mirror of
https://github.com/allaunthefox/Research-Stack.git
synced 2026-07-31 03:05:21 +00:00
This squashes all local history (768 commits) onto the scrubbed PR #90 baseline. Individual commits were lost during filter-repo corruption; the working tree content is preserved intact. Build: N/A (working tree state only)
59 lines
1.5 KiB
WebGPU Shading Language
59 lines
1.5 KiB
WebGPU Shading Language
// Q16_16 Arithmetic Theorems Verification Shader
|
|
//
|
|
// Depends on: q16_arithmetic.wgsl (shared include)
|
|
//
|
|
// This shader verifies the main arithmetic theorems from FixedPoint.lean:
|
|
// 1. mul_one: mul q one = q
|
|
// 2. div_one: div q one = q
|
|
// 3. neg_involutive: neg (neg q) = q
|
|
// 4. abs_nonNegative: (abs q).toInt >= 0
|
|
//
|
|
// Tests all 65,536 Q16_16 values in parallel on GPU.
|
|
|
|
struct ArithmeticResult {
|
|
mul_one_ok: u32,
|
|
div_one_ok: u32,
|
|
neg_involutive_ok: u32,
|
|
abs_nonNegative_ok: u32,
|
|
}
|
|
|
|
@group(0) @binding(0) var<storage, read_write> results: array<ArithmeticResult>;
|
|
|
|
const Q16_SPACE: u32 = 65536u;
|
|
|
|
@compute @workgroup_size(64)
|
|
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
|
|
let q = gid.x;
|
|
if (q >= Q16_SPACE) {
|
|
return;
|
|
}
|
|
|
|
// Lemma 1: mul_one
|
|
// mul q one = q
|
|
let mul_result = q16_mul(q, Q16_ONE);
|
|
let mul_one_ok = select(0u, 1u, mul_result == q);
|
|
|
|
// Lemma 2: div_one
|
|
// div q one = q
|
|
let div_result = q16_div(q, Q16_ONE);
|
|
let div_one_ok = select(0u, 1u, div_result == q);
|
|
|
|
// Lemma 3: neg_involutive
|
|
// neg (neg q) = q
|
|
let neg_once = q16_neg(q);
|
|
let neg_twice = q16_neg(neg_once);
|
|
let neg_involutive_ok = select(0u, 1u, neg_twice == q);
|
|
|
|
// Lemma 4: abs_nonNegative
|
|
// (abs q).toInt >= 0
|
|
let abs_q = q16_abs(q);
|
|
let abs_int = q16_to_int(abs_q);
|
|
let abs_nonNegative_ok = select(0u, 1u, abs_int >= 0);
|
|
|
|
results[q] = ArithmeticResult(
|
|
mul_one_ok,
|
|
div_one_ok,
|
|
neg_involutive_ok,
|
|
abs_nonNegative_ok
|
|
);
|
|
}
|