[PATCH 1/3] Add decompression functions for DXT1, DXT3, and DXT5
Connor McAdams
conmanx360 at gmail.com
Thu Jun 21 15:32:38 CDT 2018
This adds decompression functions for DXT1, DXT3, and DXT5 formats,
which covers all the possible DXT formats. These are used for
decompressing DXTn volume textures, and converting them to b8g8r8a8.
---
dlls/wined3d/utils.c | 321 ++++++++++++++++++++++++++++++++++++++++++++++++---
1 file changed, 305 insertions(+), 16 deletions(-)
diff --git a/dlls/wined3d/utils.c b/dlls/wined3d/utils.c
index 937c1bc..ce1ccdf 100644
--- a/dlls/wined3d/utils.c
+++ b/dlls/wined3d/utils.c
@@ -840,6 +840,305 @@ static void convert_s8_uint_d24_float(const BYTE *src, BYTE *dst, UINT src_row_p
}
}
+static void dxt5_decompress_block(const BYTE *src, BYTE *dst, UINT width, UINT height, UINT depth,
+ unsigned int x_pos, unsigned int y_pos, unsigned int z_pos, UINT dst_slice_pitch, UINT64 cur_block)
+{
+ DWORD alpha_lookup;
+ DWORD bgra;
+ DWORD temp;
+ UINT64 alpha_block, alpha_index, color_block, color_index;
+ DWORD *Dest;
+ const UINT64 *Source;
+ unsigned short color[2];
+ unsigned int i, x, y;
+ unsigned char alpha_val, color_val;
+ unsigned char alpha[8];
+ unsigned char r[4];
+ unsigned char g[4];
+ unsigned char b[4];
+
+ Source = (const UINT64 *)(src + cur_block * 16);
+ alpha_block = Source[0];
+ color_block = Source[1];
+
+ alpha[0] = alpha_block & 0xFF;
+ alpha[1] = (alpha_block >> 8) & 0xFF;
+
+ /* Generate alpha lookup values */
+ if (alpha[0] > alpha[1])
+ {
+ for (i = 0; i < 6; i++)
+ alpha[2 + i] = (((6 - i) * alpha[0]) + ((1 + i) * alpha[1])) / 7;
+ }
+ else if (alpha[0] <= alpha[1])
+ {
+ for (i = 0; i < 4; i++)
+ alpha[2 + i] = (((4 - i) * alpha[0]) + ((1 + i) * alpha[1])) / 5;
+ alpha[6] = 0;
+ alpha[7] = 255;
+ }
+
+ /* Generate color lookup values */
+ color[0] = color_block & 0xFFFF;
+ color[1] = (color_block >> 16) & 0xFFFF;
+
+ for (i = 0; i < 2; i++)
+ {
+ temp = (color[i] >> 11) * 255 + 16;
+ r[i] = (temp / 32 + temp) / 32;
+ temp = ((color[i] >> 5) & 0x3F) * 255 + 32;
+ g[i] = (temp / 64 + temp) / 64;
+ temp = (color[i] & 0x1F) * 255 + 16;
+ b[i] = (temp / 32 + temp) / 32;
+ }
+
+ for (i = 0; i < 2; i++)
+ {
+ r[2 + i] = (2 * r[0 + i] + r[1 - i]) / 3;
+ g[2 + i] = (2 * g[0 + i] + g[1 - i]) / 3;
+ b[2 + i] = (2 * b[0 + i] + b[1 - i]) / 3;
+ }
+
+ color_index = (color_block >> 32) & 0xFFFFFFFF;
+ alpha_index = (alpha_block >> 16);
+
+ Dest = (DWORD *)(dst + z_pos * dst_slice_pitch);
+
+ for (y = 0; y < 4; y++) {
+ if (y_pos + y >= height)
+ break;
+ for (x = 0; x < 4; x++) {
+ if (x_pos + x >= width)
+ break;
+
+ color_val = 0;
+ alpha_val = 0;
+ bgra = 0;
+
+ color_val = (color_index >> (y * 8));
+ color_val = (color_val >> (x * 2)) & 0x3;
+ alpha_lookup = (alpha_index >> (y * 12)) & 0xFFF;
+ alpha_val = (alpha_lookup >> (x * 3)) & 0x7;
+ bgra = ((alpha[alpha_val] << 24) | (r[color_val] << 16) | (g[color_val] << 8) | b[color_val]);
+ Dest[(y_pos + y) * width + (x_pos + x)] = bgra;
+ }
+ }
+}
+
+static void convert_dxt5_b8g8r8a8_unorm(const BYTE *src, BYTE *dst, UINT src_row_pitch, UINT src_slice_pitch,
+ UINT dst_row_pitch, UINT dst_slice_pitch, UINT width, UINT height, UINT depth)
+{
+ UINT64 current_block;
+ unsigned int x, y, z;
+
+ current_block = 0;
+
+ for (z = 0; z < depth; z++)
+ {
+ for (y = 0; y < height; y += 4)
+ {
+ for (x = 0; x < width; x += 4)
+ {
+ dxt5_decompress_block(src, dst, width, height, depth, x, y, z, dst_slice_pitch, current_block);
+ current_block++;
+ }
+ }
+ }
+}
+
+static void dxt3_decompress_block(const BYTE *src, BYTE *dst, UINT width, UINT height, UINT depth,
+ unsigned int x_pos, unsigned int y_pos, unsigned int z_pos, UINT dst_slice_pitch, UINT64 cur_block)
+{
+ DWORD bgra;
+ DWORD temp;
+ UINT64 alpha_block, alpha_lookup, color_block, color_index;
+ DWORD *Dest;
+ const UINT64 *Source;
+ unsigned short color[2];
+ unsigned int i, x, y;
+ unsigned char alpha_val, color_val;
+ unsigned char r[4];
+ unsigned char g[4];
+ unsigned char b[4];
+
+ Source = (const UINT64 *)(src + cur_block * 16);
+ alpha_block = Source[0];
+ color_block = Source[1];
+
+ /* Generate color lookup values */
+ color[0] = color_block & 0xFFFF;
+ color[1] = (color_block >> 16) & 0xFFFF;
+
+ for (i = 0; i < 2; i++)
+ {
+ temp = (color[i] >> 11) * 255 + 16;
+ r[i] = (temp / 32 + temp) / 32;
+ temp = ((color[i] >> 5) & 0x3F) * 255 + 32;
+ g[i] = (temp / 64 + temp) / 64;
+ temp = (color[i] & 0x1F) * 255 + 16;
+ b[i] = (temp / 32 + temp) / 32;
+ }
+
+ for (i = 0; i < 2; i++)
+ {
+ r[2 + i] = (2 * r[0 + i] + r[1 - i]) / 3;
+ g[2 + i] = (2 * g[0 + i] + g[1 - i]) / 3;
+ b[2 + i] = (2 * b[0 + i] + b[1 - i]) / 3;
+ }
+
+ color_index = (color_block >> 32) & 0xFFFFFFFF;
+ Dest = (DWORD *)(dst + z_pos * dst_slice_pitch);
+
+ for (y = 0; y < 4; y++) {
+ if (y_pos + y >= height)
+ break;
+ for (x = 0; x < 4; x++) {
+ if (x_pos + x >= width)
+ break;
+
+ color_val = 0;
+ alpha_val = 0;
+ bgra = 0;
+
+ color_val = (color_index >> (y * 8));
+ color_val = (color_val >> (x * 2)) & 0x3;
+
+ alpha_lookup = (alpha_block >> (y * 16)) & 0xFFFF;
+ alpha_val = (alpha_lookup >> (x * 4)) & 0xF;
+ temp = alpha_val * 255 + 8;
+ alpha_val = (temp / 16 + temp) / 16;
+
+ bgra = ((alpha_val << 24) | (r[color_val] << 16) | (g[color_val] << 8) | b[color_val]);
+ Dest[(y_pos + y) * width + (x_pos + x)] = bgra;
+ }
+ }
+}
+
+static void convert_dxt3_b8g8r8a8_unorm(const BYTE *src, BYTE *dst, UINT src_row_pitch, UINT src_slice_pitch,
+ UINT dst_row_pitch, UINT dst_slice_pitch, UINT width, UINT height, UINT depth)
+{
+ UINT64 current_block;
+ unsigned int x, y, z;
+
+ current_block = 0;
+
+ for (z = 0; z < depth; z++)
+ {
+ for (y = 0; y < height; y += 4)
+ {
+ for (x = 0; x < width; x += 4)
+ {
+ dxt3_decompress_block(src, dst, width, height, depth, x, y, z, dst_slice_pitch, current_block);
+ current_block++;
+ }
+ }
+ }
+}
+
+static void dxt1_decompress_block(const BYTE *src, BYTE *dst, UINT width, UINT height, UINT depth,
+ unsigned int x_pos, unsigned int y_pos, unsigned int z_pos, UINT dst_slice_pitch, UINT64 cur_block)
+{
+ DWORD bgra;
+ DWORD temp;
+ UINT64 color_block, color_index;
+ DWORD *Dest;
+ const UINT64 *Source;
+ unsigned short color[2];
+ unsigned int i, x, y;
+ unsigned char color_val;
+ unsigned char alpha;
+ unsigned char r[4];
+ unsigned char g[4];
+ unsigned char b[4];
+ unsigned char use_alpha;
+
+ Source = (const UINT64 *)(src + cur_block * 8);
+ color_block = Source[0];
+
+ /* Generate color lookup values */
+ color[0] = color_block & 0xFFFF;
+ color[1] = (color_block >> 16) & 0xFFFF;
+
+ for (i = 0; i < 2; i++)
+ {
+ temp = (color[i] >> 11) * 255 + 16;
+ r[i] = (temp / 32 + temp) / 32;
+ temp = ((color[i] >> 5) & 0x3F) * 255 + 32;
+ g[i] = (temp / 64 + temp) / 64;
+ temp = (color[i] & 0x1F) * 255 + 16;
+ b[i] = (temp / 32 + temp) / 32;
+ }
+
+ if (color[0] > color[1])
+ {
+ for (i = 0; i < 2; i++)
+ {
+ r[2 + i] = (2 * r[0 + i] + r[1 - i]) / 3;
+ g[2 + i] = (2 * g[0 + i] + g[1 - i]) / 3;
+ b[2 + i] = (2 * b[0 + i] + b[1 - i]) / 3;
+ }
+ use_alpha = 0;
+ }
+ else if (color[0] <= color[1])
+ {
+ r[2] = (r[0] + r[1]) / 2;
+ g[2] = (g[0] + g[1]) / 2;
+ b[2] = (b[0] + b[1]) / 2;
+
+ r[3] = 0;
+ g[3] = 0;
+ b[3] = 0;
+
+ use_alpha = 1;
+ }
+
+ color_index = (color_block >> 32) & 0xFFFFFFFF;
+ Dest = (DWORD *)(dst + z_pos * dst_slice_pitch);
+
+ for (y = 0; y < 4; y++) {
+ if (y_pos + y >= height)
+ break;
+ for (x = 0; x < 4; x++) {
+ if (x_pos + x >= width)
+ break;
+
+ color_val = 0;
+ bgra = 0;
+
+ color_val = (color_index >> (y * 8));
+ color_val = (color_val >> (x * 2)) & 0x3;
+ if (color_val == 3 && use_alpha == 1)
+ alpha = 0;
+ else
+ alpha = 255;
+
+ bgra = ((alpha << 24) | (r[color_val] << 16) | (g[color_val] << 8) | b[color_val]);
+ Dest[(y_pos + y) * width + (x_pos + x)] = bgra;
+ }
+ }
+}
+
+static void convert_dxt1_b8g8r8a8_unorm(const BYTE *src, BYTE *dst, UINT src_row_pitch, UINT src_slice_pitch,
+ UINT dst_row_pitch, UINT dst_slice_pitch, UINT width, UINT height, UINT depth)
+{
+ UINT64 current_block;
+ unsigned int x, y, z;
+
+ current_block = 0;
+
+ for (z = 0; z < depth; z++)
+ {
+ for (y = 0; y < height; y += 4)
+ {
+ for (x = 0; x < width; x += 4)
+ {
+ dxt1_decompress_block(src, dst, width, height, depth, x, y, z, dst_slice_pitch, current_block);
+ current_block++;
+ }
+ }
+ }
+}
+
static void x8_d24_unorm_upload(const BYTE *src, BYTE *dst,
unsigned int src_row_pitch, unsigned int src_slice_pitch,
unsigned int dst_row_pitch, unsigned int dst_slice_pitch,
@@ -1118,27 +1417,27 @@ static const struct wined3d_format_texture_info format_texture_info[] =
GL_RGBA, GL_UNSIGNED_BYTE, 0,
WINED3DFMT_FLAG_TEXTURE | WINED3DFMT_FLAG_POSTPIXELSHADER_BLENDING | WINED3DFMT_FLAG_FILTERING
| WINED3DFMT_FLAG_SRGB_READ | WINED3DFMT_FLAG_COMPRESSED,
- EXT_TEXTURE_COMPRESSION_S3TC, NULL},
+ EXT_TEXTURE_COMPRESSION_S3TC, convert_dxt1_b8g8r8a8_unorm},
{WINED3DFMT_DXT2, GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT, 0,
GL_RGBA, GL_UNSIGNED_BYTE, 0,
WINED3DFMT_FLAG_TEXTURE | WINED3DFMT_FLAG_POSTPIXELSHADER_BLENDING | WINED3DFMT_FLAG_FILTERING
| WINED3DFMT_FLAG_SRGB_READ | WINED3DFMT_FLAG_COMPRESSED,
- EXT_TEXTURE_COMPRESSION_S3TC, NULL},
+ EXT_TEXTURE_COMPRESSION_S3TC, convert_dxt3_b8g8r8a8_unorm},
{WINED3DFMT_DXT3, GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT, 0,
GL_RGBA, GL_UNSIGNED_BYTE, 0,
WINED3DFMT_FLAG_TEXTURE | WINED3DFMT_FLAG_POSTPIXELSHADER_BLENDING | WINED3DFMT_FLAG_FILTERING
| WINED3DFMT_FLAG_SRGB_READ | WINED3DFMT_FLAG_COMPRESSED,
- EXT_TEXTURE_COMPRESSION_S3TC, NULL},
+ EXT_TEXTURE_COMPRESSION_S3TC, convert_dxt3_b8g8r8a8_unorm},
{WINED3DFMT_DXT4, GL_COMPRESSED_RGBA_S3TC_DXT5_EXT, GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT, 0,
GL_RGBA, GL_UNSIGNED_BYTE, 0,
WINED3DFMT_FLAG_TEXTURE | WINED3DFMT_FLAG_POSTPIXELSHADER_BLENDING | WINED3DFMT_FLAG_FILTERING
| WINED3DFMT_FLAG_SRGB_READ | WINED3DFMT_FLAG_COMPRESSED,
- EXT_TEXTURE_COMPRESSION_S3TC, NULL},
+ EXT_TEXTURE_COMPRESSION_S3TC, convert_dxt5_b8g8r8a8_unorm},
{WINED3DFMT_DXT5, GL_COMPRESSED_RGBA_S3TC_DXT5_EXT, GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT, 0,
GL_RGBA, GL_UNSIGNED_BYTE, 0,
WINED3DFMT_FLAG_TEXTURE | WINED3DFMT_FLAG_POSTPIXELSHADER_BLENDING | WINED3DFMT_FLAG_FILTERING
| WINED3DFMT_FLAG_SRGB_READ | WINED3DFMT_FLAG_COMPRESSED,
- EXT_TEXTURE_COMPRESSION_S3TC, NULL},
+ EXT_TEXTURE_COMPRESSION_S3TC, convert_dxt5_b8g8r8a8_unorm},
{WINED3DFMT_BC1_UNORM, GL_COMPRESSED_RGBA_S3TC_DXT1_EXT, GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT, 0,
GL_RGBA, GL_UNSIGNED_BYTE, 0,
WINED3DFMT_FLAG_TEXTURE | WINED3DFMT_FLAG_POSTPIXELSHADER_BLENDING | WINED3DFMT_FLAG_FILTERING
@@ -3394,17 +3693,7 @@ static void apply_format_fixups(struct wined3d_adapter *adapter, struct wined3d_
* for dx9 GPUs support it, some do not, so not supporting DXTn volumes is OK for d3d9.
*
* Note that GL_NV_texture_compression_vtc adds this functionality to OpenGL, but the
- * block layout is not compatible with the one used by d3d. See volume_dxt5_test. */
- idx = get_format_idx(WINED3DFMT_DXT1);
- gl_info->formats[idx].flags[WINED3D_GL_RES_TYPE_TEX_3D] &= ~WINED3DFMT_FLAG_TEXTURE;
- idx = get_format_idx(WINED3DFMT_DXT2);
- gl_info->formats[idx].flags[WINED3D_GL_RES_TYPE_TEX_3D] &= ~WINED3DFMT_FLAG_TEXTURE;
- idx = get_format_idx(WINED3DFMT_DXT3);
- gl_info->formats[idx].flags[WINED3D_GL_RES_TYPE_TEX_3D] &= ~WINED3DFMT_FLAG_TEXTURE;
- idx = get_format_idx(WINED3DFMT_DXT4);
- gl_info->formats[idx].flags[WINED3D_GL_RES_TYPE_TEX_3D] &= ~WINED3DFMT_FLAG_TEXTURE;
- idx = get_format_idx(WINED3DFMT_DXT5);
- gl_info->formats[idx].flags[WINED3D_GL_RES_TYPE_TEX_3D] &= ~WINED3DFMT_FLAG_TEXTURE;
+ * block layout is not compatible with the one used by d3d. See volume_dxtn_test. */
idx = get_format_idx(WINED3DFMT_BC1_UNORM);
gl_info->formats[idx].flags[WINED3D_GL_RES_TYPE_TEX_3D] &= ~WINED3DFMT_FLAG_TEXTURE;
idx = get_format_idx(WINED3DFMT_BC1_UNORM_SRGB);
--
2.7.4
More information about the wine-devel
mailing list