Merge pull request #218 from paolobarbolini/more-bounds

Remove even more bounds checks

Author: HeroicKatora

src/arch/mod.rs

@@ -20,7 +20,7 @@ pub fn get_color_convert_line_ycbcr() -> Option<unsafe fn(&[u8], &[u8], &[u8], &
 }
 
 /// Arch-specific implementation of 8x8 IDCT.
-pub fn get_dequantize_and_idct_block_8x8() -> Option<unsafe fn(&[i16], &[u16; 64], usize, &mut [u8])>
+pub fn get_dequantize_and_idct_block_8x8() -> Option<unsafe fn(&[i16; 64], &[u16; 64], usize, &mut [u8])>
 {
     #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
     #[allow(unsafe_code)]

src/arch/ssse3.rs

@@ -122,12 +122,11 @@ unsafe fn transpose8(data: &mut [__m128i; 8]) {
 #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
 #[target_feature(enable = "ssse3")]
 pub unsafe fn dequantize_and_idct_block_8x8(
-    coefficients: &[i16],
+    coefficients: &[i16; 64],
     quantization_table: &[u16; 64],
     output_linestride: usize,
     output: &mut [u8],
 ) {
-    assert!(coefficients.len() >= 64);
     // The loop below will write to positions [output_linestride * i, output_linestride * i + 8)
     // for 0<=i<8. Thus, the last accessed position is at an offset of output_linestrade * 7 + 7,
     // and if that position is in-bounds, so are all other accesses.

src/huffman.rs

@@ -213,8 +213,9 @@ impl HuffmanTable {
             let bits_remaining = LUT_BITS - size;
             let start = (huffcode[i] << bits_remaining) as usize;
 
-            for j in 0 .. 1 << bits_remaining {
-                lut[start + j] = (values[i], size);
+            let val = (values[i], size);
+            for b in &mut lut[start..][..1 << bits_remaining] {
+                *b = val;
             }
         }
 

src/idct.rs

@@ -508,10 +508,11 @@ fn dequantize_and_idct_block_4x4(
         let x0 = x0 + Wrapping(1 << (FINAL_BITS - 1)) + Wrapping(128 << FINAL_BITS);
         let x2 = x2 + Wrapping(1 << (FINAL_BITS - 1)) + Wrapping(128 << FINAL_BITS);
 
-        output[i * output_linestride + 0] = stbi_clamp((x0 + t2) >> FINAL_BITS);
-        output[i * output_linestride + 3] = stbi_clamp((x0 - t2) >> FINAL_BITS);
-        output[i * output_linestride + 1] = stbi_clamp((x2 + t0) >> FINAL_BITS);
-        output[i * output_linestride + 2] = stbi_clamp((x2 - t0) >> FINAL_BITS);
+        let output = &mut output[i * output_linestride..][..4];
+        output[0] = stbi_clamp((x0 + t2) >> FINAL_BITS);
+        output[3] = stbi_clamp((x0 - t2) >> FINAL_BITS);
+        output[1] = stbi_clamp((x2 + t0) >> FINAL_BITS);
+        output[2] = stbi_clamp((x2 - t0) >> FINAL_BITS);
     }
 }
 

src/upsampler.rs

@@ -181,7 +181,10 @@ impl Upsample for UpsamplerH1V2 {
         let input_near = &input[row_near as usize * row_stride ..];
         let input_far = &input[row_far as usize * row_stride ..];
 
-        for i in 0 .. output_width {
+        let output = &mut output[..output_width];
+        let input_near = &input_near[..output_width];
+        let input_far = &input_far[..output_width];
+        for i in 0..output_width {
             output[i] = ((3 * input_near[i] as u32 + input_far[i] as u32 + 2) >> 2) as u8;
         }
     }