all public symbols
$(CC) -c xxhash.c
$(CC) -DXXH_NAMESPACE=TEST_ -c xxhash.c -o xxhash2.o
$(CC) xxhash.o xxhash2.o xxhsum.c $(XXHSUM_SPLIT_SRCS) -o xxhsum2 # will fail if one namespace missing (symbol collision)
$(RM) *.o xxhsum2 # clean
MD2ROFF ?= ronn
MD2ROFF_FLAGS ?= --roff --warnings --manual="User Commands" --organization="xxhsum $(XXHSUM_VERSION)"
xxhsum.1: xxhsum.1.md xxhash.h
cat $< | $(MD2ROFF) $(MD2ROFF_FLAGS) | $(SED) -n '/^\.\\\".*/!p' > $@
.PHONY: man
man: xxhsum.1 ## generate man page from markdown source
.PHONY: clean-man
clean-man:
$(RM) xxhsum.1
.PHONY: preview-man
preview-man: man
man ./xxhsum.1
.PHONY: test
test: DEBUGFLAGS += -DXXH_DEBUGLEVEL=1
test: all namespaceTest check test-xxhsum-c c90test test-tools
.PHONY: test-inline
test-inline:
$(MAKE) -C tests test_multiInclude
.PHONY: test-all
test-all: CFLAGS += -Werror
test-all: test test32 clangtest cxxtest usan test-inline listL120 trailingWhitespace test-unicode
.PHONY: test-tools
test-tools:
CFLAGS=-Werror $(MAKE) -C tests/bench
CFLAGS=-Werror $(MAKE) -C tests/collisions
.PHONY: listL120
listL120: # extract lines >= 120 characters in *.{c,h}, by Takayuki Matsuoka (note: $$, for Makefile compatibility)
find . -type f -name '*.c' -o -name '*.h' | while read -r filename; do awk 'length > 120 {print FILENAME "(" FNR "): " $$0}' $$filename; done
.PHONY: trailingWhitespace
trailingWhitespace:
! $(GREP) -E "`printf '[ \\t]$$'`" xxhsum.1 *.c *.h LICENSE Makefile cmake_unofficial/CMakeLists.txt
# =========================================================
# make install is validated only for the following targets
# =========================================================
ifneq (,$(filter Linux Darwin GNU/kFreeBSD GNU Haiku OpenBSD FreeBSD NetBSD DragonFly SunOS CYGWIN% , $(UNAME)))
DESTDIR ?=
# directory variables: GNU conventions prefer lowercase
# see https://www.gnu.org/prep/standards/html_node/Makefile-Conventions.html
# support both lower and uppercase (BSD), use uppercase in script
prefix ?= /usr/local
PREFIX ?= $(prefix)
exec_prefix ?= $(PREFIX)
EXEC_PREFIX ?= $(exec_prefix)
libdir ?= $(EXEC_PREFIX)/lib
LIBDIR ?= $(libdir)
includedir ?= $(PREFIX)/include
INCLUDEDIR ?= $(includedir)
bindir ?= $(EXEC_PREFIX)/bin
BINDIR ?= $(bindir)
datarootdir ?= $(PREFIX)/share
mandir ?= $(datarootdir)/man
man1dir ?= $(mandir)/man1
ifneq (,$(filter $(UNAME),FreeBSD NetBSD DragonFly))
PKGCONFIGDIR ?= $(PREFIX)/libdata/pkgconfig
else
PKGCONFIGDIR ?= $(LIBDIR)/pkgconfig
endif
ifneq (,$(filter $(UNAME),OpenBSD FreeBSD NetBSD DragonFly SunOS))
MANDIR ?= $(PREFIX)/man/man1
else
MANDIR ?= $(man1dir)
endif
ifneq (,$(filter $(UNAME),SunOS))
INSTALL ?= ginstall
else
INSTALL ?= install
endif
INSTALL_PROGRAM ?= $(INSTALL)
INSTALL_DATA ?= $(INSTALL) -m 644
PCLIBDIR ?= $(shell echo "$(LIBDIR)" | $(SED) -n $(SED_ERE_OPT) -e "s@^$(EXEC_PREFIX)(/|$$)@@p")
PCINCDIR ?= $(shell echo "$(INCLUDEDIR)" | $(SED) -n $(SED_ERE_OPT) -e "s@^$(PREFIX)(/|$$)@@p")
PCEXECDIR?= $(if $(filter $(PREFIX),$(EXEC_PREFIX)),$$\{prefix\},$(EXEC_PREFIX))
ifeq (,$(PCLIBDIR))
# Additional prefix check is required, since the empty string is technically a
# valid PCLIBDIR
ifeq (,$(shell echo "$(LIBDIR)" | $(SED) -n $(SED_ERE_OPT) -e "\\@^$(EXEC_PREFIX)(/|$$)@ p"))
$(error configured libdir ($(LIBDIR)) is outside of exec_prefix ($(EXEC_PREFIX)), can't generate pkg-config file)
endif
endif
ifeq (,$(PCINCDIR))
# Additional prefix check is required, since the empty string is technically a
# valid PCINCDIR
ifeq (,$(shell echo "$(INCLUDEDIR)" | $(SED) -n $(SED_ERE_OPT) -e "\\@^$(PREFIX)(/|$$)@ p"))
$(error configured includedir ($(INCLUDEDIR)) is outside of prefix ($(PREFIX)), can't generate pkg-config file)
endif
endif
libxxhash.pc: libxxhash.pc.in
@echo creating pkgconfig
$(Q)$(SED) $(SED_ERE_OPT) -e 's|@PREFIX@|$(PREFIX)|' \
-e 's|@EXECPREFIX@|$(PCEXECDIR)|' \
-e 's|@LIBDIR@|$(PCLIBDIR)|' \
-e 's|@INCLUDEDIR@|$(PCINCDIR)|' \
-e 's|@VERSION@|$(LIBVER)|' \
$< > $@
.PHONY: install
install: lib libxxhash.pc xxhsum ## install libraries, CLI, links and man page
@echo Installing libxxhash
$(Q)$(INSTALL) -d -m 755 $(DESTDIR)$(LIBDIR)
$(Q)$(INSTALL_DATA) libxxhash.a $(DESTDIR)$(LIBDIR)
$(Q)$(INSTALL_PROGRAM) $(LIBXXH) $(DESTDIR)$(LIBDIR)
$(Q)ln -sf $(LIBXXH) $(DESTDIR)$(LIBDIR)/libxxhash.$(SHARED_EXT_MAJOR)
$(Q)ln -sf $(LIBXXH) $(DESTDIR)$(LIBDIR)/libxxhash.$(SHARED_EXT)
$(Q)$(INSTALL) -d -m 755 $(DESTDIR)$(INCLUDEDIR) # includes
$(Q)$(INSTALL_DATA) xxhash.h $(DESTDIR)$(INCLUDEDIR)
$(Q)$(INSTALL_DATA) xxh3.h $(DESTDIR)$(INCLUDEDIR) # for compatibility, will be removed in v0.9.0
ifeq ($(DISPATCH),1)
$(Q)$(INSTALL_DATA) xxh_x86dispatch.h $(DESTDIR)$(INCLUDEDIR)
endif
@echo Installing pkgconfig
$(Q)$(INSTALL) -d -m 755 $(DESTDIR)$(PKGCONFIGDIR)/
$(Q)$(INSTALL_DATA) libxxhash.pc $(DESTDIR)$(PKGCONFIGDIR)/
@echo Installing xxhsum
$(Q)$(INSTALL) -d -m 755 $(DESTDIR)$(BINDIR)/ $(DESTDIR)$(MANDIR)/
$(Q)$(INSTALL_PROGRAM) xxhsum $(DESTDIR)$(BINDIR)/xxhsum
$(Q)ln -sf xxhsum $(DESTDIR)$(BINDIR)/xxh32sum
$(Q)ln -sf xxhsum $(DESTDIR)$(BINDIR)/xxh64sum
$(Q)ln -sf xxhsum $(DESTDIR)$(BINDIR)/xxh128sum
@echo Installing man pages
$(Q)$(INSTALL_DATA) xxhsum.1 $(DESTDIR)$(MANDIR)/xxhsum.1
$(Q)ln -sf xxhsum.1 $(DESTDIR)$(MANDIR)/xxh32sum.1
$(Q)ln -sf xxhsum.1 $(DESTDIR)$(MANDIR)/xxh64sum.1
$(Q)ln -sf xxhsum.1 $(DESTDIR)$(MANDIR)/xxh128sum.1
@echo xxhash installation completed
.PHONY: uninstall
uninstall: ## uninstall libraries, CLI, links and man page
$(Q)$(RM) $(DESTDIR)$(LIBDIR)/libxxhash.a
$(Q)$(RM) $(DESTDIR)$(LIBDIR)/libxxhash.$(SHARED_EXT)
$(Q)$(RM) $(DESTDIR)$(LIBDIR)/libxxhash.$(SHARED_EXT_MAJOR)
$(Q)$(RM) $(DESTDIR)$(LIBDIR)/$(LIBXXH)
$(Q)$(RM) $(DESTDIR)$(INCLUDEDIR)/xxhash.h
$(Q)$(RM) $(DESTDIR)$(INCLUDEDIR)/xxh3.h
$(Q)$(RM) $(DESTDIR)$(INCLUDEDIR)/xxh_x86dispatch.h
$(Q)$(RM) $(DESTDIR)$(PKGCONFIGDIR)/libxxhash.pc
$(Q)$(RM) $(DESTDIR)$(BINDIR)/xxh32sum
$(Q)$(RM) $(DESTDIR)$(BINDIR)/xxh64sum
$(Q)$(RM) $(DESTDIR)$(BINDIR)/xxh128sum
$(Q)$(RM) $(DESTDIR)$(BINDIR)/xxhsum
$(Q)$(RM) $(DESTDIR)$(MANDIR)/xxh32sum.1
$(Q)$(RM) $(DESTDIR)$(MANDIR)/xxh64sum.1
$(Q)$(RM) $(DESTDIR)$(MANDIR)/xxh128sum.1
$(Q)$(RM) $(DESTDIR)$(MANDIR)/xxhsum.1
@echo xxhsum successfully uninstalled
endif

235
third_party/xxhash/README.md vendored
Unescape View File

@ -1,235 +0,0 @@
xxHash - Extremely fast hash algorithm
======================================
xxHash is an Extremely fast Hash algorithm, running at RAM speed limits.
It successfully completes the [SMHasher](https://code.google.com/p/smhasher/wiki/SMHasher) test suite
which evaluates collision, dispersion and randomness qualities of hash functions.
Code is highly portable, and hashes are identical across all platforms (little / big endian).
|Branch |Status |
|------------|---------|
|dev | [![Build Status](https://travis-ci.org/Cyan4973/xxHash.svg?branch=dev)](https://travis-ci.org/Cyan4973/xxHash?branch=dev) |
Benchmarks
-------------------------
The reference system uses an Intel i7-9700K cpu, and runs Ubuntu x64 20.04.
The [open source benchmark program] is compiled with `clang` v10.0 using `-O3` flag.
| Hash Name | Width | Bandwidth (GB/s) | Small Data Velocity | Quality | Comment |
| --------- | ----- | ---------------- | ----- | --- | --- |
| __XXH3__ (SSE2) | 64 | 31.5 GB/s | 133.1 | 10
| __XXH128__ (SSE2) | 128 | 29.6 GB/s | 118.1 | 10
| _RAM sequential read_ | N/A | 28.0 GB/s | N/A | N/A | _for reference_
| City64 | 64 | 22.0 GB/s | 76.6 | 10
| T1ha2 | 64 | 22.0 GB/s | 99.0 | 9 | Slightly worse [collisions]
| City128 | 128 | 21.7 GB/s | 57.7 | 10
| __XXH64__ | 64 | 19.4 GB/s | 71.0 | 10
| SpookyHash | 64 | 19.3 GB/s | 53.2 | 10
| Mum | 64 | 18.0 GB/s | 67.0 | 9 | Slightly worse [collisions]
| __XXH32__ | 32 | 9.7 GB/s | 71.9 | 10
| City32 | 32 | 9.1 GB/s | 66.0 | 10
| Murmur3 | 32 | 3.9 GB/s | 56.1 | 10
| SipHash | 64 | 3.0 GB/s | 43.2 | 10
| FNV64 | 64 | 1.2 GB/s | 62.7 | 5 | Poor avalanche properties
| Blake2 | 256 | 1.1 GB/s | 5.1 | 10 | Cryptographic
| SHA1 | 160 | 0.8 GB/s | 5.6 | 10 | Cryptographic but broken
| MD5 | 128 | 0.6 GB/s | 7.8 | 10 | Cryptographic but broken
[open source benchmark program]: https://github.com/Cyan4973/xxHash/tree/release/tests/bench
[collisions]: https://github.com/Cyan4973/xxHash/wiki/Collision-ratio-comparison#collision-study
note 1: Small data velocity is a _rough_ evaluation of algorithm's efficiency on small data. For more detailed analysis, please refer to next paragraph.
note 2: some algorithms feature _faster than RAM_ speed. In which case, they can only reach their full speed when input data is already in CPU cache (L3 or better). Otherwise, they max out on RAM speed limit.
### Small data
Performance on large data is only one part of the picture.
Hashing is also very useful in constructions like hash tables and bloom filters.
In these use cases, it's frequent to hash a lot of small data (starting at a few bytes).
Algorithm's performance can be very different for such scenarios, since parts of the algorithm,
such as initialization or finalization, become fixed cost.
The impact of branch mis-prediction also becomes much more present.
XXH3 has been designed for excellent performance on both long and small inputs,
which can be observed in the following graph:
![XXH3, latency, random size](https://user-images.githubusercontent.com/750081/61976089-aedeab00-af9f-11e9-9239-e5375d6c080f.png)
For a more detailed analysis, visit the wiki :
https://github.com/Cyan4973/xxHash/wiki/Performance-comparison#benchmarks-concentrating-on-small-data-
Quality
-------------------------
Speed is not the only property that matters.
Produced hash values must respect excellent dispersion and randomness properties,
so that any sub-section of it can be used to maximally spread out a table or index,
as well as reduce the amount of collisions to the minimal theoretical level, following the [birthday paradox].
`xxHash` has been tested with Austin Appleby's excellent SMHasher test suite,
and passes all tests, ensuring reasonable quality levels.
It also passes extended tests from [newer forks of SMHasher], featuring additional scenarios and conditions.
Finally, xxHash provides its own [massive collision tester](https://github.com/Cyan4973/xxHash/tree/dev/tests/collisions),
able to generate and compare billions of hash to test the limits of 64-bit hash algorithms.
On this front too, xxHash features good results, in line with the [birthday paradox].
A more detailed analysis is documented [in the wiki](https://github.com/Cyan4973/xxHash/wiki/Collision-ratio-comparison).
[birthday paradox]: https://en.wikipedia.org/wiki/Birthday_problem
[newer forks of SMHasher]: https://github.com/rurban/smhasher
### Build modifiers
The following macros can be set at compilation time to modify libxxhash's behavior. They are generally disabled by default.
- `XXH_INLINE_ALL`: Make all functions `inline`, with implementations being directly included within `xxhash.h`.
Inlining functions is beneficial for speed on small keys.
It's _extremely effective_ when key length is expressed as _a compile time constant_,
with performance improvements observed in the +200% range .
See [this article](https://fastcompression.blogspot.com/2018/03/xxhash-for-small-keys-impressive-power.html) for details.
- `XXH_PRIVATE_API`: same outcome as `XXH_INLINE_ALL`. Still available for legacy support.
The name underlines that `XXH_*` symbols will not be exported.
- `XXH_NAMESPACE`: Prefixes all symbols with the value of `XXH_NAMESPACE`.
This macro can only use compilable character set.
Useful to evade symbol naming collisions,
in case of multiple inclusions of xxHash's source code.
Client applications still use the regular function names,
as symbols are automatically translated through `xxhash.h`.
- `XXH_FORCE_MEMORY_ACCESS`: The default method `0` uses a portable `memcpy()` notation.
Method `1` uses a gcc-specific `packed` attribute, which can provide better performance for some targets.
Method `2` forces unaligned reads, which is not standards compliant, but might sometimes be the only way to extract better read performance.
Method `3` uses a byteshift operation, which is best for old compilers which don't inline `memcpy()` or big-endian systems without a byteswap instruction
- `XXH_FORCE_ALIGN_CHECK`: Use a faster direct read path when input is aligned.
This option can result in dramatic performance improvement when input to hash is aligned on 32 or 64-bit boundaries,
when running on architectures unable to load memory from unaligned addresses, or suffering a performance penalty from it.
It is (slightly) detrimental on platform with good unaligned memory access performance (same instruction for both aligned and unaligned accesses).
This option is automatically disabled on `x86`, `x64` and `aarch64`, and enabled on all other platforms.
- `XXH_VECTOR` : manually select a vector instruction set (default: auto-selected at compilation time). Available instruction sets are `XXH_SCALAR`, `XXH_SSE2`, `XXH_AVX2`, `XXH_AVX512`, `XXH_NEON` and `XXH_VSX`. Compiler may require additional flags to ensure proper support (for example, `gcc` on linux will require `-mavx2` for AVX2, and `-mavx512f` for AVX512).
- `XXH_NO_PREFETCH` : disable prefetching. XXH3 only.
- `XXH_PREFETCH_DIST` : select prefecting distance. XXH3 only.
- `XXH_NO_INLINE_HINTS`: By default, xxHash uses `__attribute__((always_inline))` and `__forceinline` to improve performance at the cost of code size.
Defining this macro to 1 will mark all internal functions as `static`, allowing the compiler to decide whether to inline a function or not.
This is very useful when optimizing for smallest binary size,
and is automatically defined when compiling with `-O0`, `-Os`, `-Oz`, or `-fno-inline` on GCC and Clang.
This may also increase performance depending on compiler and architecture.
- `XXH_REROLL`: Reduces the size of the generated code by not unrolling some loops.
Impact on performance may vary, depending on platform and algorithm.
- `XXH_ACCEPT_NULL_INPUT_POINTER`: if set to `1`, when input is a `NULL` pointer,
xxHash'd result is the same as a zero-length input
(instead of a dereference segfault).
Adds one branch at the beginning of each hash.
- `XXH_STATIC_LINKING_ONLY`: gives access to the state declaration for static allocation.
Incompatible with dynamic linking, due to risks of ABI changes.
- `XXH_NO_LONG_LONG`: removes compilation of algorithms relying on 64-bit types (XXH3 and XXH64). Only XXH32 will be compiled.
Useful for targets (architectures and compilers) without 64-bit support.
- `XXH_IMPORT`: MSVC specific: should only be defined for dynamic linking, as it prevents linkage errors.
- `XXH_CPU_LITTLE_ENDIAN`: By default, endianess is determined by a runtime test resolved at compile time.
If, for some reason, the compiler cannot simplify the runtime test, it can cost performance.
It's possible to skip auto-detection and simply state that the architecture is little-endian by setting this macro to 1.
Setting it to 0 states big-endian.
For the Command Line Interface `xxhsum`, the following environment variables can also be set :
- `DISPATCH=1` : use `xxh_x86dispatch.c`, to automatically select between `scalar`, `sse2`, `avx2` or `avx512` instruction set at runtime, depending on local host. This option is only valid for `x86`/`x64` systems.
### Building xxHash - Using vcpkg
You can download and install xxHash using the [vcpkg](https://github.com/Microsoft/vcpkg) dependency manager:
git clone https://github.com/Microsoft/vcpkg.git
cd vcpkg
./bootstrap-vcpkg.sh
./vcpkg integrate install
./vcpkg install xxhash
The xxHash port in vcpkg is kept up to date by Microsoft team members and community contributors. If the version is out of date, please [create an issue or pull request](https://github.com/Microsoft/vcpkg) on the vcpkg repository.
### Example
The simplest example calls xxhash 64-bit variant as a one-shot function
generating a hash value from a single buffer, and invoked from a C/C++ program:
```C
#include "xxhash.h"
(...)
XXH64_hash_t hash = XXH64(buffer, size, seed);
}
```
Streaming variant is more involved, but makes it possible to provide data incrementally:
```C
#include "stdlib.h" /* abort() */
#include "xxhash.h"
XXH64_hash_t calcul_hash_streaming(FileHandler fh)
{
/* create a hash state */
XXH64_state_t* const state = XXH64_createState();
if (state==NULL) abort();
size_t const bufferSize = SOME_SIZE;
void* const buffer = malloc(bufferSize);
if (buffer==NULL) abort();
/* Initialize state with selected seed */
XXH64_hash_t const seed = 0; /* or any other value */
if (XXH64_reset(state, seed) == XXH_ERROR) abort();
/* Feed the state with input data, any size, any number of times */
(...)
while ( /* some data left */ ) {
size_t const length = get_more_data(buffer, bufferSize, fh);
if (XXH64_update(state, buffer, length) == XXH_ERROR) abort();
(...)
}
(...)
/* Produce the final hash value */
XXH64_hash_t const hash = XXH64_digest(state);
/* State could be re-used; but in this example, it is simply freed */
free(buffer);
XXH64_freeState(state);
return hash;
}
```
### License
The library files `xxhash.c` and `xxhash.h` are BSD licensed.
The utility `xxhsum` is GPL licensed.
### Other programming languages
Beyond the C reference version,
xxHash is also available from many different programming languages,
thanks to great contributors.
They are [listed here](http://www.xxhash.com/#other-languages).
### Packaging status
Many distributions bundle a package manager
which allows easy xxhash installation as both a `libxxhash` library
and `xxhsum` command line interface.
[![Packaging status](https://repology.org/badge/vertical-allrepos/xxhash.svg)](https://repology.org/project/xxhash/versions)
### Special Thanks
- Takayuki Matsuoka, aka @t-mat, for creating `xxhsum -c` and great support during early xxh releases
- Mathias Westerdahl, aka @JCash, for introducing the first version of `XXH64`
- Devin Hussey, aka @easyaspi314, for incredible low-level optimizations on `XXH3` and `XXH128`

111
third_party/xxhash/appveyor.yml vendored
Unescape View File

@ -1,111 +0,0 @@
#---------------------------------#
# general configuration #
#---------------------------------#
version: 1.0.{build}
max_jobs: 2
#---------------------------------#
# environment configuration #
#---------------------------------#
clone_depth: 2
environment:
matrix:
- COMPILER: "visual"
ARCH: "x64"
TEST_XXHSUM: "true"
- COMPILER: "visual"
ARCH: "x64"
APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2017
TEST_XXHSUM: "true"
- COMPILER: "visual"
ARCH: "Win32"
TEST_XXHSUM: "true"
- COMPILER: "visual"
ARCH: "Win32"
APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2013
TEST_XXHSUM: "true"
- COMPILER: "visual"
ARCH: "ARM"
- COMPILER: "visual"
ARCH: "ARM64"
APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2017
# note: ARM64 is not available with Visual Studio 14 2015, which is default for Appveyor
- COMPILER: "gcc"
PLATFORM: "mingw64"
- COMPILER: "gcc"
PLATFORM: "mingw32"
- COMPILER: "gcc"
PLATFORM: "clang"
install:
- ECHO Installing %COMPILER% %PLATFORM% %ARCH%
- MKDIR bin
- if [%COMPILER%]==[gcc] SET PATH_ORIGINAL=%PATH%
- if [%COMPILER%]==[gcc] (
SET "PATH_MINGW32=c:\MinGW\bin;c:\MinGW\usr\bin" &&
SET "PATH_MINGW64=c:\msys64\mingw64\bin;c:\msys64\usr\bin" &&
COPY C:\MinGW\bin\mingw32-make.exe C:\MinGW\bin\make.exe &&
COPY C:\MinGW\bin\gcc.exe C:\MinGW\bin\cc.exe
)
#---------------------------------#
# build configuration #
#---------------------------------#
build_script:
- if [%PLATFORM%]==[mingw32] SET PATH=%PATH_MINGW32%;%PATH_ORIGINAL%
- if [%PLATFORM%]==[mingw64] SET PATH=%PATH_MINGW64%;%PATH_ORIGINAL%
- if [%PLATFORM%]==[clang] SET PATH=%PATH_MINGW64%;%PATH_ORIGINAL%
- ECHO ***
- ECHO Building %COMPILER% %PLATFORM% %ARCH%
- ECHO ***
- if [%COMPILER%]==[gcc] (
if [%PLATFORM%]==[clang] (
clang -v
) ELSE (
gcc -v
)
)
- if [%COMPILER%]==[gcc] (
echo ----- &&
make -v &&
echo ----- &&
if not [%PLATFORM%]==[clang] (
if [%PLATFORM%]==[mingw32] ( SET CPPFLAGS=-DPOOL_MT=0 ) &&
make -B clean test MOREFLAGS=-Werror
) ELSE (
SET CXXFLAGS=--std=c++14 &&
make -B clean test CC=clang CXX=clang++ MOREFLAGS="--target=x86_64-w64-mingw32 -Werror -Wno-pass-failed" NO_C90_TEST=true
) &&
make -C tests/bench
)
# note 1: strict c90 tests with clang fail, due to (erroneous) presence on `inline` keyword in some included system file
# note 2: multi-threading code doesn't work with mingw32, disabled through POOL_MT=0
# note 3: clang requires C++14 to compile sort because its own code contains c++14-only code
- if [%COMPILER%]==[visual] (
cd cmake_unofficial &&
cmake . -DCMAKE_BUILD_TYPE=Release -A %ARCH% -DXXHASH_C_FLAGS="/WX" &&
cmake --build . --config Release
)
#---------------------------------#
# tests configuration #
#---------------------------------#
test_script:
# note: can only run x86 and x64 binaries on Appveyor
# note: if %COMPILER%==gcc, xxhsum was already tested within `make test`
- if [%TEST_XXHSUM%]==[true] (
ECHO *** &&
ECHO Testing %COMPILER% %PLATFORM% %ARCH% &&
ECHO *** &&
cd Release &&
xxhsum.exe -bi1 &&
ECHO ------- xxhsum tested -------
)
#---------------------------------#
# artifacts configuration #
#---------------------------------#
# none yet

12
third_party/xxhash/cmake_unofficial/.gitignore vendored
Unescape View File

@ -1,12 +0,0 @@
# cmake artifacts
CMakeCache.txt
CMakeFiles
Makefile
cmake_install.cmake
# make compilation results
*.dylib
*.a

173
third_party/xxhash/cmake_unofficial/CMakeLists.txt vendored
Unescape View File

@ -1,173 +0,0 @@
# To the extent possible under law, the author(s) have dedicated all
# copyright and related and neighboring rights to this software to
# the public domain worldwide. This software is distributed without
# any warranty.
#
# For details, see <https://creativecommons.org/publicdomain/zero/1.0/>.
cmake_minimum_required (VERSION 2.8.12 FATAL_ERROR)
set(XXHASH_DIR "${CMAKE_CURRENT_SOURCE_DIR}/..")
file(STRINGS "${XXHASH_DIR}/xxhash.h" XXHASH_VERSION_MAJOR REGEX "^#define XXH_VERSION_MAJOR +([0-9]+) *$")
string(REGEX REPLACE "^#define XXH_VERSION_MAJOR +([0-9]+) *$" "\\1" XXHASH_VERSION_MAJOR "${XXHASH_VERSION_MAJOR}")
file(STRINGS "${XXHASH_DIR}/xxhash.h" XXHASH_VERSION_MINOR REGEX "^#define XXH_VERSION_MINOR +([0-9]+) *$")
string(REGEX REPLACE "^#define XXH_VERSION_MINOR +([0-9]+) *$" "\\1" XXHASH_VERSION_MINOR "${XXHASH_VERSION_MINOR}")
file(STRINGS "${XXHASH_DIR}/xxhash.h" XXHASH_VERSION_RELEASE REGEX "^#define XXH_VERSION_RELEASE +([0-9]+) *$")
string(REGEX REPLACE "^#define XXH_VERSION_RELEASE +([0-9]+) *$" "\\1" XXHASH_VERSION_RELEASE "${XXHASH_VERSION_RELEASE}")
set(XXHASH_VERSION_STRING "${XXHASH_VERSION_MAJOR}.${XXHASH_VERSION_MINOR}.${XXHASH_VERSION_RELEASE}")
set(XXHASH_LIB_VERSION ${XXHASH_VERSION_STRING})
set(XXHASH_LIB_SOVERSION "${XXHASH_VERSION_MAJOR}")
mark_as_advanced(XXHASH_VERSION_MAJOR XXHASH_VERSION_MINOR XXHASH_VERSION_RELEASE XXHASH_VERSION_STRING XXHASH_LIB_VERSION XXHASH_LIB_SOVERSION)
if("${CMAKE_VERSION}" VERSION_LESS "3.13")
#message(WARNING "CMake ${CMAKE_VERSION} has no CMP0077 policy: options will erase uncached/untyped normal vars!")
else()
cmake_policy (SET CMP0077 NEW)
endif()
if("${CMAKE_VERSION}" VERSION_LESS "3.0")
project(xxHash C)
else()
cmake_policy (SET CMP0048 NEW)
project(xxHash
VERSION ${XXHASH_VERSION_STRING}
LANGUAGES C)
endif()
if(NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
set(CMAKE_BUILD_TYPE "Release" CACHE STRING "Project build type" FORCE)
set_property(CACHE CMAKE_BUILD_TYPE
PROPERTY STRINGS "Debug" "Release" "RelWithDebInfo" "MinSizeRel")
endif()
if(NOT CMAKE_CONFIGURATION_TYPES)
message(STATUS "xxHash build type: ${CMAKE_BUILD_TYPE}")
endif()
option(BUILD_SHARED_LIBS "Build shared library" ON)
option(XXHASH_BUILD_XXHSUM "Build the xxhsum binary" ON)
# If XXHASH is being bundled in another project, we don't want to
# install anything. However, we want to let people override this, so
# we'll use the XXHASH_BUNDLED_MODE variable to let them do that; just
# set it to OFF in your project before you add_subdirectory(xxhash/cmake_unofficial).
if(NOT DEFINED XXHASH_BUNDLED_MODE)
if("${PROJECT_SOURCE_DIR}" STREQUAL "${CMAKE_SOURCE_DIR}")
set(XXHASH_BUNDLED_MODE OFF)
else()
set(XXHASH_BUNDLED_MODE ON)
endif()
endif()
set(XXHASH_BUNDLED_MODE ${XXHASH_BUNDLED_MODE} CACHE BOOL "" FORCE)
mark_as_advanced(XXHASH_BUNDLED_MODE)
# Allow people to choose whether to build shared or static libraries
# via the BUILD_SHARED_LIBS option unless we are in bundled mode, in
# which case we always use static libraries.
include(CMakeDependentOption)
CMAKE_DEPENDENT_OPTION(BUILD_SHARED_LIBS "Build shared libraries" ON "NOT XXHASH_BUNDLED_MODE" OFF)
# libxxhash
add_library(xxhash "${XXHASH_DIR}/xxhash.c")
add_library(${PROJECT_NAME}::xxhash ALIAS xxhash)
target_include_directories(xxhash
PUBLIC
$<BUILD_INTERFACE:${XXHASH_DIR}>
$<INSTALL_INTERFACE:include/>)
if (BUILD_SHARED_LIBS)
target_compile_definitions(xxhash PUBLIC XXH_EXPORT)
endif ()
set_target_properties(xxhash PROPERTIES
SOVERSION "${XXHASH_VERSION_STRING}"
VERSION "${XXHASH_VERSION_STRING}")
if(XXHASH_BUILD_XXHSUM)
set(XXHSUM_DIR "${XXHASH_DIR}/cli")
# xxhsum
add_executable(xxhsum "${XXHASH_DIR}/xxhsum.c"
"${XXHSUM_DIR}/xsum_os_specific.c"
"${XXHSUM_DIR}/xsum_output.c"
"${XXHSUM_DIR}/xsum_sanity_check.c"
)
add_executable(${PROJECT_NAME}::xxhsum ALIAS xxhsum)
target_link_libraries(xxhsum PRIVATE xxhash)
target_include_directories(xxhsum PRIVATE "${XXHASH_DIR}")
endif(XXHASH_BUILD_XXHSUM)
# Extra warning flags
include (CheckCCompilerFlag)
if (XXHASH_C_FLAGS)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${XXHASH_C_FLAGS}")
endif()
foreach (flag
-Wall -Wextra -Wcast-qual -Wcast-align -Wshadow
-Wstrict-aliasing=1 -Wswitch-enum -Wdeclaration-after-statement
-Wstrict-prototypes -Wundef)
# Because https://gcc.gnu.org/wiki/FAQ#wnowarning
string(REGEX REPLACE "\\-Wno\\-(.+)" "-W\\1" flag_to_test "${flag}")
string(REGEX REPLACE "[^a-zA-Z0-9]+" "_" test_name "CFLAG_${flag_to_test}")
check_c_compiler_flag("${ADD_COMPILER_FLAGS_PREPEND} ${flag_to_test}" ${test_name})
if(${test_name})
set(CMAKE_C_FLAGS "${flag} ${CMAKE_C_FLAGS}")
endif()
unset(test_name)
unset(flag_to_test)
endforeach (flag)
if(NOT XXHASH_BUNDLED_MODE)
include(GNUInstallDirs)
install(TARGETS xxhash
EXPORT xxHashTargets
RUNTIME DESTINATION "${CMAKE_INSTALL_BINDIR}"
LIBRARY DESTINATION "${CMAKE_INSTALL_LIBDIR}"
ARCHIVE DESTINATION "${CMAKE_INSTALL_LIBDIR}")
install(FILES "${XXHASH_DIR}/xxhash.h"
DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}")
install(FILES "${XXHASH_DIR}/xxh3.h"
DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}")
if(XXHASH_BUILD_XXHSUM)
install(TARGETS xxhsum
EXPORT xxHashTargets
RUNTIME DESTINATION "${CMAKE_INSTALL_BINDIR}")
install(FILES "${XXHASH_DIR}/xxhsum.1"
DESTINATION "${CMAKE_INSTALL_MANDIR}/man1")
endif(XXHASH_BUILD_XXHSUM)
include(CMakePackageConfigHelpers)
set(xxHash_VERSION_CONFIG "${PROJECT_BINARY_DIR}/xxHashConfigVersion.cmake")
set(xxHash_PROJECT_CONFIG "${PROJECT_BINARY_DIR}/xxHashConfig.cmake")
set(xxHash_TARGETS_CONFIG "${PROJECT_BINARY_DIR}/xxHashTargets.cmake")
set(xxHash_CONFIG_INSTALL_DIR "${CMAKE_INSTALL_LIBDIR}/cmake/xxHash/")
write_basic_package_version_file(${xxHash_VERSION_CONFIG}
VERSION ${XXHASH_VERSION_STRING}
COMPATIBILITY AnyNewerVersion)
configure_package_config_file(
${PROJECT_SOURCE_DIR}/xxHashConfig.cmake.in
${xxHash_PROJECT_CONFIG}
INSTALL_DESTINATION ${xxHash_CONFIG_INSTALL_DIR})
if("${CMAKE_VERSION}" VERSION_LESS "3.0")
set(XXHASH_EXPORT_SET xxhash)
if(XXHASH_BUILD_XXHSUM)
set(XXHASH_EXPORT_SET ${XXHASH_EXPORT_SET} xxhsum)
endif()
export(TARGETS ${XXHASH_EXPORT_SET}
FILE ${xxHash_TARGETS_CONFIG}
NAMESPACE ${PROJECT_NAME}::)
else()
export(EXPORT xxHashTargets
FILE ${xxHash_TARGETS_CONFIG}
NAMESPACE ${PROJECT_NAME}::)
endif()
install(FILES ${xxHash_PROJECT_CONFIG} ${xxHash_VERSION_CONFIG}
DESTINATION ${xxHash_CONFIG_INSTALL_DIR})
install(EXPORT xxHashTargets
DESTINATION ${xxHash_CONFIG_INSTALL_DIR}
NAMESPACE ${PROJECT_NAME}::)
endif(NOT XXHASH_BUNDLED_MODE)

36
third_party/xxhash/cmake_unofficial/README.md vendored
Unescape View File

@ -1,36 +0,0 @@
## Usage
### Way 1: import targets
Build xxHash targets:
cd </path/to/xxHash/>
mkdir build
cd build
cmake ../cmake_unofficial [options]
cmake --build .
cmake --build . --target install #optional
Where possible options are:
- `-DXXHASH_BUILD_ENABLE_INLINE_API=<ON|OFF>`: adds xxhash.c for the `-DXXH_INLINE_ALL` api. ON by default.
- `-DXXHASH_BUILD_XXHSUM=<ON|OFF>`: build the command line binary. ON by default
- `-DBUILD_SHARED_LIBS=<ON|OFF>`: build dynamic library. ON by default.
- `-DCMAKE_INSTALL_PREFIX=<path>`: use custom install prefix path.
Add lines into downstream CMakeLists.txt:
find_package(xxHash 0.7 CONFIG REQUIRED)
...
target_link_libraries(MyTarget PRIVATE xxHash::xxhash)
### Way 2: Add subdirectory
Add lines into downstream CMakeLists.txt:
option(BUILD_SHARED_LIBS "Build shared libs" OFF) #optional
...
set(XXHASH_BUILD_ENABLE_INLINE_API OFF) #optional
set(XXHASH_BUILD_XXHSUM OFF) #optional
add_subdirectory(</path/to/xxHash/cmake_unofficial/> </path/to/xxHash/build/> EXCLUDE_FROM_ALL)
...
target_link_libraries(MyTarget PRIVATE xxHash::xxhash)

4
third_party/xxhash/cmake_unofficial/xxHashConfig.cmake.in vendored
Unescape View File

@ -1,4 +0,0 @@
@PACKAGE_INIT@
include(${CMAKE_CURRENT_LIST_DIR}/xxHashTargets.cmake)

9
third_party/xxhash/doc/README.md vendored
Unescape View File

@ -1,9 +0,0 @@
xxHash Specification
=======================
This directory contains material defining the xxHash algorithm.
It's described in [this specification document](xxhash_spec.md).
The algorithm is also be illustrated by a [simple educational library](https://github.com/easyaspi314/xxhash-clean),
written by @easyaspi314 and designed for readability
(as opposed to the reference library which is designed for speed).

206
third_party/xxhash/doc/xxhash.cry vendored
Unescape View File

@ -1,206 +0,0 @@
module xxhash where
/**
* The 32-bit variant of xxHash. The first argument is the sequence
* of L bytes to hash. The second argument is a seed value.
*/
XXH32 : {L} (fin L) => [L][8] -> [32] -> [32]
XXH32 input seed = XXH32_avalanche acc1
where (stripes16 # stripes4 # stripes1) = input
accR = foldl XXH32_rounds (XXH32_init seed) (split stripes16 : [L/16][16][8])
accL = `(L % 2^^32) + if (`L:Integer) < 16
then seed + PRIME32_5
else XXH32_converge accR
acc4 = foldl XXH32_digest4 accL (split stripes4 : [(L%16)/4][4][8])
acc1 = foldl XXH32_digest1 acc4 (stripes1 : [L%4][8])
/**
* The 64-bit variant of xxHash. The first argument is the sequence
* of L bytes to hash. The second argument is a seed value.
*/
XXH64 : {L} (fin L) => [L][8] -> [64] -> [64]
XXH64 input seed = XXH64_avalanche acc1
where (stripes32 # stripes8 # stripes4 # stripes1) = input
accR = foldl XXH64_rounds (XXH64_init seed) (split stripes32 : [L/32][32][8])
accL = `(L % 2^^64) + if (`L:Integer) < 32
then seed + PRIME64_5
else XXH64_converge accR
acc8 = foldl XXH64_digest8 accL (split stripes8 : [(L%32)/8][8][8])
acc4 = foldl XXH64_digest4 acc8 (split stripes4 : [(L%8)/4][4][8])
acc1 = foldl XXH64_digest1 acc4 (stripes1 : [L%4][8])
private
//Utility functions
/**
* Combines a sequence of bytes into a word using the little-endian
* convention.
*/
toLE bytes = join (reverse bytes)
//32-bit xxHash helper functions
//32-bit prime number constants
PRIME32_1 = 0x9E3779B1 : [32]
PRIME32_2 = 0x85EBCA77 : [32]
PRIME32_3 = 0xC2B2AE3D : [32]
PRIME32_4 = 0x27D4EB2F : [32]
PRIME32_5 = 0x165667B1 : [32]
/**
* The property shows that the hexadecimal representation of the
* PRIME32 constants is the same as the binary representation.
*/
property PRIME32s_as_bits_correct =
(PRIME32_1 == 0b10011110001101110111100110110001) /\
(PRIME32_2 == 0b10000101111010111100101001110111) /\
(PRIME32_3 == 0b11000010101100101010111000111101) /\
(PRIME32_4 == 0b00100111110101001110101100101111) /\
(PRIME32_5 == 0b00010110010101100110011110110001)
/**
* This function initializes the four internal accumulators of XXH32.
*/
XXH32_init : [32] -> [4][32]
XXH32_init seed = [acc1, acc2, acc3, acc4]
where acc1 = seed + PRIME32_1 + PRIME32_2
acc2 = seed + PRIME32_2
acc3 = seed + 0
acc4 = seed - PRIME32_1
/**
* This processes a single lane of the main round function of XXH32.
*/
XXH32_round : [32] -> [32] -> [32]
XXH32_round accN laneN = ((accN + laneN * PRIME32_2) <<< 13) * PRIME32_1
/**
* This is the main round function of XXH32 and processes a stripe,
* i.e. 4 lanes with 4 bytes each.
*/
XXH32_rounds : [4][32] -> [16][8] -> [4][32]
XXH32_rounds accs stripe =
[ XXH32_round accN (toLE laneN) | accN <- accs | laneN <- split stripe ]
/**
* This function combines the four lane accumulators into a single
* 32-bit value.
*/
XXH32_converge : [4][32] -> [32]
XXH32_converge [acc1, acc2, acc3, acc4] =
(acc1 <<< 1) + (acc2 <<< 7) + (acc3 <<< 12) + (acc4 <<< 18)
/**
* This function digests a four byte lane
*/
XXH32_digest4 : [32] -> [4][8] -> [32]
XXH32_digest4 acc lane = ((acc + toLE lane * PRIME32_3) <<< 17) * PRIME32_4
/**
* This function digests a single byte lane
*/
XXH32_digest1 : [32] -> [8] -> [32]
XXH32_digest1 acc lane = ((acc + (0 # lane) * PRIME32_5) <<< 11) * PRIME32_1
/**
* This function ensures that all input bits have a chance to impact
* any bit in the output digest, resulting in an unbiased
* distribution.
*/
XXH32_avalanche : [32] -> [32]
XXH32_avalanche acc0 = acc5
where acc1 = acc0 ^ (acc0 >> 15)
acc2 = acc1 * PRIME32_2
acc3 = acc2 ^ (acc2 >> 13)
acc4 = acc3 * PRIME32_3
acc5 = acc4 ^ (acc4 >> 16)
//64-bit xxHash helper functions
//64-bit prime number constants
PRIME64_1 = 0x9E3779B185EBCA87 : [64]
PRIME64_2 = 0xC2B2AE3D27D4EB4F : [64]
PRIME64_3 = 0x165667B19E3779F9 : [64]
PRIME64_4 = 0x85EBCA77C2B2AE63 : [64]
PRIME64_5 = 0x27D4EB2F165667C5 : [64]
/**
* The property shows that the hexadecimal representation of the
* PRIME64 constants is the same as the binary representation.
*/
property PRIME64s_as_bits_correct =
(PRIME64_1 == 0b1001111000110111011110011011000110000101111010111100101010000111) /\
(PRIME64_2 == 0b1100001010110010101011100011110100100111110101001110101101001111) /\
(PRIME64_3 == 0b0001011001010110011001111011000110011110001101110111100111111001) /\
(PRIME64_4 == 0b1000010111101011110010100111011111000010101100101010111001100011) /\
(PRIME64_5 == 0b0010011111010100111010110010111100010110010101100110011111000101)
/**
* This function initializes the four internal accumulators of XXH64.
*/
XXH64_init : [64] -> [4][64]
XXH64_init seed = [acc1, acc2, acc3, acc4]
where acc1 = seed + PRIME64_1 + PRIME64_2
acc2 = seed + PRIME64_2
acc3 = seed + 0
acc4 = seed - PRIME64_1
/**
* This processes a single lane of the main round function of XXH64.
*/
XXH64_round : [64] -> [64] -> [64]
XXH64_round accN laneN = ((accN + laneN * PRIME64_2) <<< 31) * PRIME64_1
/**
* This is the main round function of XXH64 and processes a stripe,
* i.e. 4 lanes with 8 bytes each.
*/
XXH64_rounds : [4][64] -> [32][8] -> [4][64]
XXH64_rounds accs stripe =
[ XXH64_round accN (toLE laneN) | accN <- accs | laneN <- split stripe ]
/**
* This is a helper function, used to merge the four lane accumulators.
*/
mergeAccumulator : [64] -> [64] -> [64]
mergeAccumulator acc accN = (acc ^ XXH64_round 0 accN) * PRIME64_1 + PRIME64_4
/**
* This function combines the four lane accumulators into a single
* 64-bit value.
*/
XXH64_converge : [4][64] -> [64]
XXH64_converge [acc1, acc2, acc3, acc4] =
foldl mergeAccumulator ((acc1 <<< 1) + (acc2 <<< 7) + (acc3 <<< 12) + (acc4 <<< 18)) [acc1, acc2, acc3, acc4]
/**
* This function digests an eight byte lane
*/
XXH64_digest8 : [64] -> [8][8] -> [64]
XXH64_digest8 acc lane = ((acc ^ XXH64_round 0 (toLE lane)) <<< 27) * PRIME64_1 + PRIME64_4
/**
* This function digests a four byte lane
*/
XXH64_digest4 : [64] -> [4][8] -> [64]
XXH64_digest4 acc lane = ((acc ^ (0 # toLE lane) * PRIME64_1) <<< 23) * PRIME64_2 + PRIME64_3
/**
* This function digests a single byte lane
*/
XXH64_digest1 : [64] -> [8] -> [64]
XXH64_digest1 acc lane = ((acc ^ (0 # lane) * PRIME64_5) <<< 11) * PRIME64_1
/**
* This function ensures that all input bits have a chance to impact
* any bit in the output digest, resulting in an unbiased
* distribution.
*/
XXH64_avalanche : [64] -> [64]
XXH64_avalanche acc0 = acc5
where acc1 = acc0 ^ (acc0 >> 33)
acc2 = acc1 * PRIME64_2
acc3 = acc2 ^ (acc2 >> 29)
acc4 = acc3 * PRIME64_3
acc5 = acc4 ^ (acc4 >> 32)

317
third_party/xxhash/doc/xxhash_spec.md vendored
Unescape View File

@ -1,317 +0,0 @@
xxHash fast digest algorithm
======================
### Notices
Copyright (c) Yann Collet
Permission is granted to copy and distribute this document
for any purpose and without charge,
including translations into other languages
and incorporation into compilations,
provided that the copyright notice and this notice are preserved,
and that any substantive changes or deletions from the original
are clearly marked.
Distribution of this document is unlimited.
### Version
0.1.1 (10/10/18)
Table of Contents
---------------------
- [Introduction](#introduction)
- [XXH32 algorithm description](#xxh32-algorithm-description)
- [XXH64 algorithm description](#xxh64-algorithm-description)
- [Performance considerations](#performance-considerations)
- [Reference Implementation](#reference-implementation)
Introduction
----------------
This document describes the xxHash digest algorithm for both 32-bit and 64-bit variants, named `XXH32` and `XXH64`. The algorithm takes an input a message of arbitrary length and an optional seed value, then produces an output of 32 or 64-bit as "fingerprint" or "digest".
xxHash is primarily designed for speed. It is labeled non-cryptographic, and is not meant to avoid intentional collisions (same digest for 2 different messages), or to prevent producing a message with a predefined digest.
XXH32 is designed to be fast on 32-bit machines.
XXH64 is designed to be fast on 64-bit machines.
Both variants produce different output.
However, a given variant shall produce exactly the same output, irrespective of the cpu / os used. In particular, the result remains identical whatever the endianness and width of the cpu is.
### Operation notations
All operations are performed modulo {32,64} bits. Arithmetic overflows are expected.
`XXH32` uses 32-bit modular operations. `XXH64` uses 64-bit modular operations.
- `+`: denotes modular addition
- `*`: denotes modular multiplication
- `X <<< s`: denotes the value obtained by circularly shifting (rotating) `X` left by `s` bit positions.
- `X >> s`: denotes the value obtained by shifting `X` right by s bit positions. Upper `s` bits become `0`.
- `X xor Y`: denotes the bit-wise XOR of `X` and `Y` (same width).
XXH32 Algorithm Description
-------------------------------------
### Overview
We begin by supposing that we have a message of any length `L` as input, and that we wish to find its digest. Here `L` is an arbitrary nonnegative integer; `L` may be zero. The following steps are performed to compute the digest of the message.
The algorithm collect and transform input in _stripes_ of 16 bytes. The transforms are stored inside 4 "accumulators", each one storing an unsigned 32-bit value. Each accumulator can be processed independently in parallel, speeding up processing for cpu with multiple execution units.
The algorithm uses 32-bits addition, multiplication, rotate, shift and xor operations. Many operations require some 32-bits prime number constants, all defined below:
static const u32 PRIME32_1 = 0x9E3779B1U; // 0b10011110001101110111100110110001
static const u32 PRIME32_2 = 0x85EBCA77U; // 0b10000101111010111100101001110111
static const u32 PRIME32_3 = 0xC2B2AE3DU; // 0b11000010101100101010111000111101
static const u32 PRIME32_4 = 0x27D4EB2FU; // 0b00100111110101001110101100101111
static const u32 PRIME32_5 = 0x165667B1U; // 0b00010110010101100110011110110001
These constants are prime numbers, and feature a good mix of bits 1 and 0, neither too regular, nor too dissymmetric. These properties help dispersion capabilities.
### Step 1. Initialize internal accumulators
Each accumulator gets an initial value based on optional `seed` input. Since the `seed` is optional, it can be `0`.
u32 acc1 = seed + PRIME32_1 + PRIME32_2;
u32 acc2 = seed + PRIME32_2;
u32 acc3 = seed + 0;
u32 acc4 = seed - PRIME32_1;
#### Special case: input is less than 16 bytes
When the input is too small (< 16 bytes), the algorithm will not process any stripes. Consequently, it will not make use of parallel accumulators.
In this case, a simplified initialization is performed, using a single accumulator:
u32 acc = seed + PRIME32_5;
The algorithm then proceeds directly to step 4.
### Step 2. Process stripes
A stripe is a contiguous segment of 16 bytes.
It is evenly divided into 4 _lanes_, of 4 bytes each.
The first lane is used to update accumulator 1, the second lane is used to update accumulator 2, and so on.
Each lane read its associated 32-bit value using __little-endian__ convention.
For each {lane, accumulator}, the update process is called a _round_, and applies the following formula:
accN = accN + (laneN * PRIME32_2);
accN = accN <<< 13;
accN = accN * PRIME32_1;
This shuffles the bits so that any bit from input _lane_ impacts several bits in output _accumulator_. All operations are performed modulo 2^32.
Input is consumed one full stripe at a time. Step 2 is looped as many times as necessary to consume the whole input, except for the last remaining bytes which cannot form a stripe (< 16 bytes).
When that happens, move to step 3.
### Step 3. Accumulator convergence
All 4 lane accumulators from the previous steps are merged to produce a single remaining accumulator of the same width (32-bit). The associated formula is as follows:
acc = (acc1 <<< 1) + (acc2 <<< 7) + (acc3 <<< 12) + (acc4 <<< 18);
### Step 4. Add input length
The input total length is presumed known at this stage. This step is just about adding the length to accumulator, so that it participates to final mixing.
acc = acc + (u32)inputLength;
Note that, if input length is so large that it requires more than 32-bits, only the lower 32-bits are added to the accumulator.
### Step 5. Consume remaining input
There may be up to 15 bytes remaining to consume from the input.
The final stage will digest them according to following pseudo-code:
while (remainingLength >= 4) {
lane = read_32bit_little_endian(input_ptr);
acc = acc + lane * PRIME32_3;
acc = (acc <<< 17) * PRIME32_4;
input_ptr += 4; remainingLength -= 4;
}
while (remainingLength >= 1) {
lane = read_byte(input_ptr);
acc = acc + lane * PRIME32_5;
acc = (acc <<< 11) * PRIME32_1;
input_ptr += 1; remainingLength -= 1;
}
This process ensures that all input bytes are present in the final mix.
### Step 6. Final mix (avalanche)
The final mix ensures that all input bits have a chance to impact any bit in the output digest, resulting in an unbiased distribution. This is also called avalanche effect.
acc = acc xor (acc >> 15);
acc = acc * PRIME32_2;
acc = acc xor (acc >> 13);
acc = acc * PRIME32_3;
acc = acc xor (acc >> 16);
### Step 7. Output
The `XXH32()` function produces an unsigned 32-bit value as output.
For systems which require to store and/or display the result in binary or hexadecimal format, the canonical format is defined to reproduce the same value as the natural decimal format, hence follows __big-endian__ convention (most significant byte first).
XXH64 Algorithm Description
-------------------------------------
### Overview
`XXH64`'s algorithm structure is very similar to `XXH32` one. The major difference is that `XXH64` uses 64-bit arithmetic, speeding up memory transfer for 64-bit compliant systems, but also relying on cpu capability to efficiently perform 64-bit operations.
The algorithm collects and transforms input in _stripes_ of 32 bytes. The transforms are stored inside 4 "accumulators", each one storing an unsigned 64-bit value. Each accumulator can be processed independently in parallel, speeding up processing for cpu with multiple execution units.
The algorithm uses 64-bit addition, multiplication, rotate, shift and xor operations. Many operations require some 64-bit prime number constants, all defined below:
static const u64 PRIME64_1 = 0x9E3779B185EBCA87ULL; // 0b1001111000110111011110011011000110000101111010111100101010000111
static const u64 PRIME64_2 = 0xC2B2AE3D27D4EB4FULL; // 0b1100001010110010101011100011110100100111110101001110101101001111
static const u64 PRIME64_3 = 0x165667B19E3779F9ULL; // 0b0001011001010110011001111011000110011110001101110111100111111001
static const u64 PRIME64_4 = 0x85EBCA77C2B2AE63ULL; // 0b1000010111101011110010100111011111000010101100101010111001100011
static const u64 PRIME64_5 = 0x27D4EB2F165667C5ULL; // 0b0010011111010100111010110010111100010110010101100110011111000101
These constants are prime numbers, and feature a good mix of bits 1 and 0, neither too regular, nor too dissymmetric. These properties help dispersion capabilities.
### Step 1. Initialise internal accumulators
Each accumulator gets an initial value based on optional `seed` input. Since the `seed` is optional, it can be `0`.
u64 acc1 = seed + PRIME64_1 + PRIME64_2;
u64 acc2 = seed + PRIME64_2;
u64 acc3 = seed + 0;
u64 acc4 = seed - PRIME64_1;
#### Special case: input is less than 32 bytes
When the input is too small (< 32 bytes), the algorithm will not process any stripes. Consequently, it will not make use of parallel accumulators.
In this case, a simplified initialization is performed, using a single accumulator:
u64 acc = seed + PRIME64_5;
The algorithm then proceeds directly to step 4.
### Step 2. Process stripes
A stripe is a contiguous segment of 32 bytes.
It is evenly divided into 4 _lanes_, of 8 bytes each.
The first lane is used to update accumulator 1, the second lane is used to update accumulator 2, and so on.
Each lane read its associated 64-bit value using __little-endian__ convention.
For each {lane, accumulator}, the update process is called a _round_, and applies the following formula:
round(accN,laneN):
accN = accN + (laneN * PRIME64_2);
accN = accN <<< 31;
return accN * PRIME64_1;
This shuffles the bits so that any bit from input _lane_ impacts several bits in output _accumulator_. All operations are performed modulo 2^64.
Input is consumed one full stripe at a time. Step 2 is looped as many times as necessary to consume the whole input, except for the last remaining bytes which cannot form a stripe (< 32 bytes).
When that happens, move to step 3.
### Step 3. Accumulator convergence
All 4 lane accumulators from previous steps are merged to produce a single remaining accumulator of same width (64-bit). The associated formula is as follows.
Note that accumulator convergence is more complex than 32-bit variant, and requires to define another function called _mergeAccumulator()_:
mergeAccumulator(acc,accN):
acc = acc xor round(0, accN);
acc = acc * PRIME64_1;
return acc + PRIME64_4;
which is then used in the convergence formula:
acc = (acc1 <<< 1) + (acc2 <<< 7) + (acc3 <<< 12) + (acc4 <<< 18);
acc = mergeAccumulator(acc, acc1);
acc = mergeAccumulator(acc, acc2);
acc = mergeAccumulator(acc, acc3);
acc = mergeAccumulator(acc, acc4);
### Step 4. Add input length
The input total length is presumed known at this stage. This step is just about adding the length to accumulator, so that it participates to final mixing.
acc = acc + inputLength;
### Step 5. Consume remaining input
There may be up to 31 bytes remaining to consume from the input.
The final stage will digest them according to following pseudo-code:
while (remainingLength >= 8) {
lane = read_64bit_little_endian(input_ptr);
acc = acc xor round(0, lane);
acc = (acc <<< 27) * PRIME64_1;
acc = acc + PRIME64_4;
input_ptr += 8; remainingLength -= 8;
}
if (remainingLength >= 4) {
lane = read_32bit_little_endian(input_ptr);
acc = acc xor (lane * PRIME64_1);
acc = (acc <<< 23) * PRIME64_2;
acc = acc + PRIME64_3;
input_ptr += 4; remainingLength -= 4;
}
while (remainingLength >= 1) {
lane = read_byte(input_ptr);
acc = acc xor (lane * PRIME64_5);
acc = (acc <<< 11) * PRIME64_1;
input_ptr += 1; remainingLength -= 1;
}
This process ensures that all input bytes are present in the final mix.
### Step 6. Final mix (avalanche)
The final mix ensures that all input bits have a chance to impact any bit in the output digest, resulting in an unbiased distribution. This is also called avalanche effect.
acc = acc xor (acc >> 33);
acc = acc * PRIME64_2;
acc = acc xor (acc >> 29);
acc = acc * PRIME64_3;
acc = acc xor (acc >> 32);
### Step 7. Output
The `XXH64()` function produces an unsigned 64-bit value as output.
For systems which require to store and/or display the result in binary or hexadecimal format, the canonical format is defined to reproduce the same value as the natural decimal format, hence follows __big-endian__ convention (most significant byte first).
Performance considerations
----------------------------------
The xxHash algorithms are simple and compact to implement. They provide a system independent "fingerprint" or digest of a message of arbitrary length.
The algorithm allows input to be streamed and processed in multiple steps. In such case, an internal buffer is needed to ensure data is presented to the algorithm in full stripes.
On 64-bit systems, the 64-bit variant `XXH64` is generally faster to compute, so it is a recommended variant, even when only 32-bit are needed.
On 32-bit systems though, positions are reversed: `XXH64` performance is reduced, due to its usage of 64-bit arithmetic. `XXH32` becomes a faster variant.
Reference Implementation
----------------------------------------
A reference library written in C is available at https://www.xxhash.com.
The web page also links to multiple other implementations written in many different languages.
It links to the [github project page](https://github.com/Cyan4973/xxHash) where an [issue board](https://github.com/Cyan4973/xxHash/issues) can be used for further public discussions on the topic.
Version changes
--------------------
v0.7.3: Minor fixes
v0.1.1: added a note on rationale for selection of constants
v0.1.0: initial release

15
third_party/xxhash/libxxhash.pc.in vendored
Unescape View File

@ -1,15 +0,0 @@
# xxHash - Extremely fast hash algorithm
# Copyright (C) 2012-2020, Yann Collet, Facebook
# BSD 2-Clause License (https://www.opensource.org/licenses/bsd-license.php)
prefix=@PREFIX@
exec_prefix=@EXECPREFIX@
includedir=${prefix}/@INCLUDEDIR@
libdir=${exec_prefix}/@LIBDIR@
Name: xxhash
Description: extremely fast hash algorithm
URL: http://www.xxhash.com/
Version: @VERSION@
Libs: -L${libdir} -lxxhash
Cflags: -I${includedir}

55
third_party/xxhash/xxh3.h vendored
Unescape View File

@ -1,55 +0,0 @@
/*
* xxHash - Extremely Fast Hash algorithm
* Development source file for `xxh3`
* Copyright (C) 2019-2020 Yann Collet
*
* BSD 2-Clause License (https://www.opensource.org/licenses/bsd-license.php)
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following disclaimer
* in the documentation and/or other materials provided with the
* distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* You can contact the author at:
* - xxHash homepage: https://www.xxhash.com
* - xxHash source repository: https://github.com/Cyan4973/xxHash
*/
/*
* Note: This file used to host the source code of XXH3_* variants.
* during the development period.
* The source code is now properly integrated within xxhash.h.
*
* xxh3.h is no longer useful,
* but it is still provided for compatibility with source code
* which used to include it directly.
*
* Programs are now highly discourage to include xxh3.h.
* Include `xxhash.h` instead, which is the officially supported interface.
*
* In the future, xxh3.h will start to generate warnings, then errors,
* then it will be removed from source package and from include directory.
*/
/* Simulate the same impact as including the old xxh3.h source file */
#define XXH_INLINE_ALL
#include "xxhash.h"

770
third_party/xxhash/xxh_x86dispatch.c vendored
Unescape View File

@ -1,770 +0,0 @@
/*
* xxHash - Extremely Fast Hash algorithm
* Copyright (C) 2020 Yann Collet
*
* BSD 2-Clause License (https://www.opensource.org/licenses/bsd-license.php)
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following disclaimer
* in the documentation and/or other materials provided with the
* distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* You can contact the author at:
* - xxHash homepage: https://www.xxhash.com
* - xxHash source repository: https://github.com/Cyan4973/xxHash
*/
/*!
* @file xxh_x86dispatch.c
*
* Automatic dispatcher code for the @ref xxh3_family on x86-based targets.
*
* Optional add-on.
*
* **Compile this file with the default flags for your target.** Do not compile
* with flags like `-mavx*`, `-march=native`, or `/arch:AVX*`, there will be
* an error. See @ref XXH_X86DISPATCH_ALLOW_AVX for details.
*
* @defgroup dispatch x86 Dispatcher
* @{
*/
#if defined (__cplusplus)
extern "C" {
#endif
#if !(defined(__x86_64__) || defined(__i386__) || defined(_M_IX86) || defined(_M_X64))
# error "Dispatching is currently only supported on x86 and x86_64."
#endif
/*!
* @def XXH_X86DISPATCH_ALLOW_AVX
* @brief Disables the AVX sanity check.
*
* Don't compile xxh_x86dispatch.c with options like `-mavx*`, `-march=native`,
* or `/arch:AVX*`. It is intended to be compiled for the minimum target, and
* it selectively enables SSE2, AVX2, and AVX512 when it is needed.
*
* Using this option _globally_ allows this feature, and therefore makes it
* undefined behavior to execute on any CPU without said feature.
*
* Even if the source code isn't directly using AVX intrinsics in a function,
* the compiler can still generate AVX code from autovectorization and by
* "upgrading" SSE2 intrinsics to use the VEX prefixes (a.k.a. AVX128).
*
* Use the same flags that you use to compile the rest of the program; this
* file will safely generate SSE2, AVX2, and AVX512 without these flags.
*
* Define XXH_X86DISPATCH_ALLOW_AVX to ignore this check, and feel free to open
* an issue if there is a target in the future where AVX is a default feature.
*/
#ifdef XXH_DOXYGEN
# define XXH_X86DISPATCH_ALLOW_AVX
#endif
#if defined(__AVX__) && !defined(XXH_X86DISPATCH_ALLOW_AVX)
# error "Do not compile xxh_x86dispatch.c with AVX enabled! See the comment above."
#endif
#ifdef __has_include
# define XXH_HAS_INCLUDE(header) __has_include(header)
#else
# define XXH_HAS_INCLUDE(header) 0
#endif
/*!
* @def XXH_DISPATCH_SCALAR
* @brief Enables/dispatching the scalar code path.
*
* If this is defined to 0, SSE2 support is assumed. This reduces code size
* when the scalar path is not needed.
*
* This is automatically defined to 0 when...
* - SSE2 support is enabled in the compiler
* - Targeting x86_64
* - Targeting Android x86
* - Targeting macOS
*/
#ifndef XXH_DISPATCH_SCALAR
# if defined(__SSE2__) || (defined(_M_IX86_FP) && _M_IX86_FP >= 2) /* SSE2 on by default */ \
|| defined(__x86_64__) || defined(_M_X64) /* x86_64 */ \
|| defined(__ANDROID__) || defined(__APPLEv__) /* Android or macOS */
# define XXH_DISPATCH_SCALAR 0 /* disable */
# else
# define XXH_DISPATCH_SCALAR 1
# endif
#endif
/*!
* @def XXH_DISPATCH_AVX2
* @brief Enables/disables dispatching for AVX2.
*
* This is automatically detected if it is not defined.
* - GCC 4.7 and later are known to support AVX2, but >4.9 is required for
* to get the AVX2 intrinsics and typedefs without -mavx -mavx2.
* - Visual Studio 2013 Update 2 and later are known to support AVX2.
* - The GCC/Clang internal header `<avx2intrin.h>` is detected. While this is
* not allowed to be included directly, it still appears in the builtin
* include path and is detectable with `__has_include`.
*
* @see XXH_AVX2
*/
#ifndef XXH_DISPATCH_AVX2
# if (defined(__GNUC__) && (__GNUC__ > 4)) /* GCC 5.0+ */ \
|| (defined(_MSC_VER) && _MSC_VER >= 1900) /* VS 2015+ */ \
|| (defined(_MSC_FULL_VER) && _MSC_FULL_VER >= 180030501) /* VS 2013 Update 2 */ \
|| XXH_HAS_INCLUDE(<avx2intrin.h>) /* GCC/Clang internal header */
# define XXH_DISPATCH_AVX2 1 /* enable dispatch towards AVX2 */
# else
# define XXH_DISPATCH_AVX2 0
# endif
#endif /* XXH_DISPATCH_AVX2 */
/*!
* @def XXH_DISPATCH_AVX512
* @brief Enables/disables dispatching for AVX512.
*
* Automatically detected if one of the following conditions is met:
* - GCC 4.9 and later are known to support AVX512.
* - Visual Studio 2017 and later are known to support AVX2.
* - The GCC/Clang internal header `<avx512fintrin.h>` is detected. While this
* is not allowed to be included directly, it still appears in the builtin
* include path and is detectable with `__has_include`.
*
* @see XXH_AVX512
*/
#ifndef XXH_DISPATCH_AVX512
# if (defined(__GNUC__) \
&& (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 9))) /* GCC 4.9+ */ \
|| (defined(_MSC_VER) && _MSC_VER >= 1910) /* VS 2017+ */ \
|| XXH_HAS_INCLUDE(<avx512fintrin.h>) /* GCC/Clang internal header */
# define XXH_DISPATCH_AVX512 1 /* enable dispatch towards AVX512 */
# else
# define XXH_DISPATCH_AVX512 0
# endif
#endif /* XXH_DISPATCH_AVX512 */
/*!
* @def XXH_TARGET_SSE2
* @brief Allows a function to be compiled with SSE2 intrinsics.
*
* Uses `__attribute__((__target__("sse2")))` on GCC to allow SSE2 to be used
* even with `-mno-sse2`.
*
* @def XXH_TARGET_AVX2
* @brief Like @ref XXH_TARGET_SSE2, but for AVX2.
*
* @def XXH_TARGET_AVX512
* @brief Like @ref XXH_TARGET_SSE2, but for AVX512.
*/
#if defined(__GNUC__)
# include <emmintrin.h> /* SSE2 */
# if XXH_DISPATCH_AVX2 || XXH_DISPATCH_AVX512
# include <immintrin.h> /* AVX2, AVX512F */
# endif
# define XXH_TARGET_SSE2 __attribute__((__target__("sse2")))
# define XXH_TARGET_AVX2 __attribute__((__target__("avx2")))
# define XXH_TARGET_AVX512 __attribute__((__target__("avx512f")))
#elif defined(_MSC_VER)
# include <intrin.h>
# define XXH_TARGET_SSE2
# define XXH_TARGET_AVX2
# define XXH_TARGET_AVX512
#else
# error "Dispatching is currently not supported for your compiler."
#endif
#ifdef XXH_DISPATCH_DEBUG
/* debug logging */
# include <stdio.h>
# define XXH_debugPrint(str) { fprintf(stderr, "DEBUG: xxHash dispatch: %s \n", str); fflush(NULL); }
#else
# define XXH_debugPrint(str) ((void)0)
# undef NDEBUG /* avoid redefinition */
# define NDEBUG
#endif
#include <assert.h>
#define XXH_INLINE_ALL
#define XXH_X86DISPATCH
#include "xxhash.h"
/*
* Support both AT&T and Intel dialects
*
* GCC doesn't convert AT&T syntax to Intel syntax, and will error out if
* compiled with -masm=intel. Instead, it supports dialect switching with
* curly braces: { AT&T syntax | Intel syntax }
*
* Clang's integrated assembler automatically converts AT&T syntax to Intel if
* needed, making the dialect switching useless (it isn't even supported).
*
* Note: Comments are written in the inline assembly itself.
*/
#ifdef __clang__
# define XXH_I_ATT(intel, att) att "\n\t"
#else
# define XXH_I_ATT(intel, att) "{" att "|" intel "}\n\t"
#endif
/*!
* @internal
* @brief Runs CPUID.
*
* @param eax, ecx The parameters to pass to CPUID, %eax and %ecx respectively.
* @param abcd The array to store the result in, `{ eax, ebx, ecx, edx }`
*/
static void XXH_cpuid(xxh_u32 eax, xxh_u32 ecx, xxh_u32* abcd)
{
#if defined(_MSC_VER)
__cpuidex(abcd, eax, ecx);
#else
xxh_u32 ebx, edx;
# if defined(__i386__) && defined(__PIC__)
__asm__(
"# Call CPUID\n\t"
"#\n\t"
"# On 32-bit x86 with PIC enabled, we are not allowed to overwrite\n\t"
"# EBX, so we use EDI instead.\n\t"
XXH_I_ATT("mov edi, ebx", "movl %%ebx, %%edi")
XXH_I_ATT("cpuid", "cpuid" )
XXH_I_ATT("xchg edi, ebx", "xchgl %%ebx, %%edi")
: "=D" (ebx),
# else
__asm__(
"# Call CPUID\n\t"
XXH_I_ATT("cpuid", "cpuid")
: "=b" (ebx),
# endif
"+a" (eax), "+c" (ecx), "=d" (edx));
abcd[0] = eax;
abcd[1] = ebx;
abcd[2] = ecx;
abcd[3] = edx;
#endif
}
/*
* Modified version of Intel's guide
* https://software.intel.com/en-us/articles/how-to-detect-new-instruction-support-in-the-4th-generation-intel-core-processor-family
*/
#if XXH_DISPATCH_AVX2 || XXH_DISPATCH_AVX512
/*!
* @internal
* @brief Runs `XGETBV`.
*
* While the CPU may support AVX2, the operating system might not properly save
* the full YMM/ZMM registers.
*
* xgetbv is used for detecting this: Any compliant operating system will define
* a set of flags in the xcr0 register indicating how it saves the AVX registers.
*
* You can manually disable this flag on Windows by running, as admin:
*
* bcdedit.exe /set xsavedisable 1
*
* and rebooting. Run the same command with 0 to re-enable it.
*/
static xxh_u64 XXH_xgetbv(void)
{
#if defined(_MSC_VER)
return _xgetbv(0); /* min VS2010 SP1 compiler is required */
#else
xxh_u32 xcr0_lo, xcr0_hi;
__asm__(
"# Call XGETBV\n\t"
"#\n\t"
"# Older assemblers (e.g. macOS's ancient GAS version) don't support\n\t"
"# the XGETBV opcode, so we encode it by hand instead.\n\t"
"# See <https://github.com/asmjit/asmjit/issues/78> for details.\n\t"
".byte 0x0f, 0x01, 0xd0\n\t"
: "=a" (xcr0_lo), "=d" (xcr0_hi) : "c" (0));
return xcr0_lo | ((xxh_u64)xcr0_hi << 32);
#endif
}
#endif
#define XXH_SSE2_CPUID_MASK (1 << 26)
#define XXH_OSXSAVE_CPUID_MASK ((1 << 26) | (1 << 27))
#define XXH_AVX2_CPUID_MASK (1 << 5)
#define XXH_AVX2_XGETBV_MASK ((1 << 2) | (1 << 1))
#define XXH_AVX512F_CPUID_MASK (1 << 16)
#define XXH_AVX512F_XGETBV_MASK ((7 << 5) | (1 << 2) | (1 << 1))
/*!
* @internal
* @brief Returns the best XXH3 implementation.
*
* Runs various CPUID/XGETBV tests to try and determine the best implementation.
*
* @ret The best @ref XXH_VECTOR implementation.
* @see XXH_VECTOR_TYPES
*/
static int XXH_featureTest(void)
{
xxh_u32 abcd[4];
xxh_u32 max_leaves;
int best = XXH_SCALAR;
#if XXH_DISPATCH_AVX2 || XXH_DISPATCH_AVX512
xxh_u64 xgetbv_val;
#endif
#if defined(__GNUC__) && defined(__i386__)
xxh_u32 cpuid_supported;
__asm__(
"# For the sake of ruthless backwards compatibility, check if CPUID\n\t"
"# is supported in the EFLAGS on i386.\n\t"
"# This is not necessary on x86_64 - CPUID is mandatory.\n\t"
"# The ID flag (bit 21) in the EFLAGS register indicates support\n\t"
"# for the CPUID instruction. If a software procedure can set and\n\t"
"# clear this flag, the processor executing the procedure supports\n\t"
"# the CPUID instruction.\n\t"
"# <https://c9x.me/x86/html/file_module_x86_id_45.html>\n\t"
"#\n\t"
"# Routine is from <https://wiki.osdev.org/CPUID>.\n\t"
"# Save EFLAGS\n\t"
XXH_I_ATT("pushfd", "pushfl" )
"# Store EFLAGS\n\t"
XXH_I_ATT("pushfd", "pushfl" )
"# Invert the ID bit in stored EFLAGS\n\t"
XXH_I_ATT("xor dword ptr[esp], 0x200000", "xorl $0x200000, (%%esp)")
"# Load stored EFLAGS (with ID bit inverted)\n\t"
XXH_I_ATT("popfd", "popfl" )
"# Store EFLAGS again (ID bit may or not be inverted)\n\t"
XXH_I_ATT("pushfd", "pushfl" )
"# eax = modified EFLAGS (ID bit may or may not be inverted)\n\t"
XXH_I_ATT("pop eax", "popl %%eax" )
"# eax = whichever bits were changed\n\t"
XXH_I_ATT("xor eax, dword ptr[esp]", "xorl (%%esp), %%eax" )
"# Restore original EFLAGS\n\t"
XXH_I_ATT("popfd", "popfl" )
"# eax = zero if ID bit can't be changed, else non-zero\n\t"
XXH_I_ATT("and eax, 0x200000", "andl $0x200000, %%eax" )
: "=a" (cpuid_supported) :: "cc");
if (XXH_unlikely(!cpuid_supported)) {
XXH_debugPrint("CPUID support is not detected!");
return best;
}
#endif
/* Check how many CPUID pages we have */
XXH_cpuid(0, 0, abcd);
max_leaves = abcd[0];
/* Shouldn't happen on hardware, but happens on some QEMU configs. */
if (XXH_unlikely(max_leaves == 0)) {
XXH_debugPrint("Max CPUID leaves == 0!");
return best;
}
/* Check for SSE2, OSXSAVE and xgetbv */
XXH_cpuid(1, 0, abcd);
/*
* Test for SSE2. The check is redundant on x86_64, but it doesn't hurt.
*/
if (XXH_unlikely((abcd[3] & XXH_SSE2_CPUID_MASK) != XXH_SSE2_CPUID_MASK))
return best;
XXH_debugPrint("SSE2 support detected.");
best = XXH_SSE2;
#if XXH_DISPATCH_AVX2 || XXH_DISPATCH_AVX512
/* Make sure we have enough leaves */
if (XXH_unlikely(max_leaves < 7))
return best;
/* Test for OSXSAVE and XGETBV */
if ((abcd[2] & XXH_OSXSAVE_CPUID_MASK) != XXH_OSXSAVE_CPUID_MASK)
return best;
/* CPUID check for AVX features */
XXH_cpuid(7, 0, abcd);
xgetbv_val = XXH_xgetbv();
#if XXH_DISPATCH_AVX2
/* Validate that AVX2 is supported by the CPU */
if ((abcd[1] & XXH_AVX2_CPUID_MASK) != XXH_AVX2_CPUID_MASK)
return best;
/* Validate that the OS supports YMM registers */
if ((xgetbv_val & XXH_AVX2_XGETBV_MASK) != XXH_AVX2_XGETBV_MASK) {
XXH_debugPrint("AVX2 supported by the CPU, but not the OS.");
return best;
}
/* AVX2 supported */
XXH_debugPrint("AVX2 support detected.");
best = XXH_AVX2;
#endif
#if XXH_DISPATCH_AVX512
/* Check if AVX512F is supported by the CPU */
if ((abcd[1] & XXH_AVX512F_CPUID_MASK) != XXH_AVX512F_CPUID_MASK) {
XXH_debugPrint("AVX512F not supported by CPU");
return best;
}
/* Validate that the OS supports ZMM registers */
if ((xgetbv_val & XXH_AVX512F_XGETBV_MASK) != XXH_AVX512F_XGETBV_MASK) {
XXH_debugPrint("AVX512F supported by the CPU, but not the OS.");
return best;
}
/* AVX512F supported */
XXH_debugPrint("AVX512F support detected.");
best = XXH_AVX512;
#endif
#endif
return best;
}
/* === Vector implementations === */
/*!
* @internal
* @brief Defines the various dispatch functions.
*
* TODO: Consolidate?
*
* @param suffix The suffix for the functions, e.g. sse2 or scalar
* @param target XXH_TARGET_* or empty.
*/
#define XXH_DEFINE_DISPATCH_FUNCS(suffix, target) \
\
/* === XXH3, default variants === */ \
\
XXH_NO_INLINE target XXH64_hash_t \
XXHL64_default_##suffix(const void* XXH_RESTRICT input, size_t len) \
{ \
return XXH3_hashLong_64b_internal( \
input, len, XXH3_kSecret, sizeof(XXH3_kSecret), \
XXH3_accumulate_512_##suffix, XXH3_scrambleAcc_##suffix \
); \
} \
\
/* === XXH3, Seeded variants === */ \
\
XXH_NO_INLINE target XXH64_hash_t \
XXHL64_seed_##suffix(const void* XXH_RESTRICT input, size_t len, \
XXH64_hash_t seed) \
{ \
return XXH3_hashLong_64b_withSeed_internal( \
input, len, seed, XXH3_accumulate_512_##suffix, \
XXH3_scrambleAcc_##suffix, XXH3_initCustomSecret_##suffix \
); \
} \
\
/* === XXH3, Secret variants === */ \
\
XXH_NO_INLINE target XXH64_hash_t \
XXHL64_secret_##suffix(const void* XXH_RESTRICT input, size_t len, \
const void* secret, size_t secretLen) \
{ \
return XXH3_hashLong_64b_internal( \
input, len, secret, secretLen, \
XXH3_accumulate_512_##suffix, XXH3_scrambleAcc_##suffix \
); \
} \
\
/* === XXH3 update variants === */ \
\
XXH_NO_INLINE target XXH_errorcode \
XXH3_update_##suffix(XXH3_state_t* state, const void* input, size_t len) \
{ \
return XXH3_update(state, (const xxh_u8*)input, len, \
XXH3_accumulate_512_##suffix, XXH3_scrambleAcc_##suffix); \
} \
\
/* === XXH128 default variants === */ \
\
XXH_NO_INLINE target XXH128_hash_t \
XXHL128_default_##suffix(const void* XXH_RESTRICT input, size_t len) \
{ \
return XXH3_hashLong_128b_internal( \
input, len, XXH3_kSecret, sizeof(XXH3_kSecret), \
XXH3_accumulate_512_##suffix, XXH3_scrambleAcc_##suffix \
); \
} \
\
/* === XXH128 Secret variants === */ \
\
XXH_NO_INLINE target XXH128_hash_t \
XXHL128_secret_##suffix(const void* XXH_RESTRICT input, size_t len, \
const void* XXH_RESTRICT secret, size_t secretLen) \
{ \
return XXH3_hashLong_128b_internal( \
input, len, (const xxh_u8*)secret, secretLen, \
XXH3_accumulate_512_##suffix, XXH3_scrambleAcc_##suffix); \
} \
\
/* === XXH128 Seeded variants === */ \
\
XXH_NO_INLINE target XXH128_hash_t \
XXHL128_seed_##suffix(const void* XXH_RESTRICT input, size_t len, \
XXH64_hash_t seed) \
{ \
return XXH3_hashLong_128b_withSeed_internal(input, len, seed, \
XXH3_accumulate_512_##suffix, XXH3_scrambleAcc_##suffix, \
XXH3_initCustomSecret_##suffix); \
}
/* End XXH_DEFINE_DISPATCH_FUNCS */
#if XXH_DISPATCH_SCALAR
XXH_DEFINE_DISPATCH_FUNCS(scalar, /* nothing */)
#endif
XXH_DEFINE_DISPATCH_FUNCS(sse2, XXH_TARGET_SSE2)
#if XXH_DISPATCH_AVX2
XXH_DEFINE_DISPATCH_FUNCS(avx2, XXH_TARGET_AVX2)
#endif
#if XXH_DISPATCH_AVX512
XXH_DEFINE_DISPATCH_FUNCS(avx512, XXH_TARGET_AVX512)
#endif
#undef XXH_DEFINE_DISPATCH_FUNCS
/* ==== Dispatchers ==== */
typedef XXH64_hash_t (*XXH3_dispatchx86_hashLong64_default)(const void* XXH_RESTRICT, size_t);
typedef XXH64_hash_t (*XXH3_dispatchx86_hashLong64_withSeed)(const void* XXH_RESTRICT, size_t, XXH64_hash_t);
typedef XXH64_hash_t (*XXH3_dispatchx86_hashLong64_withSecret)(const void* XXH_RESTRICT, size_t, const void* XXH_RESTRICT, size_t);
typedef XXH_errorcode (*XXH3_dispatchx86_update)(XXH3_state_t*, const void*, size_t);
typedef struct {
XXH3_dispatchx86_hashLong64_default hashLong64_default;
XXH3_dispatchx86_hashLong64_withSeed hashLong64_seed;
XXH3_dispatchx86_hashLong64_withSecret hashLong64_secret;
XXH3_dispatchx86_update update;
} XXH_dispatchFunctions_s;
#define XXH_NB_DISPATCHES 4
/*!
* @internal
* @brief Table of dispatchers for @ref XXH3_64bits().
*
* @pre The indices must match @ref XXH_VECTOR_TYPE.
*/
static const XXH_dispatchFunctions_s XXH_kDispatch[XXH_NB_DISPATCHES] = {
#if XXH_DISPATCH_SCALAR
/* Scalar */ { XXHL64_default_scalar, XXHL64_seed_scalar, XXHL64_secret_scalar, XXH3_update_scalar },
#else
/* Scalar */ { NULL, NULL, NULL, NULL },
#endif
/* SSE2 */ { XXHL64_default_sse2, XXHL64_seed_sse2, XXHL64_secret_sse2, XXH3_update_sse2 },
#if XXH_DISPATCH_AVX2
/* AVX2 */ { XXHL64_default_avx2, XXHL64_seed_avx2, XXHL64_secret_avx2, XXH3_update_avx2 },
#else
/* AVX2 */ { NULL, NULL, NULL, NULL },
#endif
#if XXH_DISPATCH_AVX512
/* AVX512 */ { XXHL64_default_avx512, XXHL64_seed_avx512, XXHL64_secret_avx512, XXH3_update_avx512 }
#else
/* AVX512 */ { NULL, NULL, NULL, NULL }
#endif
};
/*!
* @internal
* @brief The selected dispatch table for @ref XXH3_64bits().
*/
static XXH_dispatchFunctions_s XXH_g_dispatch = { NULL, NULL, NULL, NULL };
typedef XXH128_hash_t (*XXH3_dispatchx86_hashLong128_default)(const void* XXH_RESTRICT, size_t);
typedef XXH128_hash_t (*XXH3_dispatchx86_hashLong128_withSeed)(const void* XXH_RESTRICT, size_t, XXH64_hash_t);
typedef XXH128_hash_t (*XXH3_dispatchx86_hashLong128_withSecret)(const void* XXH_RESTRICT, size_t, const void* XXH_RESTRICT, size_t);
typedef struct {
XXH3_dispatchx86_hashLong128_default hashLong128_default;
XXH3_dispatchx86_hashLong128_withSeed hashLong128_seed;
XXH3_dispatchx86_hashLong128_withSecret hashLong128_secret;
XXH3_dispatchx86_update update;
} XXH_dispatch128Functions_s;
/*!
* @internal
* @brief Table of dispatchers for @ref XXH3_128bits().
*
* @pre The indices must match @ref XXH_VECTOR_TYPE.
*/
static const XXH_dispatch128Functions_s XXH_kDispatch128[XXH_NB_DISPATCHES] = {
#if XXH_DISPATCH_SCALAR
/* Scalar */ { XXHL128_default_scalar, XXHL128_seed_scalar, XXHL128_secret_scalar, XXH3_update_scalar },
#else
/* Scalar */ { NULL, NULL, NULL, NULL },
#endif
/* SSE2 */ { XXHL128_default_sse2, XXHL128_seed_sse2, XXHL128_secret_sse2, XXH3_update_sse2 },
#if XXH_DISPATCH_AVX2
/* AVX2 */ { XXHL128_default_avx2, XXHL128_seed_avx2, XXHL128_secret_avx2, XXH3_update_avx2 },
#else
/* AVX2 */ { NULL, NULL, NULL, NULL },
#endif
#if XXH_DISPATCH_AVX512
/* AVX512 */ { XXHL128_default_avx512, XXHL128_seed_avx512, XXHL128_secret_avx512, XXH3_update_avx512 }
#else
/* AVX512 */ { NULL, NULL, NULL, NULL }
#endif
};
/*!
* @internal
* @brief The selected dispatch table for @ref XXH3_64bits().
*/
static XXH_dispatch128Functions_s XXH_g_dispatch128 = { NULL, NULL, NULL, NULL };
/*!
* @internal
* @brief Runs a CPUID check and sets the correct dispatch tables.
*/
static void XXH_setDispatch(void)
{
int vecID = XXH_featureTest();
XXH_STATIC_ASSERT(XXH_AVX512 == XXH_NB_DISPATCHES-1);
assert(XXH_SCALAR <= vecID && vecID <= XXH_AVX512);
#if !XXH_DISPATCH_SCALAR
assert(vecID != XXH_SCALAR);
#endif
#if !XXH_DISPATCH_AVX512
assert(vecID != XXH_AVX512);
#endif
#if !XXH_DISPATCH_AVX2
assert(vecID != XXH_AVX2);
#endif
XXH_g_dispatch = XXH_kDispatch[vecID];
XXH_g_dispatch128 = XXH_kDispatch128[vecID];
}
/* ==== XXH3 public functions ==== */
static XXH64_hash_t
XXH3_hashLong_64b_defaultSecret_selection(const void* input, size_t len,
XXH64_hash_t seed64, const xxh_u8* secret, size_t secretLen)
{
(void)seed64; (void)secret; (void)secretLen;
if (XXH_g_dispatch.hashLong64_default == NULL) XXH_setDispatch();
return XXH_g_dispatch.hashLong64_default(input, len);
}
XXH64_hash_t XXH3_64bits_dispatch(const void* input, size_t len)
{
return XXH3_64bits_internal(input, len, 0, XXH3_kSecret, sizeof(XXH3_kSecret), XXH3_hashLong_64b_defaultSecret_selection);
}
static XXH64_hash_t
XXH3_hashLong_64b_withSeed_selection(const void* input, size_t len,
XXH64_hash_t seed64, const xxh_u8* secret, size_t secretLen)
{
(void)secret; (void)secretLen;
if (XXH_g_dispatch.hashLong64_seed == NULL) XXH_setDispatch();
return XXH_g_dispatch.hashLong64_seed(input, len, seed64);
}
XXH64_hash_t XXH3_64bits_withSeed_dispatch(const void* input, size_t len, XXH64_hash_t seed)
{
return XXH3_64bits_internal(input, len, seed, XXH3_kSecret, sizeof(XXH3_kSecret), XXH3_hashLong_64b_withSeed_selection);
}
static XXH64_hash_t
XXH3_hashLong_64b_withSecret_selection(const void* input, size_t len,
XXH64_hash_t seed64, const xxh_u8* secret, size_t secretLen)
{
(void)seed64;
if (XXH_g_dispatch.hashLong64_secret == NULL) XXH_setDispatch();
return XXH_g_dispatch.hashLong64_secret(input, len, secret, secretLen);
}
XXH64_hash_t XXH3_64bits_withSecret_dispatch(const void* input, size_t len, const void* secret, size_t secretLen)
{
return XXH3_64bits_internal(input, len, 0, secret, secretLen, XXH3_hashLong_64b_withSecret_selection);
}
XXH_errorcode
XXH3_64bits_update_dispatch(XXH3_state_t* state, const void* input, size_t len)
{
if (XXH_g_dispatch.update == NULL) XXH_setDispatch();
return XXH_g_dispatch.update(state, (const xxh_u8*)input, len);
}
/* ==== XXH128 public functions ==== */
static XXH128_hash_t
XXH3_hashLong_128b_defaultSecret_selection(const void* input, size_t len,
XXH64_hash_t seed64, const void* secret, size_t secretLen)
{
(void)seed64; (void)secret; (void)secretLen;
if (XXH_g_dispatch128.hashLong128_default == NULL) XXH_setDispatch();
return XXH_g_dispatch128.hashLong128_default(input, len);
}
XXH128_hash_t XXH3_128bits_dispatch(const void* input, size_t len)
{
return XXH3_128bits_internal(input, len, 0, XXH3_kSecret, sizeof(XXH3_kSecret), XXH3_hashLong_128b_defaultSecret_selection);
}
static XXH128_hash_t
XXH3_hashLong_128b_withSeed_selection(const void* input, size_t len,
XXH64_hash_t seed64, const void* secret, size_t secretLen)
{
(void)secret; (void)secretLen;
if (XXH_g_dispatch128.hashLong128_seed == NULL) XXH_setDispatch();
return XXH_g_dispatch128.hashLong128_seed(input, len, seed64);
}
XXH128_hash_t XXH3_128bits_withSeed_dispatch(const void* input, size_t len, XXH64_hash_t seed)
{
return XXH3_128bits_internal(input, len, seed, XXH3_kSecret, sizeof(XXH3_kSecret), XXH3_hashLong_128b_withSeed_selection);
}
static XXH128_hash_t
XXH3_hashLong_128b_withSecret_selection(const void* input, size_t len,
XXH64_hash_t seed64, const void* secret, size_t secretLen)
{
(void)seed64;
if (XXH_g_dispatch128.hashLong128_secret == NULL) XXH_setDispatch();
return XXH_g_dispatch128.hashLong128_secret(input, len, secret, secretLen);
}
XXH128_hash_t XXH3_128bits_withSecret_dispatch(const void* input, size_t len, const void* secret, size_t secretLen)
{
return XXH3_128bits_internal(input, len, 0, secret, secretLen, XXH3_hashLong_128b_withSecret_selection);
}
XXH_errorcode
XXH3_128bits_update_dispatch(XXH3_state_t* state, const void* input, size_t len)
{
if (XXH_g_dispatch128.update == NULL) XXH_setDispatch();
return XXH_g_dispatch128.update(state, (const xxh_u8*)input, len);
}
#if defined (__cplusplus)
}
#endif
/*! @} */

86
third_party/xxhash/xxh_x86dispatch.h vendored
Unescape View File

@ -1,86 +0,0 @@
/*
* xxHash - XXH3 Dispatcher for x86-based targets
* Copyright (C) 2020 Yann Collet
*
* BSD 2-Clause License (https://www.opensource.org/licenses/bsd-license.php)
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following disclaimer
* in the documentation and/or other materials provided with the
* distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* You can contact the author at:
* - xxHash homepage: https://www.xxhash.com
* - xxHash source repository: https://github.com/Cyan4973/xxHash
*/
#ifndef XXH_X86DISPATCH_H_13563687684
#define XXH_X86DISPATCH_H_13563687684
#include "xxhash.h" /* XXH64_hash_t, XXH3_state_t */
#if defined (__cplusplus)
extern "C" {
#endif
XXH_PUBLIC_API XXH64_hash_t XXH3_64bits_dispatch(const void* input, size_t len);
XXH_PUBLIC_API XXH64_hash_t XXH3_64bits_withSeed_dispatch(const void* input, size_t len, XXH64_hash_t seed);
XXH_PUBLIC_API XXH64_hash_t XXH3_64bits_withSecret_dispatch(const void* input, size_t len, const void* secret, size_t secretLen);
XXH_PUBLIC_API XXH_errorcode XXH3_64bits_update_dispatch(XXH3_state_t* state, const void* input, size_t len);
XXH_PUBLIC_API XXH128_hash_t XXH3_128bits_dispatch(const void* input, size_t len);
XXH_PUBLIC_API XXH128_hash_t XXH3_128bits_withSeed_dispatch(const void* input, size_t len, XXH64_hash_t seed);
XXH_PUBLIC_API XXH128_hash_t XXH3_128bits_withSecret_dispatch(const void* input, size_t len, const void* secret, size_t secretLen);
XXH_PUBLIC_API XXH_errorcode XXH3_128bits_update_dispatch(XXH3_state_t* state, const void* input, size_t len);
#if defined (__cplusplus)
}
#endif
/* automatic replacement of XXH3 functions.
* can be disabled by setting XXH_DISPATCH_DISABLE_REPLACE */
#ifndef XXH_DISPATCH_DISABLE_REPLACE
# undef XXH3_64bits
# define XXH3_64bits XXH3_64bits_dispatch
# undef XXH3_64bits_withSeed
# define XXH3_64bits_withSeed XXH3_64bits_withSeed_dispatch
# undef XXH3_64bits_withSecret
# define XXH3_64bits_withSecret XXH3_64bits_withSecret_dispatch
# undef XXH3_64bits_update
# define XXH3_64bits_update XXH3_64bits_update_dispatch
# undef XXH128
# define XXH128 XXH3_128bits_withSeed_dispatch
# define XXH3_128bits XXH3_128bits_dispatch
# undef XXH3_128bits
# define XXH3_128bits XXH3_128bits_dispatch
# undef XXH3_128bits_withSeed
# define XXH3_128bits_withSeed XXH3_128bits_withSeed_dispatch
# undef XXH3_128bits_withSecret
# define XXH3_128bits_withSecret XXH3_128bits_withSecret_dispatch
# undef XXH3_128bits_update
# define XXH3_128bits_update XXH3_128bits_update_dispatch
#endif /* XXH_DISPATCH_DISABLE_REPLACE */
#endif /* XXH_X86DISPATCH_H_13563687684 */

43
third_party/xxhash/xxhash.c vendored
Unescape View File

@ -1,43 +0,0 @@
/*
* xxHash - Extremely Fast Hash algorithm
* Copyright (C) 2012-2020 Yann Collet
*
* BSD 2-Clause License (https://www.opensource.org/licenses/bsd-license.php)
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following disclaimer
* in the documentation and/or other materials provided with the
* distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* You can contact the author at:
* - xxHash homepage: https://www.xxhash.com
* - xxHash source repository: https://github.com/Cyan4973/xxHash
*/
/*
* xxhash.c instantiates functions defined in xxhash.h
*/
#define XXH_STATIC_LINKING_ONLY /* access advanced declarations */
#define XXH_IMPLEMENTATION /* access definitions */
#include "xxhash.h"

252
third_party/xxhash/xxhash.h vendored
Unescape View File

@ -266,7 +266,7 @@ extern "C" {
***************************************/
#define XXH_VERSION_MAJOR 0
#define XXH_VERSION_MINOR 8
#define XXH_VERSION_RELEASE 0
#define XXH_VERSION_RELEASE 1
#define XXH_VERSION_NUMBER (XXH_VERSION_MAJOR *100*100 + XXH_VERSION_MINOR *100 + XXH_VERSION_RELEASE)
/*!
@ -275,7 +275,7 @@ extern "C" {
* This is only useful when xxHash is compiled as a shared library, as it is
* independent of the version defined in the header.
*
* @return `XXH_VERSION_NUMBER` as of when the function was compiled.
* @return `XXH_VERSION_NUMBER` as of when the libray was compiled.
*/
XXH_PUBLIC_API unsigned XXH_versionNumber (void);
@ -1394,6 +1394,27 @@ static void* XXH_memcpy(void* dest, const void* src, size_t size)
/* note: use after variable declarations */
#define XXH_STATIC_ASSERT(c) do { enum { XXH_sa = 1/(int)(!!(c)) }; } while (0)
/*!
* @internal
* @def XXH_COMPILER_GUARD(var)
* @brief Used to prevent unwanted optimizations for @p var.
*
* It uses an empty GCC inline assembly statement with a register constraint
* which forces @p var into a general purpose register (eg eax, ebx, ecx
* on x86) and marks it as modified.
*
* This is used in a few places to avoid unwanted autovectorization (e.g.
* XXH32_round()). All vectorization we want is explicit via intrinsics,
* and _usually_ isn't wanted elsewhere.
*
* We also use it to prevent unwanted constant folding for AArch64 in
* XXH3_initCustomSecret_scalar().
*/
#ifdef __GNUC__
# define XXH_COMPILER_GUARD(var) __asm__ __volatile__("" : "+r" (var))
#else
# define XXH_COMPILER_GUARD(var) ((void)0)
#endif
/* *************************************
* Basic Types
@ -1703,11 +1724,12 @@ XXH_PUBLIC_API unsigned XXH_versionNumber (void) { return XXH_VERSION_NUMBER; }
* @ingroup impl
* @{
*/
static const xxh_u32 XXH_PRIME32_1 = 0x9E3779B1U; /*!< 0b10011110001101110111100110110001 */
static const xxh_u32 XXH_PRIME32_2 = 0x85EBCA77U; /*!< 0b10000101111010111100101001110111 */
static const xxh_u32 XXH_PRIME32_3 = 0xC2B2AE3DU; /*!< 0b11000010101100101010111000111101 */
static const xxh_u32 XXH_PRIME32_4 = 0x27D4EB2FU; /*!< 0b00100111110101001110101100101111 */
static const xxh_u32 XXH_PRIME32_5 = 0x165667B1U; /*!< 0b00010110010101100110011110110001 */
/* #define instead of static const, to be used as initializers */
#define XXH_PRIME32_1 0x9E3779B1U /*!< 0b10011110001101110111100110110001 */
#define XXH_PRIME32_2 0x85EBCA77U /*!< 0b10000101111010111100101001110111 */
#define XXH_PRIME32_3 0xC2B2AE3DU /*!< 0b11000010101100101010111000111101 */
#define XXH_PRIME32_4 0x27D4EB2FU /*!< 0b00100111110101001110101100101111 */
#define XXH_PRIME32_5 0x165667B1U /*!< 0b00010110010101100110011110110001 */
#ifdef XXH_OLD_NAMES
# define PRIME32_1 XXH_PRIME32_1
@ -1733,13 +1755,12 @@ static xxh_u32 XXH32_round(xxh_u32 acc, xxh_u32 input)
acc += input * XXH_PRIME32_2;
acc = XXH_rotl32(acc, 13);
acc *= XXH_PRIME32_1;
#if defined(__GNUC__) && defined(__SSE4_1__) && !defined(XXH_ENABLE_AUTOVECTORIZE)
#if (defined(__SSE4_1__) || defined(__aarch64__)) && !defined(XXH_ENABLE_AUTOVECTORIZE)
/*
* UGLY HACK:
* This inline assembly hack forces acc into a normal register. This is the
* only thing that prevents GCC and Clang from autovectorizing the XXH32
* loop (pragmas and attributes don't work for some reason) without globally
* disabling SSE4.1.
* A compiler fence is the only thing that prevents GCC and Clang from
* autovectorizing the XXH32 loop (pragmas and attributes don't work for some
* reason) without globally disabling SSE4.1.
*
* The reason we want to avoid vectorization is because despite working on
* 4 integers at a time, there are multiple factors slowing XXH32 down on
@ -1764,22 +1785,11 @@ static xxh_u32 XXH32_round(xxh_u32 acc, xxh_u32 input)
* can load data, while v3 can multiply. SSE forces them to operate
* together.
*
* How this hack works:
* __asm__("" // Declare an assembly block but don't declare any instructions
* : // However, as an Input/Output Operand,
* "+r" // constrain a read/write operand (+) as a general purpose register (r).
* (acc) // and set acc as the operand
* );
*
* Because of the 'r', the compiler has promised that seed will be in a
* general purpose register and the '+' says that it will be 'read/write',
* so it has to assume it has changed. It is like volatile without all the
* loads and stores.
*
* Since the argument has to be in a normal register (not an SSE register),
* each time XXH32_round is called, it is impossible to vectorize.
* This is also enabled on AArch64, as Clang autovectorizes it incorrectly
* and it is pointless writing a NEON implementation that is basically the
* same speed as scalar for XXH32.
*/
__asm__("" : "+r" (acc));
XXH_COMPILER_GUARD(acc);
#endif
return acc;
}
@ -1910,7 +1920,7 @@ XXH32_finalize(xxh_u32 h32, const xxh_u8* ptr, size_t len, XXH_alignment align)
XXH_FORCE_INLINE xxh_u32
XXH32_endian_align(const xxh_u8* input, size_t len, xxh_u32 seed, XXH_alignment align)
{
const xxh_u8* bEnd = input + len;
const xxh_u8* bEnd = input ? input + len : NULL;
xxh_u32 h32;
#if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
@ -2134,35 +2144,6 @@ typedef XXH64_hash_t xxh_u64;
# define U64 xxh_u64
#endif
/*!
* XXH_REROLL_XXH64:
* Whether to reroll the XXH64_finalize() loop.
*
* Just like XXH32, we can unroll the XXH64_finalize() loop. This can be a
* performance gain on 64-bit hosts, as only one jump is required.
*
* However, on 32-bit hosts, because arithmetic needs to be done with two 32-bit
* registers, and 64-bit arithmetic needs to be simulated, it isn't beneficial
* to unroll. The code becomes ridiculously large (the largest function in the
* binary on i386!), and rerolling it saves anywhere from 3kB to 20kB. It is
* also slightly faster because it fits into cache better and is more likely
* to be inlined by the compiler.
*
* If XXH_REROLL is defined, this is ignored and the loop is always rerolled.
*/
#ifndef XXH_REROLL_XXH64
# if (defined(__ILP32__) || defined(_ILP32)) /* ILP32 is often defined on 32-bit GCC family */ \
|| !(defined(__x86_64__) || defined(_M_X64) || defined(_M_AMD64) /* x86-64 */ \
|| defined(_M_ARM64) || defined(__aarch64__) || defined(__arm64__) /* aarch64 */ \
|| defined(__PPC64__) || defined(__PPC64LE__) || defined(__ppc64__) || defined(__powerpc64__) /* ppc64 */ \
|| defined(__mips64__) || defined(__mips64)) /* mips64 */ \
|| (!defined(SIZE_MAX) || SIZE_MAX < ULLONG_MAX) /* check limits */
# define XXH_REROLL_XXH64 1
# else
# define XXH_REROLL_XXH64 0
# endif
#endif /* !defined(XXH_REROLL_XXH64) */
#if (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==3))
/*
* Manual byteshift. Best for old compilers which don't inline memcpy.
@ -2285,11 +2266,12 @@ XXH_readLE64_align(const void* ptr, XXH_alignment align)
* @ingroup impl
* @{
*/
static const xxh_u64 XXH_PRIME64_1 = 0x9E3779B185EBCA87ULL; /*!< 0b1001111000110111011110011011000110000101111010111100101010000111 */
static const xxh_u64 XXH_PRIME64_2 = 0xC2B2AE3D27D4EB4FULL; /*!< 0b1100001010110010101011100011110100100111110101001110101101001111 */
static const xxh_u64 XXH_PRIME64_3 = 0x165667B19E3779F9ULL; /*!< 0b0001011001010110011001111011000110011110001101110111100111111001 */
static const xxh_u64 XXH_PRIME64_4 = 0x85EBCA77C2B2AE63ULL; /*!< 0b1000010111101011110010100111011111000010101100101010111001100011 */
static const xxh_u64 XXH_PRIME64_5 = 0x27D4EB2F165667C5ULL; /*!< 0b0010011111010100111010110010111100010110010101100110011111000101 */
/* #define rather that static const, to be used as initializers */
#define XXH_PRIME64_1 0x9E3779B185EBCA87ULL /*!< 0b1001111000110111011110011011000110000101111010111100101010000111 */
#define XXH_PRIME64_2 0xC2B2AE3D27D4EB4FULL /*!< 0b1100001010110010101011100011110100100111110101001110101101001111 */
#define XXH_PRIME64_3 0x165667B19E3779F9ULL /*!< 0b0001011001010110011001111011000110011110001101110111100111111001 */
#define XXH_PRIME64_4 0x85EBCA77C2B2AE63ULL /*!< 0b1000010111101011110010100111011111000010101100101010111001100011 */
#define XXH_PRIME64_5 0x27D4EB2F165667C5ULL /*!< 0b0010011111010100111010110010111100010110010101100110011111000101 */
#ifdef XXH_OLD_NAMES
# define PRIME64_1 XXH_PRIME64_1
@ -2331,126 +2313,26 @@ static xxh_u64 XXH64_avalanche(xxh_u64 h64)
static xxh_u64
XXH64_finalize(xxh_u64 h64, const xxh_u8* ptr, size_t len, XXH_alignment align)
{
#define XXH_PROCESS1_64 do { \
h64 ^= (*ptr++) * XXH_PRIME64_5; \
h64 = XXH_rotl64(h64, 11) * XXH_PRIME64_1; \
} while (0)
#define XXH_PROCESS4_64 do { \
h64 ^= (xxh_u64)(XXH_get32bits(ptr)) * XXH_PRIME64_1; \
ptr += 4; \
h64 = XXH_rotl64(h64, 23) * XXH_PRIME64_2 + XXH_PRIME64_3; \
} while (0)
#define XXH_PROCESS8_64 do { \
xxh_u64 const k1 = XXH64_round(0, XXH_get64bits(ptr)); \
ptr += 8; \
h64 ^= k1; \
h64 = XXH_rotl64(h64,27) * XXH_PRIME64_1 + XXH_PRIME64_4; \
} while (0)
/* Rerolled version for 32-bit targets is faster and much smaller. */
if (XXH_REROLL || XXH_REROLL_XXH64) {
len &= 31;
while (len >= 8) {
XXH_PROCESS8_64;
xxh_u64 const k1 = XXH64_round(0, XXH_get64bits(ptr));
ptr += 8;
h64 ^= k1;
h64 = XXH_rotl64(h64,27) * XXH_PRIME64_1 + XXH_PRIME64_4;
len -= 8;
}
if (len >= 4) {
XXH_PROCESS4_64;
h64 ^= (xxh_u64)(XXH_get32bits(ptr)) * XXH_PRIME64_1;
ptr += 4;
h64 = XXH_rotl64(h64, 23) * XXH_PRIME64_2 + XXH_PRIME64_3;
len -= 4;
}
while (len > 0) {
XXH_PROCESS1_64;
h64 ^= (*ptr++) * XXH_PRIME64_5;
h64 = XXH_rotl64(h64, 11) * XXH_PRIME64_1;
--len;
}
return XXH64_avalanche(h64);
} else {
switch(len & 31) {
case 24: XXH_PROCESS8_64;
/* fallthrough */
case 16: XXH_PROCESS8_64;
/* fallthrough */
case 8: XXH_PROCESS8_64;
return XXH64_avalanche(h64);
case 28: XXH_PROCESS8_64;
/* fallthrough */
case 20: XXH_PROCESS8_64;
/* fallthrough */
case 12: XXH_PROCESS8_64;
/* fallthrough */
case 4: XXH_PROCESS4_64;
return XXH64_avalanche(h64);
case 25: XXH_PROCESS8_64;
/* fallthrough */
case 17: XXH_PROCESS8_64;
/* fallthrough */
case 9: XXH_PROCESS8_64;
XXH_PROCESS1_64;
return XXH64_avalanche(h64);
case 29: XXH_PROCESS8_64;
/* fallthrough */
case 21: XXH_PROCESS8_64;
/* fallthrough */
case 13: XXH_PROCESS8_64;
/* fallthrough */
case 5: XXH_PROCESS4_64;
XXH_PROCESS1_64;
return XXH64_avalanche(h64);
case 26: XXH_PROCESS8_64;
/* fallthrough */
case 18: XXH_PROCESS8_64;
/* fallthrough */
case 10: XXH_PROCESS8_64;
XXH_PROCESS1_64;
XXH_PROCESS1_64;
return XXH64_avalanche(h64);
case 30: XXH_PROCESS8_64;
/* fallthrough */
case 22: XXH_PROCESS8_64;
/* fallthrough */
case 14: XXH_PROCESS8_64;
/* fallthrough */
case 6: XXH_PROCESS4_64;
XXH_PROCESS1_64;
XXH_PROCESS1_64;
return XXH64_avalanche(h64);
case 27: XXH_PROCESS8_64;
/* fallthrough */
case 19: XXH_PROCESS8_64;
/* fallthrough */
case 11: XXH_PROCESS8_64;
XXH_PROCESS1_64;
XXH_PROCESS1_64;
XXH_PROCESS1_64;
return XXH64_avalanche(h64);
case 31: XXH_PROCESS8_64;
/* fallthrough */
case 23: XXH_PROCESS8_64;
/* fallthrough */
case 15: XXH_PROCESS8_64;
/* fallthrough */
case 7: XXH_PROCESS4_64;
/* fallthrough */
case 3: XXH_PROCESS1_64;
/* fallthrough */
case 2: XXH_PROCESS1_64;
/* fallthrough */
case 1: XXH_PROCESS1_64;
/* fallthrough */
case 0: return XXH64_avalanche(h64);
}
}
/* impossible to reach */
XXH_ASSERT(0);
return 0; /* unreachable, but some compilers complain without it */
}
#ifdef XXH_OLD_NAMES
@ -2466,7 +2348,7 @@ XXH64_finalize(xxh_u64 h64, const xxh_u8* ptr, size_t len, XXH_alignment align)
XXH_FORCE_INLINE xxh_u64
XXH64_endian_align(const xxh_u8* input, size_t len, xxh_u64 seed, XXH_alignment align)
{
const xxh_u8* bEnd = input + len;
const xxh_u8* bEnd = input ? input + len : NULL;
xxh_u64 h64;
#if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
@ -2664,7 +2546,7 @@ XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src
return XXH_readBE64(src);
}
#ifndef XXH_NO_XXH3
/* *********************************************************************
* XXH3
@ -2679,7 +2561,9 @@ XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src
/* === Compiler specifics === */
#if defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* >= C99 */
#if ((defined(sun) || defined(__sun)) && __cplusplus) /* Solaris includes __STDC_VERSION__ with C++. Tested with GCC 5.5 */
# define XXH_RESTRICT /* disable */
#elif defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* >= C99 */
# define XXH_RESTRICT restrict
#else
/* Note: it might be useful to define __restrict or __restrict__ for some C++ compilers */
@ -3441,7 +3325,7 @@ XXH3_len_4to8_64b(const xxh_u8* input, size_t len, const xxh_u8* secret, XXH64_h
{
XXH_ASSERT(input != NULL);
XXH_ASSERT(secret != NULL);
XXH_ASSERT(4 <= len && len < 8);
XXH_ASSERT(4 <= len && len <= 8);
seed ^= (xxh_u64)XXH_swap32((xxh_u32)seed) << 32;
{ xxh_u32 const input1 = XXH_readLE32(input);
xxh_u32 const input2 = XXH_readLE32(input + len - 4);
@ -3457,7 +3341,7 @@ XXH3_len_9to16_64b(const xxh_u8* input, size_t len, const xxh_u8* secret, XXH64_
{
XXH_ASSERT(input != NULL);
XXH_ASSERT(secret != NULL);
XXH_ASSERT(8 <= len && len <= 16);
XXH_ASSERT(9 <= len && len <= 16);
{ xxh_u64 const bitflip1 = (XXH_readLE64(secret+24) ^ XXH_readLE64(secret+32)) + seed;
xxh_u64 const bitflip2 = (XXH_readLE64(secret+40) ^ XXH_readLE64(secret+48)) - seed;
xxh_u64 const input_lo = XXH_readLE64(input) ^ bitflip1;
@ -3527,7 +3411,7 @@ XXH_FORCE_INLINE xxh_u64 XXH3_mix16B(const xxh_u8* XXH_RESTRICT input,
* GCC generates much better scalar code than Clang for the rest of XXH3,
* which is why finding a more optimal codepath is an interest.
*/
__asm__ ("" : "+r" (seed64));
XXH_COMPILER_GUARD(seed64);
#endif
{ xxh_u64 const input_lo = XXH_readLE64(input);
xxh_u64 const input_hi = XXH_readLE64(input+8);
@ -3871,12 +3755,8 @@ XXH_FORCE_INLINE XXH_TARGET_AVX2 void XXH3_initCustomSecret_avx2(void* XXH_RESTR
* On GCC & Clang, marking 'dest' as modified will cause the compiler:
* - do not extract the secret from sse registers in the internal loop
* - use less common registers, and avoid pushing these reg into stack
* The asm hack causes Clang to assume that XXH3_kSecretPtr aliases with
* customSecret, and on aarch64, this prevented LDP from merging two
* loads together for free. Putting the loads together before the stores
* properly generates LDP.
*/
__asm__("" : "+r" (dest));
XXH_COMPILER_GUARD(dest);
# endif
/* GCC -O2 need unroll loop manually */
@ -3985,7 +3865,7 @@ XXH_FORCE_INLINE XXH_TARGET_SSE2 void XXH3_initCustomSecret_sse2(void* XXH_RESTR
* - do not extract the secret from sse registers in the internal loop
* - use less common registers, and avoid pushing these reg into stack
*/
__asm__("" : "+r" (dest));
XXH_COMPILER_GUARD(dest);
# endif
for (i=0; i < nbRounds; ++i) {
@ -4231,7 +4111,7 @@ XXH3_initCustomSecret_scalar(void* XXH_RESTRICT customSecret, xxh_u64 seed64)
* without hack: 2654.4 MB/s
* with hack: 3202.9 MB/s
*/
__asm__("" : "+r" (kSecretPtr));
XXH_COMPILER_GUARD(kSecretPtr);
#endif
/*
* Note: in debug mode, this overrides the asm optimization
@ -4396,7 +4276,7 @@ XXH3_mergeAccs(const xxh_u64* XXH_RESTRICT acc, const xxh_u8* XXH_RESTRICT secre
* without hack: 2063.7 MB/s
* with hack: 2560.7 MB/s
*/
__asm__("" : "+r" (result64));
XXH_COMPILER_GUARD(result64);
#endif
}
@ -5432,6 +5312,8 @@ XXH128_hashFromCanonical(const XXH128_canonical_t* src)
#endif /* XXH_NO_LONG_LONG */
#endif /* XXH_NO_XXH3 */
/*!
* @}
*/

Write Preview
Loading…
Cancel
Save