1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
// SPDX-License-Identifier: BSD-3-Clause
// Copyright (c) 2026 Fernando Sahmkow
#pragma once
/// @file txtr_internal.h
/// @brief On-disk layout, surface-format table, and mip arithmetic for the
/// Overwatch TXTR container.
///
/// Every table here is transcribed from the Overwatch client's own code rather
/// than inferred from sample files; see docs/OW_TXTR_FORMAT.md for the
/// functions each one came from.
///
/// Internal header -- not part of the public include path.
#include <algorithm>
#include <cstring>
#include <optional>
#include <span>
#include <whiteout/common_types.h>
#include <whiteout/textures/texture.h>
#include <whiteout/textures/txtr/types.h>
namespace whiteout::textures::txtr {
// ============================================================================
// Container constants
// ============================================================================
/// Smallest header the client will read fields out of.
static constexpr u32 TXTR_HEADER_SIZE = 0x20;
/// Rejection thresholds the client applies before it will touch a header.
static constexpr u8 TXTR_MAX_MIP_COUNT = 0x10;
static constexpr u8 TXTR_MAX_FORMAT_CODE = 0x71;
#pragma pack(push, 1)
/// The fields of the `004` header the client actually reads. Bytes 0x13-0x1F
/// are untouched by every function on the texture path; the inline pixel block
/// starts at 0x20.
struct Header {
u16 flags; ///< Dimension and behaviour bits.
u8 mipCount; ///< Total mip levels; rejected above 0x10.
u8 format; ///< Surface format code; rejected at 0x72 and above.
u8 surfaces; ///< Depth for 3D, array size or face count otherwise.
u8 unknown05; ///< Stored in the runtime object, never read back.
u8 payloadCount; ///< Payload chain length; rejected above 4.
u8 unknown07; ///< `== 1` sets runtime flag 0x4000.
u16 width; ///< Mip 0 width; rejected when zero.
u16 height; ///< Mip 0 height; rejected when zero.
u32 inlineSize; ///< Pixel bytes of the whole chain, not of the block at 0x20.
u8 headerMipCount; ///< Mips resident inline; rejected above mipCount.
u8 streamMask; ///< 4-bit mask driving the quality-based mip skip.
u8 unknown12; ///< Stored in the runtime object, never read back.
};
/// The prefix each payload starts with, including the header's inline block
/// whenever `payloadCount` is non-zero.
struct PayloadPrefix {
u32 mipsFromSmallest; ///< First covered mip is `mipCount - this - 1`.
u32 mipCount; ///< Consecutive mips this payload carries.
u32 dataSize; ///< Byte count of pixel data.
u32 dataOffset; ///< Offset of that data from the payload's own start.
};
#pragma pack(pop)
static_assert(sizeof(Header) == 0x13, "TXTR Header covers bytes 0x00-0x12");
static_assert(sizeof(PayloadPrefix) == kPayloadPrefixSize, "TXTR PayloadPrefix must be 16 bytes");
// ============================================================================
// Surface formats
// ============================================================================
/// How a surface format compresses, as the client's own format table reports
/// it. Kinds 8-10 only appear in console builds and have no BCn equivalent.
enum class Compression : u32 {
None = 0,
BC1 = 1,
BC2 = 2,
BC3 = 3,
BC4 = 4,
BC5 = 5,
BC6H = 6,
BC7 = 7,
ConsoleA = 8,
ConsoleB = 9,
ConsoleC = 10,
};
/// Everything the size arithmetic needs about one surface format.
struct FormatInfo {
Compression compression = Compression::None;
u32 bitsPerPixel = 0; ///< For compressed formats, bits per *pixel*, not per block.
u32 blockWidth = 1;
u32 blockHeight = 1;
u32 blockBytes = 0; ///< Only meaningful when compressed.
};
/// Compression kind and bits-per-pixel for a raw format code.
///
/// Codes 90, 93 and 94 are deliberately absent: the client's table has no
/// entry for them either, so a header naming one cannot be loaded.
inline std::optional<FormatInfo> txtr_format_info(u8 code) {
const auto uncompressed = [](u32 bpp) { return FormatInfo{Compression::None, bpp, 1, 1, 0}; };
const auto block = [](Compression c, u32 bpp, u32 dim, u32 bytes) {
return FormatInfo{c, bpp, dim, dim, bytes};
};
if (code >= 1 && code <= 4)
return uncompressed(128);
if (code >= 5 && code <= 8)
return uncompressed(96);
if (code >= 9 && code <= 22)
return uncompressed(64);
if ((code >= 23 && code <= 48) || (code >= 68 && code <= 70) || code == 88 || code == 89 ||
code == 91 || code == 92)
return uncompressed(32);
if ((code >= 49 && code <= 60) || code == 86 || code == 87)
return uncompressed(16);
if (code >= 61 && code <= 66)
return uncompressed(8);
if (code == 67)
return uncompressed(1);
if (code == 113)
return uncompressed(12);
if (code >= 71 && code <= 73)
return block(Compression::BC1, 4, 4, 8);
if (code >= 74 && code <= 76)
return block(Compression::BC2, 8, 4, 16);
if (code >= 77 && code <= 79)
return block(Compression::BC3, 8, 4, 16);
if (code >= 80 && code <= 82)
return block(Compression::BC4, 4, 4, 8);
if (code >= 83 && code <= 85)
return block(Compression::BC5, 8, 4, 16);
if (code >= 95 && code <= 97)
return block(Compression::BC6H, 8, 4, 16);
if (code >= 98 && code <= 100)
return block(Compression::BC7, 8, 4, 16);
// Console-only compressed formats. Recognised so the parser can name them
// in its error message, but not decodable here.
if (code >= 101 && code <= 103)
return block(Compression::BC1, 4, 4, 8);
if (code >= 104 && code <= 106)
return block(Compression::ConsoleA, 8, 4, 16);
if (code >= 107 && code <= 109)
return block(Compression::ConsoleB, 4, 6, 16);
if (code >= 110 && code <= 112)
return block(Compression::ConsoleC, 2, 8, 16);
return std::nullopt;
}
/// How a surface format maps onto the library's own pixel formats.
struct FormatMapping {
PixelFormat format = PixelFormat::RGBA8;
bool isSrgb = false;
bool isBgra = false; ///< Needs an R/B swap after the copy.
bool forceOpaque = false; ///< X8 alpha channel; fill it with 0xFF.
};
/// Map a raw format code onto a library PixelFormat, or nullopt when the
/// library has no equivalent encoding.
///
/// TYPELESS and UINT variants share a layout with the UNORM form and map to the
/// same library format. SNORM and SINT are left out: PixelFormat carries no
/// signedness, so they would decode as unsigned and render wrong. Overwatch
/// ships none of them.
inline std::optional<FormatMapping> txtr_format_mapping(u8 code) {
const auto info = txtr_format_info(code);
if (!info)
return std::nullopt;
const u32 dxgi = dxgiFormatFor(code);
switch (info->compression) {
case Compression::BC1:
if (code > 100)
return std::nullopt; // Console re-use of the BC1 kind.
return FormatMapping{PixelFormat::BC1, dxgi == 72, false, false};
case Compression::BC2:
return FormatMapping{PixelFormat::BC2, dxgi == 75, false, false};
case Compression::BC3:
return FormatMapping{PixelFormat::BC3, dxgi == 78, false, false};
case Compression::BC4:
return FormatMapping{PixelFormat::BC4, false, false, false};
case Compression::BC5:
return FormatMapping{PixelFormat::BC5, false, false, false};
case Compression::BC6H:
return FormatMapping{PixelFormat::BC6H, false, false, false};
case Compression::BC7:
return FormatMapping{PixelFormat::BC7, dxgi == 99, false, false};
case Compression::ConsoleA:
case Compression::ConsoleB:
case Compression::ConsoleC:
return std::nullopt;
case Compression::None:
break;
}
switch (dxgi) {
case 1: // R32G32B32A32_TYPELESS
case 2: // R32G32B32A32_FLOAT
return FormatMapping{PixelFormat::RGBA32F, false, false, false};
case 9: // R16G16B16A16_TYPELESS
case 10: // R16G16B16A16_FLOAT
return FormatMapping{PixelFormat::RGBA16F, false, false, false};
case 34: // R16G16_FLOAT
return FormatMapping{PixelFormat::RG16F, false, false, false};
case 54: // R16_FLOAT
return FormatMapping{PixelFormat::R16F, false, false, false};
case 11: // R16G16B16A16_UNORM
case 12: // R16G16B16A16_UINT
return FormatMapping{PixelFormat::RGBA16, false, false, false};
case 15: // R32G32_TYPELESS
case 16: // R32G32_FLOAT
return FormatMapping{PixelFormat::RG32F, false, false, false};
case 27: // R8G8B8A8_TYPELESS
case 28: // R8G8B8A8_UNORM
case 30: // R8G8B8A8_UINT
return FormatMapping{PixelFormat::RGBA8, false, false, false};
case 29: // R8G8B8A8_UNORM_SRGB
return FormatMapping{PixelFormat::RGBA8, true, false, false};
case 33: // R16G16_TYPELESS
case 35: // R16G16_UNORM
case 36: // R16G16_UINT
return FormatMapping{PixelFormat::RG16, false, false, false};
case 39: // R32_TYPELESS
case 41: // R32_FLOAT
return FormatMapping{PixelFormat::R32F, false, false, false};
case 48: // R8G8_TYPELESS
case 49: // R8G8_UNORM
case 50: // R8G8_UINT
return FormatMapping{PixelFormat::RG8, false, false, false};
case 53: // R16_TYPELESS
case 56: // R16_UNORM
case 57: // R16_UINT
return FormatMapping{PixelFormat::R16, false, false, false};
case 60: // R8_TYPELESS
case 61: // R8_UNORM
case 62: // R8_UINT
case 65: // A8_UNORM
return FormatMapping{PixelFormat::R8, false, false, false};
case 87: // B8G8R8A8_UNORM
case 90: // B8G8R8A8_TYPELESS
return FormatMapping{PixelFormat::RGBA8, false, true, false};
case 88: // B8G8R8X8_UNORM
return FormatMapping{PixelFormat::RGBA8, false, true, true};
case 91: // B8G8R8A8_UNORM_SRGB
return FormatMapping{PixelFormat::RGBA8, true, true, false};
default:
return std::nullopt;
}
}
// ============================================================================
// Mip arithmetic
// ============================================================================
/// Byte layout of one mip level, matching the client's own computation.
struct MipInfo {
u32 width = 0; ///< Block-aligned width.
u32 height = 0; ///< Block-aligned height.
u32 depth = 1; ///< Slice count at this level.
u32 rowPitch = 0; ///< Bytes per row of blocks (or of pixels).
u32 slicePitch = 0; ///< Bytes per depth slice.
u64 size = 0; ///< slicePitch * depth -- what the payload advances by.
};
/// Size and pitch of mip @p mip of a surface whose level 0 is
/// @p width x @p height x @p depth.
///
/// Dimensions are clamped up to one block before alignment, so a 1x1 mip of a
/// BCn surface still occupies a full 4x4 block.
inline MipInfo txtr_mip_info(const FormatInfo& fi, u32 width, u32 height, u32 depth, u32 mip) {
const u32 bw = std::max(fi.blockWidth, 1u);
const u32 bh = std::max(fi.blockHeight, 1u);
const u32 w = std::max(width >> mip, bw);
const u32 h = std::max(height >> mip, bh);
MipInfo out;
out.width = ((w + bw - 1) / bw) * bw;
out.height = ((h + bh - 1) / bh) * bh;
out.depth = std::max(depth >> mip, 1u);
if (fi.compression == Compression::None) {
out.rowPitch = (fi.bitsPerPixel * out.width) / 8;
out.slicePitch = out.height * out.rowPitch;
} else {
out.rowPitch = fi.blockBytes * (out.width / bw);
out.slicePitch = out.rowPitch * (out.height / bh);
}
out.size = static_cast<u64>(out.slicePitch) * out.depth;
return out;
}
// ============================================================================
// Header helpers
// ============================================================================
/// Read the fixed header fields out of @p buffer. Returns nullopt when the
/// buffer is too short to hold them.
inline std::optional<Header> txtr_read_header(std::span<const u8> buffer) {
if (buffer.size() < TXTR_HEADER_SIZE)
return std::nullopt;
Header h{};
std::memcpy(&h, buffer.data(), sizeof(Header));
return h;
}
/// Apply exactly the checks `teTextureResource_Create` applies.
inline bool txtr_header_is_valid(const Header& h) {
if (h.format > TXTR_MAX_FORMAT_CODE)
return false;
if (h.mipCount == 0 || h.mipCount > TXTR_MAX_MIP_COUNT)
return false;
if (h.width == 0 || h.height == 0)
return false;
if (h.payloadCount > kMaxPayloadCount)
return false;
if (h.headerMipCount > h.mipCount)
return false;
return true;
}
/// Derive the surface topology from the flags word. Exactly one dimension bit
/// must be set; anything else is a header the client would refuse to draw.
inline std::optional<TxtrDimension> txtr_dimension(u16 flags) {
const bool isArray = hasFlag(flags, TxtrFlags::Array);
switch (flags & 0x000F) {
case 0x0001:
return isArray ? TxtrDimension::Texture1DArray : TxtrDimension::Texture1D;
case 0x0002:
return isArray ? TxtrDimension::Texture2DArray : TxtrDimension::Texture2D;
case 0x0004:
return TxtrDimension::Texture3D; // No array form exists for volumes.
case 0x0008:
return isArray ? TxtrDimension::TextureCubeArray : TxtrDimension::TextureCube;
default:
return std::nullopt;
}
}
/// TXTR carries no magic, so detection has to lean entirely on the header
/// being self-consistent and naming a format we can actually decode.
inline bool looks_like_txtr(std::span<const u8> buffer) {
const auto h = txtr_read_header(buffer);
if (!h || !txtr_header_is_valid(*h))
return false;
if (!txtr_dimension(h->flags))
return false;
if (!txtr_format_mapping(h->format))
return false;
if (h->surfaces == 0)
return false;
// The field at 0x0C sizes the whole texture rather than this file, so it
// cannot bound the buffer — a streamed texture's header is a fraction of it.
return buffer.size() >= static_cast<u64>(kInlineDataOffset);
}
/// Swap R and B for a tightly packed run of BGRA8 <-> RGBA8 pixels.
inline void txtr_swizzle_bgra(u8* pixels, u64 pixelCount, bool forceOpaque) {
for (u64 i = 0; i < pixelCount; ++i) {
u8* p = pixels + i * 4;
std::swap(p[0], p[2]);
if (forceOpaque)
p[3] = 0xFF;
}
}
} // namespace whiteout::textures::txtr