Tutorial: for expressions

In this document we explain in detail the implementation of routine to compute a CRC32 checksum on a single 8-bit input. We don't discuss the algorithm here, only the language features necessary to implement the algorithm.

Refer to the full implementation while following this document.

Function Prototype

The signature and first line of this function should look familiar enough now, but the for construct is new: DSLX provides a means of iterating a fixed number of times within a function.

A DSLX for loop has the following structure:

1. The loop signature: this consists of three elements: 1. An (index, <accumulator vars>) tuple. The index holds the current iteration number, and the accumulator vars are user-specified data carried into the current iteration. 2. The type specification for the index/accumulators tuple. Note that the index type can be controlled by the user (i.e., doesn't have to be u32, but it should be able to hold all possible loop index values). 3. An iterable, either the range() or enumerate() expressions, either of which dictates the number of iterations of the loop to complete.
2. The loop body: this has the same general form as a DSLX function. Particularly noteworthy is that the loop body ends by stating the "return" value. In a for loop, this "return" value is either used as the input to the next iteration of the loop (for non-terminal iterations) or as the result of the entire expression (for the terminal iteration).

For this specific for loop, the index variable is unused, so we assign it to _. This indicates to the DSLX frontend that "the variable is unused but that's ok"; a trailing name after the underscore is also allowed to provide additional context. The accumulator consists of a single variable crc. Both index and accumulator are of type u32. The iterable range expression specifies that the loop should execute 8 times.

  // 8 rounds of updates.
  for (_, crc): (u32, u32) in range(u32:8) {

At the end of the loop, the calculated value is being assigned to the accumulator crc - the last expression in the loop body is assigned to the accumulator:

    let mask: u32 = -(crc & u32:1);
    (crc >> u32:1) ^ (polynomial & mask)

Finally, the accumulator's initial value is being passed to the for expression as a parameter. This can be confusing, especially when compared to other languages, where the init value typically is provided at or near the top of a loop.

}(crc)

Since the for loop is the last expression in the function, it's also the function's return value, but in other contexts, it could be assigned to a variable and used elsewhere. In general, the result of a for expression can be used in the same manner as any other expression's result.