C++ map 模板到其他模板，然后到元组

Question

I'm working on some low level drivers and I thought I'd use C++ templates to spice it up.

In Verilog, you can define bitmasks using [h:l] notation where h is the high bit, l is the low bit. I have a data structure already defined for this:

uint16_t mask(uint8_t h, uint8_t l) {
  return h << 8 + l;
}

I'd like to memoize these bitmasks. Say I want to modify bits [7:4] of register R, if I've already read in R before hand I can just modify the old value of R before writing it, rather than sending out a new read request to my device.

To find the number of bytes needed for each memoization, I already have a function

constexpr uint8_t requiredBytes(uint16_t mask)

which returns the number of bytes needed to hold the mask.

Is there a way to static-time declare a memoization buffer for ease of analysis? (this will run on embedded platforms)

I assume I need a pipeline like below:

template <uint16_t... masks>
(1) V
Apply requiredBytes() to each mask in the template
(2) V
Convert each value from requirdBytes to a type (requiredBytes() = 1 => uint8_t, ...)
(3) V
Convert this list of types into a std::tuple<>

I know I can do transformation 2 using a series of template specializations, which leads to my question, how can we implement transformations 1 and 3?

Answer 1

I presume that your step 2, which you've described as already a done deal, is something along the lines of

typename bytesType<n>::type

Where the template parameter n is the number of bytes, and this gives you the type via an appropriate specialization. If so, then this seems to be something along the lines of a straightforward pack expansion:

template <uint16_t ...masks>
using tuple_mask=std::tuple<typename bytesType<requiredBytes(masks)>::type...>;

requiredBytes() is a constexpr, this provides the parameter to bytesType , which gives you the type. Then, pack-expand it, and feed it into a std::tuple declaration.