Optimize slices and builders comparation (for addresses)

[imartemy1524]

Right now comparison between slice and builder require one to convert builder to slice using ENDC CTOS (600+ gas), and then comparing to slices using SDEQ.

But this task can be done using 2 HASHEXT_SHA256 calls, for example:

int equal_slice_and_builder_bits?(slice s, builder b) asm """
OVER SBITS                             // a b bits
8 PUSHINT MOD 8 PUSHINT SUBR           // a b bits_to_add
TUCK STZEROES 1 PUSHINT HASHEXT_SHA256 // a bits_to_add hashb
ROT NEWC STSLICE                       // bits_to_add hashb b_a
ROT STZEROES 1 PUSHINT HASHEXT_SHA256 // hashb hasha
EQUAL
""";

By running function above, one can compare the sha256 hashes of bits of slice a and builder b, and it consumes less than 500 gas.

This is applicable for verifying the addresses of child contracts: instead of converting builder to a cell, then to a slice (so-called Address in tact) it is cheaper to compare the initial builder with slice using TVM HASHEXT_SHA256 function.

[pyAndr3w]

Good suggestion, but your variant is not safe, because it writes bits to the existing Builder. This action can lead to builder overflow.

I propose a safer variant:

int equal_slice_and_builder_bits?(slice s, builder b) asm """
  NEWC 8 PUSHINT s3 PUSH SBITS
  8 PUSHINT MOD SUB
  STZEROES
  TUCK 2 PUSHINT HASHEXT_SHA256
  -ROT 2 PUSHINT HASHEXT_SHA256
  EQUAL
""";

This version creates a new builder with NEWC instead of modifying the existing one, avoiding potential overflow issues. According to my tests, this approach is also more efficient: 16 steps and 335 gas compared to the original 19 steps and 395 gas.