s******n
发帖数: 3946 | 1 他的意思是假设是operator重载
++i先做++再放在stack上，i++则先复制一份copy到stack上再做++，多了一份复制（假
设编译无优化）
大家看有道理吗？
I did a real test on arm compiler turn off optimization:
the O0 code is exactly the same, except that post operator++()(int dummy)
has an extra dummy parameter which is required by c++ to identify the
difference of prefix and postfix.
2nd, even I change the Test& operator++() into Test operator++(), it still generates the same code.
printf("%d \n", 10+ (abc++).value);
sub r3, fp, #8
mov r0, r3
mov r1, #0 --> the dummy parameter of postfix
bl _ZN4TestppEi --> call the overload function
mov r3, r0
ldr r3, [r3, #0]
add r3, r3, #10
ldr r0, .L12
mov r1, r3
bl printf
printf("%d \n", 20+ (++abc).value);
sub r3, fp, #8
mov r0, r3
bl _ZN4TestppEv --> call the overload function
mov r3, r0
ldr r3, [r3, #0]
add r3, r3, #20
ldr r0, .L12
mov r1, r3
bl printf
c++ source code
#include
class Test {
public:
Test(int v):value(v){};
Test():value(0){};
int value;
int operator()();
Test& operator++();
Test& operator++(int postVersion);
};
Test& Test::operator++(){
value++;
return *this;
}
Test& Test::operator++(int postVersion){
value++;
return *this;
}
int Test::operator()() {
return value;
}
int main(int argc, char** argv)
{
Test abc(100);
printf("%d \n", 10+ (abc++).value);
printf("%d \n", 20+ (++abc).value);
} | h**********y
发帖数: 1293 | | h**********l
发帖数: 6342 | 3 。。。。
sure yes
【在 s******n 的大作中提到】
: 他的意思是假设是operator重载
: ++i先做++再放在stack上，i++则先复制一份copy到stack上再做++，多了一份复制（假
: 设编译无优化）
: 大家看有道理吗？
: I did a real test on arm compiler turn off optimization:
: the O0 code is exactly the same, except that post operator++()(int dummy)
: has an extra dummy parameter which is required by c++ to identify the
: difference of prefix and postfix.
: 2nd, even I change the Test& operator++() into Test operator++(), it still generates the same code.
: printf("%d \n", 10+ (abc++).value);
| s******n
发帖数: 3946 | 4 具体展开一下，大致汇编代码是怎么样的，两种情况下？
【在 h**********y 的大作中提到】
: 这是基本常识。
| h**********l
发帖数: 6342 | 5 你不是自己都说了吗？
基本类型也是一样的
【在 s******n 的大作中提到】
: 具体展开一下，大致汇编代码是怎么样的，两种情况下？
| H***e
发帖数: 476 | 6 G?
他们不问这种的吧？
【在 s******n 的大作中提到】
: 他的意思是假设是operator重载
: ++i先做++再放在stack上，i++则先复制一份copy到stack上再做++，多了一份复制（假
: 设编译无优化）
: 大家看有道理吗？
: I did a real test on arm compiler turn off optimization:
: the O0 code is exactly the same, except that post operator++()(int dummy)
: has an extra dummy parameter which is required by c++ to identify the
: difference of prefix and postfix.
: 2nd, even I change the Test& operator++() into Test operator++(), it still generates the same code.
: printf("%d \n", 10+ (abc++).value);
| s******n
发帖数: 3946 | 7 不是G
【在 H***e 的大作中提到】
: G?
: 他们不问这种的吧？
| m***n
发帖数: 2154 | 8 hehe ,简单的i++和++i 区别不大，但如果是对象operator重载，区别很大，呵呵 | l***i
发帖数: 1309 | 9 give me an example that i++ is better. In general i++ needs to save i in a
temp variable and then incr i because you need to return that temp value. My
understanding is that ++i is always better than i++. | d**********o
发帖数: 279 | 10 这个除了优化些， 而且避免中断产生的race problem，因为前者是atom 操作。 | | | p*i
发帖数: 411 | 11 If i is an integer or something, i++ is merely a single instruction on intel
architectures.
My
【在 l***i 的大作中提到】
: give me an example that i++ is better. In general i++ needs to save i in a
: temp variable and then incr i because you need to return that temp value. My
: understanding is that ++i is always better than i++.
| s******n
发帖数: 3946 | 12 the O0 code is exactly the same, except that post operator++()(int dummy)
has an extra dummy parameter which is required by c++ to identify the
difference of prefix and postfix.
2nd, even I change the Test& operator++() into Test operator++(), it still generates the same code.
printf("%d \n", 10+ (abc++).value);
sub r3, fp, #8
mov r0, r3
mov r1, #0 --> the dummy parameter of postfix
bl _ZN4TestppEi
mov r3, r0
ldr r3, [r3, #0]
add r3, r3, #10
ldr r0, .L12
mov r1, r3
bl printf
printf("%d \n", 20+ (++abc).value);
sub r3, fp, #8
mov r0, r3
bl _ZN4TestppEv
mov r3, r0
ldr r3, [r3, #0]
add r3, r3, #20
ldr r0, .L12
mov r1, r3
bl printf
c++ source code
#include
class Test {
public:
Test(int v):value(v){};
Test():value(0){};
int value;
int operator()();
Test& operator++();
Test& operator++(int postVersion);
};
Test& Test::operator++(){
value++;
return *this;
}
Test& Test::operator++(int postVersion){
value++;
return *this;
}
int Test::operator()() {
return value;
}
int main(int argc, char** argv)
{
Test abc(100);
printf("%d \n", 10+ (abc++).value);
printf("%d \n", 20+ (++abc).value);
} | d****z
发帖数: 314 | 13 ....yes
【在 s******n 的大作中提到】
: 他的意思是假设是operator重载
: ++i先做++再放在stack上，i++则先复制一份copy到stack上再做++，多了一份复制（假
: 设编译无优化）
: 大家看有道理吗？
: I did a real test on arm compiler turn off optimization:
: the O0 code is exactly the same, except that post operator++()(int dummy)
: has an extra dummy parameter which is required by c++ to identify the
: difference of prefix and postfix.
: 2nd, even I change the Test& operator++() into Test operator++(), it still generates the same code.
: printf("%d \n", 10+ (abc++).value);
| s******o
发帖数: 2233 | 14 c++ FAQ 13.15
http://www.parashift.com/c++-faq-lite/operator-overloading.html
"++i is sometimes faster than, and is never slower than, i++"
still generates the same code.
【在 s******n 的大作中提到】
: 他的意思是假设是operator重载
: ++i先做++再放在stack上，i++则先复制一份copy到stack上再做++，多了一份复制（假
: 设编译无优化）
: 大家看有道理吗？
: I did a real test on arm compiler turn off optimization:
: the O0 code is exactly the same, except that post operator++()(int dummy)
: has an extra dummy parameter which is required by c++ to identify the
: difference of prefix and postfix.
: 2nd, even I change the Test& operator++() into Test operator++(), it still generates the same code.
: printf("%d \n", 10+ (abc++).value);
| a****l
发帖数: 8211 | 15 都已经假设编译器无优化了,还谈什么效率？这种问题真是可笑。有优化的条件下,这种
问题是蔑视编译器设计者的智商。
still generates the same code.
【在 s******n 的大作中提到】
: 他的意思是假设是operator重载
: ++i先做++再放在stack上，i++则先复制一份copy到stack上再做++，多了一份复制（假
: 设编译无优化）
: 大家看有道理吗？
: I did a real test on arm compiler turn off optimization:
: the O0 code is exactly the same, except that post operator++()(int dummy)
: has an extra dummy parameter which is required by c++ to identify the
: difference of prefix and postfix.
: 2nd, even I change the Test& operator++() into Test operator++(), it still generates the same code.
: printf("%d \n", 10+ (abc++).value);
| S**I
发帖数: 15689 | 16 Of course you get the same assembly: you overloaded both ++ operators in the
exact same way. Post-increment operator should be overloaded as following:
Test Test::operator++(int postVersion){
Test temp (*this);
value++;
return temp;
}
still generates the same code.
postfix
【在 s******n 的大作中提到】
: the O0 code is exactly the same, except that post operator++()(int dummy)
: has an extra dummy parameter which is required by c++ to identify the
: difference of prefix and postfix.
: 2nd, even I change the Test& operator++() into Test operator++(), it still generates the same code.
: printf("%d \n", 10+ (abc++).value);
: sub r3, fp, #8
: mov r0, r3
: mov r1, #0 --> the dummy parameter of postfix
: bl _ZN4TestppEi
: mov r3, r0
| s******n
发帖数: 3946 | 17 这个是正解了！
the
【在 S**I 的大作中提到】
: Of course you get the same assembly: you overloaded both ++ operators in the
: exact same way. Post-increment operator should be overloaded as following:
: Test Test::operator++(int postVersion){
: Test temp (*this);
: value++;
: return temp;
: }
:
:
: still generates the same code.
| a********m
发帖数: 15480 | 18 恩。lz代码错了。。。。
the
【在 S**I 的大作中提到】
: Of course you get the same assembly: you overloaded both ++ operators in the
: exact same way. Post-increment operator should be overloaded as following:
: Test Test::operator++(int postVersion){
: Test temp (*this);
: value++;
: return temp;
: }
:
:
: still generates the same code.
| h********e
发帖数: 1972 | 19 LZ代码完全写错了嘛。。。。i++肯定会有一份多余的拷贝的。。 |
|
