[英]How to Abstract SIMD code to handle different datatypes without a performance hit
I have been writing code for performing matrix operations. 我一直在编写用于执行矩阵运算的代码。 Initially it was just for x86 and now am porting it to different architectures.
最初它仅用于x86,现在正将其移植到不同的体系结构。 Also, I want it to support different data types other than float.
另外,我希望它支持除float以外的其他数据类型。
Consider the following code for adding to float arrays 考虑以下代码以添加到浮点数组
void add(float *a, float *b, float *dst, int len)
{
int k = 0;
for(; k + 8 < len; k += 8,a += 8, b += 8, dst+= 8){
__m256 x = _mm256_load_ps(a);
__m256 y = _mm256_load_ps(b);
__m256 z = _mm256_add_ps(x, y);
_mm256_store_ps(dst, z);
}
}
Here is what I have thought of to improve the code to support several platforms and data types. 这就是我考虑过的改进代码以支持多种平台和数据类型的想法。
For different the data types, I was going to change the function to template function 对于不同的数据类型,我打算将功能更改为模板功能
For simd instructions I thought of having macros that rename all architecture specific intrinsic functions to generic simd instructions such as SIMD_ADD . 对于simd指令,我想到了将所有特定于体系结构的固有函数重命名为通用simd指令(例如SIMD_ADD)的宏。 Problem is that different data-types require different intrinsic functions and the return type of the intrinsic is dependent on the data type too.
问题在于,不同的数据类型需要不同的内在函数,并且内在函数的返回类型也取决于数据类型。
Also if I were to write a subtract function, I would end up copying most of the code just to replace SIMD_ADD macro for SIMD_SUB macro. 同样,如果我要编写一个减法函数,我最终将复制大部分代码,只是将SIMD_ADD宏替换为SIMD_SUB宏。 Is their a neat way such that the I dont have to repeat the same code for all element wise operations such multiplication, division and subtraction ?
它们是否是一种整洁的方式,以便我不必对所有元素明智的操作(例如乘法,除法和减法)重复相同的代码?
How would one tackle points 2 & 3 without abstracting to the extent of affecting the code's performance ? 在不抽象影响代码性能的程度的情况下,如何解决第2点和第3点?
I ended up having template classes for simd instructions with a specialization for each datatype. 我最终得到了simd指令的模板类,每种数据类型都有专门的类。 Unfortunately the compiler will not inline it automatically therefore you must use the compiler specific attributes to force it to inline
不幸的是,编译器不会自动内联它,因此您必须使用编译器特定的属性来强制它内联
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.