From: Chris Duncan <chris@codecow.com>
Date: Thu, 1 Oct 2026 00:23:48 +0000 (-0700)
Subject: Combine load and splat.
X-Git-Url: https://git.codecow.com/?a=commitdiff_plain;h=1c0e94cd5ac0ed57801b65c628be57b129ba314e;p=nano25519.git

Combine load and splat.
---

diff --git a/src/assembly/ed25519/fe.ts b/src/assembly/ed25519/fe.ts
index 17964f4..46dfba8 100644
--- a/src/assembly/ed25519/fe.ts
+++ b/src/assembly/ed25519/fe.ts
@@ -303,8 +303,7 @@ export function fe_mul (h: FieldElement, f: FieldElement, g: FieldElement): void
 	const g89xx_19: v128 = i32x4.mul(g89xx, v19)
 
 	// f[0]
-	let fi: i32 = load<i32>(f_ptr, 0)
-	let fv: v128 = i32x4.splat(fi)
+	let fv: v128 = v128.load_splat<i32>(f_ptr, 0 << 2)
 
 	let h01: v128 = i64x2.extmul_low_i32x4_s(fv, g0123)
 	let h23: v128 = i64x2.extmul_high_i32x4_s(fv, g0123)
@@ -313,8 +312,7 @@ export function fe_mul (h: FieldElement, f: FieldElement, g: FieldElement): void
 	let h89: v128 = i64x2.extmul_low_i32x4_s(fv, g89xx)
 
 	// f[1]
-	fi = load<i32>(f_ptr, 4)
-	fv = i32x4.splat(fi)
+	fv = v128.load_splat<i32>(f_ptr, 1 << 2)
 	fv = i32x4.add(fv, v128.and(fv, odds))
 
 	let h12: v128 = i64x2.extmul_low_i32x4_s(fv, g0123)
@@ -324,8 +322,7 @@ export function fe_mul (h: FieldElement, f: FieldElement, g: FieldElement): void
 	let h90: v128 = i64x2.extmul_low_i32x4_s(fv, v128.bitselect(g89xx, g89xx_19, m))
 
 	// f[2]
-	fi = load<i32>(f_ptr, 8)
-	fv = i32x4.splat(fi)
+	fv = v128.load_splat<i32>(f_ptr, 2 << 2)
 
 	let t0: v128 = i64x2.extmul_low_i32x4_s(fv, g0123)
 	let t1: v128 = i64x2.extmul_high_i32x4_s(fv, g0123)
@@ -340,8 +337,7 @@ export function fe_mul (h: FieldElement, f: FieldElement, g: FieldElement): void
 	h01 = i64x2.add(h01, t4)
 
 	// f[3]
-	fi = load<i32>(f_ptr, 12)
-	fv = i32x4.splat(fi)
+	fv = v128.load_splat<i32>(f_ptr, 3 << 2)
 	fv = i32x4.add(fv, v128.and(fv, odds))
 
 	t0 = i64x2.extmul_low_i32x4_s(fv, g0123)
@@ -357,8 +353,7 @@ export function fe_mul (h: FieldElement, f: FieldElement, g: FieldElement): void
 	h12 = i64x2.add(h12, t4)
 
 	// f[4]
-	fi = load<i32>(f_ptr, 16)
-	fv = i32x4.splat(fi)
+	fv = v128.load_splat<i32>(f_ptr, 4 << 2)
 
 	t0 = i64x2.extmul_low_i32x4_s(fv, g0123)
 	t1 = i64x2.extmul_high_i32x4_s(fv, g0123)
@@ -373,8 +368,7 @@ export function fe_mul (h: FieldElement, f: FieldElement, g: FieldElement): void
 	h23 = i64x2.add(h23, t4)
 
 	// f[5]
-	fi = load<i32>(f_ptr, 20)
-	fv = i32x4.splat(fi)
+	fv = v128.load_splat<i32>(f_ptr, 5 << 2)
 	fv = i32x4.add(fv, v128.and(fv, odds))
 
 	t0 = i64x2.extmul_low_i32x4_s(fv, g0123)
@@ -390,8 +384,7 @@ export function fe_mul (h: FieldElement, f: FieldElement, g: FieldElement): void
 	h34 = i64x2.add(h34, t4)
 
 	// f[6]
-	fi = load<i32>(f_ptr, 24)
-	fv = i32x4.splat(fi)
+	fv = v128.load_splat<i32>(f_ptr, 6 << 2)
 
 	t0 = i64x2.extmul_low_i32x4_s(fv, g0123)
 	t1 = i64x2.extmul_high_i32x4_s(fv, g0123)
@@ -406,8 +399,7 @@ export function fe_mul (h: FieldElement, f: FieldElement, g: FieldElement): void
 	h45 = i64x2.add(h45, t4)
 
 	// f[7]
-	fi = load<i32>(f_ptr, 28)
-	fv = i32x4.splat(fi)
+	fv = v128.load_splat<i32>(f_ptr, 7 << 2)
 	fv = i32x4.add(fv, v128.and(fv, odds))
 
 	t0 = i64x2.extmul_low_i32x4_s(fv, g0123)
@@ -423,8 +415,7 @@ export function fe_mul (h: FieldElement, f: FieldElement, g: FieldElement): void
 	h56 = i64x2.add(h56, t4)
 
 	// f[8]
-	fi = load<i32>(f_ptr, 32)
-	fv = i32x4.splat(fi)
+	fv = v128.load_splat<i32>(f_ptr, 8 << 2)
 
 	t0 = i64x2.extmul_low_i32x4_s(fv, g0123)
 	t1 = i64x2.extmul_high_i32x4_s(fv, g0123_19)
@@ -439,8 +430,7 @@ export function fe_mul (h: FieldElement, f: FieldElement, g: FieldElement): void
 	h67 = i64x2.add(h67, t4)
 
 	// f[9]
-	fi = load<i32>(f_ptr, 36)
-	fv = i32x4.splat(fi)
+	fv = v128.load_splat<i32>(f_ptr, 9 << 2)
 	fv = i32x4.add(fv, v128.and(fv, odds))
 
 	t0 = i64x2.extmul_low_i32x4_s(fv, v128.bitselect(g0123, g0123_19, m))