Support for SIMD operations for RGG1555

Description

Up to 3x the performance for RGG1555 with SIMD operations. The current implementation is hardcoded for 16 bytes at a time (128 bits). Still requires some testing and validation for AVX2 and AVX512.

Fixes #45 (closed)

src/ppu.rs

@@ -2095,6 +2095,18 @@ impl Ppu {
     ///
     /// This method should provide the same results as the SIMD implementation.
     pub fn rgb888_to_rgb1555_scalar(rgb888_pixels: &[u8], rgb1555_pixels: &mut [u8]) {
+        assert!(
+            rgb888_pixels.len() % 3 == 0,
+            "Length of rgb888_pixels must be a multiple of 3"
+        );
+        assert!(
+            rgb1555_pixels.len() % 2 == 0,
+            "Length of rgb1555_pixels must be a multiple of 2"
+        );
+        assert!(
+            rgb888_pixels.len() / 3 == rgb1555_pixels.len() / 2,
+            "Length of rgb1555_pixels must be two thirds the length of rgb888_pixels"
+        );
         for index in 0..rgb888_pixels.len() / RGB_SIZE {
             let (r, g, b) = (
                 rgb888_pixels[index * RGB_SIZE],
@@ -2132,6 +2144,10 @@ impl Ppu {
             rgb1555_pixels.len() % 2 == 0,
             "Length of rgb1555_pixels must be a multiple of 2"
         );
+        assert!(
+            rgb888_pixels.len() / 3 == rgb1555_pixels.len() / 2,
+            "Length of rgb1555_pixels must be two thirds the length of rgb888_pixels"
+        );
 
         let num_pixels = rgb888_pixels.len() / 3;
         let simd_chunks = num_pixels / SIMD_WIDTH;