From ca7ad58236db090f48740a13abbb6b850f1df4ee Mon Sep 17 00:00:00 2001 From: BlueCube3310 <53150244+BlueCube3310@users.noreply.github.com> Date: Thu, 7 Nov 2024 11:42:39 +0100 Subject: [PATCH] bcdec: Fix decompresssing mipmaps not divisible by 4 --- modules/bcdec/image_decompress_bcdec.cpp | 113 +++++++++++++++++------ 1 file changed, 87 insertions(+), 26 deletions(-) diff --git a/modules/bcdec/image_decompress_bcdec.cpp b/modules/bcdec/image_decompress_bcdec.cpp index c76470e3cc6..8a62cd2a4e9 100644 --- a/modules/bcdec/image_decompress_bcdec.cpp +++ b/modules/bcdec/image_decompress_bcdec.cpp @@ -47,7 +47,6 @@ inline void bcdec_bc6h_half_u(const void *compressedBlock, void *decompressedBlo static void decompress_image(BCdecFormat format, const void *src, void *dst, const uint64_t width, const uint64_t height) { const uint8_t *src_blocks = reinterpret_cast(src); uint8_t *dec_blocks = reinterpret_cast(dst); - uint64_t src_pos = 0, dst_pos = 0; #define DECOMPRESS_LOOP(func, block_size, color_bytesize, color_components) \ for (uint64_t y = 0; y < height; y += 4) { \ @@ -59,34 +58,96 @@ static void decompress_image(BCdecFormat format, const void *src, void *dst, con dst_pos += 3 * width * color_bytesize; \ } - switch (format) { - case BCdec_BC1: { - DECOMPRESS_LOOP(bcdec_bc1, BCDEC_BC1_BLOCK_SIZE, 4, 4) - } break; - case BCdec_BC2: { - DECOMPRESS_LOOP(bcdec_bc2, BCDEC_BC2_BLOCK_SIZE, 4, 4) - } break; - case BCdec_BC3: { - DECOMPRESS_LOOP(bcdec_bc3, BCDEC_BC3_BLOCK_SIZE, 4, 4) - } break; - case BCdec_BC4: { - DECOMPRESS_LOOP(bcdec_bc4, BCDEC_BC4_BLOCK_SIZE, 1, 1) - } break; - case BCdec_BC5: { - DECOMPRESS_LOOP(bcdec_bc5, BCDEC_BC5_BLOCK_SIZE, 2, 2) - } break; - case BCdec_BC6U: { - DECOMPRESS_LOOP(bcdec_bc6h_half_u, BCDEC_BC6H_BLOCK_SIZE, 6, 3) - } break; - case BCdec_BC6S: { - DECOMPRESS_LOOP(bcdec_bc6h_half_s, BCDEC_BC6H_BLOCK_SIZE, 6, 3) - } break; - case BCdec_BC7: { - DECOMPRESS_LOOP(bcdec_bc7, BCDEC_BC7_BLOCK_SIZE, 4, 4) - } break; +#define DECOMPRESS_LOOP_SAFE(func, block_size, color_bytesize, color_components, output) \ + for (uint64_t y = 0; y < height; y += 4) { \ + for (uint64_t x = 0; x < width; x += 4) { \ + const uint32_t yblock = MIN(height - y, 4ul); \ + const uint32_t xblock = MIN(width - x, 4ul); \ + \ + const bool incomplete = yblock < 4 && xblock < 4; \ + uint8_t *dec_out = incomplete ? output : &dec_blocks[y * 4 * width + x * color_bytesize]; \ + \ + func(&src_blocks[src_pos], dec_out, 4 * color_components); \ + src_pos += block_size; \ + \ + if (incomplete) { \ + for (uint32_t cy = 0; cy < yblock; cy++) { \ + for (uint32_t cx = 0; cx < xblock; cx++) { \ + memcpy(&dec_blocks[(y + cy) * 4 * width + (x + cx) * color_bytesize], &output[cy * 4 + cx * color_bytesize], color_bytesize); \ + } \ + } \ + } \ + } \ + } + + if (width % 4 != 0 || height % 4 != 0) { + uint64_t src_pos = 0; + + uint8_t r8_output[4 * 4]; + uint8_t rg8_output[4 * 4 * 2]; + uint8_t rgba8_output[4 * 4 * 4]; + uint8_t rgbh_output[4 * 4 * 6]; + + switch (format) { + case BCdec_BC1: { + DECOMPRESS_LOOP_SAFE(bcdec_bc1, BCDEC_BC1_BLOCK_SIZE, 4, 4, rgba8_output) + } break; + case BCdec_BC2: { + DECOMPRESS_LOOP_SAFE(bcdec_bc2, BCDEC_BC2_BLOCK_SIZE, 4, 4, rgba8_output) + } break; + case BCdec_BC3: { + DECOMPRESS_LOOP_SAFE(bcdec_bc3, BCDEC_BC3_BLOCK_SIZE, 4, 4, rgba8_output) + } break; + case BCdec_BC4: { + DECOMPRESS_LOOP_SAFE(bcdec_bc4, BCDEC_BC4_BLOCK_SIZE, 1, 1, r8_output) + } break; + case BCdec_BC5: { + DECOMPRESS_LOOP_SAFE(bcdec_bc5, BCDEC_BC5_BLOCK_SIZE, 2, 2, rg8_output) + } break; + case BCdec_BC6U: { + DECOMPRESS_LOOP_SAFE(bcdec_bc6h_half_u, BCDEC_BC6H_BLOCK_SIZE, 6, 3, rgbh_output) + } break; + case BCdec_BC6S: { + DECOMPRESS_LOOP_SAFE(bcdec_bc6h_half_s, BCDEC_BC6H_BLOCK_SIZE, 6, 3, rgbh_output) + } break; + case BCdec_BC7: { + DECOMPRESS_LOOP_SAFE(bcdec_bc7, BCDEC_BC7_BLOCK_SIZE, 4, 4, rgba8_output) + } break; + } + + } else { + uint64_t src_pos = 0, dst_pos = 0; + + switch (format) { + case BCdec_BC1: { + DECOMPRESS_LOOP(bcdec_bc1, BCDEC_BC1_BLOCK_SIZE, 4, 4) + } break; + case BCdec_BC2: { + DECOMPRESS_LOOP(bcdec_bc2, BCDEC_BC2_BLOCK_SIZE, 4, 4) + } break; + case BCdec_BC3: { + DECOMPRESS_LOOP(bcdec_bc3, BCDEC_BC3_BLOCK_SIZE, 4, 4) + } break; + case BCdec_BC4: { + DECOMPRESS_LOOP(bcdec_bc4, BCDEC_BC4_BLOCK_SIZE, 1, 1) + } break; + case BCdec_BC5: { + DECOMPRESS_LOOP(bcdec_bc5, BCDEC_BC5_BLOCK_SIZE, 2, 2) + } break; + case BCdec_BC6U: { + DECOMPRESS_LOOP(bcdec_bc6h_half_u, BCDEC_BC6H_BLOCK_SIZE, 6, 3) + } break; + case BCdec_BC6S: { + DECOMPRESS_LOOP(bcdec_bc6h_half_s, BCDEC_BC6H_BLOCK_SIZE, 6, 3) + } break; + case BCdec_BC7: { + DECOMPRESS_LOOP(bcdec_bc7, BCDEC_BC7_BLOCK_SIZE, 4, 4) + } break; + } } #undef DECOMPRESS_LOOP +#undef DECOMPRESS_LOOP_SAFE } void image_decompress_bcdec(Image *p_image) {