Refactor setting stack size in regex code

[davidwendt]

This PR is the first of several to cleanup the regex strings code. The regex code employs multiple stack sizes to perform its matching based on the number of instructions in the given regex pattern. The current implementation allocates the stack for the regex code. This PR moves this down into the regex functions themselves. This helps simplify the interface and reduce compile time a bit.

[nvdbaranec]

Question about the transition between "LARGE" stack size and "global memory". It seems like there's a hole between RX_LARGE_INSTS (920) and MAX_STACK_INSTS (1000). It looks like this code can get into the RX_STACK_ANY case as soon as the # of instructions is > 920. But the code for allocating the global memory doesn't kick in until 1000 instructions:

  if (insts_count > MAX_STACK_INSTS) {
    auto relist_alloc_size = relist::alloc_size(insts_count);
    rlm_size               = relist_alloc_size * 2L * strings_count;  // reljunk has 2 relist ptrs
  }

Is there some kind of rounding thing going on here where it will always be some safe multiple of instructions?

[davidwendt]

I think you are right. There appears to be an overlap where too much memory is being allocated. I should change the threshold check to use RX_LARGE_INSTS instead of MAX_STACK_INSTS and then get rid of the latter.

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (branch-21.08@606d854). Click here to learn what that means.
The diff coverage is n/a.

@@               Coverage Diff               @@
##             branch-21.08    #8358   +/-   ##
===============================================
  Coverage                ?   82.83%           
===============================================
  Files                   ?      109           
  Lines                   ?    17901           
  Branches                ?        0           
===============================================
  Hits                    ?    14828           
  Misses                  ?     3073           
  Partials                ?        0           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 606d854...5aef511. Read the comment docs.

[devavret]

Didn't review the entire logic but now I understand the change. Code quality looks good.

[davidwendt]

@gpucibot merge