zoukankan      html  css  js  c++  java
  • 使用纯粹的ABAP位操作实现两个整数相加

    Recently I came across this very funny picture and I would like to share with you.

    This picture shows totally five different approaches to implement “a + b”, and how your brain reacts when you read these code

    From this picture we know that for most guys, they will be crazy when they read the source code of the last solution as the code is really difficult to understand compared with all other four solutions.

    The idea of the last solution is to achieve the add operation via bitwise operation &, | and ^. See detail of these three operations via wikipedia.
    You might be interested with how ABAP can play around with these bitwise operation from two of my blogs:

    Back to the last solution to implement a + b using bitwise operation, the code actually conveys the idea how the Adder is designed in electronics. See detail design in Wikipedia.

    The idea is to use & to determine whether there is a carry about the add result of current bit, and use | to store the add result of current bit.

    For example, how 2 + 3 = 5 works via bitwise operation:

    And here below is my implementation using ABAP( Only integers which >= 0 are supported ).

    REPORT zint.
    
    PARAMETERS: a TYPE int4 OBLIGATORY DEFAULT 100,
                b TYPE int4 OBLIGATORY DEFAULT 100.
    
    DATA: threshold TYPE int4.
    
    FORM add USING a TYPE int4 b TYPE int4 CHANGING cv_result TYPE int4.
      DATA: n TYPE int4 VALUE 0,
            c TYPE int4 VALUE 0.
    
      DATA: i TYPE int4 VALUE 1.
      DATA: boolean_a TYPE abap_bool,
            boolean_b TYPE abap_bool,
            _a        TYPE int4,
            _b        TYPE int4,
            aa TYPE int4,
            bb TYPE int4.
    
      DATA(wrapper_one) = zcl_integer=>value_of( 1 ).
      DATA(wrapper_c) = zcl_integer=>value_of( c ).
    
      aa = a.
      bb = b.
      WHILE i < threshold.
        DATA(wrapper_a) = zcl_integer=>value_of( aa ).
        DATA(wrapper_b) = zcl_integer=>value_of( bb ).
        boolean_a = boolc( wrapper_a->and( wrapper_one )->get_raw_value( ) EQ 1 ).
        boolean_b = boolc( wrapper_b->and( wrapper_one )->get_raw_value( ) EQ 1 ).
    
        _a = COND int4( WHEN boolean_a EQ abap_true THEN 1 ELSE 0 ).
        _b = COND int4( WHEN boolean_b EQ abap_true THEN 1 ELSE 0 ).
        wrapper_a = zcl_integer=>value_of( _a ).
        wrapper_b = zcl_integer=>value_of( _b ).
        wrapper_c = zcl_integer=>value_of( c ).
        DATA(_n_wrapper) = wrapper_a->xor( wrapper_b )->xor( wrapper_c ).
        DATA(b_or_c) = wrapper_b->or( wrapper_c ).
        DATA(b_and_c) = wrapper_b->and( wrapper_c ).
        DATA(_c_wrapper) = wrapper_a->and( b_or_c )->or( b_and_c ).
        c = _c_wrapper->get_raw_value( ).
        DATA(_n_i0_wrapper) = zcl_integer=>value_of( COND int4( WHEN _n_wrapper->get_raw_value( ) > 0 THEN i ELSE 0 ) ).
        DATA(wrapper_n) = zcl_integer=>value_of( n ).
        n = wrapper_n->or( _n_i0_wrapper )->get_raw_value( ).
    
        wrapper_a = zcl_integer=>value_of( aa ).
        aa = wrapper_a->shift_right( )->get_raw_value( ).
    
        wrapper_b = zcl_integer=>value_of( bb ).
        bb = wrapper_b->shift_right( )->get_raw_value( ).
    
        cv_result = n.
    
        DATA(wrapper_i) = zcl_integer=>value_of( i ).
        wrapper_i->shift_left( ).
    
        i = wrapper_i->get_raw_value( ).
    
      ENDWHILE.
    
    ENDFORM.
    
    START-OF-SELECTION.
      DATA: i TYPE int4.
      threshold = ipow( base = 2 exp = 30 ).
      PERFORM add USING a b CHANGING i.
    
      WRITE: / i.
    
    

    This solution might not make too much sense from business perspective but it at least help me refresh what I have learned in my university.

    要获取更多Jerry的原创文章,请关注公众号"汪子熙":

  • 相关阅读:
    [源码解析] PyTorch 流水线并行实现 (4)--前向计算
    [源码解析] PyTorch 流水线并行实现 (3)--切分数据和运行时系统
    [源码解析] PyTorch 流水线并行实现 (2)--如何划分模型
    [源码解析] PyTorch 流水线并行实现 (1)--基础知识
    [源码解析] 深度学习分布式训练框架 horovod (21) --- 之如何恢复训练
    [源码解析] 深度学习流水线并行 PipeDream(6)--- 1F1B策略
    [源码解析] 深度学习流水线并行 PipeDream(5)--- 通信模块
    [源码解析] 深度学习流水线并行 PipeDream(4)--- 运行时引擎
    6.耐心——想象力决定生产力
    5.权利——自由不等于免费
  • 原文地址:https://www.cnblogs.com/sap-jerry/p/13576500.html
Copyright © 2011-2022 走看看