nodejs · lucshi · Oct 20, 2022 · Oct 20, 2022 · Oct 21, 2022
diff --git a/deps/base64/base64.gyp b/deps/base64/base64.gyp
@@ -46,13 +46,15 @@
             'HAVE_SSE42=1',
             'HAVE_AVX=1',
             'HAVE_AVX2=1',
+            'HAVE_AVX512=1',
           ],
           'dependencies': [
             'base64_ssse3',
             'base64_sse41',
             'base64_sse42',
             'base64_avx',
             'base64_avx2',
+            'base64_avx512',
           ],
         }, {
           'sources': [
@@ -61,6 +63,7 @@
             'base64/lib/arch/sse42/codec.c',
             'base64/lib/arch/avx/codec.c',
             'base64/lib/arch/avx2/codec.c',
+            'base64/lib/arch/avx512/codec.c',
           ],
         }],
       ],
@@ -162,6 +165,30 @@
       ],
     },
 
+    {
+      'target_name': 'base64_avx512',
+      'type': 'static_library',
+      'include_dirs': [ 'base64/include', 'base64/lib' ],
+      'sources': [ 'base64/lib/arch/avx512/codec.c' ],
+      'defines': [ 'BASE64_STATIC_DEFINE', 'HAVE_AVX512=1' ],
+      'conditions': [
+        [ 'OS!="win"', {
+          'cflags': [ '-mavx512vl', '-mavx512vbmi' ],
+          'xcode_settings': {
+            'OTHER_CFLAGS': [ '-mavx512vl', '-mavx512vbmi' ]
+          },
+        }, {
+          'msvs_settings': {
+            'VCCLCompilerTool': {
+              'AdditionalOptions': [
+                '/arch:AVX512'
+              ],
+            },
+          },
+        }],
+      ],
+    },
+
     {
       'target_name': 'base64_neon32',
       'type': 'static_library',

diff --git a/deps/base64/base64/CMakeLists.txt b/deps/base64/base64/CMakeLists.txt
@@ -17,7 +17,7 @@ if (POLICY CMP0127)
     cmake_policy(SET CMP0127 NEW)
 endif()
 
-project(base64 LANGUAGES C VERSION 0.4.0)
+project(base64 LANGUAGES C VERSION 0.5.0)
 
 include(GNUInstallDirs)
 include(CMakeDependentOption)
@@ -62,6 +62,8 @@ cmake_dependent_option(BASE64_WITH_AVX "add AVX codepath" ON ${_IS_X86} OFF)
 add_feature_info(AVX BASE64_WITH_AVX "add AVX codepath")
 cmake_dependent_option(BASE64_WITH_AVX2 "add AVX 2 codepath" ON ${_IS_X86} OFF)
 add_feature_info(AVX2 BASE64_WITH_AVX2 "add AVX2 codepath")
+cmake_dependent_option(BASE64_WITH_AVX512 "add AVX 512 codepath" ON ${_IS_X86} OFF)
+add_feature_info(AVX2 BASE64_WITH_AVX512 "add AVX512 codepath")
 
 cmake_dependent_option(BASE64_WITH_NEON32 "add NEON32 codepath" OFF _TARGET_ARCH_arm OFF)
 add_feature_info(NEON32 BASE64_WITH_NEON32 "add NEON32 codepath")
@@ -118,6 +120,7 @@ add_library(base64
     lib/arch/sse42/codec.c
     lib/arch/avx/codec.c
     lib/arch/avx2/codec.c
+    lib/arch/avx512/codec.c
 
     lib/arch/neon32/codec.c
     lib/arch/neon64/codec.c
@@ -206,6 +209,7 @@ if (_TARGET_ARCH STREQUAL "x86" OR _TARGET_ARCH STREQUAL "x64")
     configure_codec(SSE42 __SSSE4_2__)
     configure_codec(AVX)
     configure_codec(AVX2)
+    configure_codec(AVX512)
 
 elseif (_TARGET_ARCH STREQUAL "arm")
     set(BASE64_NEON32_CFLAGS "${COMPILE_FLAGS_NEON32}" CACHE STRING "the NEON32 compile flags (for 'lib/arch/neon32/codec.c')")

diff --git a/deps/base64/base64/LICENSE b/deps/base64/base64/LICENSE
@@ -1,7 +1,7 @@
 Copyright (c) 2005-2007, Nick Galbreath
-Copyright (c) 2013-2019, Alfred Klomp
-Copyright (c) 2015-2017, Wojciech Mula
+Copyright (c) 2015-2018, Wojciech Muła
 Copyright (c) 2016-2017, Matthieu Darbois
+Copyright (c) 2013-2022, Alfred Klomp
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without

diff --git a/deps/base64/base64/Makefile b/deps/base64/base64/Makefile
@@ -4,6 +4,7 @@ CFLAGS += -std=c99 -O3 -Wall -Wextra -pedantic
 OBJCOPY ?= objcopy
 
 OBJS = \
+  lib/arch/avx512/codec.o \
   lib/arch/avx2/codec.o \
   lib/arch/generic/codec.o \
   lib/arch/neon32/codec.o \
@@ -16,6 +17,7 @@ OBJS = \
   lib/codec_choose.o \
   lib/tables/tables.o
 
+HAVE_AVX512 = 0
 HAVE_AVX2   = 0
 HAVE_NEON32 = 0
 HAVE_NEON64 = 0
@@ -26,6 +28,9 @@ HAVE_AVX    = 0
 
 # The user should supply compiler flags for the codecs they want to build.
 # Check which codecs we're going to include:
+ifdef AVX512_CFLAGS
+  HAVE_AVX512 = 1
+endif
 ifdef AVX2_CFLAGS
   HAVE_AVX2 = 1
 endif
@@ -64,7 +69,8 @@ lib/libbase64.o: $(OBJS)
 	$(OBJCOPY) --keep-global-symbols=lib/exports.txt $@
 
 lib/config.h:
-	@echo "#define HAVE_AVX2   $(HAVE_AVX2)"    > $@
+	@echo "#define HAVE_AVX512 $(HAVE_AVX512)"  > $@
+	@echo "#define HAVE_AVX2   $(HAVE_AVX2)"   >> $@
 	@echo "#define HAVE_NEON32 $(HAVE_NEON32)" >> $@
 	@echo "#define HAVE_NEON64 $(HAVE_NEON64)" >> $@
 	@echo "#define HAVE_SSSE3  $(HAVE_SSSE3)"  >> $@
@@ -75,6 +81,7 @@ lib/config.h:
 $(OBJS): lib/config.h
 $(OBJS): CFLAGS += -Ilib
 
+lib/arch/avx512/codec.o: CFLAGS += $(AVX512_CFLAGS)
 lib/arch/avx2/codec.o:   CFLAGS += $(AVX2_CFLAGS)
 lib/arch/neon32/codec.o: CFLAGS += $(NEON32_CFLAGS)
 lib/arch/neon64/codec.o: CFLAGS += $(NEON64_CFLAGS)

diff --git a/deps/base64/base64/README.md b/deps/base64/base64/README.md
@@ -3,7 +3,7 @@
 [![Build Status](https://github.com/aklomp/base64/actions/workflows/test.yml/badge.svg)](https://github.com/aklomp/base64/actions/workflows/test.yml)
 
 This is an implementation of a base64 stream encoding/decoding library in C99
-with SIMD (AVX2, NEON, AArch64/NEON, SSSE3, SSE4.1, SSE4.2, AVX) and
+with SIMD (AVX2, AVX512, NEON, AArch64/NEON, SSSE3, SSE4.1, SSE4.2, AVX) and
 [OpenMP](http://www.openmp.org) acceleration. It also contains wrapper functions
 to encode/decode simple length-delimited strings. This library aims to be:
 
@@ -19,6 +19,10 @@ will pick an optimized codec that lets it encode/decode 12 or 24 bytes at a
 time, which gives a speedup of four or more times compared to the "plain"
 bytewise codec.
 
+AVX512 support is only for encoding at present, utilizing the AVX512 VL and VBMI
+instructions. Decoding part reused AVX2 implementations. For CPUs later than
+Cannonlake (manufactured in 2018) supports these instructions.
+
 NEON support is hardcoded to on or off at compile time, because portable
 runtime feature detection is unavailable on ARM.
 
@@ -59,6 +63,9 @@ optimizations described by Wojciech Muła in a
 [articles](http://0x80.pl/notesen/2016-01-17-sse-base64-decoding.html).
 His own code is [here](https://github.com/WojciechMula/toys/tree/master/base64).
 
+The AVX512 encoder is based on code from Wojciech Muła's
+[base64simd](https://github.com/WojciechMula/base64simd) library.
+
 The OpenMP implementation was added by Ferry Toth (@htot) from [Exalon Delft](http://www.exalondelft.nl).
 
 ## Building
@@ -76,8 +83,8 @@ To compile just the "plain" library without SIMD codecs, type:
 make lib/libbase64.o
 ```
 
-Optional SIMD codecs can be included by specifying the `AVX2_CFLAGS`, `NEON32_CFLAGS`, `NEON64_CFLAGS`,
-`SSSE3_CFLAGS`, `SSE41_CFLAGS`, `SSE42_CFLAGS` and/or `AVX_CFLAGS` environment variables.
+Optional SIMD codecs can be included by specifying the `AVX2_CFLAGS`, `AVX512_CFLAGS`,
+`NEON32_CFLAGS`, `NEON64_CFLAGS`, `SSSE3_CFLAGS`, `SSE41_CFLAGS`, `SSE42_CFLAGS` and/or `AVX_CFLAGS` environment variables.
 A typical build invocation on x86 looks like this:
 
 ```sh
@@ -93,6 +100,15 @@ Example:
 AVX2_CFLAGS=-mavx2 make
 ```
 
+### AVX512
+
+To build and include the AVX512 codec, set the `AVX512_CFLAGS` environment variable to a value that will turn on AVX512 support in your compiler, typically `-mavx512vl -mavx512vbmi`.
+Example:
+
+```sh
+AVX512_CFLAGS="-mavx512vl -mavx512vbmi" make
+```
+
 The codec will only be used if runtime feature detection shows that the target machine supports AVX2.
 
 ### SSSE3
@@ -208,6 +224,7 @@ Mainly there for testing purposes, this is also useful on ARM where the only way
 The following constants can be used:
 
 - `BASE64_FORCE_AVX2`
+- `BASE64_FORCE_AVX512`
 - `BASE64_FORCE_NEON32`
 - `BASE64_FORCE_NEON64`
 - `BASE64_FORCE_PLAIN`
@@ -434,7 +451,7 @@ x86 processors
 | i7-4770 @ 3.4 GHz DDR1600 OPENMP 4 thread | 4884\*    | 7099\*    | 4917\*    | 7057\*    | 4799\*  | 7143\*  | 4902\*   | 7219\*   |
 | i7-4770 @ 3.4 GHz DDR1600 OPENMP 8 thread | 5212\*    | 8849\*    | 5284\*    | 9099\*    | 5289\*  | 9220\*  | 4849\*   | 9200\*   |
 | i7-4870HQ @ 2.5 GHz                       | 1471\*    | 3066\*    | 6721\*    | 6962\*    | 7015\*  | 8267\*  | 8328\*   | 11576\*  |
-| i5-4590S @ 3.0 GHz                        | 3356      | 3197      | 4363      | 6104      | 4243    | 6233    | 4160     | 6344     |
+| i5-4590S @ 3.0 GHz                        | 3356      | 3197      | 4363      | 6104      | 4243\*  | 6233    | 4160\*   | 6344     |
 | Xeon X5570 @ 2.93 GHz                     | 2161      | 1508      | 3160      | 3915      | -       | -       | -        | -        |
 | Pentium4 @ 3.4 GHz                        | 896       | 740       | -         | -         | -       | -       | -        | -        |
 | Atom N270                                 | 243       | 266       | 508       | 387       | -       | -       | -        | -        |

diff --git a/deps/base64/base64/cmake/Modules/TargetSIMDInstructionSet.cmake b/deps/base64/base64/cmake/Modules/TargetSIMDInstructionSet.cmake
@@ -21,6 +21,7 @@ macro(define_SIMD_compile_flags)
         set(COMPILE_FLAGS_SSE42 "-msse4.2")
         set(COMPILE_FLAGS_AVX "-mavx")
         set(COMPILE_FLAGS_AVX2 "-mavx2")
+        set(COMPILE_FLAGS_AVX512 "-mavx512vl -mavx512vbmi")
 
         #arm
         set(COMPILE_FLAGS_NEON32 "-mfpu=neon")
@@ -30,5 +31,6 @@ macro(define_SIMD_compile_flags)
         set(COMPILE_FLAGS_SSE42 " ")
         set(COMPILE_FLAGS_AVX "/arch:AVX")
         set(COMPILE_FLAGS_AVX2 "/arch:AVX2")
+        set(COMPILE_FLAGS_AVX512 "/arch:AVX512")
     endif()
 endmacro(define_SIMD_compile_flags)
diff --git a/deps/base64/base64/cmake/config.h.in b/deps/base64/base64/cmake/config.h.in
@@ -16,6 +16,9 @@
 #cmakedefine01 BASE64_WITH_AVX2
 #define HAVE_AVX2 BASE64_WITH_AVX2
 
+#cmakedefine01 BASE64_WITH_AVX512
+#define HAVE_AVX512 BASE64_WITH_AVX512
+
 #cmakedefine01 BASE64_WITH_NEON32
 #define HAVE_NEON32 BASE64_WITH_NEON32
 

diff --git a/deps/base64/base64/include/libbase64.h b/deps/base64/base64/include/libbase64.h
@@ -53,6 +53,7 @@ extern "C" {
 #define BASE64_FORCE_SSE41	(1 << 5)
 #define BASE64_FORCE_SSE42	(1 << 6)
 #define BASE64_FORCE_AVX	(1 << 7)
+#define BASE64_FORCE_AVX512	(1 << 8)
 
 struct base64_state {
 	int eof;

diff --git a/deps/base64/base64/lib/arch/avx/codec.c b/deps/base64/base64/lib/arch/avx/codec.c
@@ -11,19 +11,43 @@
 #if HAVE_AVX
 #include <immintrin.h>
 
+// Only enable inline assembly on supported compilers.
+#ifndef BASE64_AVX_USE_ASM
+# if defined(__GNUC__) || defined(__clang__)
+#  define BASE64_AVX_USE_ASM 1
+# else
+#  define BASE64_AVX_USE_ASM 0
+# endif
+#endif
+
 #include "../ssse3/dec_reshuffle.c"
 #include "../ssse3/dec_loop.c"
-#include "../ssse3/enc_translate.c"
-#include "../ssse3/enc_reshuffle.c"
-#include "../ssse3/enc_loop.c"
+
+#if BASE64_AVX_USE_ASM
+# include "enc_loop_asm.c"
+#else
+# include "../ssse3/enc_translate.c"
+# include "../ssse3/enc_reshuffle.c"
+# include "../ssse3/enc_loop.c"
+#endif
 
 #endif	// HAVE_AVX
 
 BASE64_ENC_FUNCTION(avx)
 {
 #if HAVE_AVX
 	#include "../generic/enc_head.c"
+
+	// For supported compilers, use a hand-optimized inline assembly
+	// encoder. Otherwise fall back on the SSSE3 encoder, but compiled with
+	// AVX flags to generate better optimized AVX code.
+
+#if BASE64_AVX_USE_ASM
+	enc_loop_avx(&s, &slen, &o, &olen);
+#else
 	enc_loop_ssse3(&s, &slen, &o, &olen);
+#endif
+
 	#include "../generic/enc_tail.c"
 #else
 	BASE64_ENC_STUB