From 874a1a7813dcc3c34301359e4c3c0e0d731a2c70 Mon Sep 17 00:00:00 2001
From: Chris Duncan <chris@zoso.dev>
Date: Wed, 8 Jan 2025 08:35:46 -0800
Subject: [PATCH] Inline G mix 0 of round 0.

---
 src/lib/nano-pow/shaders/gpu-compute.ts | 82 ++++++++++++++++++++++++-
 1 file changed, 81 insertions(+), 1 deletion(-)

diff --git a/src/lib/nano-pow/shaders/gpu-compute.ts b/src/lib/nano-pow/shaders/gpu-compute.ts
index 7081a34..5d839f1 100644
--- a/src/lib/nano-pow/shaders/gpu-compute.ts
+++ b/src/lib/nano-pow/shaders/gpu-compute.ts
@@ -209,8 +209,88 @@ fn main(
 	* ROUND 0.1
 	* G(&v0, &v1, &v8, &v9, &v16, &v17, &v24, &v25, m0, m1, m2, m3);
 	*/
+	// a = a + b;
+	o0 = v0 + v8;
+	o1 = v1 + v9;
+	if (v0 > 0xFFFFFFFFu - v8) {
+		o1 = o1 + 1u;
+	}
+	v0 = o0;
+	v1 = o1;
+
+	// a = a + m[sigma[r][2*i+0]];
+	o0 = v0 + m0;
+	o1 = v1 + m1;
+	if (v0 > 0xFFFFFFFFu - m0) {
+		o1 = o1 + 1u;
+	}
+	v0 = o0;
+	v1 = o1;
+
+	// d = rotr64(d ^ a, 32);
+	xor0 = v24 ^ v0;
+	xor1 = v25 ^ v1;
+	v24 = xor1;
+	v25 = xor0;
+
+	// c = c + d;
+	o0 = v16 + v24;
+	o1 = v17 + v25;
+	if (v16 > 0xFFFFFFFFu - v24) {
+		o1 = o1 + 1u;
+	}
+	v16 = o0;
+	v17 = o1;
+
+	// b = rotr64(b ^ c, 24);
+	xor0 = v8 ^ v16;
+	xor1 = v9 ^ v17;
+	v8 = (xor0 >> 24u) ^ (xor1 << 8u);
+	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
+
+	// a = a + b;
+	o0 = v0 + v8;
+	o1 = v1 + v9;
+	if (v0 > 0xFFFFFFFFu - v8) {
+		o1 = o1 + 1u;
+	}
+	v0 = o0;
+	v1 = o1;
+
+	// a = a + m[sigma[r][2*i+1]];
+	o0 = v0 + m2;
+	o1 = v1 + m3;
+	if (v0 > 0xFFFFFFFFu - m2) {
+		o1 = o1 + 1u;
+	}
+	v0 = o0;
+	v1 = o1;
+
+	// d = rotr64(d ^ a, 16)
+	xor0 = v24 ^ v0;
+	xor1 = v25 ^ v1;
+	v24 = (xor0 >> 16u) ^ (xor1 << 16u);
+	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
+
+	// c = c + d;
+	o0 = v16 + v24;
+	o1 = v17 + v25;
+	if (v16 > 0xFFFFFFFFu - v24) {
+		o1 = o1 + 1u;
+	}
+	v16 = o0;
+	v17 = o1;
+
+	// b = rotr64(b ^ c, 63)
+	xor0 = v8 ^ v16;
+	xor1 = v9 ^ v17;
+	v8 = (xor1 >> 31u) ^ (xor0 << 1u);
+	v9 = (xor0 >> 31u) ^ (xor1 << 1u);
+
+
+
+	/* ROUND 0.2 */
 
-	G(&v0, &v1, &v8, &v9, &v16, &v17, &v24, &v25, m0, m1, m2, m3);
 	G(&v2, &v3, &v10, &v11, &v18, &v19, &v26, &v27, m4, m5, m6, m7);
 	G(&v4, &v5, &v12, &v13, &v20, &v21, &v28, &v29, m8, m9, 0u, 0u);
 	G(&v6, &v7, &v14, &v15, &v22, &v23, &v30, &v31, 0u, 0u, 0u, 0u);
-- 
2.34.1