Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

transfer-lamports: Improve asm performance, update table #12

Merged
merged 1 commit into from
Nov 4, 2024
Merged

transfer-lamports: Improve asm performance, update table #12

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 13 additions & 8 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -165,20 +165,25 @@ the amount given by a little-endian u64 in instruction data.
| Rust | 459 |
| Zig | 44 |
| C | 104 |
| Assembly | 31 |
| Assembly | 30 |
| Rust (pinocchio) | 32 |

This one starts to get interesting since it requires parsing the instruction
input. Since the assembly version knows exactly where to find everything, it can
be hyper-optimized.
be hyper-optimized. The pinocchio version performs very closely to the assembly
implementation!

* CPI: allocates a PDA given by the seed "You pass butter" and a bump seed in
the instruction data. This requires a call to `create_program_address` to check
the address and `invoke_signed` to CPI to the system program.

| Language | CU Usage |
| --- | --- |
| Rust | 3662 |
| Zig | 2825 |
| C | 3122 |
| Rust (pinocchio) | 2816 |
| Language | CU Usage | CU Usage (minus syscalls) |
| --- | --- | --- |
| Rust | 3662 | 1162 |
| Zig | 2825 | 325 |
| C | 3122 | 622 |
| Rust (pinocchio) | 2816 | 316 |

Note: `create_program_address` consumes 1500 CUs, and `invoke` consumes 1000, so
we can subtract 2500 CUs from each program to see the actual cost of the program
logic.
8 changes: 4 additions & 4 deletions transfer-lamports/asm/main.s
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,10 +12,9 @@ entrypoint:
add64 r4, 8 + 8 + 32 + 32 + 8 + 8 + 10240 + 8 # calculate end of account data
add64 r4, r3
mov64 r5, r4 # check how much padding we need to add
and64 r5, -8 # clear low bits
and64 r4, -8 # clear low bits
jeq r5, r4, 1 # no low bits set, jump ahead
add64 r4, 8 # add 8 for truncation if needed
and64 r4, -8 # clear low bits

ldxb r5, [r4 + 0] # get second account
jne r5, 0xff, error # we don't allow duplicates
Expand All @@ -25,11 +24,12 @@ entrypoint:
add64 r7, 8 + 32 + 32 + 8 + 8 + 10240 + 8 # calculate end of account data
add64 r7, r6
mov64 r8, r7 # check how much padding we need to add
and64 r8, -8 # clear low bits
and64 r7, -8 # clear low bits
jeq r8, r7, 1 # no low bits set, jump ahead
add64 r7, 8 # add 8 for truncation if low bits are set

ldxdw r8, [r7 + 0] # get instruction data size
jne r8, 0x08, error # need 8 bytes of instruction data
jne r8, 8, error # need 8 bytes of instruction data
ldxdw r8, [r7 + 8] # get instruction data as little-endian u64

sub64 r2, r8 # subtract lamports
Expand Down