在stdint.h中有uint24_t的定义被注释掉

基本上，它们不是标准的，不是便携式的。但你可以用它们



      以下为原文

    Basically, that they're not standard, not portable. But you may use them, indeed

我不会评论它们，把它们放进你的代码中。它们是多么有用，取决于这个项目。



      以下为原文

    I would not comment them,  Put them in your code.
How usefull they are depends on the project.

使用24位ADC UIT32 32 T可能会更方便地调整增益/偏移，而不会立即溢出。无论如何，没有反对使用UTIN 24t来保存RAM的问题。



      以下为原文

    With a 24 bit ADC uint32_t might come handier to adjust gain/offset without the danger of immediate overflow.
Anyway - no objections to use uint24_t to preserve RAM.

正如NKurzman所说，不要修改编译器头文件。如果你这样做，那么你的代码依赖于一个修改的编译器。至于使用它们是否是一个好主意，我会说不，除非你没有其他的选择（也就是说，如果没有它们，你将用完代码空间或RAM）。使用它们将你锁定在一个特定的编译器上，当你决定移动到PIC24时，你就必须重写代码。



      以下为原文

    As NKurzman said, do not modify the compiler header files.  If you do, then your code depends on a modified compiler.
 
As to whether it's a good idea to use them, I would say no, unless you have no other option (i.e., you would run out of code space or RAM without them).  Using them locks you in to a particular compiler, and when you decide you need to move to a PIC24, you'll have to rewrite the code.

如果Pababiy是一个问题，您总是可以自由编译条件，或者将UTI2424T算术端口移植到另一个编译器。如果你敢



      以下为原文

    If portabiliy is an issue, you are always free to compile conditionally - or to port uint24_t arithmetics to another compiler. If you dare

HI，在8位计算机上，数学可能需要大量的指令周期。根据所使用的算法，所使用的时间可能取决于参数的实际值，或者对于参数的数据类型是恒定的。UTIN 32×T类型的参数的乘积可能占用UTIN 24T乘法时间的两倍，这可能需要UTIN 16×T乘法所需的时间的两倍。对于编译器可能存在不同的算法。PIC18F有一个8位的硬件乘法器。使用硬件乘法器，用于大数据类型的小值，可能比迭代移位和加法算法花费更多的时间，这是没有硬件乘法器使用的。但是对于大的值，硬件获胜。如果你在多M上进行数学运算。易用性和时序是很重要的，UTI2424T类型可能有很大帮助。在其他一些线程中，做RMS和平均滤波器的计算，有一些技巧，在PIC16和PIC18上有很好的性能。在长线程的末尾有一个示例项目：HTTP://www. McCHIP.COM/FUMMS/FIDEP。OST/845 54 1另一个长线程：HTTP://www. McCHIP.COM/FUMMS/FUNDSPE/964。88.如果您在自由模式下使用XC8编译器，通过将程序裁剪到实际数据中可能会获得很多好处，如果可以进行PRO模式优化，也许会有很多收获。使用H时间，在代码中进行时间测量，或切换引脚，并使用示波器或逻辑分析仪测量时序。迈西尔



      以下为原文

    Hi,
On a 8 bit computer, mathematics may take a lot of instruction cycles.
Depending on the algorithm used, time used may depend on actual values of arguments, 
or be constant for the data type of argument.
Roughly, time used for mathematic operations, may increase with square of size of argument type,
multiplication of uint32_t type arguments may take about twice the time of uint24_t multiplication,
which again may take about two times the time needed for uint16_t multiplication.
 
There may be different algorithms available to the compiler. PIC18F have a 8 bit hardware multiplier available.
Using the hardware multiplier for small values in large data type, may take more time than the iterative shift and add algorithm, that is used without hardware multiplier.
but for large values, hardware wins.
 
If you are doing mathematics on multiple measurements, and timing is important, uint24_t type may help a lot.
 
In some other threads, doing RMS and averaging filter calculations, there are some tricks used, giving good performance on PIC16 and PIC18.
There is example project in the end of the long thread: http://www.microchip.com/forums/FindPost/845541
Another long thread: http://www.microchip.com/forums/FindPost/964778 
 
If you are using XC8 compiler in free mode, there may be a lot to gain by tailoring the program to actual data to be processed,
If Pro mode optimization is available, maybe that may catch up a lot.
 
To know where, and how much time is used, do time measurements in code,
or toggle a pin, and measure timing using a oscilloscope or logic analyzer.
 
   Mysil

IRC，从XC8 V1.42开始，非标准的24位类型在lt；Stdit.h & gt中不再可用；但我认为您仍然可以使用“签名短长”和“无符号短长”类型，所以如果您愿意，可以使用它们。也就是说，Microchip已经决定不再支持24位类型，这意味着即使你可以声明“短长”类型变量，24位的数学函数可能是不可用的。ED作为XC8中的32位“长”型。



      以下为原文

    IIRC, starting with XC8 v1.42 the non-standard 24-bit types featured in are no longer available; but I think you can still use "signed short long" and "unsigned short long" types, so you can typedef them if you like. That said, Microchip has decided to stop supporting 24-bit types, which means 24-bit math functions might not be available even if you can declare "short long" type variables.
 
That is, "short long" might one day be implemented as the "long" type, just like "long long" has been implemented as a 32-bit "long" type in XC8.

由于C是第一次由ANSI在1988标准化，所以长时间至少保证了32位，因为它小于32位会破坏大量的代码。它不会发生。XC8中的长时间现在是空扩展。XC8不主张支持C99（其中A）。这个名称的64位类型最初是标准的，支持GCC扩展。将完整的64位整数支持添加到XC8将需要一些内部更改来处理最大的内置类型的大小的倍增（4字节是围绕编译器的大部分构建的假设）。它可能在XC8 V2.0的路线图上。实际上，使用64位整数数学也会显著地增加代码大小-每个分配/加/减将需要大约32倍的代码的两倍；乘法和除法将再次变大。



      以下为原文

   
 
It won't.
long has been guaranteed to be at least 32 bits for nearly 3 decades, since C was first standardized by ANSI in 1988.
Making it be less than 32 bits would break vast amounts of code.
It won't happen.
 

 
long long in XC8 is currently a null extension.
XC8 does not claim to support C99 (in which a 64 bit type of that name was first made standard, endorsing the GCC extension).
Adding full 64 bit integer support to XC8 would require some internal changes to handle the doubling in size of the largest builtin type (4 bytes is an assumption around which much of the compiler is built). It might possibly be on the roadmap for XC8 v2.0.
Actually using 64 bit integer math would also significantly increase code size - every assign/add/subtract would require about twice as much code as for 32 bit; multiply and divide would be bigger again.

我想他指的是相反的马克，“短长”只会被当作“长”，即两个32位。



      以下为原文

    I think he meant the opposite Mark,
that "short long" would just be treated as "long",
i.e. both 32 bits.
 

嗨，1和0，你从哪里得到这个：“Microchip已经决定停止支持24位类型”？使“短长”24位整数类型的改变或消失将打破高技术与GX.XC8编译器的兼容性和性能的良好历史。32位浮点数学可能已经接收到超过24位浮点库的更新，而这些更新库没有相同的更新。这不是原因。另外，在32位浮点计算中需要24位的数学运算，其中尾数是24位，而UTI2424TIS易于阅读和理解，我同意类型定义不属于STDIN。HealthHead是为了支持兼容性和可移植性。根据定义，将定义保留为注释，例如，当需要时，可以在项目代码中定义类型。例如，Mysil。



      以下为原文

    Hi,
1and0, where do you have this from: 'Microchip has decided to stop supporting 24-bit types'?
 
Making 'short long' 24 bit integer type change or disappear would break the good history of compatibility and performance in HI-TECH -> XC8 compilers.
 
32 bit floating point math may have received updates that outperform 24 bit float libraries that haven't had the same updates.
That is Not a reason to discard the 24 bit integer type.
besides, 24 bit math is needed in 32 bit float calculations, where the mantissa is 24 bits.
 
While: uint24_t  
is easy to read and understand, I agree that the type definition do not belong in stdint.h
Standard headers are meant to support compatibility and portability according to the standards.
Keeping the definitions as comments, is o.k. as example of how the types may be defined in project code when wanted.
 
Regards,
   Mysil

是的，这就是我所说的，“短长”有一天会被当作32位“长”来对待。马克和我似乎总是在某个地方沟通不畅。：（



      以下为原文

   
Yes, that is what I said, that "short long" would one day be treated as 32-bit "long". Mark and I seems to always have miscommunication somewhere. :(

也许是强调微笑：



      以下为原文

   
Maybe it is the accent Smile:

由于24位整数类型“It24.t”和“Unt24.t”不再可用，并且“签名短长”和“无符号短长”仍然允许，这些“短长”类型变量有多少位？我不是C标准专家，我在这里引用你（我可能是错误的）“长隆”已经被保证至少有64位3年了。因为“长长”类型在XC8中至少不为64位，直到添加了64位整数支持，我认为编译器应该生成一个错误。当使用时。相反，XC8编译器默默地把“长隆”当作32位“长”，这似乎并不正确。



      以下为原文

     
Since 24-bit integer types "int24_t" and "uint24_t" are no longer available, and "signed short long" and "unsigned short long" are still allowed, how many bits would these "short long" type variables be?
 

I'm no C standard expert and am quoting you here (I might be wrong) "long long" "has been guaranteed to be at least [64] bits for nearly 3 decades." Since "long long" types are not at least 64 bits in XC8, and until 64-bit integer support is added, the compiler in my opinion should generate an error when it is used. Instead, the XC8 compiler silently treats "long long" as a 32-bit "long" -- that just does not seem right.
 

我在这里做了一个巨大的假设，因为“因特24t”和“Unt2424t”类型不再可用了。我可能是错的。我现在找不到线程，Microchip正在考虑移除对24位浮点的支持。那么，对于24位整数的移除，有什么区别呢？不给24位浮点库提供相同的更新。例如，24位Sqt（）函数要比32位Sqt（）慢很多。IILC，XC8中的32位Sqt（）用于接管20K的TIN，现在它需要2K TIN。



      以下为原文

   
I'm making a huge assumption there as "int24_t" and "uint24_t" types are no longer available -- I might be wrong.
 

I cannot find the thread now, where Microchip is thinking to remove support for 24-bit float; so what is the difference with removing support for 24-bit integer.
 

That is because they are not giving the same updates to the 24-bit float libraries. For example, the 24-bit sqrt() function is a LOT slower than the 32-bit sqrt().  IIRC, the 32-bit sqrt() in XC8 used to take over 20K Tcy, now it takes ~2K Tcy.
 
 

可能是这样。英语不是我的第一语言。微笑：



      以下为原文

   
It could be. English is not my first language. Smile:

嗨，所有类型定义都是C语言的原生类型，或者是编译器提供的扩展。C语言已知的数据类型是：“签名”和“未签名”“char”、“短”和“长”int……对于一些编译器，而不是xc8，有m个。Ay是一系列的翻译和重新翻译的形式，Untu32 32插入，或者从.lt.Stdit.h & gt中删除一个Type，只要编译器能够提供定义中使用的类型，就意味着什么。对语言和标准而言，就像“long long int”是一个扩展，直到它被包含在标准中的一些修订中。附加了我的函数用于无符号16位、24位和32位平方根计算。连同一个乘法例程一起使用。问候，Mysil。
MyuMul1632.c（2.10 KB）-下载119次



      以下为原文

    Hi,
All type definitions in are aliases of types that are native of the C programming language , or are extensions provided by the compiler. 
The data types known by the C language are: 'signed' and 'unsigned' 'char', 'short' and 'long' 'int' ...
For some compilers, but not XC8, 
there may be a series of translations and retranslations of the form __uint32_t
 
Inserting, or removing a typedef from means nothing to the compiler, as long as the compiler is able to provide the type that is used in the definition.
 
The 'short long int' type in HI-TECH C and XC8 is an extension to the language, and the standards,
just as 'long long int' was an extension, until it was included in the standard in some revision.
 
Attached are my functions for unsigned 16 bit, 24 bit and 32 bit square root calculation.
Together with a multiplication routine used.
 
Regards,
   Mysil
   Attachment(s)

isqrt.c (6.29 KB) - downloaded 106 times
myUmul1632.c (2.10 KB) - downloaded 119 times

感谢整型SQL（）函数。你有这些功能的执行时间吗？IIRC，一个16位平方根例程在组装中花费110的时间。



      以下为原文

   
Thank you for the integer sqrt() functions. Do you have the execution time of these functions? IIRC, a 16-bit square root routine takes ~110 Tcy in assembly.
 

这里是Microchip考虑删除24位浮点的位置。因此，也许Microchip也在考虑删除24位整数。也许Mark可以发光并给出这24位类型的未来。



      以下为原文

   
Here is the post where Microchip is considering removing the 24-bit float. So, maybe Microchip is also considering removing the 24-bit integer. Perhaps Mark can shed light and give a heads up on the future of these 24-bit types.