Mmr packing #841
Mmr packing #841
Conversation
hackaugusto
force-pushed
the
hacka-mmr-packing
branch
from
9338943
to
79ba9e1
Compare
stdlib/asm/collections/mmr.masm
Outdated
| padw padw padw | ||
| exec.native::hash_even_memory_addresses | ||
| exec.native::digest_output | ||
| # => [HASH, peaks_end, peaks_end, ...] |
There was a problem hiding this comment.
I'm modifying the code as follows:
- [done] Also push this to the advice map, that way pack works better as an inverse operation of unpack.
- My initial idea was to have another procedure, I guess we can still have it if anyone finds that useful, but from the naming pack/unpack should be the inverse of each other, so one read and the other writes to the advice map
- [done] I'm going to change the padding to make the length a word
- My initial idea was to just push padding as necessary to make use of
adv.mem, but that would make the format of the input and output different, which can be super confusing. I think it is better to make them consistent. Edit: It would be even more confusing than I thought, I assumed I could just push data to the end, but that is not howadv.memworks, it reads a memory range, so the padding would appear on the position I added in this PR.
- My initial idea was to just push padding as necessary to make use of
There was a problem hiding this comment.
humm, adv.mem only accepts immediate parameters, I have to fix that before continuing here.
done: #844
hackaugusto
force-pushed
the
hacka-mmr-packing
branch
2 times, most recently
from
8e42ad4
to
4420b91
Compare
|
marking this as a draft, first have to fix |
hackaugusto
force-pushed
the
hacka-mmr-packing
branch
4 times, most recently
from
550b151
to
76da1b8
Compare
stdlib/asm/collections/mmr.masm
Outdated
| padw padw padw | ||
| exec.native::hash_even_memory_addresses | ||
| exec.native::digest_output | ||
| # => [HASH, peaks_end, peaks_end, peaks_start, num_leaves, ...] |
There was a problem hiding this comment.
If adv.mem was defined with start_address and end_address instead of number_of_words, I could reuse peaks_end and maybe save a few cycles.
I haven't changed it because I'm not sure if that would be best in general.
There was a problem hiding this comment.
in the end it saved just a single cycle.
note: moving the "control word" with swapw uses the same number of cycles.
hackaugusto
force-pushed
the
hacka-mmr-packing
branch
from
76da1b8
to
84519a8
Compare
hackaugusto
force-pushed
the
hacka-mmr-packing
branch
from
84519a8
to
045d790
Compare
stdlib/asm/crypto/hashes/native.masm
Outdated
| @@ -9,7 +26,7 @@ | |||
| #! Cycles: 4 + 3 * words, where `words` is the `start_addr - end_addr - 1` | |||
| #! | |||
| #! Where `A` is the capacity word that will be used by the hashing function, and `B'` the hash output. | |||
| proc.hash_even_memory_addresses | |||
| export.hash_even_memory_addresses | |||
There was a problem hiding this comment.
nit: the addresses don't have to be even - right? (it's the number of words that needs to be even?) Maybe rename this to something like hash_memory_even?
There was a problem hiding this comment.
yep, that is right. changed the name!
hackaugusto
force-pushed
the
hacka-mmr-packing
branch
4 times, most recently
from
16c61cb
to
fba711a
Compare
hackaugusto
force-pushed
the
hacka-mmr-packing
branch
from
fba711a
to
a1f17d5
Compare
Describe your changes
Review/Merge after
#834#844This adds the procedure to pack a Mmr.
Checklist before requesting a review
nextaccording to naming convention.