File size: 20,547 Bytes
7a0c684 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 |
"""
Advanced Texture and Buffer Management for Virtual GPU
Features:
- Mipmapped textures with automatic generation
- Multiple texture formats and compression
- Advanced filtering modes (bilinear, trilinear, anisotropic)
- Modern buffer types (uniform, storage, indirect)
- Full framebuffer support with MSAA
- Memory-efficient texture streaming
- Cache-aware memory layout
"""
import numpy as np
from enum import Enum, auto
import cv2 # for mipmap generation
from typing import List, Tuple, Optional, Union, Dict
import zlib # for texture compression
import time # for access tracking
from typing import List, Tuple, Optional, Union
import zlib # for texture compression
class TextureFormat(Enum):
R8 = auto() # 8-bit red channel
RG8 = auto() # 8-bit red-green
RGB8 = auto() # 8-bit RGB
RGBA8 = auto() # 8-bit RGBA
R16F = auto() # 16-bit float red
RGBA16F = auto() # 16-bit float RGBA
R32F = auto() # 32-bit float red
RGBA32F = auto() # 32-bit float RGBA
BC1 = auto() # Block compression (DXT1)
BC3 = auto() # Block compression (DXT5)
class FilterMode(Enum):
NEAREST = auto()
BILINEAR = auto()
TRILINEAR = auto()
ANISOTROPIC = auto()
class WrapMode(Enum):
REPEAT = auto()
CLAMP = auto()
MIRROR = auto()
class MipLevel:
def __init__(self, width: int, height: int, data: np.ndarray):
self.width = width
self.height = height
self.data = data
self.cache_hits = 0 # For cache analysis
class Texture:
def __init__(self, width: int, height: int, format: TextureFormat = TextureFormat.RGBA8,
filter_mode: FilterMode = FilterMode.BILINEAR,
wrap_mode: WrapMode = WrapMode.REPEAT,
generate_mipmaps: bool = True,
aniso_level: int = 1):
self.width = width
self.height = height
self.format = format
self.filter_mode = filter_mode
self.wrap_mode = wrap_mode
self.aniso_level = min(max(1, aniso_level), 16)
# Initialize main texture data
self.channels = self._get_channel_count()
self.dtype = self._get_data_type()
self.data = np.zeros((height, width, self.channels), dtype=self.dtype)
# Mipmap chain
self.mipmaps: List[MipLevel] = []
if generate_mipmaps:
self._generate_mipchain()
# Compression state
self.compressed = False
self.compressed_data = None
# Cache for sampling coordinates
self.sample_cache = {}
self.cache_size_limit = 1024 # Adjust based on memory constraints
def _get_channel_count(self) -> int:
return {
TextureFormat.R8: 1,
TextureFormat.RG8: 2,
TextureFormat.RGB8: 3,
TextureFormat.RGBA8: 4,
TextureFormat.R16F: 1,
TextureFormat.RGBA16F: 4,
TextureFormat.R32F: 1,
TextureFormat.RGBA32F: 4,
TextureFormat.BC1: 3,
TextureFormat.BC3: 4
}[self.format]
def _get_data_type(self) -> np.dtype:
if self.format in [TextureFormat.R8, TextureFormat.RG8, TextureFormat.RGB8, TextureFormat.RGBA8]:
return np.uint8
elif self.format in [TextureFormat.R16F, TextureFormat.RGBA16F]:
return np.float16
return np.float32
def _generate_mipchain(self):
"""Generate complete mipmap chain using box filtering"""
self.mipmaps.clear()
current = self.data
while current.shape[0] > 1 and current.shape[1] > 1:
# Box filter downsampling
next_mip = cv2.resize(current,
(current.shape[1]//2, current.shape[0]//2),
interpolation=cv2.INTER_LINEAR)
self.mipmaps.append(MipLevel(next_mip.shape[1], next_mip.shape[0], next_mip))
current = next_mip
def compress(self):
"""Compress texture data using format-specific compression"""
if self.format in [TextureFormat.BC1, TextureFormat.BC3]:
# Block compression would go here
# For now, use basic zlib compression
self.compressed_data = zlib.compress(self.data.tobytes())
self.compressed = True
def decompress(self):
"""Decompress texture data"""
if self.compressed and self.compressed_data:
raw_data = zlib.decompress(self.compressed_data)
self.data = np.frombuffer(raw_data, dtype=self.dtype).reshape(
self.height, self.width, self.channels)
self.compressed = False
def upload(self, img: np.ndarray, generate_mipmaps: bool = True):
"""Upload new texture data and optionally regenerate mipmaps"""
assert img.shape == self.data.shape, f"Shape mismatch: expected {self.data.shape}, got {img.shape}"
self.data[:] = img
if generate_mipmaps:
self._generate_mipchain()
# Clear sample cache
self.sample_cache.clear()
def _sample_nearest(self, u: float, v: float, mip_level: int = 0) -> np.ndarray:
"""Nearest neighbor sampling"""
if mip_level >= len(self.mipmaps):
data = self.data
else:
data = self.mipmaps[mip_level].data
# Apply wrap mode
if self.wrap_mode == WrapMode.REPEAT:
u = u % 1.0
v = v % 1.0
elif self.wrap_mode == WrapMode.CLAMP:
u = min(max(u, 0), 1)
v = min(max(v, 0), 1)
x = min(max(int(u * (data.shape[1]-1)), 0), data.shape[1]-1)
y = min(max(int(v * (data.shape[0]-1)), 0), data.shape[0]-1)
return data[y, x]
def _sample_bilinear(self, u: float, v: float, mip_level: int = 0) -> np.ndarray:
"""Bilinear texture sampling"""
# Check cache first
cache_key = (u, v, mip_level)
if cache_key in self.sample_cache:
return self.sample_cache[cache_key]
if mip_level >= len(self.mipmaps):
data = self.data
else:
data = self.mipmaps[mip_level].data
# Apply wrap mode
if self.wrap_mode == WrapMode.REPEAT:
u = u % 1.0
v = v % 1.0
elif self.wrap_mode == WrapMode.CLAMP:
u = min(max(u, 0), 1)
v = min(max(v, 0), 1)
# Calculate sample coordinates
x = u * (data.shape[1] - 1)
y = v * (data.shape[0] - 1)
x0, y0 = int(x), int(y)
x1, y1 = min(x0 + 1, data.shape[1] - 1), min(y0 + 1, data.shape[0] - 1)
# Calculate interpolation weights
wx = x - x0
wy = y - y0
# Sample four nearest texels
c00 = data[y0, x0]
c10 = data[y0, x1]
c01 = data[y1, x0]
c11 = data[y1, x1]
# Bilinear interpolation
result = (c00 * (1-wx) * (1-wy) +
c10 * wx * (1-wy) +
c01 * (1-wx) * wy +
c11 * wx * wy)
# Cache result
if len(self.sample_cache) < self.cache_size_limit:
self.sample_cache[cache_key] = result
return result
def sample(self, u: float, v: float) -> np.ndarray:
"""Sample texture with current filter mode"""
if self.compressed:
self.decompress()
# Calculate mip level for trilinear/anisotropic
mip_level = 0
if self.filter_mode in [FilterMode.TRILINEAR, FilterMode.ANISOTROPIC]:
# Basic mip level selection based on texture coordinate derivatives
# In a real GPU, this would use screen-space derivatives
mip_level = min(len(self.mipmaps)-1, int(max(0, -np.log2(max(u, v)))))
# Apply filtering mode
if self.filter_mode == FilterMode.NEAREST:
return self._sample_nearest(u, v, mip_level)
elif self.filter_mode == FilterMode.BILINEAR:
return self._sample_bilinear(u, v, mip_level)
elif self.filter_mode == FilterMode.TRILINEAR:
# Interpolate between two mip levels
if mip_level < len(self.mipmaps):
low_mip = self._sample_bilinear(u, v, mip_level)
high_mip = self._sample_bilinear(u, v, mip_level + 1)
factor = max(0, -np.log2(max(u, v))) - mip_level
return low_mip * (1-factor) + high_mip * factor
return self._sample_bilinear(u, v, mip_level)
else: # ANISOTROPIC
# Simple anisotropic approximation - average multiple samples
samples = []
for i in range(self.aniso_level):
offset = i / (self.aniso_level - 1) - 0.5
samples.append(self._sample_bilinear(u + offset*0.001, v, mip_level))
return np.mean(samples, axis=0)
class BufferType(Enum):
VERTEX = auto() # Vertex data
INDEX = auto() # Index data
UNIFORM = auto() # Uniform buffer
STORAGE = auto() # Storage buffer
INDIRECT = auto() # Indirect draw commands
class Buffer:
def __init__(self, data: np.ndarray, buffer_type: BufferType,
dynamic: bool = False, map_write: bool = False):
"""
Initialize buffer with specific type and usage flags
Args:
data: Initial buffer data
buffer_type: Type of buffer (vertex, index, uniform, etc)
dynamic: Whether buffer will be frequently updated
map_write: Whether buffer should be mappable for CPU writes
"""
self.buffer_type = buffer_type
self.dynamic = dynamic
self.map_write = map_write
self.mapped = False
# Main storage
self.data = np.array(data)
# Shadow buffer for mapped writes
self.shadow_buffer = None if not map_write else np.array(data)
# Cache alignment and striding
self.stride = self._calculate_stride()
self.aligned_size = self._align_size(self.data.nbytes)
# Usage tracking
self.access_count = 0
self.last_access = 0
def _calculate_stride(self) -> int:
"""Calculate optimal stride for the buffer type"""
base_stride = self.data.itemsize * self.data.shape[-1]
# Align to 16 bytes for modern GPUs
return ((base_stride + 15) // 16) * 16
def _align_size(self, size: int) -> int:
"""Align buffer size to GPU-friendly boundaries"""
return ((size + 255) // 256) * 256
def upload(self, data: np.ndarray):
"""Upload new data to buffer"""
assert data.shape == self.data.shape, f"Shape mismatch: expected {self.data.shape}, got {data.shape}"
if self.mapped:
raise RuntimeError("Cannot upload to mapped buffer")
self.data[:] = data
self.access_count += 1
self.last_access = time.time()
def map(self) -> Optional[np.ndarray]:
"""Map buffer for CPU access"""
if not self.map_write:
raise RuntimeError("Buffer not created with map_write flag")
if self.mapped:
raise RuntimeError("Buffer already mapped")
self.mapped = True
self.shadow_buffer[:] = self.data
return self.shadow_buffer
def unmap(self):
"""Unmap buffer and apply changes"""
if not self.mapped:
raise RuntimeError("Buffer not mapped")
self.data[:] = self.shadow_buffer
self.mapped = False
self.access_count += 1
self.last_access = time.time()
def get(self) -> np.ndarray:
"""Get buffer data"""
if self.mapped:
raise RuntimeError("Cannot read from mapped buffer")
self.access_count += 1
self.last_access = time.time()
return self.data
class AttachmentType(Enum):
COLOR = auto() # Color attachment
DEPTH = auto() # Depth attachment
STENCIL = auto() # Stencil attachment
DEPTH_STENCIL = auto() # Combined depth-stencil
class MSAASamples(Enum):
MSAA_1X = 1 # No multisampling
MSAA_2X = 2 # 2x multisampling
MSAA_4X = 4 # 4x multisampling
MSAA_8X = 8 # 8x multisampling
class Attachment:
def __init__(self, width: int, height: int,
attachment_type: AttachmentType,
format: TextureFormat = TextureFormat.RGBA8,
samples: MSAASamples = MSAASamples.MSAA_1X):
self.width = width
self.height = height
self.attachment_type = attachment_type
self.format = format
self.samples = samples
# Initialize storage based on type and format
channels = 4 if format in [TextureFormat.RGBA8, TextureFormat.RGBA16F, TextureFormat.RGBA32F] else 1
if samples == MSAASamples.MSAA_1X:
self.data = np.zeros((height, width, channels), dtype=self._get_dtype())
else:
self.data = np.zeros((height, width, samples.value, channels), dtype=self._get_dtype())
def _get_dtype(self) -> np.dtype:
if self.format in [TextureFormat.R8, TextureFormat.RGBA8]:
return np.uint8
elif self.format in [TextureFormat.R16F, TextureFormat.RGBA16F]:
return np.float16
return np.float32
class Framebuffer:
def __init__(self, width: int, height: int,
color_formats: List[TextureFormat] = [TextureFormat.RGBA8],
depth_format: Optional[TextureFormat] = TextureFormat.R32F,
stencil_format: Optional[TextureFormat] = TextureFormat.R8,
samples: MSAASamples = MSAASamples.MSAA_1X):
"""
Initialize framebuffer with multiple render targets, depth, and stencil
Args:
width: Framebuffer width
height: Framebuffer height
color_formats: List of formats for color attachments
depth_format: Format for depth attachment (None to disable)
stencil_format: Format for stencil attachment (None to disable)
samples: MSAA sample count
"""
self.width = width
self.height = height
self.samples = samples
# Create color attachments
self.color_attachments = [
Attachment(width, height, AttachmentType.COLOR, format, samples)
for format in color_formats
]
# Create depth attachment
self.depth_attachment = (
Attachment(width, height, AttachmentType.DEPTH, depth_format, samples)
if depth_format else None
)
# Create stencil attachment
self.stencil_attachment = (
Attachment(width, height, AttachmentType.STENCIL, stencil_format, samples)
if stencil_format else None
)
# Active color attachment
self.active_color = 0
# State tracking
self.clear_values = {
AttachmentType.COLOR: np.zeros(4),
AttachmentType.DEPTH: 1.0,
AttachmentType.STENCIL: 0
}
def bind(self, color_attachment: int = 0):
"""Bind specific color attachment for writing"""
assert 0 <= color_attachment < len(self.color_attachments)
self.active_color = color_attachment
def set_clear_values(self, color=None, depth=None, stencil=None):
"""Set clear values for attachments"""
if color is not None:
self.clear_values[AttachmentType.COLOR] = np.array(color)
if depth is not None:
self.clear_values[AttachmentType.DEPTH] = depth
if stencil is not None:
self.clear_values[AttachmentType.STENCIL] = stencil
def clear(self, color=True, depth=True, stencil=True):
"""Clear specified attachments"""
if color:
for attachment in self.color_attachments:
if self.samples == MSAASamples.MSAA_1X:
attachment.data.fill(self.clear_values[AttachmentType.COLOR])
else:
for s in range(self.samples.value):
attachment.data[..., s, :] = self.clear_values[AttachmentType.COLOR]
if depth and self.depth_attachment:
if self.samples == MSAASamples.MSAA_1X:
self.depth_attachment.data.fill(self.clear_values[AttachmentType.DEPTH])
else:
for s in range(self.samples.value):
self.depth_attachment.data[..., s, 0] = self.clear_values[AttachmentType.DEPTH]
if stencil and self.stencil_attachment:
if self.samples == MSAASamples.MSAA_1X:
self.stencil_attachment.data.fill(self.clear_values[AttachmentType.STENCIL])
else:
for s in range(self.samples.value):
self.stencil_attachment.data[..., s, 0] = self.clear_values[AttachmentType.STENCIL]
def write_color(self, x: int, y: int, color: np.ndarray, sample: int = 0):
"""Write color value to current attachment"""
attachment = self.color_attachments[self.active_color]
if self.samples == MSAASamples.MSAA_1X:
attachment.data[y, x] = color
else:
attachment.data[y, x, sample] = color
def write_depth(self, x: int, y: int, depth: float, sample: int = 0):
"""Write depth value"""
if self.depth_attachment:
if self.samples == MSAASamples.MSAA_1X:
self.depth_attachment.data[y, x] = depth
else:
self.depth_attachment.data[y, x, sample] = depth
def write_stencil(self, x: int, y: int, stencil: int, sample: int = 0):
"""Write stencil value"""
if self.stencil_attachment:
if self.samples == MSAASamples.MSAA_1X:
self.stencil_attachment.data[y, x] = stencil
else:
self.stencil_attachment.data[y, x, sample] = stencil
def read_color(self, x: int, y: int, attachment: int = 0) -> np.ndarray:
"""Read color value (resolves MSAA if needed)"""
attachment = self.color_attachments[attachment]
if self.samples == MSAASamples.MSAA_1X:
return attachment.data[y, x]
# Resolve MSAA by averaging samples
return np.mean(attachment.data[y, x], axis=0)
def read_depth(self, x: int, y: int) -> float:
"""Read depth value (resolves MSAA if needed)"""
if not self.depth_attachment:
return 1.0
if self.samples == MSAASamples.MSAA_1X:
return self.depth_attachment.data[y, x]
return np.mean(self.depth_attachment.data[y, x])
def read_stencil(self, x: int, y: int) -> int:
"""Read stencil value (uses first sample in MSAA)"""
if not self.stencil_attachment:
return 0
if self.samples == MSAASamples.MSAA_1X:
return self.stencil_attachment.data[y, x]
return self.stencil_attachment.data[y, x, 0]
def resolve_msaa(self) -> List[np.ndarray]:
"""Resolve MSAA framebuffer to non-MSAA textures"""
if self.samples == MSAASamples.MSAA_1X:
return [attachment.data for attachment in self.color_attachments]
resolved = []
for attachment in self.color_attachments:
# Average across samples
resolved.append(np.mean(attachment.data, axis=2))
return resolved
|