120 (#56)

wanghenshui · Jun 30, 2023 · 641ae17 · 641ae17
1 parent dbaa359
commit 641ae17
Show file tree

Hide file tree

Showing 3 changed files with 257 additions and 16 deletions.
diff --git a/README.md b/README.md
@@ -26,7 +26,7 @@ RSS使用仓库的release RSS [链接](https://github.com/wanghenshui/cppweeklyn
 - 杨杰
 - 阿刚
 - MSK
-- 不语 x11
+- 不语 x13
 - ywp
 - 范天泷
 - 罗盛波
@@ -63,7 +63,7 @@ RSS使用仓库的release RSS [链接](https://github.com/wanghenshui/cppweeklyn
 |                       |                       |                       |                       | [095](./posts/095.md) |[096](./posts/096.md) | [097](./posts/097.md)| [098](./posts/098.md) | [099](./posts/099.md) | [100](./posts/100.md) |
 | --------------------- | --------------------- | --------------------- | --------------------- | --------------------- | --------------------- | --------------------- | --------------------- | --------------------- | --------------------- |
 |  [101](./posts/101.md) | [102](./posts/102.md) | [103](./posts/103.md) | [104](./posts/104.md) | [105](./posts/105.md)  | [106](./posts/106.md) | [107](./posts/107.md) | [108](./posts/108.md) | [109](./posts/109.md) | [110](./posts/110.md) |
-|  [111](./posts/111.md) | [112](./posts/112.md) | [113](./posts/113.md) | [114](./posts/114.md)  | [115](./posts/115.md) | [116](./posts/116.md) | [117](./posts/117.md) | [118](./posts/118.md) | [119](./posts/119.md) |  |
+|  [111](./posts/111.md) | [112](./posts/112.md) | [113](./posts/113.md) | [114](./posts/114.md)  | [115](./posts/115.md) | [116](./posts/116.md) | [117](./posts/117.md) | [118](./posts/118.md) | [119](./posts/119.md) | [120](./posts/120.md) |
 
 
 

diff --git a/posts/119.md b/posts/119.md
@@ -176,24 +176,29 @@ int foo (int x, int y) {
 }
 ```
 
-生成的汇编更简单
+生成的汇编更简单 gcc 13.1 c++23 O2
 
 ```asm
-@@ -8,11 +8,7 @@ _Z3fooii:
-.cfi_startproc
-xorl    %eax, %eax
-cmpl    %esi, %edi
--    je    .L1
--    setg    %al
--    movzbl    %al, %eax
--    leal    -1(%rax,%rax), %eax
--.L1:
-+    setne    %al
-ret
-.cfi_endproc
- .LFE0:
+foo(int, int):
+  xorl %eax, %eax
+  cmpl %esi, %edi
+  setne %al
+  ret
 ```
 
+注释掉assume的效果
+
+```asm
+foo(int, int):
+  xorl %eax, %eax
+  cmpl %esi, %edi
+  je .L1
+  setg %al
+  movzbl %al, %eax
+  leal -1(%rax,%rax), %eax
+.L1:
+  ret
+```
 - [C++20 Coroutines Driving a Job System](https://poniesandlight.co.uk/reflect/coroutines_job_system/)
 
 代码在这里 https://github.com/tgfrerer/pal_tasks/tree/main

diff --git a/posts/120.md b/posts/120.md
@@ -0,0 +1,236 @@
+---
+layout: post
+title:  第120期
+---
+# C++ 中文周刊 第120期
+
+
+[周刊项目地址](https://github.com/wanghenshui/cppweeklynews)
+
+公众号
+
+<img src="https://wanghenshui.github.io/cppweeklynews/assets/code.png" alt=""  width="30%">
+
+
+
+RSS https://github.com/wanghenshui/cppweeklynews/releases.atom
+
+欢迎投稿，推荐或自荐文章/软件/资源等
+
+请[提交 issue](https://github.com/wanghenshui/cppweeklynews/issues)
+
+感谢不语赞助
+
+---
+
+## 资讯
+
+标准委员会动态/ide/编译器信息放在这里
+
+[编译器信息最新动态推荐关注hellogcc公众号 本周更新 2023-06-28 第208期 ](https://mp.weixin.qq.com/s/NBrrr23WAZRzvamvtdZ0Yg)
+
+
+## 文章
+
+- [Did you know about gnu::vector_size extension?](https://github.com/tip-of-the-week/cpp/blob/master/tips/336.md)
+
+```cpp
+int main() {
+  using v4si = int [[gnu::vector_size(4 * sizeof(int))]];
+
+  v4si a = {1, 2, 3, 4};
+  v4si b = {4, 3, 2, 1};
+  v4si c;
+
+  c = a + b;
+  std::cout << c[0] << c[1] << c[2] << c[3]; // prints 5555
+}
+```
+
+这玩意是给simd方便的。看gcc样例
+
+```c
+#include <immintrin.h>
+
+typedef unsigned char u8x16 __attribute__ ((vector_size (16)));
+typedef unsigned int  u32x4 __attribute__ ((vector_size (16)));
+
+typedef union {
+        __m128i mm;
+        u8x16   u8;
+        u32x4   u32;
+} v128;
+
+v128 x, y = { 0 };
+memcpy (&x, ptr, sizeof x);
+y.u8  += 0x80;
+x.mm  = _mm_adds_epu8 (x.mm, y.mm);
+x.u32 &= 0xffffff;
+
+/* Instead of a variable, a compound literal may be used to pass the
+   return value of an intrinsic call to a function expecting the union: */
+v128 foo (v128);
+x = foo ((v128) {_mm_adds_epu8 (x.mm, y.mm)});
+```
+
+- [移植MySQL 8.0的函数跟踪库到rocksdb](https://zhuanlan.zhihu.com/p/638545766)
+
+这个debug宏很好用，方便阅读，代码在这里 https://github.com/nosql-cn/AxeDB
+
+不过既然已经编译了，通过clangd 应该也能抓到堆栈。不过静态的堆栈没有这种运行时堆栈有意义，运行时的更直观一些
+
+- [Finite State Machine with std::variant](https://www.cppstories.com/2023/finite-state-machines-variant-cpp/)
+
+知乎有人翻译成中文了，挺好。 https://zhuanlan.zhihu.com/p/639886110
+
+原理就是`std::variant` + `std::visit`
+
+
+- [C++ standard library ABI compatibility](https://maskray.me/blog/2023-06-25-c++-standard-library-abi-compatibility)
+
+看不懂
+
+- [Valgrind - A Framework for Heavyweight Dynamic Binary Instrumentation 论文翻译](https://zhuanlan.zhihu.com/p/640105594)
+
+了解一波
+
+- [从源码吃透共识协议：braft 日志复制](https://zhuanlan.zhihu.com/p/635963776)
+
+代码走读，感兴趣的可以看看
+
+- [Polymorphic types aren’t trivially relocatable](https://quuxplusone.github.io/blog/2023/06/24/polymorphic-types-arent-trivially-relocatable/)
+
+还是讲他的relocatable提案
+
+- [AddressSanitizer continue_on_error](https://devblogs.microsoft.com/cppblog/addresssanitizer-continue_on_error/)
+
+介绍vs上体验AddressSanitizer新特性COE(continue_on_error)
+
+用 ` -fsanitizer=address` 同时设置
+```bash
+    set ASAN_OPTIONS=continue_on_error=1
+    set ASAN_OPTIONS=continue_on_error=2
+```
+
+- [Dynamic bit shuffle using AVX-512](https://lemire.me/blog/2023/06/29/dynamic-bit-shuffle-using-avx-512/)
+
+打散一个64字节的数的场景
+
+```cpp
+uint64_t w = some value;
+uint8_t indexes[64] = {63, 62, 61, 60, 59, 58, 57, 56, 55, 54, 53, 52, 51,
+                       50, 49, 48, 47, 46, 45, 44, 43, 42, 41, 40, 39, 38,
+                       37, 36, 35, 34, 33, 32, 31, 30, 29, 28, 27, 26, 25,
+                       24, 23, 22, 21, 20, 19, 18, 17, 16, 15, 14, 13, 12,
+                       11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0};
+bit_shuffle(w, indexes); // returns a reversed version 
+```
+
+可能的实现
+
+```cpp
+uint64_t slow_bit_shuffle(uint64_t w, uint8_t indexes[64]) {
+  uint64_t out{};
+  for (size_t i = 0; i < 64; i++) {
+    bool bit_set = w & (uint64_t(1) << indexes[i]);
+    out |= (uint64_t(bit_set) << i);
+  }
+  return out;
+}
+```
+考虑avx512
+
+```cpp
+uint64_t bit_shuffle(uint64_t w, uint8_t indexes[64]) {
+  __mmask64 as_mask = _cvtu64_mask64(w);
+  __m512i as_vec_register =
+  _mm512_maskz_mov_epi8(as_mask, _mm512_set1_epi8(0xFF));
+  __m512i as_vec_register_shuf =
+  _mm512_permutexvar_epi8(_mm512_loadu_si512(indexes), as_vec_register);
+  return _cvtmask64_u64(_mm512_movepi8_mask(as_vec_register_shuf));
+}
+```
+
+快点
+
+```c++
+uint64_t faster_bit_shuffle(uint64_t w, uint8_t indexes[64]) {
+  __m512i as_vec_register = _mm512_set1_epi64(w);
+  __mmask64 as_mask = _mm512_bitshuffle_epi64_mask(as_vec_register,
+     _mm512_loadu_si512(indexes));
+  return _cvtmask64_u64(as_mask);
+}
+```
+
+- [Hiding Memory Latency With In-Order CPU Cores OR How Compilers Optimize Your Code](https://johnnysswlab.com/hiding-memory-latency-with-in-order-cpu-cores-or-how-compilers-optimize-your-code/)
+
+
+有些芯片性能太好掩盖了一些代码优化的潜力，讲了几个场景 循环展开/pipeline对于老芯片带来的性能提升。观点挺有意思
+
+这个博客 反复介绍过多次  https://johnnysswlab.com/ 推荐收藏没事看看
+
+- [My way to get into conferences](https://www.sandordargo.com/blog/2023/06/28/how-to-get-into-conferences)
+
+讲自己的演讲经历，如何演讲，找话题找组织，如何保证感兴趣，等等
+
+话说，想组织一个c++讨论演讲，怎么才能组织起人来看，找到演讲者呢。
+
+之前看祁宇(qicosmos)搞过，太正式了其实。类似meetingcpp这种网络会议直播模式就行
+
+大家给给点子，我想整一个。没话题其实可以以国外视频转述一遍。加深理解。
+
+- [How to wait for multiple C++ coroutines to complete before propagating failure, unhelpful lambda](https://devblogs.microsoft.com/oldnewthing/20230627-00/?p=108375)
+- [How to wait for multiple C++ coroutines to complete before propagating failure, false hope](https://devblogs.microsoft.com/oldnewthing/20230628-00/?p=108377)
+- [How to wait for multiple C++ coroutines to complete before propagating failure, peeling away at a tuple](https://devblogs.microsoft.com/oldnewthing/20230629-00/?p=108380)
+
+raymond chen真能写。我一个没看
+
+- [60 terrible tips for a C++ developer](https://pvs-studio.com/en/blog/posts/cpp/1053/)
+
+讲一些用法，尽量避免。没啥说的。有些有点搞笑。这里就不列举了
+## 视频
+
+- [C++ Weekly - Ep 382 - The Static Initialization Order Fiasco and C++20's constinit](https://www.youtube.com/watch?v=rEwijXgC_Kg&ab_channel=C%2B%2BWeeklyWithJasonTurner)
+
+讲constinit的。没啥说的。能用就用
+
+- [[MUC++] Ivica Bogosavljevic - Introduction to Hardware Efficiency](https://www.youtube.com/watch?v=mJJ5lTXvptQ&ab_channel=MUCplusplus)
+
+这个还是很值得一看的，这个哥们是性能调优专家，有个博客搜索权重挺高 https://johnnysswlab.com/
+
+这个改天我转b站上。或者谁有空给传一下
+
+- [Timur Doumler: C++ and Safety](https://www.youtube.com/watch?v=WsswOOTxlIM&ab_channel=SwedenCpp)
+
+这个是技术债了，异常问题异常安全等等，我看不进去。这里标记个TODO改天整理一下观点（或者找chatgpt老师提取一下。。。）
+
+- [Low-Latency Trading Systems in C++: Templated Meta-State Machines in HFT - Jason McGuiness - ACCU 23](https://www.youtube.com/watch?v=FnMfhWiSweo&ab_channel=ACCUConference)
+
+代码https://sourceforge.net/p/libjmmcg/git-repo/ci/ACCUConf2023/tree/libjmmcg/
+
+挺能high的，没看完 PPT https://github.com/wanghenshui/wanghenshui.github.io/blob/master/assets/ACCUConf2023_JMMcG_e0d2d_PRESENTED.pdf
+
+- [Applied C++20 Coroutines - Jim Pascoe - ACCU 2023](https://www.youtube.com/watch?v=ltesWt-92xw&ab_channel=ACCUConference)
+
+ppt https://jamespascoe.github.io/accu2023/
+
+看个乐
+## 开源项目需要人手
+
+- [asteria](https://github.com/lhmouse/asteria) 一个脚本语言，可嵌入，长期找人，希望胖友们帮帮忙，也可以加群384042845和作者对线
+- [Unilang](https://github.com/linuxdeepin/unilang) deepin的一个通用编程语言，点子有点意思，也缺人，感兴趣的可以github讨论区或者deepin论坛看一看。这里也挂着长期推荐了
+- [gcc-mcf](https://gcc-mcf.lhmouse.com/) 懂的都懂
+
+
+## 新项目介绍/版本更新
+
+- vulkan.cpp支持module了，看看怎么封装的？  https://github.com/KhronosGroup/Vulkan-Hpp/blob/main/vulkan/vulkan.cppm
+- https://github.com/bkryza/clang-uml  支持用clang直接画图！不过得编译，用compile database.json来生成。不知道大项目的效果。感觉周末可以试验一下 话说plant uml太难看了，不能直接生成mermaid么
+- https://github.com/AMDResearch/omnitrace amd 性能工具。仅限linux可用
+---
+
+
+
+[本文永久链接](https://wanghenshui.github.io/cppweeklynews/posts/120.html)
+
+如果有疑问评论最好在上面链接到评论区里评论，这样方便搜索，微信公众号有点封闭/知乎吞评论