SIMD-0496: Integer Inflation Rewards#496
SIMD-0496: Integer Inflation Rewards#496ksn6 wants to merge 3 commits intosolana-foundation:mainfrom
Conversation
|
Hello ksn6! Welcome to the SIMD process. By opening this PR you are affirming that your SIMD has been thoroughly discussed and vetted in the SIMD discussion section. The SIMD PR section should only be used to submit a final technical specification for review. If your design / idea still needs discussion, please close this PR and create a new discussion here. This PR requires the following approvals before it can be merged:
Once all requirements are met, you can merge this PR by commenting |
| type: Core | ||
| status: Idea | ||
| created: 2026-03-18 | ||
| feature: (to be assigned upon acceptance) |
There was a problem hiding this comment.
might as well generate the feature key now and paste the pubkey here
|
|
||
| For mainnet's `Inflation::default()`: | ||
|
|
||
| | Field | f64 | Scaled u128 | |
There was a problem hiding this comment.
I'm not getting these values when I run f64_to_scaled on my dev box. This is what I get:
0.08 | 92233720368547760
0.015 | 17293822569102704
0.85 | 979983278915819904
0.05 | 57646075230342352
|
|
||
| #### `fixed_ln`: natural log via `atanh` series | ||
|
|
||
| For `x` in `[S/2, S]` (i.e. real values in [0.5, 1.0]): |
There was a problem hiding this comment.
nit : Inflation::new_fixed and Inflation::new_pico use taper = 1.0 so we should define fixed_ln(0) as well
| remainder = year_nanos % NANOS_PER_YEAR | ||
|
|
||
| int_part = fixed_pow(decay_base_scaled, full_years) | ||
| frac_part = fixed_exp(remainder * fixed_ln(decay_base_scaled) / NANOS_PER_YEAR) |
There was a problem hiding this comment.
i'm assuming decay_base_scaled = (1 - taper)^year?
It would be good to define all these variables in terms of the 4 initially provided, lil confusing as is
| ### Validator and foundation rates | ||
|
|
||
| ```text | ||
| foundation_share = total * foundation_scaled / S (if year_nanos < foundation_term_nanos) |
There was a problem hiding this comment.
nit: here too would be good to explicitly define foundation_term_nanos in terms of Inflation.foundation_term w/ rounding rule
2 participants
Overview
Specifies the replacement of f64 arithmetic in the inflation reward pipeline with fixed-point u128 integer math, gated behind a feature flag.
Implementation: anza-xyz/agave#10335
Motivation
IEEE-754 floating-point is non-deterministic across CPU architectures. The current inflation reward calculation uses f64 throughout, which is a latent consensus-splitting bug as the validator fleet diversifies. This SIMD specifies exact integer arithmetic that all implementations must replicate bit-for-bit.