That night I stopped looking at 2.9 seconds, I watched the MacBook Pro fan roar as Browser Wallet generated a Private Transfer Proof for a Phoenix transaction...
first time 11s, second time 3.8s, third time 2.9s.
beautiful!
but going from 11 to 2.9 is a 73.6% reduction, meaning Warm Start is nearly 3.8 times faster than Cold Start.
I only paid attention to the gap between the two states.
real users aren’t going to keep Browser Client open all day just so Parameter Caching is always ready, are they?
close the tab, switch networks, battery drops, open it again... WASM Module loads, Initialization Parameters initialize, Local Proof Generation runs again.
Peak Memory Usage 1.2GB isn’t that alarming on Mac.
move to a Smartphone with 4GB RAM, and that alone already takes up 30% of total RAM.
honestly, this is the Client-side Performance I actually want to examine in Hedger.
I like Dusk pushing Privacy down to the client with Lightweight Circuit.
a Private Transaction where the Proof stays on the user’s device feels right to I.
but being right doesn’t mean it’s easy to live with.
30 Cold Starts at 11s is 330s.
30 times at 2.9s is only 87s.
a difference of 243s... more than 4 minutes.
I want Browser Benchmark to run under Poor Network, unstable 4G, CPU throttling and Older Android Device, not just powerful machines with Cached Parameters.
if Batch Aggregation or Multi-transfer Aggregation can bundle multiple transfers into a Single Proof, I prefer that over forcing Proof Generation down by another few tenths of a second.
reducing time sounds good.
reducing the number of times you have to wait is what actually matters.
Testnet gives I the right to be patient.
Mainnet lets technology in without making I notice it.
for I, the best Privacy is the kind that turns cryptography into background noise... it runs, but no one has to think about it.
if you had to choose, would you rather Dusk show off faster Proofs, or prove that even the worst Cold Start is still light enough for real-world use?
#dusk $DUSK @Dusk
first time 11s, second time 3.8s, third time 2.9s.
beautiful!
but going from 11 to 2.9 is a 73.6% reduction, meaning Warm Start is nearly 3.8 times faster than Cold Start.
I only paid attention to the gap between the two states.
real users aren’t going to keep Browser Client open all day just so Parameter Caching is always ready, are they?
close the tab, switch networks, battery drops, open it again... WASM Module loads, Initialization Parameters initialize, Local Proof Generation runs again.
Peak Memory Usage 1.2GB isn’t that alarming on Mac.
move to a Smartphone with 4GB RAM, and that alone already takes up 30% of total RAM.
honestly, this is the Client-side Performance I actually want to examine in Hedger.
I like Dusk pushing Privacy down to the client with Lightweight Circuit.
a Private Transaction where the Proof stays on the user’s device feels right to I.
but being right doesn’t mean it’s easy to live with.
30 Cold Starts at 11s is 330s.
30 times at 2.9s is only 87s.
a difference of 243s... more than 4 minutes.
I want Browser Benchmark to run under Poor Network, unstable 4G, CPU throttling and Older Android Device, not just powerful machines with Cached Parameters.
if Batch Aggregation or Multi-transfer Aggregation can bundle multiple transfers into a Single Proof, I prefer that over forcing Proof Generation down by another few tenths of a second.
reducing time sounds good.
reducing the number of times you have to wait is what actually matters.
Testnet gives I the right to be patient.
Mainnet lets technology in without making I notice it.
for I, the best Privacy is the kind that turns cryptography into background noise... it runs, but no one has to think about it.
if you had to choose, would you rather Dusk show off faster Proofs, or prove that even the worst Cold Start is still light enough for real-world use?
#dusk $DUSK @Dusk