From 3a4783cfa11fe563468f63339142d359fb6395ae Mon Sep 17 00:00:00 2001
From: adrien <adrien@bouvais.lu>
Date: Sat, 25 Apr 2026 18:47:13 +0200
Subject: [PATCH] Benchmark SIMD vs Native

---
 src/Scalar.zig    |  2 --
 src/Vector.zig    |  1 +
 src/benchmark.zig | 77 ++++++++++++++++++++++++++++++++++++++++++++---
 src/main.zig      | 10 +++---
 4 files changed, 79 insertions(+), 11 deletions(-)

diff --git a/src/Scalar.zig b/src/Scalar.zig
index ecc35f6..34877cb 100644
--- a/src/Scalar.zig
+++ b/src/Scalar.zig
@@ -7,8 +7,6 @@ const UnitScale = Scales.UnitScale;
 const Dimensions = @import("Dimensions.zig");
 const Dimension = Dimensions.Dimension;
 
-// ---------------------------------------------------------------------------
-
 /// A dimensioned scalar value. `T` is the numeric type, `d` the dimension exponents, `s` the SI scales.
 /// All dimension and unit tracking is resolved at comptime — zero runtime overhead.
 pub fn Scalar(comptime T: type, comptime d_opt: Dimensions.ArgOpts, comptime s_opt: Scales.ArgOpts) type {
diff --git a/src/Vector.zig b/src/Vector.zig
index 2e2bdee..62f9325 100644
--- a/src/Vector.zig
+++ b/src/Vector.zig
@@ -74,6 +74,7 @@ pub fn Vector(comptime len: usize, comptime Q: type) type {
             }
             return res;
         }
+
         /// Element-wise subtraction. Dimensions must match; scales resolve to the finer of the two.
         pub inline fn sub(self: Self, rhs: anytype) Vector(len, Scalar(
             T,
diff --git a/src/benchmark.zig b/src/benchmark.zig
index 2d9d5d0..1d77d0e 100644
--- a/src/benchmark.zig
+++ b/src/benchmark.zig
@@ -11,14 +11,25 @@ pub fn main(init: std.process.Init) !void {
 
     io = init.io;
 
-    try bench_Scalar(&stdout_writer.interface);
+    try vectorSIMDvsNative(f64, &stdout_writer.interface);
     try stdout_writer.flush();
-    try bench_vsNative(&stdout_writer.interface);
+    try vectorSIMDvsNative(f32, &stdout_writer.interface);
     try stdout_writer.flush();
-    try bench_crossTypeVsNative(&stdout_writer.interface);
+    try vectorSIMDvsNative(i32, &stdout_writer.interface);
     try stdout_writer.flush();
-    try bench_Vector(&stdout_writer.interface);
+    try vectorSIMDvsNative(i64, &stdout_writer.interface);
     try stdout_writer.flush();
+    try vectorSIMDvsNative(i128, &stdout_writer.interface);
+    try stdout_writer.flush();
+
+    // try bench_Scalar(&stdout_writer.interface);
+    // try stdout_writer.flush();
+    // try bench_vsNative(&stdout_writer.interface);
+    // try stdout_writer.flush();
+    // try bench_crossTypeVsNative(&stdout_writer.interface);
+    // try stdout_writer.flush();
+    // try bench_Vector(&stdout_writer.interface);
+    // try stdout_writer.flush();
 }
 
 fn getTime() Io.Timestamp {
@@ -478,3 +489,61 @@ fn bench_Vector(writer: *std.Io.Writer) !void {
     }
     try writer.print("└──────────────────┴──────┴─────────┴─────────┴─────────┘\n", .{});
 }
+
+fn vectorSIMDvsNative(comptime T: type, writer: *std.Io.Writer) !void {
+    const iterations: u64 = 100_000;
+    const lens = [_]u32{ 1, 2, 3, 4, 5, 10, 100, 1_000, 10_000 };
+
+    try writer.print("\nSIMD Speedup Analysis: {s}\n", .{@typeName(T)});
+    try writer.print("┌────────────┬────────────┬────────────┬────────────┐\n", .{});
+    try writer.print("│ Vector Len │ Scalar (ns)│ Vector (ns)│ Speedup    │\n", .{});
+    try writer.print("├────────────┼────────────┼────────────┼────────────┤\n", .{});
+
+    inline for (lens) |vector_len| {
+        // --- Scalar Test ---
+        var scalar_val: T = 10;
+        const start_scalar = getTime();
+
+        var i: u64 = 0;
+        while (i < iterations * vector_len) : (i += 1) {
+            if (comptime @typeInfo(T) == .int)
+                scalar_val = scalar_val +% 1
+            else
+                scalar_val = scalar_val + 1;
+        }
+        const scalar_time = start_scalar.durationTo(getTime()).toNanoseconds();
+
+        // --- Vector Test ---
+        var vector_val: @Vector(vector_len, T) = @splat(20);
+        const start_vector = getTime();
+
+        i = 0;
+        const increment: @Vector(vector_len, T) = @splat(1);
+        while (i < iterations) : (i += 1) {
+            if (comptime @typeInfo(T) == .int)
+                vector_val = vector_val +% increment
+            else
+                vector_val = vector_val + increment;
+        }
+        const vector_time = start_vector.durationTo(getTime()).toNanoseconds();
+
+        // --- Results ---
+        const s_float = @as(f64, @floatFromInt(scalar_time));
+        const v_float = @as(f64, @floatFromInt(vector_time));
+
+        // Speedup = ScalarTime / VectorTime.
+        // > 1.0 means SIMD is faster.
+        const speedup = if (vector_time > 0) s_float / v_float else 0;
+
+        try writer.print("│ {d:<10} │ {d:>10} │ {d:>10} │ {d:>9.2}x │\n", .{
+            vector_len,
+            scalar_time,
+            vector_time,
+            speedup,
+        });
+
+        std.mem.doNotOptimizeAway(scalar_val);
+        std.mem.doNotOptimizeAway(vector_val);
+    }
+    try writer.print("└────────────┴────────────┴────────────┴────────────┘\n", .{});
+}
diff --git a/src/main.zig b/src/main.zig
index 015d36f..eab001d 100644
--- a/src/main.zig
+++ b/src/main.zig
@@ -8,9 +8,9 @@ pub const Base = @import("Base.zig");
 
 test {
     _ = @import("Scalar.zig");
-    _ = @import("Vector.zig");
-    _ = @import("Dimensions.zig");
-    _ = @import("Scales.zig");
-    _ = @import("Base.zig");
-    _ = @import("helper.zig");
+    // _ = @import("Vector.zig");
+    // _ = @import("Dimensions.zig");
+    // _ = @import("Scales.zig");
+    // _ = @import("Base.zig");
+    // _ = @import("helper.zig");
 }