Update requirements.txt

.gitignore

@@ -0,0 +1,2 @@
+__pycache__/
+venv/

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "diff-gaussian-rasterization"]
+	path = diff-gaussian-rasterization
+	url = https://github.com/ashawkey/diff-gaussian-rasterization.git

app.py

@@ -1,172 +1,86 @@
 import os
-import numpy as np
-import torch
-import torch.nn.functional as F
-import torchvision.transforms.functional as TF
-from safetensors.torch import load_file
-import rembg
-import gradio as gr
 
-# download checkpoints
-from huggingface_hub import hf_hub_download
-ckpt_path = hf_hub_download(repo_id="ashawkey/LGM", filename="model_fp16.safetensors")
+import gradio as gr
+import torch
+from diffusers import DiffusionPipeline
 
 try:
     import diff_gaussian_rasterization
 except ImportError:
     os.system("pip install ./diff-gaussian-rasterization")
 
-import kiui
-from kiui.op import recenter
-
-from core.options import Options
-from core.models import LGM
-from mvdream.pipeline_mvdream import MVDreamPipeline
-
-IMAGENET_DEFAULT_MEAN = (0.485, 0.456, 0.406)
-IMAGENET_DEFAULT_STD = (0.229, 0.224, 0.225)
-
-TMP_DIR = '/tmp'
+TMP_DIR = "/tmp"
 os.makedirs(TMP_DIR, exist_ok=True)
 
-# opt = tyro.cli(AllConfigs)
-opt = Options(
-    input_size=256,
-    up_channels=(1024, 1024, 512, 256, 128), # one more decoder
-    up_attention=(True, True, True, False, False),
-    splat_size=128,
-    output_size=512, # render & supervise Gaussians at a higher resolution.
-    batch_size=8,
-    num_views=8,
-    gradient_accumulation_steps=1,
-    mixed_precision='bf16',
-    resume=ckpt_path,
-)
-
-# model
-model = LGM(opt)
-
-# resume pretrained checkpoint
-if opt.resume is not None:
-    if opt.resume.endswith('safetensors'):
-        ckpt = load_file(opt.resume, device='cpu')
-    else:
-        ckpt = torch.load(opt.resume, map_location='cpu')
-    model.load_state_dict(ckpt, strict=False)
-    print(f'[INFO] Loaded checkpoint from {opt.resume}')
-else:
-    print(f'[WARN] model randomly initialized, are you sure?')
-
-device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
-model = model.half().to(device)
-model.eval()
-
-tan_half_fov = np.tan(0.5 * np.deg2rad(opt.fovy))
-proj_matrix = torch.zeros(4, 4, dtype=torch.float32, device=device)
-proj_matrix[0, 0] = -1 / tan_half_fov
-proj_matrix[1, 1] = -1 / tan_half_fov
-proj_matrix[2, 2] = - (opt.zfar + opt.znear) / (opt.zfar - opt.znear)
-proj_matrix[3, 2] = - (opt.zfar * opt.znear) / (opt.zfar - opt.znear)
-proj_matrix[2, 3] = 1
-
-# load dreams
-pipe_text = MVDreamPipeline.from_pretrained(
-    'ashawkey/mvdream-sd2.1-diffusers', # remote weights
+
+image_pipeline = DiffusionPipeline.from_pretrained(
+    "ashawkey/imagedream-ipmv-diffusers",
+    custom_pipeline="dylanebert/multi_view_diffusion",
     torch_dtype=torch.float16,
     trust_remote_code=True,
-    # local_files_only=True,
-)
-pipe_text = pipe_text.to(device)
+).to("cuda")
+
 
-pipe_image = MVDreamPipeline.from_pretrained(
-    "ashawkey/imagedream-ipmv-diffusers", # remote weights
+splat_pipeline = DiffusionPipeline.from_pretrained(
+    "dylanebert/LGM",
+    custom_pipeline="dylanebert/LGM",
     torch_dtype=torch.float16,
     trust_remote_code=True,
-    # local_files_only=True,
-)
-pipe_image = pipe_image.to(device)
+).to("cuda")
 
-# load rembg
-bg_remover = rembg.new_session()
 
-# process function
 def run(input_image):
-    prompt_neg = "ugly, blurry, pixelated obscure, unnatural colors, poor lighting, dull, unclear, cropped, lowres, low quality, artifacts, duplicate"
-
-    # seed
-    kiui.seed_everything(42)
-
-    output_ply_path = os.path.join(TMP_DIR, 'output.ply')
-
-    input_image = np.array(input_image) # uint8
-    # bg removal
-    carved_image = rembg.remove(input_image, session=bg_remover) # [H, W, 4]
-    mask = carved_image[..., -1] > 0
-    image = recenter(carved_image, mask, border_ratio=0.2)
-    image = image.astype(np.float32) / 255.0
-    image = image[..., :3] * image[..., 3:4] + (1 - image[..., 3:4])
-    mv_image = pipe_image("", image, negative_prompt=prompt_neg, num_inference_steps=30, guidance_scale=5.0,  elevation=0)
-
-    # generate gaussians
-    input_image = np.stack([mv_image[1], mv_image[2], mv_image[3], mv_image[0]], axis=0) # [4, 256, 256, 3], float32
-    input_image = torch.from_numpy(input_image).permute(0, 3, 1, 2).float().to(device) # [4, 3, 256, 256]
-    input_image = F.interpolate(input_image, size=(opt.input_size, opt.input_size), mode='bilinear', align_corners=False)
-    input_image = TF.normalize(input_image, IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD)
-
-    rays_embeddings = model.prepare_default_rays(device, elevation=0)
-    input_image = torch.cat([input_image, rays_embeddings], dim=1).unsqueeze(0) # [1, 4, 9, H, W]
-
-    with torch.no_grad():
-        with torch.autocast(device_type='cuda', dtype=torch.float16):
-            # generate gaussians
-            gaussians = model.forward_gaussians(input_image)
-        
-        # save gaussians
-        model.gs.save_ply(gaussians, output_ply_path)
-    
+    input_image = input_image.astype("float32") / 255.0
+    images = image_pipeline(
+        "", input_image, guidance_scale=5, num_inference_steps=30, elevation=0
+    )
+    images = (images * 255).astype("uint8")
+    gaussians = splat_pipeline(images)
+    output_ply_path = os.path.join(TMP_DIR, "output.ply")
+    splat_pipeline.save_ply(gaussians, output_ply_path)
     return output_ply_path
 
-# gradio UI
 
-_TITLE = '''LGM Mini'''
+_TITLE = """LGM Mini"""
 
-_DESCRIPTION = '''
+_DESCRIPTION = """
 <div>
 A lightweight version of <a href="https://huggingface.co/spaces/ashawkey/LGM">LGM: Large Multi-View Gaussian Model for High-Resolution 3D Content Creation</a>.
 </div>
-'''
+"""
 
-css = '''
+css = """
 #duplicate-button {
     margin: auto;
     color: white;
     background: #1565c0;
     border-radius: 100vh;
 }
-'''
+"""
 
 block = gr.Blocks(title=_TITLE, css=css)
 with block:
-    gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
+    gr.DuplicateButton(
+        value="Duplicate Space for private use", elem_id="duplicate-button"
+    )
 
     with gr.Row():
         with gr.Column(scale=1):
-            gr.Markdown('# ' + _TITLE)
+            gr.Markdown("# " + _TITLE)
     gr.Markdown(_DESCRIPTION)
-    
-    with gr.Row(variant='panel'):
+
+    with gr.Row(variant="panel"):
         with gr.Column(scale=1):
             # input image
-            input_image = gr.Image(label="image", type='pil', height=320)
+            input_image = gr.Image(label="image", type="numpy", height=320)
             # gen button
             button_gen = gr.Button("Generate")
 
-        
         with gr.Column(scale=1):
             output_splat = gr.Model3D(label="3D Gaussians")
 
         button_gen.click(fn=run, inputs=[input_image], outputs=[output_splat])
-    
+
     gr.Examples(
         examples=[
             "data_test/frog_sweater.jpg",
@@ -180,7 +94,7 @@ with block:
         outputs=[output_splat],
         fn=lambda x: run(input_image=x),
         cache_examples=True,
-        label='Image-to-3D Examples'
+        label="Image-to-3D Examples",
     )
-    
+
 block.queue().launch(debug=True, share=True)

core/attention.py

@@ -1,156 +0,0 @@
-# Copyright (c) Meta Platforms, Inc. and affiliates.
-#
-# This source code is licensed under the Apache License, Version 2.0
-# found in the LICENSE file in the root directory of this source tree.
-
-# References:
-#   https://github.com/facebookresearch/dino/blob/master/vision_transformer.py
-#   https://github.com/rwightman/pytorch-image-models/tree/master/timm/models/vision_transformer.py
-
-import os
-import warnings
-
-from torch import Tensor
-from torch import nn
-
-XFORMERS_ENABLED = os.environ.get("XFORMERS_DISABLED") is None
-try:
-    if XFORMERS_ENABLED:
-        from xformers.ops import memory_efficient_attention, unbind
-
-        XFORMERS_AVAILABLE = True
-        warnings.warn("xFormers is available (Attention)")
-    else:
-        warnings.warn("xFormers is disabled (Attention)")
-        raise ImportError
-except ImportError:
-    XFORMERS_AVAILABLE = False
-    warnings.warn("xFormers is not available (Attention)")
-
-
-class Attention(nn.Module):
-    def __init__(
-        self,
-        dim: int,
-        num_heads: int = 8,
-        qkv_bias: bool = False,
-        proj_bias: bool = True,
-        attn_drop: float = 0.0,
-        proj_drop: float = 0.0,
-    ) -> None:
-        super().__init__()
-        self.num_heads = num_heads
-        head_dim = dim // num_heads
-        self.scale = head_dim**-0.5
-
-        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
-        self.attn_drop = nn.Dropout(attn_drop)
-        self.proj = nn.Linear(dim, dim, bias=proj_bias)
-        self.proj_drop = nn.Dropout(proj_drop)
-
-    def forward(self, x: Tensor) -> Tensor:
-        B, N, C = x.shape
-        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
-
-        q, k, v = qkv[0] * self.scale, qkv[1], qkv[2]
-        attn = q @ k.transpose(-2, -1)
-
-        attn = attn.softmax(dim=-1)
-        attn = self.attn_drop(attn)
-
-        x = (attn @ v).transpose(1, 2).reshape(B, N, C)
-        x = self.proj(x)
-        x = self.proj_drop(x)
-        return x
-
-
-class MemEffAttention(Attention):
-    def forward(self, x: Tensor, attn_bias=None) -> Tensor:
-        if not XFORMERS_AVAILABLE:
-            if attn_bias is not None:
-                raise AssertionError("xFormers is required for using nested tensors")
-            return super().forward(x)
-
-        B, N, C = x.shape
-        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads)
-
-        q, k, v = unbind(qkv, 2)
-
-        x = memory_efficient_attention(q, k, v, attn_bias=attn_bias)
-        x = x.reshape([B, N, C])
-
-        x = self.proj(x)
-        x = self.proj_drop(x)
-        return x
-
-
-class CrossAttention(nn.Module):
-    def __init__(
-        self,
-        dim: int,
-        dim_q: int,
-        dim_k: int,
-        dim_v: int,
-        num_heads: int = 8,
-        qkv_bias: bool = False,
-        proj_bias: bool = True,
-        attn_drop: float = 0.0,
-        proj_drop: float = 0.0,
-    ) -> None:
-        super().__init__()
-        self.dim = dim
-        self.num_heads = num_heads
-        head_dim = dim // num_heads
-        self.scale = head_dim**-0.5
-
-        self.to_q = nn.Linear(dim_q, dim, bias=qkv_bias)
-        self.to_k = nn.Linear(dim_k, dim, bias=qkv_bias)
-        self.to_v = nn.Linear(dim_v, dim, bias=qkv_bias)
-        self.attn_drop = nn.Dropout(attn_drop)
-        self.proj = nn.Linear(dim, dim, bias=proj_bias)
-        self.proj_drop = nn.Dropout(proj_drop)
-
-    def forward(self, q: Tensor, k: Tensor, v: Tensor) -> Tensor:
-        # q: [B, N, Cq]
-        # k: [B, M, Ck]
-        # v: [B, M, Cv]
-        # return: [B, N, C]
-
-        B, N, _ = q.shape
-        M = k.shape[1]
-        
-        q = self.scale * self.to_q(q).reshape(B, N, self.num_heads, self.dim // self.num_heads).permute(0, 2, 1, 3) # [B, nh, N, C/nh]
-        k = self.to_k(k).reshape(B, M, self.num_heads, self.dim // self.num_heads).permute(0, 2, 1, 3) # [B, nh, M, C/nh]
-        v = self.to_v(v).reshape(B, M, self.num_heads, self.dim // self.num_heads).permute(0, 2, 1, 3) # [B, nh, M, C/nh]
-
-        attn = q @ k.transpose(-2, -1) # [B, nh, N, M]
-
-        attn = attn.softmax(dim=-1) # [B, nh, N, M]
-        attn = self.attn_drop(attn)
-
-        x = (attn @ v).transpose(1, 2).reshape(B, N, -1) # [B, nh, N, M] @ [B, nh, M, C/nh] --> [B, nh, N, C/nh] --> [B, N, nh, C/nh] --> [B, N, C]
-        x = self.proj(x)
-        x = self.proj_drop(x)
-        return x
-
-
-class MemEffCrossAttention(CrossAttention):
-    def forward(self, q: Tensor, k: Tensor, v: Tensor, attn_bias=None) -> Tensor:
-        if not XFORMERS_AVAILABLE:
-            if attn_bias is not None:
-                raise AssertionError("xFormers is required for using nested tensors")
-            return super().forward(x)
-
-        B, N, _ = q.shape
-        M = k.shape[1]
-
-        q = self.scale * self.to_q(q).reshape(B, N, self.num_heads, self.dim // self.num_heads) # [B, N, nh, C/nh]
-        k = self.to_k(k).reshape(B, M, self.num_heads, self.dim // self.num_heads) # [B, M, nh, C/nh]
-        v = self.to_v(v).reshape(B, M, self.num_heads, self.dim // self.num_heads) # [B, M, nh, C/nh]
-
-        x = memory_efficient_attention(q, k, v, attn_bias=attn_bias)
-        x = x.reshape(B, N, -1)
-
-        x = self.proj(x)
-        x = self.proj_drop(x)
-        return x

core/gs.py

@@ -1,193 +0,0 @@
-import os
-import numpy as np
-
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-
-from diff_gaussian_rasterization import (
-    GaussianRasterizationSettings,
-    GaussianRasterizer,
-)
-
-from core.options import Options
-
-import kiui
-
-class GaussianRenderer:
-    def __init__(self, opt: Options):
-        
-        self.opt = opt
-        self.bg_color = torch.tensor([1, 1, 1], dtype=torch.float32, device="cuda")
-        
-        # intrinsics
-        self.tan_half_fov = np.tan(0.5 * np.deg2rad(self.opt.fovy))
-        self.proj_matrix = torch.zeros(4, 4, dtype=torch.float32)
-        self.proj_matrix[0, 0] = 1 / self.tan_half_fov
-        self.proj_matrix[1, 1] = 1 / self.tan_half_fov
-        self.proj_matrix[2, 2] = (opt.zfar + opt.znear) / (opt.zfar - opt.znear)
-        self.proj_matrix[3, 2] = - (opt.zfar * opt.znear) / (opt.zfar - opt.znear)
-        self.proj_matrix[2, 3] = 1
-        
-    def render(self, gaussians, cam_view, cam_view_proj, cam_pos, bg_color=None, scale_modifier=1):
-        # gaussians: [B, N, 14]
-        # cam_view, cam_view_proj: [B, V, 4, 4]
-        # cam_pos: [B, V, 3]
-
-        device = gaussians.device
-        B, V = cam_view.shape[:2]
-
-        # loop of loop...
-        images = []
-        alphas = []
-        for b in range(B):
-
-            # pos, opacity, scale, rotation, shs
-            means3D = gaussians[b, :, 0:3].contiguous().float()
-            opacity = gaussians[b, :, 3:4].contiguous().float()
-            scales = gaussians[b, :, 4:7].contiguous().float()
-            rotations = gaussians[b, :, 7:11].contiguous().float()
-            rgbs = gaussians[b, :, 11:].contiguous().float() # [N, 3]
-
-            for v in range(V):
-                
-                # render novel views
-                view_matrix = cam_view[b, v].float()
-                view_proj_matrix = cam_view_proj[b, v].float()
-                campos = cam_pos[b, v].float()
-
-                raster_settings = GaussianRasterizationSettings(
-                    image_height=self.opt.output_size,
-                    image_width=self.opt.output_size,
-                    tanfovx=self.tan_half_fov,
-                    tanfovy=self.tan_half_fov,
-                    bg=self.bg_color if bg_color is None else bg_color,
-                    scale_modifier=scale_modifier,
-                    viewmatrix=view_matrix,
-                    projmatrix=view_proj_matrix,
-                    sh_degree=0,
-                    campos=campos,
-                    prefiltered=False,
-                    debug=False,
-                )
-
-                rasterizer = GaussianRasterizer(raster_settings=raster_settings)
-
-                # Rasterize visible Gaussians to image, obtain their radii (on screen).
-                rendered_image, radii, rendered_depth, rendered_alpha = rasterizer(
-                    means3D=means3D,
-                    means2D=torch.zeros_like(means3D, dtype=torch.float32, device=device),
-                    shs=None,
-                    colors_precomp=rgbs,
-                    opacities=opacity,
-                    scales=scales,
-                    rotations=rotations,
-                    cov3D_precomp=None,
-                )
-
-                rendered_image = rendered_image.clamp(0, 1)
-
-                images.append(rendered_image)
-                alphas.append(rendered_alpha)
-
-        images = torch.stack(images, dim=0).view(B, V, 3, self.opt.output_size, self.opt.output_size)
-        alphas = torch.stack(alphas, dim=0).view(B, V, 1, self.opt.output_size, self.opt.output_size)
-
-        return {
-            "image": images, # [B, V, 3, H, W]
-            "alpha": alphas, # [B, V, 1, H, W]
-        }
-
-
-    def save_ply(self, gaussians, path, compatible=True):
-        # gaussians: [B, N, 14]
-        # compatible: save pre-activated gaussians as in the original paper
-
-        assert gaussians.shape[0] == 1, 'only support batch size 1'
-
-        from plyfile import PlyData, PlyElement
-
-        os.makedirs(os.path.dirname(path), exist_ok=True)
-     
-        means3D = gaussians[0, :, 0:3].contiguous().float()
-        opacity = gaussians[0, :, 3:4].contiguous().float()
-        scales = gaussians[0, :, 4:7].contiguous().float()
-        rotations = gaussians[0, :, 7:11].contiguous().float()
-        shs = gaussians[0, :, 11:].unsqueeze(1).contiguous().float() # [N, 1, 3]
-
-        # prune by opacity
-        mask = opacity.squeeze(-1) >= 0.005
-        means3D = means3D[mask]
-        opacity = opacity[mask]
-        scales = scales[mask]
-        rotations = rotations[mask]
-        shs = shs[mask]
-
-        # invert activation to make it compatible with the original ply format
-        if compatible:
-            opacity = kiui.op.inverse_sigmoid(opacity)
-            scales = torch.log(scales + 1e-8)
-            shs = (shs - 0.5) / 0.28209479177387814
-
-        xyzs = means3D.detach().cpu().numpy()
-        f_dc = shs.detach().transpose(1, 2).flatten(start_dim=1).contiguous().cpu().numpy()
-        opacities = opacity.detach().cpu().numpy()
-        scales = scales.detach().cpu().numpy()
-        rotations = rotations.detach().cpu().numpy()
-
-        l = ['x', 'y', 'z']
-        # All channels except the 3 DC
-        for i in range(f_dc.shape[1]):
-            l.append('f_dc_{}'.format(i))
-        l.append('opacity')
-        for i in range(scales.shape[1]):
-            l.append('scale_{}'.format(i))
-        for i in range(rotations.shape[1]):
-            l.append('rot_{}'.format(i))
-
-        dtype_full = [(attribute, 'f4') for attribute in l]
-
-        elements = np.empty(xyzs.shape[0], dtype=dtype_full)
-        attributes = np.concatenate((xyzs, f_dc, opacities, scales, rotations), axis=1)
-        elements[:] = list(map(tuple, attributes))
-        el = PlyElement.describe(elements, 'vertex')
-
-        PlyData([el]).write(path)
-    
-    def load_ply(self, path, compatible=True):
-
-        from plyfile import PlyData, PlyElement
-
-        plydata = PlyData.read(path)
-
-        xyz = np.stack((np.asarray(plydata.elements[0]["x"]),
-                        np.asarray(plydata.elements[0]["y"]),
-                        np.asarray(plydata.elements[0]["z"])),  axis=1)
-        print("Number of points at loading : ", xyz.shape[0])
-
-        opacities = np.asarray(plydata.elements[0]["opacity"])[..., np.newaxis]
-
-        shs = np.zeros((xyz.shape[0], 3))
-        shs[:, 0] = np.asarray(plydata.elements[0]["f_dc_0"])
-        shs[:, 1] = np.asarray(plydata.elements[0]["f_dc_1"])
-        shs[:, 2] = np.asarray(plydata.elements[0]["f_dc_2"])
-
-        scale_names = [p.name for p in plydata.elements[0].properties if p.name.startswith("scale_")]
-        scales = np.zeros((xyz.shape[0], len(scale_names)))
-        for idx, attr_name in enumerate(scale_names):
-            scales[:, idx] = np.asarray(plydata.elements[0][attr_name])
-
-        rot_names = [p.name for p in plydata.elements[0].properties if p.name.startswith("rot_")]
-        rots = np.zeros((xyz.shape[0], len(rot_names)))
-        for idx, attr_name in enumerate(rot_names):
-            rots[:, idx] = np.asarray(plydata.elements[0][attr_name])
-          
-        gaussians = np.concatenate([xyz, opacities, scales, rots, shs], axis=1)
-        gaussians = torch.from_numpy(gaussians).float() # cpu
-
-        if compatible:
-            gaussians[..., 3:4] = torch.sigmoid(gaussians[..., 3:4])
-            gaussians[..., 4:7] = torch.exp(gaussians[..., 4:7])
-            gaussians[..., 11:] = 0.28209479177387814 * gaussians[..., 11:] + 0.5
-
-        return gaussians

core/models.py

@@ -1,185 +0,0 @@
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-import numpy as np
-
-import kiui
-from kiui.lpips import LPIPS
-
-from core.unet import UNet
-from core.options import Options
-from core.gs import GaussianRenderer
-
-
-class LGM(nn.Module):
-    def __init__(
-        self,
-        opt: Options,
-    ):
-        super().__init__()
-
-        self.opt = opt
-
-        # unet
-        self.unet = UNet(
-            9, 14, 
-            down_channels=self.opt.down_channels,
-            down_attention=self.opt.down_attention,
-            mid_attention=self.opt.mid_attention,
-            up_channels=self.opt.up_channels,
-            up_attention=self.opt.up_attention,
-        )
-
-        # last conv
-        self.conv = nn.Conv2d(14, 14, kernel_size=1) # NOTE: maybe remove it if train again
-
-        # Gaussian Renderer
-        self.gs = GaussianRenderer(opt)
-
-        # activations...
-        self.pos_act = lambda x: x.clamp(-1, 1)
-        self.scale_act = lambda x: 0.1 * F.softplus(x)
-        self.opacity_act = lambda x: torch.sigmoid(x)
-        self.rot_act = F.normalize
-        self.rgb_act = lambda x: 0.5 * torch.tanh(x) + 0.5 # NOTE: may use sigmoid if train again
-
-        # LPIPS loss
-        if self.opt.lambda_lpips > 0:
-            self.lpips_loss = LPIPS(net='vgg')
-            self.lpips_loss.requires_grad_(False)
-
-
-    def state_dict(self, **kwargs):
-        # remove lpips_loss
-        state_dict = super().state_dict(**kwargs)
-        for k in list(state_dict.keys()):
-            if 'lpips_loss' in k:
-                del state_dict[k]
-        return state_dict
-
-
-    def prepare_default_rays(self, device, elevation=0):
-        
-        from kiui.cam import orbit_camera
-        from core.utils import get_rays
-
-        cam_poses = np.stack([
-            orbit_camera(elevation, 0, radius=self.opt.cam_radius),
-            orbit_camera(elevation, 90, radius=self.opt.cam_radius),
-            orbit_camera(elevation, 180, radius=self.opt.cam_radius),
-            orbit_camera(elevation, 270, radius=self.opt.cam_radius),
-        ], axis=0) # [4, 4, 4]
-        cam_poses = torch.from_numpy(cam_poses)
-
-        rays_embeddings = []
-        for i in range(cam_poses.shape[0]):
-            rays_o, rays_d = get_rays(cam_poses[i], self.opt.input_size, self.opt.input_size, self.opt.fovy) # [h, w, 3]
-            rays_plucker = torch.cat([torch.cross(rays_o, rays_d, dim=-1), rays_d], dim=-1) # [h, w, 6]
-            rays_embeddings.append(rays_plucker)
-
-            ## visualize rays for plotting figure
-            # kiui.vis.plot_image(rays_d * 0.5 + 0.5, save=True)
-
-        rays_embeddings = torch.stack(rays_embeddings, dim=0).permute(0, 3, 1, 2).contiguous().to(device) # [V, 6, h, w]
-        
-        return rays_embeddings
-        
-
-    def forward_gaussians(self, images):
-        # images: [B, 4, 9, H, W]
-        # return: Gaussians: [B, dim_t]
-
-        B, V, C, H, W = images.shape
-        images = images.view(B*V, C, H, W)
-
-        x = self.unet(images) # [B*4, 14, h, w]
-        x = self.conv(x) # [B*4, 14, h, w]
-
-        x = x.reshape(B, 4, 14, self.opt.splat_size, self.opt.splat_size)
-        
-        ## visualize multi-view gaussian features for plotting figure
-        # tmp_alpha = self.opacity_act(x[0, :, 3:4])
-        # tmp_img_rgb = self.rgb_act(x[0, :, 11:]) * tmp_alpha + (1 - tmp_alpha)
-        # tmp_img_pos = self.pos_act(x[0, :, 0:3]) * 0.5 + 0.5
-        # kiui.vis.plot_image(tmp_img_rgb, save=True)
-        # kiui.vis.plot_image(tmp_img_pos, save=True)
-
-        x = x.permute(0, 1, 3, 4, 2).reshape(B, -1, 14)
-        
-        pos = self.pos_act(x[..., 0:3]) # [B, N, 3]
-        opacity = self.opacity_act(x[..., 3:4])
-        scale = self.scale_act(x[..., 4:7])
-        rotation = self.rot_act(x[..., 7:11])
-        rgbs = self.rgb_act(x[..., 11:])
-
-        rot_matrix = torch.tensor([[1.0, 0.0, 0.0, 0.0],
-                                   [0.0, -1.0, 0.0, 0.0],
-                                   [0.0, 0.0, -1.0, 0.0],
-                                   [0.0, 0.0, 0.0, 1.0]], dtype=torch.float32, device=images.device)
-
-        pos_4d = torch.cat([pos, torch.ones_like(pos[..., :1])], dim=-1)
-        pos = torch.matmul(pos_4d, rot_matrix) # [B, N, 4]
-        pos = pos[..., :3]
-
-        rotation = torch.matmul(rotation, rot_matrix)
-
-        gaussians = torch.cat([pos, opacity, scale, rotation, rgbs], dim=-1) # [B, N, 14]
-        
-        return gaussians
-
-    
-    def forward(self, data, step_ratio=1):
-        # data: output of the dataloader
-        # return: loss
-
-        results = {}
-        loss = 0
-
-        images = data['input'] # [B, 4, 9, h, W], input features
-        
-        # use the first view to predict gaussians
-        gaussians = self.forward_gaussians(images) # [B, N, 14]
-
-        results['gaussians'] = gaussians
-
-        # random bg for training
-        if self.training:
-            bg_color = torch.rand(3, dtype=torch.float32, device=gaussians.device)
-        else:
-            bg_color = torch.ones(3, dtype=torch.float32, device=gaussians.device)
-
-        # use the other views for rendering and supervision
-        results = self.gs.render(gaussians, data['cam_view'], data['cam_view_proj'], data['cam_pos'], bg_color=bg_color)
-        pred_images = results['image'] # [B, V, C, output_size, output_size]
-        pred_alphas = results['alpha'] # [B, V, 1, output_size, output_size]
-
-        results['images_pred'] = pred_images
-        results['alphas_pred'] = pred_alphas
-
-        gt_images = data['images_output'] # [B, V, 3, output_size, output_size], ground-truth novel views
-        gt_masks = data['masks_output'] # [B, V, 1, output_size, output_size], ground-truth masks
-
-        gt_images = gt_images * gt_masks + bg_color.view(1, 1, 3, 1, 1) * (1 - gt_masks)
-
-        loss_mse = F.mse_loss(pred_images, gt_images) + F.mse_loss(pred_alphas, gt_masks)
-        loss = loss + loss_mse
-
-        if self.opt.lambda_lpips > 0:
-            loss_lpips = self.lpips_loss(
-                # gt_images.view(-1, 3, self.opt.output_size, self.opt.output_size) * 2 - 1,
-                # pred_images.view(-1, 3, self.opt.output_size, self.opt.output_size) * 2 - 1,
-                # downsampled to at most 256 to reduce memory cost
-                F.interpolate(gt_images.view(-1, 3, self.opt.output_size, self.opt.output_size) * 2 - 1, (256, 256), mode='bilinear', align_corners=False), 
-                F.interpolate(pred_images.view(-1, 3, self.opt.output_size, self.opt.output_size) * 2 - 1, (256, 256), mode='bilinear', align_corners=False),
-            ).mean()
-            results['loss_lpips'] = loss_lpips
-            loss = loss + self.opt.lambda_lpips * loss_lpips
-            
-        results['loss'] = loss
-
-        # metric
-        with torch.no_grad():
-            psnr = -10 * torch.log10(torch.mean((pred_images.detach() - gt_images) ** 2))
-            results['psnr'] = psnr
-
-        return results

core/options.py

@@ -1,120 +0,0 @@
-import tyro
-from dataclasses import dataclass
-from typing import Tuple, Literal, Dict, Optional
-
-
-@dataclass
-class Options:
-    ### model
-    # Unet image input size
-    input_size: int = 256
-    # Unet definition
-    down_channels: Tuple[int] = (64, 128, 256, 512, 1024, 1024)
-    down_attention: Tuple[bool] = (False, False, False, True, True, True)
-    mid_attention: bool = True
-    up_channels: Tuple[int] = (1024, 1024, 512, 256)
-    up_attention: Tuple[bool] = (True, True, True, False)
-    # Unet output size, dependent on the input_size and U-Net structure!
-    splat_size: int = 64
-    # gaussian render size
-    output_size: int = 256
-    
-    ### dataset
-    # data mode (only support s3 now)
-    data_mode: Literal['s3'] = 's3'
-    # fovy of the dataset
-    fovy: float = 49.1
-    # camera near plane
-    znear: float = 0.5
-    # camera far plane
-    zfar: float = 2.5
-    # number of all views (input + output)
-    num_views: int = 12
-    # number of views
-    num_input_views: int = 4
-    # camera radius
-    cam_radius: float = 1.5 # to better use [-1, 1]^3 space
-    # num workers
-    num_workers: int = 8
-
-    ### training
-    # workspace
-    workspace: str = './workspace'
-    # resume 
-    resume: Optional[str] = None
-    # batch size (per-GPU)
-    batch_size: int = 8
-    # gradient accumulation
-    gradient_accumulation_steps: int = 1
-    # training epochs
-    num_epochs: int = 30
-    # lpips loss weight
-    lambda_lpips: float = 1.0
-    # gradient clip
-    gradient_clip: float = 1.0
-    # mixed precision
-    mixed_precision: str = 'bf16'
-    # learning rate
-    lr: float = 4e-4
-    # augmentation prob for grid distortion
-    prob_grid_distortion: float = 0.5
-    # augmentation prob for camera jitter
-    prob_cam_jitter: float = 0.5
-
-    ### testing
-    # test image path
-    test_path: Optional[str] = None
-
-    ### misc
-    # nvdiffrast backend setting
-    force_cuda_rast: bool = False
-    # render fancy video with gaussian scaling effect
-    fancy_video: bool = False
-    
-
-# all the default settings
-config_defaults: Dict[str, Options] = {}
-config_doc: Dict[str, str] = {}
-
-config_doc['lrm'] = 'the default settings for LGM'
-config_defaults['lrm'] = Options()
-
-config_doc['small'] = 'small model with lower resolution Gaussians'
-config_defaults['small'] = Options(
-    input_size=256,
-    splat_size=64,
-    output_size=256,
-    batch_size=8,
-    gradient_accumulation_steps=1,
-    mixed_precision='bf16',
-)
-
-config_doc['big'] = 'big model with higher resolution Gaussians'
-config_defaults['big'] = Options(
-    input_size=256,
-    up_channels=(1024, 1024, 512, 256, 128), # one more decoder
-    up_attention=(True, True, True, False, False),
-    splat_size=128,
-    output_size=512, # render & supervise Gaussians at a higher resolution.
-    batch_size=8,
-    num_views=8,
-    gradient_accumulation_steps=1,
-    mixed_precision='bf16',
-)
-
-config_doc['tiny'] = 'tiny model for ablation'
-config_defaults['tiny'] = Options(
-    input_size=256, 
-    down_channels=(32, 64, 128, 256, 512),
-    down_attention=(False, False, False, False, True),
-    up_channels=(512, 256, 128),
-    up_attention=(True, False, False, False),
-    splat_size=64,
-    output_size=256,
-    batch_size=16,
-    num_views=8,
-    gradient_accumulation_steps=1,
-    mixed_precision='bf16',
-)
-
-AllConfigs = tyro.extras.subcommand_type_from_defaults(config_defaults, config_doc)

core/provider_objaverse.py

@@ -1,172 +0,0 @@
-import os
-import cv2
-import random
-import numpy as np
-
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-import torchvision.transforms.functional as TF
-from torch.utils.data import Dataset
-
-import kiui
-from core.options import Options
-from core.utils import get_rays, grid_distortion, orbit_camera_jitter
-
-IMAGENET_DEFAULT_MEAN = (0.485, 0.456, 0.406)
-IMAGENET_DEFAULT_STD = (0.229, 0.224, 0.225)
-
-
-class ObjaverseDataset(Dataset):
-
-    def _warn(self):
-        raise NotImplementedError('this dataset is just an example and cannot be used directly, you should modify it to your own setting! (search keyword TODO)')
-
-    def __init__(self, opt: Options, training=True):
-        
-        self.opt = opt
-        self.training = training
-
-        # TODO: remove this barrier
-        self._warn()
-
-        # TODO: load the list of objects for training
-        self.items = []
-        with open('TODO: file containing the list', 'r') as f:
-            for line in f.readlines():
-                self.items.append(line.strip())
-
-        # naive split
-        if self.training:
-            self.items = self.items[:-self.opt.batch_size]
-        else:
-            self.items = self.items[-self.opt.batch_size:]
-        
-        # default camera intrinsics
-        self.tan_half_fov = np.tan(0.5 * np.deg2rad(self.opt.fovy))
-        self.proj_matrix = torch.zeros(4, 4, dtype=torch.float32)
-        self.proj_matrix[0, 0] = 1 / self.tan_half_fov
-        self.proj_matrix[1, 1] = 1 / self.tan_half_fov
-        self.proj_matrix[2, 2] = (self.opt.zfar + self.opt.znear) / (self.opt.zfar - self.opt.znear)
-        self.proj_matrix[3, 2] = - (self.opt.zfar * self.opt.znear) / (self.opt.zfar - self.opt.znear)
-        self.proj_matrix[2, 3] = 1
-
-
-    def __len__(self):
-        return len(self.items)
-
-    def __getitem__(self, idx):
-
-        uid = self.items[idx]
-        results = {}
-
-        # load num_views images
-        images = []
-        masks = []
-        cam_poses = []
-        
-        vid_cnt = 0
-
-        # TODO: choose views, based on your rendering settings
-        if self.training:
-            # input views are in (36, 72), other views are randomly selected
-            vids = np.random.permutation(np.arange(36, 73))[:self.opt.num_input_views].tolist() + np.random.permutation(100).tolist()
-        else:
-            # fixed views
-            vids = np.arange(36, 73, 4).tolist() + np.arange(100).tolist()
-        
-        for vid in vids:
-
-            image_path = os.path.join(uid, 'rgb', f'{vid:03d}.png')
-            camera_path = os.path.join(uid, 'pose', f'{vid:03d}.txt')
-
-            try:
-                # TODO: load data (modify self.client here)
-                image = np.frombuffer(self.client.get(image_path), np.uint8)
-                image = torch.from_numpy(cv2.imdecode(image, cv2.IMREAD_UNCHANGED).astype(np.float32) / 255) # [512, 512, 4] in [0, 1]
-                c2w = [float(t) for t in self.client.get(camera_path).decode().strip().split(' ')]
-                c2w = torch.tensor(c2w, dtype=torch.float32).reshape(4, 4)
-            except Exception as e:
-                # print(f'[WARN] dataset {uid} {vid}: {e}')
-                continue
-            
-            # TODO: you may have a different camera system
-            # blender world + opencv cam --> opengl world & cam
-            c2w[1] *= -1
-            c2w[[1, 2]] = c2w[[2, 1]]
-            c2w[:3, 1:3] *= -1 # invert up and forward direction
-
-            # scale up radius to fully use the [-1, 1]^3 space!
-            c2w[:3, 3] *= self.opt.cam_radius / 1.5 # 1.5 is the default scale
-          
-            image = image.permute(2, 0, 1) # [4, 512, 512]
-            mask = image[3:4] # [1, 512, 512]
-            image = image[:3] * mask + (1 - mask) # [3, 512, 512], to white bg
-            image = image[[2,1,0]].contiguous() # bgr to rgb
-
-            images.append(image)
-            masks.append(mask.squeeze(0))
-            cam_poses.append(c2w)
-
-            vid_cnt += 1
-            if vid_cnt == self.opt.num_views:
-                break
-
-        if vid_cnt < self.opt.num_views:
-            print(f'[WARN] dataset {uid}: not enough valid views, only {vid_cnt} views found!')
-            n = self.opt.num_views - vid_cnt
-            images = images + [images[-1]] * n
-            masks = masks + [masks[-1]] * n
-            cam_poses = cam_poses + [cam_poses[-1]] * n
-          
-        images = torch.stack(images, dim=0) # [V, C, H, W]
-        masks = torch.stack(masks, dim=0) # [V, H, W]
-        cam_poses = torch.stack(cam_poses, dim=0) # [V, 4, 4]
-
-        # normalized camera feats as in paper (transform the first pose to a fixed position)
-        transform = torch.tensor([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, self.opt.cam_radius], [0, 0, 0, 1]], dtype=torch.float32) @ torch.inverse(cam_poses[0])
-        cam_poses = transform.unsqueeze(0) @ cam_poses  # [V, 4, 4]
-
-        images_input = F.interpolate(images[:self.opt.num_input_views].clone(), size=(self.opt.input_size, self.opt.input_size), mode='bilinear', align_corners=False) # [V, C, H, W]
-        cam_poses_input = cam_poses[:self.opt.num_input_views].clone()
-
-        # data augmentation
-        if self.training:
-            # apply random grid distortion to simulate 3D inconsistency
-            if random.random() < self.opt.prob_grid_distortion:
-                images_input[1:] = grid_distortion(images_input[1:])
-            # apply camera jittering (only to input!)
-            if random.random() < self.opt.prob_cam_jitter:
-                cam_poses_input[1:] = orbit_camera_jitter(cam_poses_input[1:])
-
-        images_input = TF.normalize(images_input, IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD)
-
-        # resize render ground-truth images, range still in [0, 1]
-        results['images_output'] = F.interpolate(images, size=(self.opt.output_size, self.opt.output_size), mode='bilinear', align_corners=False) # [V, C, output_size, output_size]
-        results['masks_output'] = F.interpolate(masks.unsqueeze(1), size=(self.opt.output_size, self.opt.output_size), mode='bilinear', align_corners=False) # [V, 1, output_size, output_size]
-
-        # build rays for input views
-        rays_embeddings = []
-        for i in range(self.opt.num_input_views):
-            rays_o, rays_d = get_rays(cam_poses_input[i], self.opt.input_size, self.opt.input_size, self.opt.fovy) # [h, w, 3]
-            rays_plucker = torch.cat([torch.cross(rays_o, rays_d, dim=-1), rays_d], dim=-1) # [h, w, 6]
-            rays_embeddings.append(rays_plucker)
-
-     
-        rays_embeddings = torch.stack(rays_embeddings, dim=0).permute(0, 3, 1, 2).contiguous() # [V, 6, h, w]
-        final_input = torch.cat([images_input, rays_embeddings], dim=1) # [V=4, 9, H, W]
-        results['input'] = final_input
-
-        # opengl to colmap camera for gaussian renderer
-        cam_poses[:, :3, 1:3] *= -1 # invert up & forward direction
-        
-        # cameras needed by gaussian rasterizer
-        cam_view = torch.inverse(cam_poses).transpose(1, 2) # [V, 4, 4]
-        cam_view_proj = cam_view @ self.proj_matrix # [V, 4, 4]
-        cam_pos = - cam_poses[:, :3, 3] # [V, 3]
-        
-        results['cam_view'] = cam_view
-        results['cam_view_proj'] = cam_view_proj
-        results['cam_pos'] = cam_pos
-
-        return results

core/unet.py

@@ -1,319 +0,0 @@
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-
-import numpy as np
-from typing import Tuple, Optional, Literal
-from functools import partial
-
-from core.attention import MemEffAttention, MemEffCrossAttention
-
-class MVAttention(nn.Module):
-    def __init__(
-        self, 
-        dim: int,
-        num_heads: int = 8,
-        qkv_bias: bool = False,
-        proj_bias: bool = True,
-        attn_drop: float = 0.0,
-        proj_drop: float = 0.0,
-        groups: int = 32,
-        eps: float = 1e-5,
-        residual: bool = True,
-        skip_scale: float = 1,
-        num_frames: int = 4, # WARN: hardcoded!
-    ):
-        super().__init__()
-
-        self.residual = residual
-        self.skip_scale = skip_scale
-        self.num_frames = num_frames
-
-        self.norm = nn.GroupNorm(num_groups=groups, num_channels=dim, eps=eps, affine=True)
-        self.attn = MemEffAttention(dim, num_heads, qkv_bias, proj_bias, attn_drop, proj_drop)
-
-    def forward(self, x):
-        # x: [B*V, C, H, W]
-        BV, C, H, W = x.shape
-        B = BV // self.num_frames # assert BV % self.num_frames == 0
-
-        res = x
-        x = self.norm(x)
-
-        x = x.reshape(B, self.num_frames, C, H, W).permute(0, 1, 3, 4, 2).reshape(B, -1, C)
-        x = self.attn(x)
-        x = x.reshape(B, self.num_frames, H, W, C).permute(0, 1, 4, 2, 3).reshape(BV, C, H, W)
-
-        if self.residual:
-            x = (x + res) * self.skip_scale
-        return x
-
-class ResnetBlock(nn.Module):
-    def __init__(
-        self,
-        in_channels: int,
-        out_channels: int,
-        resample: Literal['default', 'up', 'down'] = 'default',
-        groups: int = 32,
-        eps: float = 1e-5,
-        skip_scale: float = 1, # multiplied to output
-    ):
-        super().__init__()
-
-        self.in_channels = in_channels
-        self.out_channels = out_channels
-        self.skip_scale = skip_scale
-
-        self.norm1 = nn.GroupNorm(num_groups=groups, num_channels=in_channels, eps=eps, affine=True)
-        self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=1, padding=1)
-
-        self.norm2 = nn.GroupNorm(num_groups=groups, num_channels=out_channels, eps=eps, affine=True)
-        self.conv2 = nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=1, padding=1)
-
-        self.act = F.silu
-
-        self.resample = None
-        if resample == 'up':
-            self.resample = partial(F.interpolate, scale_factor=2.0, mode="nearest")
-        elif resample == 'down':
-            self.resample = nn.AvgPool2d(kernel_size=2, stride=2)
-        
-        self.shortcut = nn.Identity()
-        if self.in_channels != self.out_channels:
-            self.shortcut = nn.Conv2d(in_channels, out_channels, kernel_size=1, bias=True)
-
-    
-    def forward(self, x):
-        res = x
-
-        x = self.norm1(x)
-        x = self.act(x)
-
-        if self.resample:
-            res = self.resample(res)
-            x = self.resample(x)
-        
-        x = self.conv1(x)
-        x = self.norm2(x)
-        x = self.act(x)
-        x = self.conv2(x)
-
-        x = (x + self.shortcut(res)) * self.skip_scale
-
-        return x
-
-class DownBlock(nn.Module):
-    def __init__(
-        self,
-        in_channels: int,
-        out_channels: int,
-        num_layers: int = 1,
-        downsample: bool = True,
-        attention: bool = True,
-        attention_heads: int = 16,
-        skip_scale: float = 1,
-    ):
-        super().__init__()
- 
-        nets = []
-        attns = []
-        for i in range(num_layers):
-            in_channels = in_channels if i == 0 else out_channels
-            nets.append(ResnetBlock(in_channels, out_channels, skip_scale=skip_scale))
-            if attention:
-                attns.append(MVAttention(out_channels, attention_heads, skip_scale=skip_scale))
-            else:
-                attns.append(None)
-        self.nets = nn.ModuleList(nets)
-        self.attns = nn.ModuleList(attns)
-
-        self.downsample = None
-        if downsample:
-            self.downsample = nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=2, padding=1)
-
-    def forward(self, x):
-        xs = []
-
-        for attn, net in zip(self.attns, self.nets):
-            x = net(x)
-            if attn:
-                x = attn(x)
-            xs.append(x)
-
-        if self.downsample:
-            x = self.downsample(x)
-            xs.append(x)
-  
-        return x, xs
-
-
-class MidBlock(nn.Module):
-    def __init__(
-        self,
-        in_channels: int,
-        num_layers: int = 1,
-        attention: bool = True,
-        attention_heads: int = 16,
-        skip_scale: float = 1,
-    ):
-        super().__init__()
-
-        nets = []
-        attns = []
-        # first layer
-        nets.append(ResnetBlock(in_channels, in_channels, skip_scale=skip_scale))
-        # more layers
-        for i in range(num_layers):
-            nets.append(ResnetBlock(in_channels, in_channels, skip_scale=skip_scale))
-            if attention:
-                attns.append(MVAttention(in_channels, attention_heads, skip_scale=skip_scale))
-            else:
-                attns.append(None)
-        self.nets = nn.ModuleList(nets)
-        self.attns = nn.ModuleList(attns)
-        
-    def forward(self, x):
-        x = self.nets[0](x)
-        for attn, net in zip(self.attns, self.nets[1:]):
-            if attn:
-                x = attn(x)
-            x = net(x)
-        return x
-
-
-class UpBlock(nn.Module):
-    def __init__(
-        self,
-        in_channels: int,
-        prev_out_channels: int,
-        out_channels: int,
-        num_layers: int = 1,
-        upsample: bool = True,
-        attention: bool = True,
-        attention_heads: int = 16,
-        skip_scale: float = 1,
-    ):
-        super().__init__()
-
-        nets = []
-        attns = []
-        for i in range(num_layers):
-            cin = in_channels if i == 0 else out_channels
-            cskip = prev_out_channels if (i == num_layers - 1) else out_channels
-
-            nets.append(ResnetBlock(cin + cskip, out_channels, skip_scale=skip_scale))
-            if attention:
-                attns.append(MVAttention(out_channels, attention_heads, skip_scale=skip_scale))
-            else:
-                attns.append(None)
-        self.nets = nn.ModuleList(nets)
-        self.attns = nn.ModuleList(attns)
-
-        self.upsample = None
-        if upsample:
-            self.upsample = nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=1, padding=1)
-
-    def forward(self, x, xs):
-
-        for attn, net in zip(self.attns, self.nets):
-            res_x = xs[-1]
-            xs = xs[:-1]
-            x = torch.cat([x, res_x], dim=1)
-            x = net(x)
-            if attn:
-                x = attn(x)
-            
-        if self.upsample:
-            x = F.interpolate(x, scale_factor=2.0, mode='nearest')
-            x = self.upsample(x)
-
-        return x
-
-
-# it could be asymmetric!
-class UNet(nn.Module):
-    def __init__(
-        self,
-        in_channels: int = 3,
-        out_channels: int = 3,
-        down_channels: Tuple[int] = (64, 128, 256, 512, 1024),
-        down_attention: Tuple[bool] = (False, False, False, True, True),
-        mid_attention: bool = True,
-        up_channels: Tuple[int] = (1024, 512, 256),
-        up_attention: Tuple[bool] = (True, True, False),
-        layers_per_block: int = 2,
-        skip_scale: float = np.sqrt(0.5),
-    ):
-        super().__init__()
-
-        # first
-        self.conv_in = nn.Conv2d(in_channels, down_channels[0], kernel_size=3, stride=1, padding=1)
-
-        # down
-        down_blocks = []
-        cout = down_channels[0]
-        for i in range(len(down_channels)):
-            cin = cout
-            cout = down_channels[i]
-
-            down_blocks.append(DownBlock(
-                cin, cout, 
-                num_layers=layers_per_block, 
-                downsample=(i != len(down_channels) - 1), # not final layer
-                attention=down_attention[i],
-                skip_scale=skip_scale,
-            ))
-        self.down_blocks = nn.ModuleList(down_blocks)
-
-        # mid
-        self.mid_block = MidBlock(down_channels[-1], attention=mid_attention, skip_scale=skip_scale)
-
-        # up
-        up_blocks = []
-        cout = up_channels[0]
-        for i in range(len(up_channels)):
-            cin = cout
-            cout = up_channels[i]
-            cskip = down_channels[max(-2 - i, -len(down_channels))] # for assymetric
-
-            up_blocks.append(UpBlock(
-                cin, cskip, cout, 
-                num_layers=layers_per_block + 1, # one more layer for up
-                upsample=(i != len(up_channels) - 1), # not final layer
-                attention=up_attention[i],
-                skip_scale=skip_scale,
-            ))
-        self.up_blocks = nn.ModuleList(up_blocks)
-
-        # last
-        self.norm_out = nn.GroupNorm(num_channels=up_channels[-1], num_groups=32, eps=1e-5)
-        self.conv_out = nn.Conv2d(up_channels[-1], out_channels, kernel_size=3, stride=1, padding=1)
-
-
-    def forward(self, x):
-        # x: [B, Cin, H, W]
-
-        # first
-        x = self.conv_in(x)
-        
-        # down
-        xss = [x]
-        for block in self.down_blocks:
-            x, xs = block(x)
-            xss.extend(xs)
-        
-        # mid
-        x = self.mid_block(x)
-
-        # up
-        for block in self.up_blocks:
-            xs = xss[-len(block.nets):]
-            xss = xss[:-len(block.nets)]
-            x = block(x, xs)
-
-        # last
-        x = self.norm_out(x)
-        x = F.silu(x)
-        x = self.conv_out(x) # [B, Cout, H', W']
-
-        return x

core/utils.py

@@ -1,109 +0,0 @@
-import numpy as np
-
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-
-import roma
-from kiui.op import safe_normalize
-
-def get_rays(pose, h, w, fovy, opengl=True):
-
-    x, y = torch.meshgrid(
-        torch.arange(w, device=pose.device),
-        torch.arange(h, device=pose.device),
-        indexing="xy",
-    )
-    x = x.flatten()
-    y = y.flatten()
-
-    cx = w * 0.5
-    cy = h * 0.5
-
-    focal = h * 0.5 / np.tan(0.5 * np.deg2rad(fovy))
-
-    camera_dirs = F.pad(
-        torch.stack(
-            [
-                (x - cx + 0.5) / focal,
-                (y - cy + 0.5) / focal * (-1.0 if opengl else 1.0),
-            ],
-            dim=-1,
-        ),
-        (0, 1),
-        value=(-1.0 if opengl else 1.0),
-    )  # [hw, 3]
-
-    rays_d = camera_dirs @ pose[:3, :3].transpose(0, 1)  # [hw, 3]
-    rays_o = pose[:3, 3].unsqueeze(0).expand_as(rays_d) # [hw, 3]
-
-    rays_o = rays_o.view(h, w, 3)
-    rays_d = safe_normalize(rays_d).view(h, w, 3)
-
-    return rays_o, rays_d
-
-def orbit_camera_jitter(poses, strength=0.1):
-    # poses: [B, 4, 4], assume orbit camera in opengl format
-    # random orbital rotate
-
-    B = poses.shape[0]
-    rotvec_x = poses[:, :3, 1] * strength * np.pi * (torch.rand(B, 1, device=poses.device) * 2 - 1)
-    rotvec_y = poses[:, :3, 0] * strength * np.pi / 2 * (torch.rand(B, 1, device=poses.device) * 2 - 1)
-
-    rot = roma.rotvec_to_rotmat(rotvec_x) @ roma.rotvec_to_rotmat(rotvec_y)
-    R = rot @ poses[:, :3, :3]
-    T = rot @ poses[:, :3, 3:]
-
-    new_poses = poses.clone()
-    new_poses[:, :3, :3] = R
-    new_poses[:, :3, 3:] = T
-    
-    return new_poses
-
-def grid_distortion(images, strength=0.5):
-    # images: [B, C, H, W]
-    # num_steps: int, grid resolution for distortion
-    # strength: float in [0, 1], strength of distortion
-
-    B, C, H, W = images.shape
-
-    num_steps = np.random.randint(8, 17)
-    grid_steps = torch.linspace(-1, 1, num_steps)
-
-    # have to loop batch...
-    grids = []
-    for b in range(B):
-        # construct displacement
-        x_steps = torch.linspace(0, 1, num_steps) # [num_steps], inclusive
-        x_steps = (x_steps + strength * (torch.rand_like(x_steps) - 0.5) / (num_steps - 1)).clamp(0, 1) # perturb
-        x_steps = (x_steps * W).long() # [num_steps]
-        x_steps[0] = 0
-        x_steps[-1] = W
-        xs = []
-        for i in range(num_steps - 1):
-            xs.append(torch.linspace(grid_steps[i], grid_steps[i + 1], x_steps[i + 1] - x_steps[i]))
-        xs = torch.cat(xs, dim=0) # [W]
-
-        y_steps = torch.linspace(0, 1, num_steps) # [num_steps], inclusive
-        y_steps = (y_steps + strength * (torch.rand_like(y_steps) - 0.5) / (num_steps - 1)).clamp(0, 1) # perturb
-        y_steps = (y_steps * H).long() # [num_steps]
-        y_steps[0] = 0
-        y_steps[-1] = H
-        ys = []
-        for i in range(num_steps - 1):
-            ys.append(torch.linspace(grid_steps[i], grid_steps[i + 1], y_steps[i + 1] - y_steps[i]))
-        ys = torch.cat(ys, dim=0) # [H]
-
-        # construct grid
-        grid_x, grid_y = torch.meshgrid(xs, ys, indexing='xy') # [H, W]
-        grid = torch.stack([grid_x, grid_y], dim=-1) # [H, W, 2]
-
-        grids.append(grid)
-    
-    grids = torch.stack(grids, dim=0).to(images.device) # [B, H, W, 2]
-
-    # grid sample
-    images = F.grid_sample(images, grids, align_corners=False)
-
-    return images
-

diff-gaussian-rasterization

@@ -0,0 +1 @@
+Subproject commit d986da0d4cf2dfeb43b9a379b6e9fa0a7f3f7eea

diff-gaussian-rasterization/.gitignore

@@ -1,7 +0,0 @@
-build/
-diff_gaussian_rasterization.egg-info/
-dist/
-
-__pycache__
-
-*.so

diff-gaussian-rasterization/.gitmodules

@@ -1,3 +0,0 @@
-[submodule "third_party/glm"]
-	path = third_party/glm
-	url = https://github.com/g-truc/glm.git

diff-gaussian-rasterization/CMakeLists.txt

@@ -1,36 +0,0 @@
-#
-# Copyright (C) 2023, Inria
-# GRAPHDECO research group, https://team.inria.fr/graphdeco
-# All rights reserved.
-#
-# This software is free for non-commercial, research and evaluation use 
-# under the terms of the LICENSE.md file.
-#
-# For inquiries contact  george.drettakis@inria.fr
-#
-
-cmake_minimum_required(VERSION 3.20)
-
-project(DiffRast LANGUAGES CUDA CXX)
-
-set(CMAKE_CXX_STANDARD 17)
-set(CMAKE_CXX_EXTENSIONS OFF)
-set(CMAKE_CUDA_STANDARD 17)
-
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
-
-add_library(CudaRasterizer
-	cuda_rasterizer/backward.h
-	cuda_rasterizer/backward.cu
-	cuda_rasterizer/forward.h
-	cuda_rasterizer/forward.cu
-	cuda_rasterizer/auxiliary.h
-	cuda_rasterizer/rasterizer_impl.cu
-	cuda_rasterizer/rasterizer_impl.h
-	cuda_rasterizer/rasterizer.h
-)
-
-set_target_properties(CudaRasterizer PROPERTIES CUDA_ARCHITECTURES "75;86")
-
-target_include_directories(CudaRasterizer PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/cuda_rasterizer)
-target_include_directories(CudaRasterizer PRIVATE third_party/glm ${CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES})

diff-gaussian-rasterization/LICENSE.md

@@ -1,83 +0,0 @@
-Gaussian-Splatting License  
-===========================  
-
-**Inria** and **the Max Planck Institut for Informatik (MPII)** hold all the ownership rights on the *Software* named **gaussian-splatting**.  
-The *Software* is in the process of being registered with the Agence pour la Protection des  
-Programmes (APP).  
-
-The *Software* is still being developed by the *Licensor*.  
-
-*Licensor*'s goal is to allow the research community to use, test and evaluate  
-the *Software*.  
-
-## 1.  Definitions  
-
-*Licensee* means any person or entity that uses the *Software* and distributes  
-its *Work*.  
-
-*Licensor* means the owners of the *Software*, i.e Inria and MPII  
-
-*Software* means the original work of authorship made available under this  
-License ie gaussian-splatting.  
-
-*Work* means the *Software* and any additions to or derivative works of the  
-*Software* that are made available under this License.  
-
-
-## 2.  Purpose  
-This license is intended to define the rights granted to the *Licensee* by  
-Licensors under the *Software*.  
-
-## 3.  Rights granted  
-
-For the above reasons Licensors have decided to distribute the *Software*.  
-Licensors grant non-exclusive rights to use the *Software* for research purposes  
-to research users (both academic and industrial), free of charge, without right  
-to sublicense.. The *Software* may be used "non-commercially", i.e., for research  
-and/or evaluation purposes only.  
-
-Subject to the terms and conditions of this License, you are granted a  
-non-exclusive, royalty-free, license to reproduce, prepare derivative works of,  
-publicly display, publicly perform and distribute its *Work* and any resulting  
-derivative works in any form.  
-
-## 4.  Limitations  
-
-**4.1 Redistribution.** You may reproduce or distribute the *Work* only if (a) you do  
-so under this License, (b) you include a complete copy of this License with  
-your distribution, and (c) you retain without modification any copyright,  
-patent, trademark, or attribution notices that are present in the *Work*.  
-
-**4.2 Derivative Works.** You may specify that additional or different terms apply  
-to the use, reproduction, and distribution of your derivative works of the *Work*  
-("Your Terms") only if (a) Your Terms provide that the use limitation in  
-Section 2 applies to your derivative works, and (b) you identify the specific  
-derivative works that are subject to Your Terms. Notwithstanding Your Terms,  
-this License (including the redistribution requirements in Section 3.1) will  
-continue to apply to the *Work* itself.  
-
-**4.3** Any other use without of prior consent of Licensors is prohibited. Research  
-users explicitly acknowledge having received from Licensors all information  
-allowing to appreciate the adequacy between of the *Software* and their needs and  
-to undertake all necessary precautions for its execution and use.  
-
-**4.4** The *Software* is provided both as a compiled library file and as source  
-code. In case of using the *Software* for a publication or other results obtained  
-through the use of the *Software*, users are strongly encouraged to cite the  
-corresponding publications as explained in the documentation of the *Software*.  
-
-## 5.  Disclaimer  
-
-THE USER CANNOT USE, EXPLOIT OR DISTRIBUTE THE *SOFTWARE* FOR COMMERCIAL PURPOSES  
-WITHOUT PRIOR AND EXPLICIT CONSENT OF LICENSORS. YOU MUST CONTACT INRIA FOR ANY  
-UNAUTHORIZED USE: stip-sophia.transfert@inria.fr . ANY SUCH ACTION WILL  
-CONSTITUTE A FORGERY. THIS *SOFTWARE* IS PROVIDED "AS IS" WITHOUT ANY WARRANTIES  
-OF ANY NATURE AND ANY EXPRESS OR IMPLIED WARRANTIES, WITH REGARDS TO COMMERCIAL  
-USE, PROFESSIONNAL USE, LEGAL OR NOT, OR OTHER, OR COMMERCIALISATION OR  
-ADAPTATION. UNLESS EXPLICITLY PROVIDED BY LAW, IN NO EVENT, SHALL INRIA OR THE  
-AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR  
-CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE  
-GOODS OR SERVICES, LOSS OF USE, DATA, OR PROFITS OR BUSINESS INTERRUPTION)  
-HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT  
-LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING FROM, OUT OF OR  
-IN CONNECTION WITH THE *SOFTWARE* OR THE USE OR OTHER DEALINGS IN THE *SOFTWARE*.  

diff-gaussian-rasterization/README.md

@@ -1,35 +0,0 @@
-# Differential Gaussian Rasterization
-
-**NOTE**: this is a modified version to support depth & alpha rendering (both forward and backward) from the [original repository](https://github.com/graphdeco-inria/diff-gaussian-rasterization). 
-
-```python
-rendered_image, radii, rendered_depth, rendered_alpha = rasterizer(
-    means3D=means3D,
-    means2D=means2D,
-    shs=shs,
-    colors_precomp=colors_precomp,
-    opacities=opacity,
-    scales=scales,
-    rotations=rotations,
-    cov3D_precomp=cov3D_precomp,
-)
-```
-
-
-Used as the rasterization engine for the paper "3D Gaussian Splatting for Real-Time Rendering of Radiance Fields". If you can make use of it in your own research, please be so kind to cite us.
-
-<section class="section" id="BibTeX">
-  <div class="container is-max-desktop content">
-    <h2 class="title">BibTeX</h2>
-    <pre><code>@Article{kerbl3Dgaussians,
-      author       = {Kerbl, Bernhard and Kopanas, Georgios and Leimk{\"u}hler, Thomas and Drettakis, George},
-      title        = {3D Gaussian Splatting for Real-Time Radiance Field Rendering},
-      journal      = {ACM Transactions on Graphics},
-      number       = {4},
-      volume       = {42},
-      month        = {July},
-      year         = {2023},
-      url          = {https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/}
-}</code></pre>
-  </div>
-</section>

diff-gaussian-rasterization/cuda_rasterizer/auxiliary.h

@@ -1,175 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#ifndef CUDA_RASTERIZER_AUXILIARY_H_INCLUDED
-#define CUDA_RASTERIZER_AUXILIARY_H_INCLUDED
-
-#include "config.h"
-#include "stdio.h"
-
-#define BLOCK_SIZE (BLOCK_X * BLOCK_Y)
-#define NUM_WARPS (BLOCK_SIZE/32)
-
-// Spherical harmonics coefficients
-__device__ const float SH_C0 = 0.28209479177387814f;
-__device__ const float SH_C1 = 0.4886025119029199f;
-__device__ const float SH_C2[] = {
-	1.0925484305920792f,
-	-1.0925484305920792f,
-	0.31539156525252005f,
-	-1.0925484305920792f,
-	0.5462742152960396f
-};
-__device__ const float SH_C3[] = {
-	-0.5900435899266435f,
-	2.890611442640554f,
-	-0.4570457994644658f,
-	0.3731763325901154f,
-	-0.4570457994644658f,
-	1.445305721320277f,
-	-0.5900435899266435f
-};
-
-__forceinline__ __device__ float ndc2Pix(float v, int S)
-{
-	return ((v + 1.0) * S - 1.0) * 0.5;
-}
-
-__forceinline__ __device__ void getRect(const float2 p, int max_radius, uint2& rect_min, uint2& rect_max, dim3 grid)
-{
-	rect_min = {
-		min(grid.x, max((int)0, (int)((p.x - max_radius) / BLOCK_X))),
-		min(grid.y, max((int)0, (int)((p.y - max_radius) / BLOCK_Y)))
-	};
-	rect_max = {
-		min(grid.x, max((int)0, (int)((p.x + max_radius + BLOCK_X - 1) / BLOCK_X))),
-		min(grid.y, max((int)0, (int)((p.y + max_radius + BLOCK_Y - 1) / BLOCK_Y)))
-	};
-}
-
-__forceinline__ __device__ float3 transformPoint4x3(const float3& p, const float* matrix)
-{
-	float3 transformed = {
-		matrix[0] * p.x + matrix[4] * p.y + matrix[8] * p.z + matrix[12],
-		matrix[1] * p.x + matrix[5] * p.y + matrix[9] * p.z + matrix[13],
-		matrix[2] * p.x + matrix[6] * p.y + matrix[10] * p.z + matrix[14],
-	};
-	return transformed;
-}
-
-__forceinline__ __device__ float4 transformPoint4x4(const float3& p, const float* matrix)
-{
-	float4 transformed = {
-		matrix[0] * p.x + matrix[4] * p.y + matrix[8] * p.z + matrix[12],
-		matrix[1] * p.x + matrix[5] * p.y + matrix[9] * p.z + matrix[13],
-		matrix[2] * p.x + matrix[6] * p.y + matrix[10] * p.z + matrix[14],
-		matrix[3] * p.x + matrix[7] * p.y + matrix[11] * p.z + matrix[15]
-	};
-	return transformed;
-}
-
-__forceinline__ __device__ float3 transformVec4x3(const float3& p, const float* matrix)
-{
-	float3 transformed = {
-		matrix[0] * p.x + matrix[4] * p.y + matrix[8] * p.z,
-		matrix[1] * p.x + matrix[5] * p.y + matrix[9] * p.z,
-		matrix[2] * p.x + matrix[6] * p.y + matrix[10] * p.z,
-	};
-	return transformed;
-}
-
-__forceinline__ __device__ float3 transformVec4x3Transpose(const float3& p, const float* matrix)
-{
-	float3 transformed = {
-		matrix[0] * p.x + matrix[1] * p.y + matrix[2] * p.z,
-		matrix[4] * p.x + matrix[5] * p.y + matrix[6] * p.z,
-		matrix[8] * p.x + matrix[9] * p.y + matrix[10] * p.z,
-	};
-	return transformed;
-}
-
-__forceinline__ __device__ float dnormvdz(float3 v, float3 dv)
-{
-	float sum2 = v.x * v.x + v.y * v.y + v.z * v.z;
-	float invsum32 = 1.0f / sqrt(sum2 * sum2 * sum2);
-	float dnormvdz = (-v.x * v.z * dv.x - v.y * v.z * dv.y + (sum2 - v.z * v.z) * dv.z) * invsum32;
-	return dnormvdz;
-}
-
-__forceinline__ __device__ float3 dnormvdv(float3 v, float3 dv)
-{
-	float sum2 = v.x * v.x + v.y * v.y + v.z * v.z;
-	float invsum32 = 1.0f / sqrt(sum2 * sum2 * sum2);
-
-	float3 dnormvdv;
-	dnormvdv.x = ((+sum2 - v.x * v.x) * dv.x - v.y * v.x * dv.y - v.z * v.x * dv.z) * invsum32;
-	dnormvdv.y = (-v.x * v.y * dv.x + (sum2 - v.y * v.y) * dv.y - v.z * v.y * dv.z) * invsum32;
-	dnormvdv.z = (-v.x * v.z * dv.x - v.y * v.z * dv.y + (sum2 - v.z * v.z) * dv.z) * invsum32;
-	return dnormvdv;
-}
-
-__forceinline__ __device__ float4 dnormvdv(float4 v, float4 dv)
-{
-	float sum2 = v.x * v.x + v.y * v.y + v.z * v.z + v.w * v.w;
-	float invsum32 = 1.0f / sqrt(sum2 * sum2 * sum2);
-
-	float4 vdv = { v.x * dv.x, v.y * dv.y, v.z * dv.z, v.w * dv.w };
-	float vdv_sum = vdv.x + vdv.y + vdv.z + vdv.w;
-	float4 dnormvdv;
-	dnormvdv.x = ((sum2 - v.x * v.x) * dv.x - v.x * (vdv_sum - vdv.x)) * invsum32;
-	dnormvdv.y = ((sum2 - v.y * v.y) * dv.y - v.y * (vdv_sum - vdv.y)) * invsum32;
-	dnormvdv.z = ((sum2 - v.z * v.z) * dv.z - v.z * (vdv_sum - vdv.z)) * invsum32;
-	dnormvdv.w = ((sum2 - v.w * v.w) * dv.w - v.w * (vdv_sum - vdv.w)) * invsum32;
-	return dnormvdv;
-}
-
-__forceinline__ __device__ float sigmoid(float x)
-{
-	return 1.0f / (1.0f + expf(-x));
-}
-
-__forceinline__ __device__ bool in_frustum(int idx,
-	const float* orig_points,
-	const float* viewmatrix,
-	const float* projmatrix,
-	bool prefiltered,
-	float3& p_view)
-{
-	float3 p_orig = { orig_points[3 * idx], orig_points[3 * idx + 1], orig_points[3 * idx + 2] };
-
-	// Bring points to screen space
-	float4 p_hom = transformPoint4x4(p_orig, projmatrix);
-	float p_w = 1.0f / (p_hom.w + 0.0000001f);
-	float3 p_proj = { p_hom.x * p_w, p_hom.y * p_w, p_hom.z * p_w };
-	p_view = transformPoint4x3(p_orig, viewmatrix);
-
-	if (p_view.z <= 0.2f)// || ((p_proj.x < -1.3 || p_proj.x > 1.3 || p_proj.y < -1.3 || p_proj.y > 1.3)))
-	{
-		if (prefiltered)
-		{
-			printf("Point is filtered although prefiltered is set. This shouldn't happen!");
-			__trap();
-		}
-		return false;
-	}
-	return true;
-}
-
-#define CHECK_CUDA(A, debug) \
-A; if(debug) { \
-auto ret = cudaDeviceSynchronize(); \
-if (ret != cudaSuccess) { \
-std::cerr << "\n[CUDA ERROR] in " << __FILE__ << "\nLine " << __LINE__ << ": " << cudaGetErrorString(ret); \
-throw std::runtime_error(cudaGetErrorString(ret)); \
-} \
-}
-
-#endif

diff-gaussian-rasterization/cuda_rasterizer/backward.cu

@@ -1,712 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#include "backward.h"
-#include "auxiliary.h"
-#include <cooperative_groups.h>
-#include <cooperative_groups/reduce.h>
-namespace cg = cooperative_groups;
-
-// Backward pass for conversion of spherical harmonics to RGB for
-// each Gaussian.
-__device__ void computeColorFromSH(int idx, int deg, int max_coeffs, const glm::vec3* means, glm::vec3 campos, const float* shs, const bool* clamped, const glm::vec3* dL_dcolor, glm::vec3* dL_dmeans, glm::vec3* dL_dshs)
-{
-	// Compute intermediate values, as it is done during forward
-	glm::vec3 pos = means[idx];
-	glm::vec3 dir_orig = pos - campos;
-	glm::vec3 dir = dir_orig / glm::length(dir_orig);
-
-	glm::vec3* sh = ((glm::vec3*)shs) + idx * max_coeffs;
-
-	// Use PyTorch rule for clamping: if clamping was applied,
-	// gradient becomes 0.
-	glm::vec3 dL_dRGB = dL_dcolor[idx];
-	dL_dRGB.x *= clamped[3 * idx + 0] ? 0 : 1;
-	dL_dRGB.y *= clamped[3 * idx + 1] ? 0 : 1;
-	dL_dRGB.z *= clamped[3 * idx + 2] ? 0 : 1;
-
-	glm::vec3 dRGBdx(0, 0, 0);
-	glm::vec3 dRGBdy(0, 0, 0);
-	glm::vec3 dRGBdz(0, 0, 0);
-	float x = dir.x;
-	float y = dir.y;
-	float z = dir.z;
-
-	// Target location for this Gaussian to write SH gradients to
-	glm::vec3* dL_dsh = dL_dshs + idx * max_coeffs;
-
-	// No tricks here, just high school-level calculus.
-	float dRGBdsh0 = SH_C0;
-	dL_dsh[0] = dRGBdsh0 * dL_dRGB;
-	if (deg > 0)
-	{
-		float dRGBdsh1 = -SH_C1 * y;
-		float dRGBdsh2 = SH_C1 * z;
-		float dRGBdsh3 = -SH_C1 * x;
-		dL_dsh[1] = dRGBdsh1 * dL_dRGB;
-		dL_dsh[2] = dRGBdsh2 * dL_dRGB;
-		dL_dsh[3] = dRGBdsh3 * dL_dRGB;
-
-		dRGBdx = -SH_C1 * sh[3];
-		dRGBdy = -SH_C1 * sh[1];
-		dRGBdz = SH_C1 * sh[2];
-
-		if (deg > 1)
-		{
-			float xx = x * x, yy = y * y, zz = z * z;
-			float xy = x * y, yz = y * z, xz = x * z;
-
-			float dRGBdsh4 = SH_C2[0] * xy;
-			float dRGBdsh5 = SH_C2[1] * yz;
-			float dRGBdsh6 = SH_C2[2] * (2.f * zz - xx - yy);
-			float dRGBdsh7 = SH_C2[3] * xz;
-			float dRGBdsh8 = SH_C2[4] * (xx - yy);
-			dL_dsh[4] = dRGBdsh4 * dL_dRGB;
-			dL_dsh[5] = dRGBdsh5 * dL_dRGB;
-			dL_dsh[6] = dRGBdsh6 * dL_dRGB;
-			dL_dsh[7] = dRGBdsh7 * dL_dRGB;
-			dL_dsh[8] = dRGBdsh8 * dL_dRGB;
-
-			dRGBdx += SH_C2[0] * y * sh[4] + SH_C2[2] * 2.f * -x * sh[6] + SH_C2[3] * z * sh[7] + SH_C2[4] * 2.f * x * sh[8];
-			dRGBdy += SH_C2[0] * x * sh[4] + SH_C2[1] * z * sh[5] + SH_C2[2] * 2.f * -y * sh[6] + SH_C2[4] * 2.f * -y * sh[8];
-			dRGBdz += SH_C2[1] * y * sh[5] + SH_C2[2] * 2.f * 2.f * z * sh[6] + SH_C2[3] * x * sh[7];
-
-			if (deg > 2)
-			{
-				float dRGBdsh9 = SH_C3[0] * y * (3.f * xx - yy);
-				float dRGBdsh10 = SH_C3[1] * xy * z;
-				float dRGBdsh11 = SH_C3[2] * y * (4.f * zz - xx - yy);
-				float dRGBdsh12 = SH_C3[3] * z * (2.f * zz - 3.f * xx - 3.f * yy);
-				float dRGBdsh13 = SH_C3[4] * x * (4.f * zz - xx - yy);
-				float dRGBdsh14 = SH_C3[5] * z * (xx - yy);
-				float dRGBdsh15 = SH_C3[6] * x * (xx - 3.f * yy);
-				dL_dsh[9] = dRGBdsh9 * dL_dRGB;
-				dL_dsh[10] = dRGBdsh10 * dL_dRGB;
-				dL_dsh[11] = dRGBdsh11 * dL_dRGB;
-				dL_dsh[12] = dRGBdsh12 * dL_dRGB;
-				dL_dsh[13] = dRGBdsh13 * dL_dRGB;
-				dL_dsh[14] = dRGBdsh14 * dL_dRGB;
-				dL_dsh[15] = dRGBdsh15 * dL_dRGB;
-
-				dRGBdx += (
-					SH_C3[0] * sh[9] * 3.f * 2.f * xy +
-					SH_C3[1] * sh[10] * yz +
-					SH_C3[2] * sh[11] * -2.f * xy +
-					SH_C3[3] * sh[12] * -3.f * 2.f * xz +
-					SH_C3[4] * sh[13] * (-3.f * xx + 4.f * zz - yy) +
-					SH_C3[5] * sh[14] * 2.f * xz +
-					SH_C3[6] * sh[15] * 3.f * (xx - yy));
-
-				dRGBdy += (
-					SH_C3[0] * sh[9] * 3.f * (xx - yy) +
-					SH_C3[1] * sh[10] * xz +
-					SH_C3[2] * sh[11] * (-3.f * yy + 4.f * zz - xx) +
-					SH_C3[3] * sh[12] * -3.f * 2.f * yz +
-					SH_C3[4] * sh[13] * -2.f * xy +
-					SH_C3[5] * sh[14] * -2.f * yz +
-					SH_C3[6] * sh[15] * -3.f * 2.f * xy);
-
-				dRGBdz += (
-					SH_C3[1] * sh[10] * xy +
-					SH_C3[2] * sh[11] * 4.f * 2.f * yz +
-					SH_C3[3] * sh[12] * 3.f * (2.f * zz - xx - yy) +
-					SH_C3[4] * sh[13] * 4.f * 2.f * xz +
-					SH_C3[5] * sh[14] * (xx - yy));
-			}
-		}
-	}
-
-	// The view direction is an input to the computation. View direction
-	// is influenced by the Gaussian's mean, so SHs gradients
-	// must propagate back into 3D position.
-	glm::vec3 dL_ddir(glm::dot(dRGBdx, dL_dRGB), glm::dot(dRGBdy, dL_dRGB), glm::dot(dRGBdz, dL_dRGB));
-
-	// Account for normalization of direction
-	float3 dL_dmean = dnormvdv(float3{ dir_orig.x, dir_orig.y, dir_orig.z }, float3{ dL_ddir.x, dL_ddir.y, dL_ddir.z });
-
-	// Gradients of loss w.r.t. Gaussian means, but only the portion 
-	// that is caused because the mean affects the view-dependent color.
-	// Additional mean gradient is accumulated in below methods.
-	dL_dmeans[idx] += glm::vec3(dL_dmean.x, dL_dmean.y, dL_dmean.z);
-}
-
-// Backward version of INVERSE 2D covariance matrix computation
-// (due to length launched as separate kernel before other 
-// backward steps contained in preprocess)
-__global__ void computeCov2DCUDA(int P,
-	const float3* means,
-	const int* radii,
-	const float* cov3Ds,
-	const float h_x, float h_y,
-	const float tan_fovx, float tan_fovy,
-	const float* view_matrix,
-	const float* dL_dconics,
-	float3* dL_dmeans,
-	float* dL_dcov)
-{
-	auto idx = cg::this_grid().thread_rank();
-	if (idx >= P || !(radii[idx] > 0))
-		return;
-
-	// Reading location of 3D covariance for this Gaussian
-	const float* cov3D = cov3Ds + 6 * idx;
-
-	// Fetch gradients, recompute 2D covariance and relevant 
-	// intermediate forward results needed in the backward.
-	float3 mean = means[idx];
-	float3 dL_dconic = { dL_dconics[4 * idx], dL_dconics[4 * idx + 1], dL_dconics[4 * idx + 3] };
-	float3 t = transformPoint4x3(mean, view_matrix);
-	
-	const float limx = 1.3f * tan_fovx;
-	const float limy = 1.3f * tan_fovy;
-	const float txtz = t.x / t.z;
-	const float tytz = t.y / t.z;
-	t.x = min(limx, max(-limx, txtz)) * t.z;
-	t.y = min(limy, max(-limy, tytz)) * t.z;
-	
-	const float x_grad_mul = txtz < -limx || txtz > limx ? 0 : 1;
-	const float y_grad_mul = tytz < -limy || tytz > limy ? 0 : 1;
-
-	glm::mat3 J = glm::mat3(h_x / t.z, 0.0f, -(h_x * t.x) / (t.z * t.z),
-		0.0f, h_y / t.z, -(h_y * t.y) / (t.z * t.z),
-		0, 0, 0);
-
-	glm::mat3 W = glm::mat3(
-		view_matrix[0], view_matrix[4], view_matrix[8],
-		view_matrix[1], view_matrix[5], view_matrix[9],
-		view_matrix[2], view_matrix[6], view_matrix[10]);
-
-	glm::mat3 Vrk = glm::mat3(
-		cov3D[0], cov3D[1], cov3D[2],
-		cov3D[1], cov3D[3], cov3D[4],
-		cov3D[2], cov3D[4], cov3D[5]);
-
-	glm::mat3 T = W * J;
-
-	glm::mat3 cov2D = glm::transpose(T) * glm::transpose(Vrk) * T;
-
-	// Use helper variables for 2D covariance entries. More compact.
-	float a = cov2D[0][0] += 0.3f;
-	float b = cov2D[0][1];
-	float c = cov2D[1][1] += 0.3f;
-
-	float denom = a * c - b * b;
-	float dL_da = 0, dL_db = 0, dL_dc = 0;
-	float denom2inv = 1.0f / ((denom * denom) + 0.0000001f);
-
-	if (denom2inv != 0)
-	{
-		// Gradients of loss w.r.t. entries of 2D covariance matrix,
-		// given gradients of loss w.r.t. conic matrix (inverse covariance matrix).
-		// e.g., dL / da = dL / d_conic_a * d_conic_a / d_a
-		dL_da = denom2inv * (-c * c * dL_dconic.x + 2 * b * c * dL_dconic.y + (denom - a * c) * dL_dconic.z);
-		dL_dc = denom2inv * (-a * a * dL_dconic.z + 2 * a * b * dL_dconic.y + (denom - a * c) * dL_dconic.x);
-		dL_db = denom2inv * 2 * (b * c * dL_dconic.x - (denom + 2 * b * b) * dL_dconic.y + a * b * dL_dconic.z);
-
-		// Gradients of loss L w.r.t. each 3D covariance matrix (Vrk) entry, 
-		// given gradients w.r.t. 2D covariance matrix (diagonal).
-		// cov2D = transpose(T) * transpose(Vrk) * T;
-		dL_dcov[6 * idx + 0] = (T[0][0] * T[0][0] * dL_da + T[0][0] * T[1][0] * dL_db + T[1][0] * T[1][0] * dL_dc);
-		dL_dcov[6 * idx + 3] = (T[0][1] * T[0][1] * dL_da + T[0][1] * T[1][1] * dL_db + T[1][1] * T[1][1] * dL_dc);
-		dL_dcov[6 * idx + 5] = (T[0][2] * T[0][2] * dL_da + T[0][2] * T[1][2] * dL_db + T[1][2] * T[1][2] * dL_dc);
-
-		// Gradients of loss L w.r.t. each 3D covariance matrix (Vrk) entry, 
-		// given gradients w.r.t. 2D covariance matrix (off-diagonal).
-		// Off-diagonal elements appear twice --> double the gradient.
-		// cov2D = transpose(T) * transpose(Vrk) * T;
-		dL_dcov[6 * idx + 1] = 2 * T[0][0] * T[0][1] * dL_da + (T[0][0] * T[1][1] + T[0][1] * T[1][0]) * dL_db + 2 * T[1][0] * T[1][1] * dL_dc;
-		dL_dcov[6 * idx + 2] = 2 * T[0][0] * T[0][2] * dL_da + (T[0][0] * T[1][2] + T[0][2] * T[1][0]) * dL_db + 2 * T[1][0] * T[1][2] * dL_dc;
-		dL_dcov[6 * idx + 4] = 2 * T[0][2] * T[0][1] * dL_da + (T[0][1] * T[1][2] + T[0][2] * T[1][1]) * dL_db + 2 * T[1][1] * T[1][2] * dL_dc;
-	}
-	else
-	{
-		for (int i = 0; i < 6; i++)
-			dL_dcov[6 * idx + i] = 0;
-	}
-
-	// Gradients of loss w.r.t. upper 2x3 portion of intermediate matrix T
-	// cov2D = transpose(T) * transpose(Vrk) * T;
-	float dL_dT00 = 2 * (T[0][0] * Vrk[0][0] + T[0][1] * Vrk[0][1] + T[0][2] * Vrk[0][2]) * dL_da +
-		(T[1][0] * Vrk[0][0] + T[1][1] * Vrk[0][1] + T[1][2] * Vrk[0][2]) * dL_db;
-	float dL_dT01 = 2 * (T[0][0] * Vrk[1][0] + T[0][1] * Vrk[1][1] + T[0][2] * Vrk[1][2]) * dL_da +
-		(T[1][0] * Vrk[1][0] + T[1][1] * Vrk[1][1] + T[1][2] * Vrk[1][2]) * dL_db;
-	float dL_dT02 = 2 * (T[0][0] * Vrk[2][0] + T[0][1] * Vrk[2][1] + T[0][2] * Vrk[2][2]) * dL_da +
-		(T[1][0] * Vrk[2][0] + T[1][1] * Vrk[2][1] + T[1][2] * Vrk[2][2]) * dL_db;
-	float dL_dT10 = 2 * (T[1][0] * Vrk[0][0] + T[1][1] * Vrk[0][1] + T[1][2] * Vrk[0][2]) * dL_dc +
-		(T[0][0] * Vrk[0][0] + T[0][1] * Vrk[0][1] + T[0][2] * Vrk[0][2]) * dL_db;
-	float dL_dT11 = 2 * (T[1][0] * Vrk[1][0] + T[1][1] * Vrk[1][1] + T[1][2] * Vrk[1][2]) * dL_dc +
-		(T[0][0] * Vrk[1][0] + T[0][1] * Vrk[1][1] + T[0][2] * Vrk[1][2]) * dL_db;
-	float dL_dT12 = 2 * (T[1][0] * Vrk[2][0] + T[1][1] * Vrk[2][1] + T[1][2] * Vrk[2][2]) * dL_dc +
-		(T[0][0] * Vrk[2][0] + T[0][1] * Vrk[2][1] + T[0][2] * Vrk[2][2]) * dL_db;
-
-	// Gradients of loss w.r.t. upper 3x2 non-zero entries of Jacobian matrix
-	// T = W * J
-	float dL_dJ00 = W[0][0] * dL_dT00 + W[0][1] * dL_dT01 + W[0][2] * dL_dT02;
-	float dL_dJ02 = W[2][0] * dL_dT00 + W[2][1] * dL_dT01 + W[2][2] * dL_dT02;
-	float dL_dJ11 = W[1][0] * dL_dT10 + W[1][1] * dL_dT11 + W[1][2] * dL_dT12;
-	float dL_dJ12 = W[2][0] * dL_dT10 + W[2][1] * dL_dT11 + W[2][2] * dL_dT12;
-
-	float tz = 1.f / t.z;
-	float tz2 = tz * tz;
-	float tz3 = tz2 * tz;
-
-	// Gradients of loss w.r.t. transformed Gaussian mean t
-	float dL_dtx = x_grad_mul * -h_x * tz2 * dL_dJ02;
-	float dL_dty = y_grad_mul * -h_y * tz2 * dL_dJ12;
-	float dL_dtz = -h_x * tz2 * dL_dJ00 - h_y * tz2 * dL_dJ11 + (2 * h_x * t.x) * tz3 * dL_dJ02 + (2 * h_y * t.y) * tz3 * dL_dJ12;
-
-	// Account for transformation of mean to t
-	// t = transformPoint4x3(mean, view_matrix);
-	float3 dL_dmean = transformVec4x3Transpose({ dL_dtx, dL_dty, dL_dtz }, view_matrix);
-
-	// Gradients of loss w.r.t. Gaussian means, but only the portion 
-	// that is caused because the mean affects the covariance matrix.
-	// Additional mean gradient is accumulated in BACKWARD::preprocess.
-	dL_dmeans[idx] = dL_dmean;
-}
-
-// Backward pass for the conversion of scale and rotation to a 
-// 3D covariance matrix for each Gaussian. 
-__device__ void computeCov3D(int idx, const glm::vec3 scale, float mod, const glm::vec4 rot, const float* dL_dcov3Ds, glm::vec3* dL_dscales, glm::vec4* dL_drots)
-{
-	// Recompute (intermediate) results for the 3D covariance computation.
-	glm::vec4 q = rot;// / glm::length(rot);
-	float r = q.x;
-	float x = q.y;
-	float y = q.z;
-	float z = q.w;
-
-	glm::mat3 R = glm::mat3(
-		1.f - 2.f * (y * y + z * z), 2.f * (x * y - r * z), 2.f * (x * z + r * y),
-		2.f * (x * y + r * z), 1.f - 2.f * (x * x + z * z), 2.f * (y * z - r * x),
-		2.f * (x * z - r * y), 2.f * (y * z + r * x), 1.f - 2.f * (x * x + y * y)
-	);
-
-	glm::mat3 S = glm::mat3(1.0f);
-
-	glm::vec3 s = mod * scale;
-	S[0][0] = s.x;
-	S[1][1] = s.y;
-	S[2][2] = s.z;
-
-	glm::mat3 M = S * R;
-
-	const float* dL_dcov3D = dL_dcov3Ds + 6 * idx;
-
-	glm::vec3 dunc(dL_dcov3D[0], dL_dcov3D[3], dL_dcov3D[5]);
-	glm::vec3 ounc = 0.5f * glm::vec3(dL_dcov3D[1], dL_dcov3D[2], dL_dcov3D[4]);
-
-	// Convert per-element covariance loss gradients to matrix form
-	glm::mat3 dL_dSigma = glm::mat3(
-		dL_dcov3D[0], 0.5f * dL_dcov3D[1], 0.5f * dL_dcov3D[2],
-		0.5f * dL_dcov3D[1], dL_dcov3D[3], 0.5f * dL_dcov3D[4],
-		0.5f * dL_dcov3D[2], 0.5f * dL_dcov3D[4], dL_dcov3D[5]
-	);
-
-	// Compute loss gradient w.r.t. matrix M
-	// dSigma_dM = 2 * M
-	glm::mat3 dL_dM = 2.0f * M * dL_dSigma;
-
-	glm::mat3 Rt = glm::transpose(R);
-	glm::mat3 dL_dMt = glm::transpose(dL_dM);
-
-	// Gradients of loss w.r.t. scale
-	glm::vec3* dL_dscale = dL_dscales + idx;
-	dL_dscale->x = glm::dot(Rt[0], dL_dMt[0]);
-	dL_dscale->y = glm::dot(Rt[1], dL_dMt[1]);
-	dL_dscale->z = glm::dot(Rt[2], dL_dMt[2]);
-
-	dL_dMt[0] *= s.x;
-	dL_dMt[1] *= s.y;
-	dL_dMt[2] *= s.z;
-
-	// Gradients of loss w.r.t. normalized quaternion
-	glm::vec4 dL_dq;
-	dL_dq.x = 2 * z * (dL_dMt[0][1] - dL_dMt[1][0]) + 2 * y * (dL_dMt[2][0] - dL_dMt[0][2]) + 2 * x * (dL_dMt[1][2] - dL_dMt[2][1]);
-	dL_dq.y = 2 * y * (dL_dMt[1][0] + dL_dMt[0][1]) + 2 * z * (dL_dMt[2][0] + dL_dMt[0][2]) + 2 * r * (dL_dMt[1][2] - dL_dMt[2][1]) - 4 * x * (dL_dMt[2][2] + dL_dMt[1][1]);
-	dL_dq.z = 2 * x * (dL_dMt[1][0] + dL_dMt[0][1]) + 2 * r * (dL_dMt[2][0] - dL_dMt[0][2]) + 2 * z * (dL_dMt[1][2] + dL_dMt[2][1]) - 4 * y * (dL_dMt[2][2] + dL_dMt[0][0]);
-	dL_dq.w = 2 * r * (dL_dMt[0][1] - dL_dMt[1][0]) + 2 * x * (dL_dMt[2][0] + dL_dMt[0][2]) + 2 * y * (dL_dMt[1][2] + dL_dMt[2][1]) - 4 * z * (dL_dMt[1][1] + dL_dMt[0][0]);
-
-	// Gradients of loss w.r.t. unnormalized quaternion
-	float4* dL_drot = (float4*)(dL_drots + idx);
-	*dL_drot = float4{ dL_dq.x, dL_dq.y, dL_dq.z, dL_dq.w };//dnormvdv(float4{ rot.x, rot.y, rot.z, rot.w }, float4{ dL_dq.x, dL_dq.y, dL_dq.z, dL_dq.w });
-}
-
-// Backward pass of the preprocessing steps, except
-// for the covariance computation and inversion
-// (those are handled by a previous kernel call)
-template<int C>
-__global__ void preprocessCUDA(
-	int P, int D, int M,
-	const float3* means,
-	const int* radii,
-	const float* shs,
-	const bool* clamped,
-	const glm::vec3* scales,
-	const glm::vec4* rotations,
-	const float scale_modifier,
-	const float* view,
-	const float* proj,
-	const glm::vec3* campos,
-	const float3* dL_dmean2D,
-	glm::vec3* dL_dmeans,
-	float* dL_dcolor,
-	float* dL_ddepth,
-	float* dL_dcov3D,
-	float* dL_dsh,
-	glm::vec3* dL_dscale,
-	glm::vec4* dL_drot)
-{
-	auto idx = cg::this_grid().thread_rank();
-	if (idx >= P || !(radii[idx] > 0))
-		return;
-
-	float3 m = means[idx];
-
-	// Taking care of gradients from the screenspace points
-	float4 m_hom = transformPoint4x4(m, proj);
-	float m_w = 1.0f / (m_hom.w + 0.0000001f);
-
-	// Compute loss gradient w.r.t. 3D means due to gradients of 2D means
-	// from rendering procedure
-	glm::vec3 dL_dmean;
-	float mul1 = (proj[0] * m.x + proj[4] * m.y + proj[8] * m.z + proj[12]) * m_w * m_w;
-	float mul2 = (proj[1] * m.x + proj[5] * m.y + proj[9] * m.z + proj[13]) * m_w * m_w;
-	dL_dmean.x = (proj[0] * m_w - proj[3] * mul1) * dL_dmean2D[idx].x + (proj[1] * m_w - proj[3] * mul2) * dL_dmean2D[idx].y;
-	dL_dmean.y = (proj[4] * m_w - proj[7] * mul1) * dL_dmean2D[idx].x + (proj[5] * m_w - proj[7] * mul2) * dL_dmean2D[idx].y;
-	dL_dmean.z = (proj[8] * m_w - proj[11] * mul1) * dL_dmean2D[idx].x + (proj[9] * m_w - proj[11] * mul2) * dL_dmean2D[idx].y;
-
-	// That's the second part of the mean gradient. Previous computation
-	// of cov2D and following SH conversion also affects it.
-	dL_dmeans[idx] += dL_dmean;
-
-	// the w must be equal to 1 for view^T * [x,y,z,1]
-	float3 m_view = transformPoint4x3(m, view);
-
-	// Compute loss gradient w.r.t. 3D means due to gradients of depth
-	// from rendering procedure
-	glm::vec3 dL_dmean2;
-	float mul3 = view[2] * m.x + view[6] * m.y + view[10] * m.z + view[14];
-	dL_dmean2.x = (view[2] - view[3] * mul3) * dL_ddepth[idx];
-	dL_dmean2.y = (view[6] - view[7] * mul3) * dL_ddepth[idx];
-	dL_dmean2.z = (view[10] - view[11] * mul3) * dL_ddepth[idx];
-
-	// That's the third part of the mean gradient.
-	dL_dmeans[idx] += dL_dmean2;
-
-	// Compute gradient updates due to computing colors from SHs
-	if (shs)
-		computeColorFromSH(idx, D, M, (glm::vec3*)means, *campos, shs, clamped, (glm::vec3*)dL_dcolor, (glm::vec3*)dL_dmeans, (glm::vec3*)dL_dsh);
-
-	// Compute gradient updates due to computing covariance from scale/rotation
-	if (scales)
-		computeCov3D(idx, scales[idx], scale_modifier, rotations[idx], dL_dcov3D, dL_dscale, dL_drot);
-}
-
-// Backward version of the rendering procedure.
-template <uint32_t C>
-__global__ void __launch_bounds__(BLOCK_X * BLOCK_Y)
-renderCUDA(
-	const uint2* __restrict__ ranges,
-	const uint32_t* __restrict__ point_list,
-	int W, int H,
-	const float* __restrict__ bg_color,
-	const float2* __restrict__ points_xy_image,
-	const float4* __restrict__ conic_opacity,
-	const float* __restrict__ colors,
-	const float* __restrict__ depths,
-	const float* __restrict__ alphas,
-	const uint32_t* __restrict__ n_contrib,
-	const float* __restrict__ dL_dpixels,
-	const float* __restrict__ dL_dpixel_depths,
-	const float* __restrict__ dL_dalphas,
-	float3* __restrict__ dL_dmean2D,
-	float4* __restrict__ dL_dconic2D,
-	float* __restrict__ dL_dopacity,
-	float* __restrict__ dL_dcolors,
-	float* __restrict__ dL_ddepths
-)
-{
-	// We rasterize again. Compute necessary block info.
-	auto block = cg::this_thread_block();
-	const uint32_t horizontal_blocks = (W + BLOCK_X - 1) / BLOCK_X;
-	const uint2 pix_min = { block.group_index().x * BLOCK_X, block.group_index().y * BLOCK_Y };
-	const uint2 pix_max = { min(pix_min.x + BLOCK_X, W), min(pix_min.y + BLOCK_Y , H) };
-	const uint2 pix = { pix_min.x + block.thread_index().x, pix_min.y + block.thread_index().y };
-	const uint32_t pix_id = W * pix.y + pix.x;
-	const float2 pixf = { (float)pix.x, (float)pix.y };
-
-	const bool inside = pix.x < W&& pix.y < H;
-	const uint2 range = ranges[block.group_index().y * horizontal_blocks + block.group_index().x];
-
-	const int rounds = ((range.y - range.x + BLOCK_SIZE - 1) / BLOCK_SIZE);
-
-	bool done = !inside;
-	int toDo = range.y - range.x;
-
-	__shared__ int collected_id[BLOCK_SIZE];
-	__shared__ float2 collected_xy[BLOCK_SIZE];
-	__shared__ float4 collected_conic_opacity[BLOCK_SIZE];
-	__shared__ float collected_colors[C * BLOCK_SIZE];
-	__shared__ float collected_depths[BLOCK_SIZE];
-
-	// In the forward, we stored the final value for T, the
-	// product of all (1 - alpha) factors. 
-	const float T_final = inside ? (1 - alphas[pix_id]) : 0;
-	float T = T_final;
-
-	// We start from the back. The ID of the last contributing
-	// Gaussian is known from each pixel from the forward.
-	uint32_t contributor = toDo;
-	const int last_contributor = inside ? n_contrib[pix_id] : 0;
-
-	float accum_rec[C] = { 0 };
-	float dL_dpixel[C];
-	float accum_depth_rec = 0;
-	float dL_dpixel_depth;
-	float accum_alpha_rec = 0;
-	float dL_dalpha;
-	if (inside) {
-		for (int i = 0; i < C; i++)
-			dL_dpixel[i] = dL_dpixels[i * H * W + pix_id];
-		dL_dpixel_depth = dL_dpixel_depths[pix_id];
-		dL_dalpha = dL_dalphas[pix_id];
-	}
-
-	float last_alpha = 0;
-	float last_color[C] = { 0 };
-	float last_depth = 0;
-
-	// Gradient of pixel coordinate w.r.t. normalized 
-	// screen-space viewport corrdinates (-1 to 1)
-	const float ddelx_dx = 0.5 * W;
-	const float ddely_dy = 0.5 * H;
-
-	// Traverse all Gaussians
-	for (int i = 0; i < rounds; i++, toDo -= BLOCK_SIZE)
-	{
-		// Load auxiliary data into shared memory, start in the BACK
-		// and load them in revers order.
-		block.sync();
-		const int progress = i * BLOCK_SIZE + block.thread_rank();
-		if (range.x + progress < range.y)
-		{
-			const int coll_id = point_list[range.y - progress - 1];
-			collected_id[block.thread_rank()] = coll_id;
-			collected_xy[block.thread_rank()] = points_xy_image[coll_id];
-			collected_conic_opacity[block.thread_rank()] = conic_opacity[coll_id];
-			for (int i = 0; i < C; i++)
-				collected_colors[i * BLOCK_SIZE + block.thread_rank()] = colors[coll_id * C + i];
-			collected_depths[block.thread_rank()] = depths[coll_id];
-		}
-		block.sync();
-
-		// Iterate over Gaussians
-		for (int j = 0; !done && j < min(BLOCK_SIZE, toDo); j++)
-		{
-			// Keep track of current Gaussian ID. Skip, if this one
-			// is behind the last contributor for this pixel.
-			contributor--;
-			if (contributor >= last_contributor)
-				continue;
-
-			// Compute blending values, as before.
-			const float2 xy = collected_xy[j];
-			const float2 d = { xy.x - pixf.x, xy.y - pixf.y };
-			const float4 con_o = collected_conic_opacity[j];
-			const float power = -0.5f * (con_o.x * d.x * d.x + con_o.z * d.y * d.y) - con_o.y * d.x * d.y;
-			if (power > 0.0f)
-				continue;
-
-			const float G = exp(power);
-			const float alpha = min(0.99f, con_o.w * G);
-			if (alpha < 1.0f / 255.0f)
-				continue;
-
-			T = T / (1.f - alpha);
-			const float dchannel_dcolor = alpha * T;
-			const float dpixel_depth_ddepth = alpha * T;
-
-			// Propagate gradients to per-Gaussian colors and keep
-			// gradients w.r.t. alpha (blending factor for a Gaussian/pixel
-			// pair).
-			float dL_dopa = 0.0f;
-			const int global_id = collected_id[j];
-			for (int ch = 0; ch < C; ch++)
-			{
-				const float c = collected_colors[ch * BLOCK_SIZE + j];
-				// Update last color (to be used in the next iteration)
-				accum_rec[ch] = last_alpha * last_color[ch] + (1.f - last_alpha) * accum_rec[ch];
-				last_color[ch] = c;
-
-				const float dL_dchannel = dL_dpixel[ch];
-				dL_dopa += (c - accum_rec[ch]) * dL_dchannel;
-				// Update the gradients w.r.t. color of the Gaussian. 
-				// Atomic, since this pixel is just one of potentially
-				// many that were affected by this Gaussian.
-				atomicAdd(&(dL_dcolors[global_id * C + ch]), dchannel_dcolor * dL_dchannel);
-			}
-			
-			// Propagate gradients from pixel depth to opacity
-			const float c_d = collected_depths[j];
-			accum_depth_rec = last_alpha * last_depth + (1.f - last_alpha) * accum_depth_rec;
-			last_depth = c_d;
-			dL_dopa += (c_d - accum_depth_rec) * dL_dpixel_depth;
-			atomicAdd(&(dL_ddepths[global_id]), dpixel_depth_ddepth * dL_dpixel_depth);
-
-			// Propagate gradients from pixel alpha (weights_sum) to opacity
-			accum_alpha_rec = last_alpha + (1.f - last_alpha) * accum_alpha_rec;
-			dL_dopa += (1 - accum_alpha_rec) * dL_dalpha; //- (alpha - accum_alpha_rec) * dL_dalpha;
-
-			dL_dopa *= T;
-			// Update last alpha (to be used in the next iteration)
-			last_alpha = alpha;
-
-			// Account for fact that alpha also influences how much of
-			// the background color is added if nothing left to blend
-			float bg_dot_dpixel = 0;
-			for (int i = 0; i < C; i++)
-				bg_dot_dpixel += bg_color[i] * dL_dpixel[i];
-			dL_dopa += (-T_final / (1.f - alpha)) * bg_dot_dpixel;
-
-
-			// Helpful reusable temporary variables
-			const float dL_dG = con_o.w * dL_dopa;
-			const float gdx = G * d.x;
-			const float gdy = G * d.y;
-			const float dG_ddelx = -gdx * con_o.x - gdy * con_o.y;
-			const float dG_ddely = -gdy * con_o.z - gdx * con_o.y;
-
-			// Update gradients w.r.t. 2D mean position of the Gaussian
-			atomicAdd(&dL_dmean2D[global_id].x, dL_dG * dG_ddelx * ddelx_dx);
-			atomicAdd(&dL_dmean2D[global_id].y, dL_dG * dG_ddely * ddely_dy);
-
-			// Update gradients w.r.t. 2D covariance (2x2 matrix, symmetric)
-			atomicAdd(&dL_dconic2D[global_id].x, -0.5f * gdx * d.x * dL_dG);
-			atomicAdd(&dL_dconic2D[global_id].y, -0.5f * gdx * d.y * dL_dG);
-			atomicAdd(&dL_dconic2D[global_id].w, -0.5f * gdy * d.y * dL_dG);
-
-			// Update gradients w.r.t. opacity of the Gaussian
-			atomicAdd(&(dL_dopacity[global_id]), G * dL_dopa);
-		}
-	}
-}
-
-void BACKWARD::preprocess(
-	int P, int D, int M,
-	const float3* means3D,
-	const int* radii,
-	const float* shs,
-	const bool* clamped,
-	const glm::vec3* scales,
-	const glm::vec4* rotations,
-	const float scale_modifier,
-	const float* cov3Ds,
-	const float* viewmatrix,
-	const float* projmatrix,
-	const float focal_x, float focal_y,
-	const float tan_fovx, float tan_fovy,
-	const glm::vec3* campos,
-	const float3* dL_dmean2D,
-	const float* dL_dconic,
-	glm::vec3* dL_dmean3D,
-	float* dL_dcolor,
-	float* dL_ddepth,
-	float* dL_dcov3D,
-	float* dL_dsh,
-	glm::vec3* dL_dscale,
-	glm::vec4* dL_drot)
-{
-	// Propagate gradients for the path of 2D conic matrix computation. 
-	// Somewhat long, thus it is its own kernel rather than being part of 
-	// "preprocess". When done, loss gradient w.r.t. 3D means has been
-	// modified and gradient w.r.t. 3D covariance matrix has been computed.	
-	computeCov2DCUDA << <(P + 255) / 256, 256 >> > (
-		P,
-		means3D,
-		radii,
-		cov3Ds,
-		focal_x,
-		focal_y,
-		tan_fovx,
-		tan_fovy,
-		viewmatrix,
-		dL_dconic,
-		(float3*)dL_dmean3D,
-		dL_dcov3D);
-
-	// Propagate gradients for remaining steps: finish 3D mean gradients,
-	// propagate color gradients to SH (if desireD), propagate 3D covariance
-	// matrix gradients to scale and rotation.
-	preprocessCUDA<NUM_CHANNELS> << < (P + 255) / 256, 256 >> > (
-		P, D, M,
-		(float3*)means3D,
-		radii,
-		shs,
-		clamped,
-		(glm::vec3*)scales,
-		(glm::vec4*)rotations,
-		scale_modifier,
-		viewmatrix,
-		projmatrix,
-		campos,
-		(float3*)dL_dmean2D,
-		(glm::vec3*)dL_dmean3D,
-		dL_dcolor,
-		dL_ddepth,
-		dL_dcov3D,
-		dL_dsh,
-		dL_dscale,
-		dL_drot);
-}
-
-void BACKWARD::render(
-	const dim3 grid, const dim3 block,
-	const uint2* ranges,
-	const uint32_t* point_list,
-	int W, int H,
-	const float* bg_color,
-	const float2* means2D,
-	const float4* conic_opacity,
-	const float* colors,
-	const float* depths,
-	const float* alphas,
-	const uint32_t* n_contrib,
-	const float* dL_dpixels,
-	const float* dL_dpixel_depths,
-	const float* dL_dalphas,
-	float3* dL_dmean2D,
-	float4* dL_dconic2D,
-	float* dL_dopacity,
-	float* dL_dcolors,
-	float* dL_ddepths)
-{
-	renderCUDA<NUM_CHANNELS> << <grid, block >> >(
-		ranges,
-		point_list,
-		W, H,
-		bg_color,
-		means2D,
-		conic_opacity,
-		colors,
-		depths,
-		alphas,
-		n_contrib,
-		dL_dpixels,
-		dL_dpixel_depths,
-		dL_dalphas,
-		dL_dmean2D,
-		dL_dconic2D,
-		dL_dopacity,
-		dL_dcolors,
-		dL_ddepths
-		);
-}

diff-gaussian-rasterization/cuda_rasterizer/backward.h

@@ -1,70 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#ifndef CUDA_RASTERIZER_BACKWARD_H_INCLUDED
-#define CUDA_RASTERIZER_BACKWARD_H_INCLUDED
-
-#include <cuda.h>
-#include "cuda_runtime.h"
-#include "device_launch_parameters.h"
-#define GLM_FORCE_CUDA
-#include <glm/glm.hpp>
-
-namespace BACKWARD
-{
-	void render(
-		const dim3 grid, dim3 block,
-		const uint2* ranges,
-		const uint32_t* point_list,
-		int W, int H,
-		const float* bg_color,
-		const float2* means2D,
-		const float4* conic_opacity,
-		const float* colors,
-		const float* depths,
-		const float* alphas,
-		const uint32_t* n_contrib,
-		const float* dL_dpixels,
-		const float* dL_dpixel_depths,
-		const float* dL_dalphas,
-		float3* dL_dmean2D,
-		float4* dL_dconic2D,
-		float* dL_dopacity,
-		float* dL_dcolors,
-		float* dL_ddepths);
-
-	void preprocess(
-		int P, int D, int M,
-		const float3* means,
-		const int* radii,
-		const float* shs,
-		const bool* clamped,
-		const glm::vec3* scales,
-		const glm::vec4* rotations,
-		const float scale_modifier,
-		const float* cov3Ds,
-		const float* view,
-		const float* proj,
-		const float focal_x, float focal_y,
-		const float tan_fovx, float tan_fovy,
-		const glm::vec3* campos,
-		const float3* dL_dmean2D,
-		const float* dL_dconics,
-		glm::vec3* dL_dmeans,
-		float* dL_dcolor,
-		float* dL_ddepth,
-		float* dL_dcov3D,
-		float* dL_dsh,
-		glm::vec3* dL_dscale,
-		glm::vec4* dL_drot);
-}
-
-#endif

diff-gaussian-rasterization/cuda_rasterizer/config.h

@@ -1,19 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#ifndef CUDA_RASTERIZER_CONFIG_H_INCLUDED
-#define CUDA_RASTERIZER_CONFIG_H_INCLUDED
-
-#define NUM_CHANNELS 3 // Default 3, RGB
-#define BLOCK_X 16
-#define BLOCK_Y 16
-
-#endif

diff-gaussian-rasterization/cuda_rasterizer/forward.cu

@@ -1,466 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#include "forward.h"
-#include "auxiliary.h"
-#include <cooperative_groups.h>
-#include <cooperative_groups/reduce.h>
-namespace cg = cooperative_groups;
-
-// Forward method for converting the input spherical harmonics
-// coefficients of each Gaussian to a simple RGB color.
-__device__ glm::vec3 computeColorFromSH(int idx, int deg, int max_coeffs, const glm::vec3* means, glm::vec3 campos, const float* shs, bool* clamped)
-{
-	// The implementation is loosely based on code for 
-	// "Differentiable Point-Based Radiance Fields for 
-	// Efficient View Synthesis" by Zhang et al. (2022)
-	glm::vec3 pos = means[idx];
-	glm::vec3 dir = pos - campos;
-	dir = dir / glm::length(dir);
-
-	glm::vec3* sh = ((glm::vec3*)shs) + idx * max_coeffs;
-	glm::vec3 result = SH_C0 * sh[0];
-
-	if (deg > 0)
-	{
-		float x = dir.x;
-		float y = dir.y;
-		float z = dir.z;
-		result = result - SH_C1 * y * sh[1] + SH_C1 * z * sh[2] - SH_C1 * x * sh[3];
-
-		if (deg > 1)
-		{
-			float xx = x * x, yy = y * y, zz = z * z;
-			float xy = x * y, yz = y * z, xz = x * z;
-			result = result +
-				SH_C2[0] * xy * sh[4] +
-				SH_C2[1] * yz * sh[5] +
-				SH_C2[2] * (2.0f * zz - xx - yy) * sh[6] +
-				SH_C2[3] * xz * sh[7] +
-				SH_C2[4] * (xx - yy) * sh[8];
-
-			if (deg > 2)
-			{
-				result = result +
-					SH_C3[0] * y * (3.0f * xx - yy) * sh[9] +
-					SH_C3[1] * xy * z * sh[10] +
-					SH_C3[2] * y * (4.0f * zz - xx - yy) * sh[11] +
-					SH_C3[3] * z * (2.0f * zz - 3.0f * xx - 3.0f * yy) * sh[12] +
-					SH_C3[4] * x * (4.0f * zz - xx - yy) * sh[13] +
-					SH_C3[5] * z * (xx - yy) * sh[14] +
-					SH_C3[6] * x * (xx - 3.0f * yy) * sh[15];
-			}
-		}
-	}
-	result += 0.5f;
-
-	// RGB colors are clamped to positive values. If values are
-	// clamped, we need to keep track of this for the backward pass.
-	clamped[3 * idx + 0] = (result.x < 0);
-	clamped[3 * idx + 1] = (result.y < 0);
-	clamped[3 * idx + 2] = (result.z < 0);
-	return glm::max(result, 0.0f);
-}
-
-// Forward version of 2D covariance matrix computation
-__device__ float3 computeCov2D(const float3& mean, float focal_x, float focal_y, float tan_fovx, float tan_fovy, const float* cov3D, const float* viewmatrix)
-{
-	// The following models the steps outlined by equations 29
-	// and 31 in "EWA Splatting" (Zwicker et al., 2002). 
-	// Additionally considers aspect / scaling of viewport.
-	// Transposes used to account for row-/column-major conventions.
-	float3 t = transformPoint4x3(mean, viewmatrix);
-
-	const float limx = 1.3f * tan_fovx;
-	const float limy = 1.3f * tan_fovy;
-	const float txtz = t.x / t.z;
-	const float tytz = t.y / t.z;
-	t.x = min(limx, max(-limx, txtz)) * t.z;
-	t.y = min(limy, max(-limy, tytz)) * t.z;
-
-	glm::mat3 J = glm::mat3(
-		focal_x / t.z, 0.0f, -(focal_x * t.x) / (t.z * t.z),
-		0.0f, focal_y / t.z, -(focal_y * t.y) / (t.z * t.z),
-		0, 0, 0);
-
-	glm::mat3 W = glm::mat3(
-		viewmatrix[0], viewmatrix[4], viewmatrix[8],
-		viewmatrix[1], viewmatrix[5], viewmatrix[9],
-		viewmatrix[2], viewmatrix[6], viewmatrix[10]);
-
-	glm::mat3 T = W * J;
-
-	glm::mat3 Vrk = glm::mat3(
-		cov3D[0], cov3D[1], cov3D[2],
-		cov3D[1], cov3D[3], cov3D[4],
-		cov3D[2], cov3D[4], cov3D[5]);
-
-	glm::mat3 cov = glm::transpose(T) * glm::transpose(Vrk) * T;
-
-	// Apply low-pass filter: every Gaussian should be at least
-	// one pixel wide/high. Discard 3rd row and column.
-	cov[0][0] += 0.3f;
-	cov[1][1] += 0.3f;
-	return { float(cov[0][0]), float(cov[0][1]), float(cov[1][1]) };
-}
-
-// Forward method for converting scale and rotation properties of each
-// Gaussian to a 3D covariance matrix in world space. Also takes care
-// of quaternion normalization.
-__device__ void computeCov3D(const glm::vec3 scale, float mod, const glm::vec4 rot, float* cov3D)
-{
-	// Create scaling matrix
-	glm::mat3 S = glm::mat3(1.0f);
-	S[0][0] = mod * scale.x;
-	S[1][1] = mod * scale.y;
-	S[2][2] = mod * scale.z;
-
-	// Normalize quaternion to get valid rotation
-	glm::vec4 q = rot;// / glm::length(rot);
-	float r = q.x;
-	float x = q.y;
-	float y = q.z;
-	float z = q.w;
-
-	// Compute rotation matrix from quaternion
-	glm::mat3 R = glm::mat3(
-		1.f - 2.f * (y * y + z * z), 2.f * (x * y - r * z), 2.f * (x * z + r * y),
-		2.f * (x * y + r * z), 1.f - 2.f * (x * x + z * z), 2.f * (y * z - r * x),
-		2.f * (x * z - r * y), 2.f * (y * z + r * x), 1.f - 2.f * (x * x + y * y)
-	);
-
-	glm::mat3 M = S * R;
-
-	// Compute 3D world covariance matrix Sigma
-	glm::mat3 Sigma = glm::transpose(M) * M;
-
-	// Covariance is symmetric, only store upper right
-	cov3D[0] = Sigma[0][0];
-	cov3D[1] = Sigma[0][1];
-	cov3D[2] = Sigma[0][2];
-	cov3D[3] = Sigma[1][1];
-	cov3D[4] = Sigma[1][2];
-	cov3D[5] = Sigma[2][2];
-}
-
-// Perform initial steps for each Gaussian prior to rasterization.
-template<int C>
-__global__ void preprocessCUDA(int P, int D, int M,
-	const float* orig_points,
-	const glm::vec3* scales,
-	const float scale_modifier,
-	const glm::vec4* rotations,
-	const float* opacities,
-	const float* shs,
-	bool* clamped,
-	const float* cov3D_precomp,
-	const float* colors_precomp,
-	const float* viewmatrix,
-	const float* projmatrix,
-	const glm::vec3* cam_pos,
-	const int W, int H,
-	const float tan_fovx, float tan_fovy,
-	const float focal_x, float focal_y,
-	int* radii,
-	float2* points_xy_image,
-	float* depths,
-	float* cov3Ds,
-	float* rgb,
-	float4* conic_opacity,
-	const dim3 grid,
-	uint32_t* tiles_touched,
-	bool prefiltered)
-{
-	auto idx = cg::this_grid().thread_rank();
-	if (idx >= P)
-		return;
-
-	// Initialize radius and touched tiles to 0. If this isn't changed,
-	// this Gaussian will not be processed further.
-	radii[idx] = 0;
-	tiles_touched[idx] = 0;
-
-	// Perform near culling, quit if outside.
-	float3 p_view;
-	if (!in_frustum(idx, orig_points, viewmatrix, projmatrix, prefiltered, p_view))
-		return;
-
-	// Transform point by projecting
-	float3 p_orig = { orig_points[3 * idx], orig_points[3 * idx + 1], orig_points[3 * idx + 2] };
-	float4 p_hom = transformPoint4x4(p_orig, projmatrix);
-	float p_w = 1.0f / (p_hom.w + 0.0000001f);
-	float3 p_proj = { p_hom.x * p_w, p_hom.y * p_w, p_hom.z * p_w };
-
-	// If 3D covariance matrix is precomputed, use it, otherwise compute
-	// from scaling and rotation parameters. 
-	const float* cov3D;
-	if (cov3D_precomp != nullptr)
-	{
-		cov3D = cov3D_precomp + idx * 6;
-	}
-	else
-	{
-		computeCov3D(scales[idx], scale_modifier, rotations[idx], cov3Ds + idx * 6);
-		cov3D = cov3Ds + idx * 6;
-	}
-
-	// Compute 2D screen-space covariance matrix
-	float3 cov = computeCov2D(p_orig, focal_x, focal_y, tan_fovx, tan_fovy, cov3D, viewmatrix);
-
-	// Invert covariance (EWA algorithm)
-	float det = (cov.x * cov.z - cov.y * cov.y);
-	if (det == 0.0f)
-		return;
-	float det_inv = 1.f / det;
-	float3 conic = { cov.z * det_inv, -cov.y * det_inv, cov.x * det_inv };
-
-	// Compute extent in screen space (by finding eigenvalues of
-	// 2D covariance matrix). Use extent to compute a bounding rectangle
-	// of screen-space tiles that this Gaussian overlaps with. Quit if
-	// rectangle covers 0 tiles. 
-	float mid = 0.5f * (cov.x + cov.z);
-	float lambda1 = mid + sqrt(max(0.1f, mid * mid - det));
-	float lambda2 = mid - sqrt(max(0.1f, mid * mid - det));
-	float my_radius = ceil(3.f * sqrt(max(lambda1, lambda2)));
-	float2 point_image = { ndc2Pix(p_proj.x, W), ndc2Pix(p_proj.y, H) };
-	uint2 rect_min, rect_max;
-	getRect(point_image, my_radius, rect_min, rect_max, grid);
-	if ((rect_max.x - rect_min.x) * (rect_max.y - rect_min.y) == 0)
-		return;
-
-	// If colors have been precomputed, use them, otherwise convert
-	// spherical harmonics coefficients to RGB color.
-	if (colors_precomp == nullptr)
-	{
-		glm::vec3 result = computeColorFromSH(idx, D, M, (glm::vec3*)orig_points, *cam_pos, shs, clamped);
-		rgb[idx * C + 0] = result.x;
-		rgb[idx * C + 1] = result.y;
-		rgb[idx * C + 2] = result.z;
-	}
-
-	// Store some useful helper data for the next steps.
-	depths[idx] = p_view.z;
-	radii[idx] = my_radius;
-	points_xy_image[idx] = point_image;
-	// Inverse 2D covariance and opacity neatly pack into one float4
-	conic_opacity[idx] = { conic.x, conic.y, conic.z, opacities[idx] };
-	tiles_touched[idx] = (rect_max.y - rect_min.y) * (rect_max.x - rect_min.x);
-}
-
-// Main rasterization method. Collaboratively works on one tile per
-// block, each thread treats one pixel. Alternates between fetching 
-// and rasterizing data.
-template <uint32_t CHANNELS>
-__global__ void __launch_bounds__(BLOCK_X * BLOCK_Y)
-renderCUDA(
-	const uint2* __restrict__ ranges,
-	const uint32_t* __restrict__ point_list,
-	int W, int H,
-	const float2* __restrict__ points_xy_image,
-	const float* __restrict__ features,
-	const float* __restrict__ depths,
-	const float4* __restrict__ conic_opacity,
-	float* __restrict__ out_alpha,
-	uint32_t* __restrict__ n_contrib,
-	const float* __restrict__ bg_color,
-	float* __restrict__ out_color,
-	float* __restrict__ out_depth)
-{
-	// Identify current tile and associated min/max pixel range.
-	auto block = cg::this_thread_block();
-	uint32_t horizontal_blocks = (W + BLOCK_X - 1) / BLOCK_X;
-	uint2 pix_min = { block.group_index().x * BLOCK_X, block.group_index().y * BLOCK_Y };
-	uint2 pix_max = { min(pix_min.x + BLOCK_X, W), min(pix_min.y + BLOCK_Y , H) };
-	uint2 pix = { pix_min.x + block.thread_index().x, pix_min.y + block.thread_index().y };
-	uint32_t pix_id = W * pix.y + pix.x;
-	float2 pixf = { (float)pix.x, (float)pix.y };
-
-	// Check if this thread is associated with a valid pixel or outside.
-	bool inside = pix.x < W&& pix.y < H;
-	// Done threads can help with fetching, but don't rasterize
-	bool done = !inside;
-
-	// Load start/end range of IDs to process in bit sorted list.
-	uint2 range = ranges[block.group_index().y * horizontal_blocks + block.group_index().x];
-	const int rounds = ((range.y - range.x + BLOCK_SIZE - 1) / BLOCK_SIZE);
-	int toDo = range.y - range.x;
-
-	// Allocate storage for batches of collectively fetched data.
-	__shared__ int collected_id[BLOCK_SIZE];
-	__shared__ float2 collected_xy[BLOCK_SIZE];
-	__shared__ float4 collected_conic_opacity[BLOCK_SIZE];
-
-	// Initialize helper variables
-	float T = 1.0f;
-	uint32_t contributor = 0;
-	uint32_t last_contributor = 0;
-	float C[CHANNELS] = { 0 };
-	float weight = 0;
-	float D = 0;
-
-	// Iterate over batches until all done or range is complete
-	for (int i = 0; i < rounds; i++, toDo -= BLOCK_SIZE)
-	{
-		// End if entire block votes that it is done rasterizing
-		int num_done = __syncthreads_count(done);
-		if (num_done == BLOCK_SIZE)
-			break;
-
-		// Collectively fetch per-Gaussian data from global to shared
-		int progress = i * BLOCK_SIZE + block.thread_rank();
-		if (range.x + progress < range.y)
-		{
-			int coll_id = point_list[range.x + progress];
-			collected_id[block.thread_rank()] = coll_id;
-			collected_xy[block.thread_rank()] = points_xy_image[coll_id];
-			collected_conic_opacity[block.thread_rank()] = conic_opacity[coll_id];
-		}
-		block.sync();
-
-		// Iterate over current batch
-		for (int j = 0; !done && j < min(BLOCK_SIZE, toDo); j++)
-		{
-			// Keep track of current position in range
-			contributor++;
-
-			// Resample using conic matrix (cf. "Surface 
-			// Splatting" by Zwicker et al., 2001)
-			float2 xy = collected_xy[j];
-			float2 d = { xy.x - pixf.x, xy.y - pixf.y };
-			float4 con_o = collected_conic_opacity[j];
-			float power = -0.5f * (con_o.x * d.x * d.x + con_o.z * d.y * d.y) - con_o.y * d.x * d.y;
-			if (power > 0.0f)
-				continue;
-
-			// Eq. (2) from 3D Gaussian splatting paper.
-			// Obtain alpha by multiplying with Gaussian opacity
-			// and its exponential falloff from mean.
-			// Avoid numerical instabilities (see paper appendix). 
-			float alpha = min(0.99f, con_o.w * exp(power));
-			if (alpha < 1.0f / 255.0f)
-				continue;
-			float test_T = T * (1 - alpha);
-			if (test_T < 0.0001f)
-			{
-				done = true;
-				continue;
-			}
-
-			// Eq. (3) from 3D Gaussian splatting paper.
-			for (int ch = 0; ch < CHANNELS; ch++)
-				C[ch] += features[collected_id[j] * CHANNELS + ch] * alpha * T;
-			weight += alpha * T;
-			D += depths[collected_id[j]] * alpha * T;
-
-			T = test_T;
-
-			// Keep track of last range entry to update this
-			// pixel.
-			last_contributor = contributor;
-		}
-	}
-
-	// All threads that treat valid pixel write out their final
-	// rendering data to the frame and auxiliary buffers.
-	if (inside)
-	{
-		n_contrib[pix_id] = last_contributor;
-		for (int ch = 0; ch < CHANNELS; ch++)
-			out_color[ch * H * W + pix_id] = C[ch] + T * bg_color[ch];
-		out_alpha[pix_id] = weight; //1 - T;
-		out_depth[pix_id] = D;
-	}
-}
-
-void FORWARD::render(
-	const dim3 grid, dim3 block,
-	const uint2* ranges,
-	const uint32_t* point_list,
-	int W, int H,
-	const float2* means2D,
-	const float* colors,
-	const float* depths,
-	const float4* conic_opacity,
-	float* out_alpha,
-	uint32_t* n_contrib,
-	const float* bg_color,
-	float* out_color,
-	float* out_depth)
-{
-	renderCUDA<NUM_CHANNELS> << <grid, block >> > (
-		ranges,
-		point_list,
-		W, H,
-		means2D,
-		colors,
-		depths,
-		conic_opacity,
-		out_alpha,
-		n_contrib,
-		bg_color,
-		out_color,
-		out_depth);
-}
-
-void FORWARD::preprocess(int P, int D, int M,
-	const float* means3D,
-	const glm::vec3* scales,
-	const float scale_modifier,
-	const glm::vec4* rotations,
-	const float* opacities,
-	const float* shs,
-	bool* clamped,
-	const float* cov3D_precomp,
-	const float* colors_precomp,
-	const float* viewmatrix,
-	const float* projmatrix,
-	const glm::vec3* cam_pos,
-	const int W, int H,
-	const float focal_x, float focal_y,
-	const float tan_fovx, float tan_fovy,
-	int* radii,
-	float2* means2D,
-	float* depths,
-	float* cov3Ds,
-	float* rgb,
-	float4* conic_opacity,
-	const dim3 grid,
-	uint32_t* tiles_touched,
-	bool prefiltered)
-{
-	preprocessCUDA<NUM_CHANNELS> << <(P + 255) / 256, 256 >> > (
-		P, D, M,
-		means3D,
-		scales,
-		scale_modifier,
-		rotations,
-		opacities,
-		shs,
-		clamped,
-		cov3D_precomp,
-		colors_precomp,
-		viewmatrix, 
-		projmatrix,
-		cam_pos,
-		W, H,
-		tan_fovx, tan_fovy,
-		focal_x, focal_y,
-		radii,
-		means2D,
-		depths,
-		cov3Ds,
-		rgb,
-		conic_opacity,
-		grid,
-		tiles_touched,
-		prefiltered
-		);
-}

diff-gaussian-rasterization/cuda_rasterizer/forward.h

@@ -1,68 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#ifndef CUDA_RASTERIZER_FORWARD_H_INCLUDED
-#define CUDA_RASTERIZER_FORWARD_H_INCLUDED
-
-#include <cuda.h>
-#include "cuda_runtime.h"
-#include "device_launch_parameters.h"
-#define GLM_FORCE_CUDA
-#include <glm/glm.hpp>
-
-namespace FORWARD
-{
-	// Perform initial steps for each Gaussian prior to rasterization.
-	void preprocess(int P, int D, int M,
-		const float* orig_points,
-		const glm::vec3* scales,
-		const float scale_modifier,
-		const glm::vec4* rotations,
-		const float* opacities,
-		const float* shs,
-		bool* clamped,
-		const float* cov3D_precomp,
-		const float* colors_precomp,
-		const float* viewmatrix,
-		const float* projmatrix,
-		const glm::vec3* cam_pos,
-		const int W, int H,
-		const float focal_x, float focal_y,
-		const float tan_fovx, float tan_fovy,
-		int* radii,
-		float2* points_xy_image,
-		float* depths,
-		float* cov3Ds,
-		float* colors,
-		float4* conic_opacity,
-		const dim3 grid,
-		uint32_t* tiles_touched,
-		bool prefiltered);
-
-	// Main rasterization method.
-	void render(
-		const dim3 grid, dim3 block,
-		const uint2* ranges,
-		const uint32_t* point_list,
-		int W, int H,
-		const float2* points_xy_image,
-		const float* features,
-		const float* depths,
-		const float4* conic_opacity,
-		float* out_alpha,
-		uint32_t* n_contrib,
-		const float* bg_color,
-		float* out_color,
-		float* out_depth);
-}
-
-
-#endif

diff-gaussian-rasterization/cuda_rasterizer/rasterizer.h

@@ -1,94 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#ifndef CUDA_RASTERIZER_H_INCLUDED
-#define CUDA_RASTERIZER_H_INCLUDED
-
-#include <vector>
-#include <functional>
-
-namespace CudaRasterizer
-{
-	class Rasterizer
-	{
-	public:
-
-		static void markVisible(
-			int P,
-			float* means3D,
-			float* viewmatrix,
-			float* projmatrix,
-			bool* present);
-
-		static int forward(
-			std::function<char* (size_t)> geometryBuffer,
-			std::function<char* (size_t)> binningBuffer,
-			std::function<char* (size_t)> imageBuffer,
-			const int P, int D, int M,
-			const float* background,
-			const int width, int height,
-			const float* means3D,
-			const float* shs,
-			const float* colors_precomp,
-			const float* opacities,
-			const float* scales,
-			const float scale_modifier,
-			const float* rotations,
-			const float* cov3D_precomp,
-			const float* viewmatrix,
-			const float* projmatrix,
-			const float* cam_pos,
-			const float tan_fovx, float tan_fovy,
-			const bool prefiltered,
-			float* out_color,
-			float* out_depth,
-			float* out_alpha,
-			int* radii = nullptr,
-			bool debug = false);
-
-		static void backward(
-			const int P, int D, int M, int R,
-			const float* background,
-			const int width, int height,
-			const float* means3D,
-			const float* shs,
-			const float* colors_precomp,
-			const float* alphas,
-			const float* scales,
-			const float scale_modifier,
-			const float* rotations,
-			const float* cov3D_precomp,
-			const float* viewmatrix,
-			const float* projmatrix,
-			const float* campos,
-			const float tan_fovx, float tan_fovy,
-			const int* radii,
-			char* geom_buffer,
-			char* binning_buffer,
-			char* image_buffer,
-			const float* dL_dpix,
-			const float* dL_dpix_depth,
-			const float* dL_dalphas,
-			float* dL_dmean2D,
-			float* dL_dconic,
-			float* dL_dopacity,
-			float* dL_dcolor,
-			float* dL_ddepth,
-			float* dL_dmean3D,
-			float* dL_dcov3D,
-			float* dL_dsh,
-			float* dL_dscale,
-			float* dL_drot,
-			bool debug);
-	};
-};
-
-#endif

diff-gaussian-rasterization/cuda_rasterizer/rasterizer_impl.cu

@@ -1,447 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#include "rasterizer_impl.h"
-#include <iostream>
-#include <fstream>
-#include <algorithm>
-#include <numeric>
-#include <cuda.h>
-#include "cuda_runtime.h"
-#include "device_launch_parameters.h"
-#include <cub/cub.cuh>
-#include <cub/device/device_radix_sort.cuh>
-#define GLM_FORCE_CUDA
-#include <glm/glm.hpp>
-
-#include <cooperative_groups.h>
-#include <cooperative_groups/reduce.h>
-namespace cg = cooperative_groups;
-
-#include "auxiliary.h"
-#include "forward.h"
-#include "backward.h"
-
-// Helper function to find the next-highest bit of the MSB
-// on the CPU.
-uint32_t getHigherMsb(uint32_t n)
-{
-	uint32_t msb = sizeof(n) * 4;
-	uint32_t step = msb;
-	while (step > 1)
-	{
-		step /= 2;
-		if (n >> msb)
-			msb += step;
-		else
-			msb -= step;
-	}
-	if (n >> msb)
-		msb++;
-	return msb;
-}
-
-// Wrapper method to call auxiliary coarse frustum containment test.
-// Mark all Gaussians that pass it.
-__global__ void checkFrustum(int P,
-	const float* orig_points,
-	const float* viewmatrix,
-	const float* projmatrix,
-	bool* present)
-{
-	auto idx = cg::this_grid().thread_rank();
-	if (idx >= P)
-		return;
-
-	float3 p_view;
-	present[idx] = in_frustum(idx, orig_points, viewmatrix, projmatrix, false, p_view);
-}
-
-// Generates one key/value pair for all Gaussian / tile overlaps. 
-// Run once per Gaussian (1:N mapping).
-__global__ void duplicateWithKeys(
-	int P,
-	const float2* points_xy,
-	const float* depths,
-	const uint32_t* offsets,
-	uint64_t* gaussian_keys_unsorted,
-	uint32_t* gaussian_values_unsorted,
-	int* radii,
-	dim3 grid)
-{
-	auto idx = cg::this_grid().thread_rank();
-	if (idx >= P)
-		return;
-
-	// Generate no key/value pair for invisible Gaussians
-	if (radii[idx] > 0)
-	{
-		// Find this Gaussian's offset in buffer for writing keys/values.
-		uint32_t off = (idx == 0) ? 0 : offsets[idx - 1];
-		uint2 rect_min, rect_max;
-
-		getRect(points_xy[idx], radii[idx], rect_min, rect_max, grid);
-
-		// For each tile that the bounding rect overlaps, emit a 
-		// key/value pair. The key is |  tile ID  |      depth      |,
-		// and the value is the ID of the Gaussian. Sorting the values 
-		// with this key yields Gaussian IDs in a list, such that they
-		// are first sorted by tile and then by depth. 
-		for (int y = rect_min.y; y < rect_max.y; y++)
-		{
-			for (int x = rect_min.x; x < rect_max.x; x++)
-			{
-				uint64_t key = y * grid.x + x;
-				key <<= 32;
-				key |= *((uint32_t*)&depths[idx]);
-				gaussian_keys_unsorted[off] = key;
-				gaussian_values_unsorted[off] = idx;
-				off++;
-			}
-		}
-	}
-}
-
-// Check keys to see if it is at the start/end of one tile's range in 
-// the full sorted list. If yes, write start/end of this tile. 
-// Run once per instanced (duplicated) Gaussian ID.
-__global__ void identifyTileRanges(int L, uint64_t* point_list_keys, uint2* ranges)
-{
-	auto idx = cg::this_grid().thread_rank();
-	if (idx >= L)
-		return;
-
-	// Read tile ID from key. Update start/end of tile range if at limit.
-	uint64_t key = point_list_keys[idx];
-	uint32_t currtile = key >> 32;
-	if (idx == 0)
-		ranges[currtile].x = 0;
-	else
-	{
-		uint32_t prevtile = point_list_keys[idx - 1] >> 32;
-		if (currtile != prevtile)
-		{
-			ranges[prevtile].y = idx;
-			ranges[currtile].x = idx;
-		}
-	}
-	if (idx == L - 1)
-		ranges[currtile].y = L;
-}
-
-// Mark Gaussians as visible/invisible, based on view frustum testing
-void CudaRasterizer::Rasterizer::markVisible(
-	int P,
-	float* means3D,
-	float* viewmatrix,
-	float* projmatrix,
-	bool* present)
-{
-	checkFrustum << <(P + 255) / 256, 256 >> > (
-		P,
-		means3D,
-		viewmatrix, projmatrix,
-		present);
-}
-
-CudaRasterizer::GeometryState CudaRasterizer::GeometryState::fromChunk(char*& chunk, size_t P)
-{
-	GeometryState geom;
-	obtain(chunk, geom.depths, P, 128);
-	obtain(chunk, geom.clamped, P * 3, 128);
-	obtain(chunk, geom.internal_radii, P, 128);
-	obtain(chunk, geom.means2D, P, 128);
-	obtain(chunk, geom.cov3D, P * 6, 128);
-	obtain(chunk, geom.conic_opacity, P, 128);
-	obtain(chunk, geom.rgb, P * 3, 128);
-	obtain(chunk, geom.tiles_touched, P, 128);
-	cub::DeviceScan::InclusiveSum(nullptr, geom.scan_size, geom.tiles_touched, geom.tiles_touched, P);
-	obtain(chunk, geom.scanning_space, geom.scan_size, 128);
-	obtain(chunk, geom.point_offsets, P, 128);
-	return geom;
-}
-
-CudaRasterizer::ImageState CudaRasterizer::ImageState::fromChunk(char*& chunk, size_t N)
-{
-	ImageState img;
-	obtain(chunk, img.n_contrib, N, 128);
-	obtain(chunk, img.ranges, N, 128);
-	return img;
-}
-
-CudaRasterizer::BinningState CudaRasterizer::BinningState::fromChunk(char*& chunk, size_t P)
-{
-	BinningState binning;
-	obtain(chunk, binning.point_list, P, 128);
-	obtain(chunk, binning.point_list_unsorted, P, 128);
-	obtain(chunk, binning.point_list_keys, P, 128);
-	obtain(chunk, binning.point_list_keys_unsorted, P, 128);
-	cub::DeviceRadixSort::SortPairs(
-		nullptr, binning.sorting_size,
-		binning.point_list_keys_unsorted, binning.point_list_keys,
-		binning.point_list_unsorted, binning.point_list, P);
-	obtain(chunk, binning.list_sorting_space, binning.sorting_size, 128);
-	return binning;
-}
-
-// Forward rendering procedure for differentiable rasterization
-// of Gaussians.
-int CudaRasterizer::Rasterizer::forward(
-	std::function<char* (size_t)> geometryBuffer,
-	std::function<char* (size_t)> binningBuffer,
-	std::function<char* (size_t)> imageBuffer,
-	const int P, int D, int M,
-	const float* background,
-	const int width, int height,
-	const float* means3D,
-	const float* shs,
-	const float* colors_precomp,
-	const float* opacities,
-	const float* scales,
-	const float scale_modifier,
-	const float* rotations,
-	const float* cov3D_precomp,
-	const float* viewmatrix,
-	const float* projmatrix,
-	const float* cam_pos,
-	const float tan_fovx, float tan_fovy,
-	const bool prefiltered,
-	float* out_color,
-	float* out_depth,
-	float* out_alpha,
-	int* radii,
-	bool debug)
-{
-	const float focal_y = height / (2.0f * tan_fovy);
-	const float focal_x = width / (2.0f * tan_fovx);
-
-	size_t chunk_size = required<GeometryState>(P);
-	char* chunkptr = geometryBuffer(chunk_size);
-	GeometryState geomState = GeometryState::fromChunk(chunkptr, P);
-
-	if (radii == nullptr)
-	{
-		radii = geomState.internal_radii;
-	}
-
-	dim3 tile_grid((width + BLOCK_X - 1) / BLOCK_X, (height + BLOCK_Y - 1) / BLOCK_Y, 1);
-	dim3 block(BLOCK_X, BLOCK_Y, 1);
-
-	// Dynamically resize image-based auxiliary buffers during training
-	size_t img_chunk_size = required<ImageState>(width * height);
-	char* img_chunkptr = imageBuffer(img_chunk_size);
-	ImageState imgState = ImageState::fromChunk(img_chunkptr, width * height);
-
-	if (NUM_CHANNELS != 3 && colors_precomp == nullptr)
-	{
-		throw std::runtime_error("For non-RGB, provide precomputed Gaussian colors!");
-	}
-
-	// Run preprocessing per-Gaussian (transformation, bounding, conversion of SHs to RGB)
-	CHECK_CUDA(FORWARD::preprocess(
-		P, D, M,
-		means3D,
-		(glm::vec3*)scales,
-		scale_modifier,
-		(glm::vec4*)rotations,
-		opacities,
-		shs,
-		geomState.clamped,
-		cov3D_precomp,
-		colors_precomp,
-		viewmatrix, projmatrix,
-		(glm::vec3*)cam_pos,
-		width, height,
-		focal_x, focal_y,
-		tan_fovx, tan_fovy,
-		radii,
-		geomState.means2D,
-		geomState.depths,
-		geomState.cov3D,
-		geomState.rgb,
-		geomState.conic_opacity,
-		tile_grid,
-		geomState.tiles_touched,
-		prefiltered
-	), debug)
-
-	// Compute prefix sum over full list of touched tile counts by Gaussians
-	// E.g., [2, 3, 0, 2, 1] -> [2, 5, 5, 7, 8]
-	CHECK_CUDA(cub::DeviceScan::InclusiveSum(geomState.scanning_space, geomState.scan_size, geomState.tiles_touched, geomState.point_offsets, P), debug)
-
-	// Retrieve total number of Gaussian instances to launch and resize aux buffers
-	int num_rendered;
-	CHECK_CUDA(cudaMemcpy(&num_rendered, geomState.point_offsets + P - 1, sizeof(int), cudaMemcpyDeviceToHost), debug);
-
-	size_t binning_chunk_size = required<BinningState>(num_rendered);
-	char* binning_chunkptr = binningBuffer(binning_chunk_size);
-	BinningState binningState = BinningState::fromChunk(binning_chunkptr, num_rendered);
-
-	// For each instance to be rendered, produce adequate [ tile | depth ] key 
-	// and corresponding dublicated Gaussian indices to be sorted
-	duplicateWithKeys << <(P + 255) / 256, 256 >> > (
-		P,
-		geomState.means2D,
-		geomState.depths,
-		geomState.point_offsets,
-		binningState.point_list_keys_unsorted,
-		binningState.point_list_unsorted,
-		radii,
-		tile_grid)
-	CHECK_CUDA(, debug)
-
-	int bit = getHigherMsb(tile_grid.x * tile_grid.y);
-
-	// Sort complete list of (duplicated) Gaussian indices by keys
-	CHECK_CUDA(cub::DeviceRadixSort::SortPairs(
-		binningState.list_sorting_space,
-		binningState.sorting_size,
-		binningState.point_list_keys_unsorted, binningState.point_list_keys,
-		binningState.point_list_unsorted, binningState.point_list,
-		num_rendered, 0, 32 + bit), debug)
-
-	CHECK_CUDA(cudaMemset(imgState.ranges, 0, tile_grid.x * tile_grid.y * sizeof(uint2)), debug);
-
-	// Identify start and end of per-tile workloads in sorted list
-	if (num_rendered > 0)
-		identifyTileRanges << <(num_rendered + 255) / 256, 256 >> > (
-			num_rendered,
-			binningState.point_list_keys,
-			imgState.ranges);
-	CHECK_CUDA(, debug);
-
-	// Let each tile blend its range of Gaussians independently in parallel
-	const float* feature_ptr = colors_precomp != nullptr ? colors_precomp : geomState.rgb;
-	CHECK_CUDA(FORWARD::render(
-		tile_grid, block,
-		imgState.ranges,
-		binningState.point_list,
-		width, height,
-		geomState.means2D,
-		feature_ptr,
-		geomState.depths,
-		geomState.conic_opacity,
-		out_alpha,
-		imgState.n_contrib,
-		background,
-		out_color,
-		out_depth), debug);
-
-	return num_rendered;
-}
-
-// Produce necessary gradients for optimization, corresponding
-// to forward render pass
-void CudaRasterizer::Rasterizer::backward(
-	const int P, int D, int M, int R,
-	const float* background,
-	const int width, int height,
-	const float* means3D,
-	const float* shs,
-	const float* colors_precomp,
-	const float* alphas,
-	const float* scales,
-	const float scale_modifier,
-	const float* rotations,
-	const float* cov3D_precomp,
-	const float* viewmatrix,
-	const float* projmatrix,
-	const float* campos,
-	const float tan_fovx, float tan_fovy,
-	const int* radii,
-	char* geom_buffer,
-	char* binning_buffer,
-	char* img_buffer,
-	const float* dL_dpix,
-	const float* dL_dpix_depth,
-	const float* dL_dalphas,
-	float* dL_dmean2D,
-	float* dL_dconic,
-	float* dL_dopacity,
-	float* dL_dcolor,
-	float* dL_ddepth,
-	float* dL_dmean3D,
-	float* dL_dcov3D,
-	float* dL_dsh,
-	float* dL_dscale,
-	float* dL_drot,
-	bool debug)
-{
-	GeometryState geomState = GeometryState::fromChunk(geom_buffer, P);
-	BinningState binningState = BinningState::fromChunk(binning_buffer, R);
-	ImageState imgState = ImageState::fromChunk(img_buffer, width * height);
-
-	if (radii == nullptr)
-	{
-		radii = geomState.internal_radii;
-	}
-
-	const float focal_y = height / (2.0f * tan_fovy);
-	const float focal_x = width / (2.0f * tan_fovx);
-
-	const dim3 tile_grid((width + BLOCK_X - 1) / BLOCK_X, (height + BLOCK_Y - 1) / BLOCK_Y, 1);
-	const dim3 block(BLOCK_X, BLOCK_Y, 1);
-
-	// Compute loss gradients w.r.t. 2D mean position, conic matrix,
-	// opacity and RGB of Gaussians from per-pixel loss gradients.
-	// If we were given precomputed colors and not SHs, use them.
-	const float* color_ptr = (colors_precomp != nullptr) ? colors_precomp : geomState.rgb;
-	const float* depth_ptr = geomState.depths;
-	CHECK_CUDA(BACKWARD::render(
-		tile_grid,
-		block,
-		imgState.ranges,
-		binningState.point_list,
-		width, height,
-		background,
-		geomState.means2D,
-		geomState.conic_opacity,
-		color_ptr,
-		depth_ptr,
-		alphas,
-		imgState.n_contrib,
-		dL_dpix,
-		dL_dpix_depth,
-		dL_dalphas,
-		(float3*)dL_dmean2D,
-		(float4*)dL_dconic,
-		dL_dopacity,
-		dL_dcolor,
-		dL_ddepth), debug)
-
-	// Take care of the rest of preprocessing. Was the precomputed covariance
-	// given to us or a scales/rot pair? If precomputed, pass that. If not,
-	// use the one we computed ourselves.
-	const float* cov3D_ptr = (cov3D_precomp != nullptr) ? cov3D_precomp : geomState.cov3D;
-	CHECK_CUDA(BACKWARD::preprocess(P, D, M,
-		(float3*)means3D,
-		radii,
-		shs,
-		geomState.clamped,
-		(glm::vec3*)scales,
-		(glm::vec4*)rotations,
-		scale_modifier,
-		cov3D_ptr,
-		viewmatrix,
-		projmatrix,
-		focal_x, focal_y,
-		tan_fovx, tan_fovy,
-		(glm::vec3*)campos,
-		(float3*)dL_dmean2D,
-		dL_dconic,
-		(glm::vec3*)dL_dmean3D,
-		dL_dcolor,
-		dL_ddepth,
-		dL_dcov3D,
-		dL_dsh,
-		(glm::vec3*)dL_dscale,
-		(glm::vec4*)dL_drot), debug)
-}

diff-gaussian-rasterization/cuda_rasterizer/rasterizer_impl.h

@@ -1,73 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#pragma once
-
-#include <iostream>
-#include <vector>
-#include "rasterizer.h"
-#include <cuda_runtime_api.h>
-
-namespace CudaRasterizer
-{
-	template <typename T>
-	static void obtain(char*& chunk, T*& ptr, std::size_t count, std::size_t alignment)
-	{
-		std::size_t offset = (reinterpret_cast<std::uintptr_t>(chunk) + alignment - 1) & ~(alignment - 1);
-		ptr = reinterpret_cast<T*>(offset);
-		chunk = reinterpret_cast<char*>(ptr + count);
-	}
-
-	struct GeometryState
-	{
-		size_t scan_size;
-		float* depths;
-		char* scanning_space;
-		bool* clamped;
-		int* internal_radii;
-		float2* means2D;
-		float* cov3D;
-		float4* conic_opacity;
-		float* rgb;
-		uint32_t* point_offsets;
-		uint32_t* tiles_touched;
-
-		static GeometryState fromChunk(char*& chunk, size_t P);
-	};
-
-	struct ImageState
-	{
-		uint2* ranges;
-		uint32_t* n_contrib;
-
-		static ImageState fromChunk(char*& chunk, size_t N);
-	};
-
-	struct BinningState
-	{
-		size_t sorting_size;
-		uint64_t* point_list_keys_unsorted;
-		uint64_t* point_list_keys;
-		uint32_t* point_list_unsorted;
-		uint32_t* point_list;
-		char* list_sorting_space;
-
-		static BinningState fromChunk(char*& chunk, size_t P);
-	};
-
-	template<typename T> 
-	size_t required(size_t P)
-	{
-		char* size = nullptr;
-		T::fromChunk(size, P);
-		return ((size_t)size) + 128;
-	}
-};

diff-gaussian-rasterization/diff_gaussian_rasterization/__init__.py

@@ -1,224 +0,0 @@
-#
-# Copyright (C) 2023, Inria
-# GRAPHDECO research group, https://team.inria.fr/graphdeco
-# All rights reserved.
-#
-# This software is free for non-commercial, research and evaluation use 
-# under the terms of the LICENSE.md file.
-#
-# For inquiries contact  george.drettakis@inria.fr
-#
-
-from typing import NamedTuple
-import torch.nn as nn
-import torch
-from . import _C
-
-def cpu_deep_copy_tuple(input_tuple):
-    copied_tensors = [item.cpu().clone() if isinstance(item, torch.Tensor) else item for item in input_tuple]
-    return tuple(copied_tensors)
-
-def rasterize_gaussians(
-    means3D,
-    means2D,
-    sh,
-    colors_precomp,
-    opacities,
-    scales,
-    rotations,
-    cov3Ds_precomp,
-    raster_settings,
-):
-    return _RasterizeGaussians.apply(
-        means3D,
-        means2D,
-        sh,
-        colors_precomp,
-        opacities,
-        scales,
-        rotations,
-        cov3Ds_precomp,
-        raster_settings,
-    )
-
-class _RasterizeGaussians(torch.autograd.Function):
-    @staticmethod
-    def forward(
-        ctx,
-        means3D,
-        means2D,
-        sh,
-        colors_precomp,
-        opacities,
-        scales,
-        rotations,
-        cov3Ds_precomp,
-        raster_settings,
-    ):
-
-        # Restructure arguments the way that the C++ lib expects them
-        args = (
-            raster_settings.bg, 
-            means3D,
-            colors_precomp,
-            opacities,
-            scales,
-            rotations,
-            raster_settings.scale_modifier,
-            cov3Ds_precomp,
-            raster_settings.viewmatrix,
-            raster_settings.projmatrix,
-            raster_settings.tanfovx,
-            raster_settings.tanfovy,
-            raster_settings.image_height,
-            raster_settings.image_width,
-            sh,
-            raster_settings.sh_degree,
-            raster_settings.campos,
-            raster_settings.prefiltered,
-            raster_settings.debug
-        )
-
-        # Invoke C++/CUDA rasterizer
-        if raster_settings.debug:
-            cpu_args = cpu_deep_copy_tuple(args) # Copy them before they can be corrupted
-            try:
-                num_rendered, color, depth, alpha, radii, geomBuffer, binningBuffer, imgBuffer = _C.rasterize_gaussians(*args)
-            except Exception as ex:
-                torch.save(cpu_args, "snapshot_fw.dump")
-                print("\nAn error occured in forward. Please forward snapshot_fw.dump for debugging.")
-                raise ex
-        else:
-            num_rendered, color, depth, alpha, radii, geomBuffer, binningBuffer, imgBuffer = _C.rasterize_gaussians(*args)
-
-        # Keep relevant tensors for backward
-        ctx.raster_settings = raster_settings
-        ctx.num_rendered = num_rendered
-        ctx.save_for_backward(colors_precomp, means3D, scales, rotations, cov3Ds_precomp, radii, sh, geomBuffer, binningBuffer, imgBuffer, alpha)
-        return color, radii, depth, alpha
-
-    @staticmethod
-    def backward(ctx, grad_color, grad_radii, grad_depth, grad_alpha):
-
-        # Restore necessary values from context
-        num_rendered = ctx.num_rendered
-        raster_settings = ctx.raster_settings
-        colors_precomp, means3D, scales, rotations, cov3Ds_precomp, radii, sh, geomBuffer, binningBuffer, imgBuffer, alpha = ctx.saved_tensors
-
-        # Restructure args as C++ method expects them
-        args = (raster_settings.bg,
-                means3D, 
-                radii, 
-                colors_precomp, 
-                scales, 
-                rotations, 
-                raster_settings.scale_modifier, 
-                cov3Ds_precomp, 
-                raster_settings.viewmatrix, 
-                raster_settings.projmatrix, 
-                raster_settings.tanfovx, 
-                raster_settings.tanfovy, 
-                grad_color,
-                grad_depth,
-                grad_alpha,
-                sh, 
-                raster_settings.sh_degree, 
-                raster_settings.campos,
-                geomBuffer,
-                num_rendered,
-                binningBuffer,
-                imgBuffer,
-                alpha,
-                raster_settings.debug)
-
-        # Compute gradients for relevant tensors by invoking backward method
-        if raster_settings.debug:
-            cpu_args = cpu_deep_copy_tuple(args) # Copy them before they can be corrupted
-            try:
-                grad_means2D, grad_colors_precomp, grad_opacities, grad_means3D, grad_cov3Ds_precomp, grad_sh, grad_scales, grad_rotations = _C.rasterize_gaussians_backward(*args)
-            except Exception as ex:
-                torch.save(cpu_args, "snapshot_bw.dump")
-                print("\nAn error occured in backward. Writing snapshot_bw.dump for debugging.\n")
-                raise ex
-        else:
-             grad_means2D, grad_colors_precomp, grad_opacities, grad_means3D, grad_cov3Ds_precomp, grad_sh, grad_scales, grad_rotations = _C.rasterize_gaussians_backward(*args)
-
-        grads = (
-            grad_means3D,
-            grad_means2D,
-            grad_sh,
-            grad_colors_precomp,
-            grad_opacities,
-            grad_scales,
-            grad_rotations,
-            grad_cov3Ds_precomp,
-            None,
-        )
-
-        return grads
-
-class GaussianRasterizationSettings(NamedTuple):
-    image_height: int
-    image_width: int 
-    tanfovx : float
-    tanfovy : float
-    bg : torch.Tensor
-    scale_modifier : float
-    viewmatrix : torch.Tensor
-    projmatrix : torch.Tensor
-    sh_degree : int
-    campos : torch.Tensor
-    prefiltered : bool
-    debug : bool
-
-class GaussianRasterizer(nn.Module):
-    def __init__(self, raster_settings):
-        super().__init__()
-        self.raster_settings = raster_settings
-
-    def markVisible(self, positions):
-        # Mark visible points (based on frustum culling for camera) with a boolean 
-        with torch.no_grad():
-            raster_settings = self.raster_settings
-            visible = _C.mark_visible(
-                positions,
-                raster_settings.viewmatrix,
-                raster_settings.projmatrix)
-            
-        return visible
-
-    def forward(self, means3D, means2D, opacities, shs = None, colors_precomp = None, scales = None, rotations = None, cov3D_precomp = None):
-        
-        raster_settings = self.raster_settings
-
-        if (shs is None and colors_precomp is None) or (shs is not None and colors_precomp is not None):
-            raise Exception('Please provide excatly one of either SHs or precomputed colors!')
-        
-        if ((scales is None or rotations is None) and cov3D_precomp is None) or ((scales is not None or rotations is not None) and cov3D_precomp is not None):
-            raise Exception('Please provide exactly one of either scale/rotation pair or precomputed 3D covariance!')
-        
-        if shs is None:
-            shs = torch.Tensor([])
-        if colors_precomp is None:
-            colors_precomp = torch.Tensor([])
-
-        if scales is None:
-            scales = torch.Tensor([])
-        if rotations is None:
-            rotations = torch.Tensor([])
-        if cov3D_precomp is None:
-            cov3D_precomp = torch.Tensor([])
-
-        # Invoke C++/CUDA rasterization routine
-        return rasterize_gaussians(
-            means3D,
-            means2D,
-            shs,
-            colors_precomp,
-            opacities,
-            scales, 
-            rotations,
-            cov3D_precomp,
-            raster_settings, 
-        )
-

diff-gaussian-rasterization/ext.cpp

@@ -1,19 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#include <torch/extension.h>
-#include "rasterize_points.h"
-
-PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
-  m.def("rasterize_gaussians", &RasterizeGaussiansCUDA);
-  m.def("rasterize_gaussians_backward", &RasterizeGaussiansBackwardCUDA);
-  m.def("mark_visible", &markVisible);
-}

diff-gaussian-rasterization/rasterize_points.cu

@@ -1,229 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#include <math.h>
-#include <torch/extension.h>
-#include <cstdio>
-#include <sstream>
-#include <iostream>
-#include <tuple>
-#include <stdio.h>
-#include <cuda_runtime_api.h>
-#include <memory>
-#include "cuda_rasterizer/config.h"
-#include "cuda_rasterizer/rasterizer.h"
-#include <fstream>
-#include <string>
-#include <functional>
-
-std::function<char*(size_t N)> resizeFunctional(torch::Tensor& t) {
-    auto lambda = [&t](size_t N) {
-        t.resize_({(long long)N});
-		return reinterpret_cast<char*>(t.contiguous().data_ptr());
-    };
-    return lambda;
-}
-
-std::tuple<int, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor>
-RasterizeGaussiansCUDA(
-	const torch::Tensor& background,
-	const torch::Tensor& means3D,
-    const torch::Tensor& colors,
-    const torch::Tensor& opacity,
-	const torch::Tensor& scales,
-	const torch::Tensor& rotations,
-	const float scale_modifier,
-	const torch::Tensor& cov3D_precomp,
-	const torch::Tensor& viewmatrix,
-	const torch::Tensor& projmatrix,
-	const float tan_fovx, 
-	const float tan_fovy,
-    const int image_height,
-    const int image_width,
-	const torch::Tensor& sh,
-	const int degree,
-	const torch::Tensor& campos,
-	const bool prefiltered,
-	const bool debug)
-{
-  if (means3D.ndimension() != 2 || means3D.size(1) != 3) {
-    AT_ERROR("means3D must have dimensions (num_points, 3)");
-  }
-  
-  const int P = means3D.size(0);
-  const int H = image_height;
-  const int W = image_width;
-
-  auto int_opts = means3D.options().dtype(torch::kInt32);
-  auto float_opts = means3D.options().dtype(torch::kFloat32);
-
-  torch::Tensor out_color = torch::full({NUM_CHANNELS, H, W}, 0.0, float_opts);
-  torch::Tensor out_depth = torch::full({1, H, W}, 0.0, float_opts);
-  torch::Tensor out_alpha = torch::full({1, H, W}, 0.0, float_opts);
-  torch::Tensor radii = torch::full({P}, 0, means3D.options().dtype(torch::kInt32));
-  
-  torch::Device device(torch::kCUDA);
-  torch::TensorOptions options(torch::kByte);
-  torch::Tensor geomBuffer = torch::empty({0}, options.device(device));
-  torch::Tensor binningBuffer = torch::empty({0}, options.device(device));
-  torch::Tensor imgBuffer = torch::empty({0}, options.device(device));
-  std::function<char*(size_t)> geomFunc = resizeFunctional(geomBuffer);
-  std::function<char*(size_t)> binningFunc = resizeFunctional(binningBuffer);
-  std::function<char*(size_t)> imgFunc = resizeFunctional(imgBuffer);
-  
-  int rendered = 0;
-  if(P != 0)
-  {
-	  int M = 0;
-	  if(sh.size(0) != 0)
-	  {
-		M = sh.size(1);
-      }
-
-	  rendered = CudaRasterizer::Rasterizer::forward(
-	    geomFunc,
-		binningFunc,
-		imgFunc,
-	    P, degree, M,
-		background.contiguous().data<float>(),
-		W, H,
-		means3D.contiguous().data<float>(),
-		sh.contiguous().data_ptr<float>(),
-		colors.contiguous().data<float>(), 
-		opacity.contiguous().data<float>(), 
-		scales.contiguous().data_ptr<float>(),
-		scale_modifier,
-		rotations.contiguous().data_ptr<float>(),
-		cov3D_precomp.contiguous().data<float>(), 
-		viewmatrix.contiguous().data<float>(), 
-		projmatrix.contiguous().data<float>(),
-		campos.contiguous().data<float>(),
-		tan_fovx,
-		tan_fovy,
-		prefiltered,
-		out_color.contiguous().data<float>(),
-		out_depth.contiguous().data<float>(),
-		out_alpha.contiguous().data<float>(),
-		radii.contiguous().data<int>(),
-		debug);
-  }
-  return std::make_tuple(rendered, out_color, out_depth, out_alpha, radii, geomBuffer, binningBuffer, imgBuffer);
-}
-
-std::tuple<torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor>
- RasterizeGaussiansBackwardCUDA(
- 	const torch::Tensor& background,
-	const torch::Tensor& means3D,
-	const torch::Tensor& radii,
-    const torch::Tensor& colors,
-	const torch::Tensor& scales,
-	const torch::Tensor& rotations,
-	const float scale_modifier,
-	const torch::Tensor& cov3D_precomp,
-	const torch::Tensor& viewmatrix,
-    const torch::Tensor& projmatrix,
-	const float tan_fovx,
-	const float tan_fovy,
-    const torch::Tensor& dL_dout_color,
-	const torch::Tensor& dL_dout_depth,
-	const torch::Tensor& dL_dout_alpha,
-	const torch::Tensor& sh,
-	const int degree,
-	const torch::Tensor& campos,
-	const torch::Tensor& geomBuffer,
-	const int R,
-	const torch::Tensor& binningBuffer,
-	const torch::Tensor& imageBuffer,
-	const torch::Tensor& alphas,
-	const bool debug) 
-{
-  const int P = means3D.size(0);
-  const int H = dL_dout_color.size(1);
-  const int W = dL_dout_color.size(2);
-  
-  int M = 0;
-  if(sh.size(0) != 0)
-  {	
-	M = sh.size(1);
-  }
-
-  torch::Tensor dL_dmeans3D = torch::zeros({P, 3}, means3D.options());
-  torch::Tensor dL_dmeans2D = torch::zeros({P, 3}, means3D.options());
-  torch::Tensor dL_dcolors = torch::zeros({P, NUM_CHANNELS}, means3D.options());
-  torch::Tensor dL_ddepths = torch::zeros({P, 1}, means3D.options());
-  torch::Tensor dL_dconic = torch::zeros({P, 2, 2}, means3D.options());
-  torch::Tensor dL_dopacity = torch::zeros({P, 1}, means3D.options());
-  torch::Tensor dL_dcov3D = torch::zeros({P, 6}, means3D.options());
-  torch::Tensor dL_dsh = torch::zeros({P, M, 3}, means3D.options());
-  torch::Tensor dL_dscales = torch::zeros({P, 3}, means3D.options());
-  torch::Tensor dL_drotations = torch::zeros({P, 4}, means3D.options());
-  
-  if(P != 0)
-  {  
-	  CudaRasterizer::Rasterizer::backward(P, degree, M, R,
-	  background.contiguous().data<float>(),
-	  W, H, 
-	  means3D.contiguous().data<float>(),
-	  sh.contiguous().data<float>(),
-	  colors.contiguous().data<float>(),
-	  alphas.contiguous().data<float>(),
-	  scales.data_ptr<float>(),
-	  scale_modifier,
-	  rotations.data_ptr<float>(),
-	  cov3D_precomp.contiguous().data<float>(),
-	  viewmatrix.contiguous().data<float>(),
-	  projmatrix.contiguous().data<float>(),
-	  campos.contiguous().data<float>(),
-	  tan_fovx,
-	  tan_fovy,
-	  radii.contiguous().data<int>(),
-	  reinterpret_cast<char*>(geomBuffer.contiguous().data_ptr()),
-	  reinterpret_cast<char*>(binningBuffer.contiguous().data_ptr()),
-	  reinterpret_cast<char*>(imageBuffer.contiguous().data_ptr()),
-	  dL_dout_color.contiguous().data<float>(),
-	  dL_dout_depth.contiguous().data<float>(),
-	  dL_dout_alpha.contiguous().data<float>(),
-	  dL_dmeans2D.contiguous().data<float>(),
-	  dL_dconic.contiguous().data<float>(),  
-	  dL_dopacity.contiguous().data<float>(),
-	  dL_dcolors.contiguous().data<float>(),
-	  dL_ddepths.contiguous().data<float>(),
-	  dL_dmeans3D.contiguous().data<float>(),
-	  dL_dcov3D.contiguous().data<float>(),
-	  dL_dsh.contiguous().data<float>(),
-	  dL_dscales.contiguous().data<float>(),
-	  dL_drotations.contiguous().data<float>(),
-	  debug);
-  }
-
-  return std::make_tuple(dL_dmeans2D, dL_dcolors, dL_dopacity, dL_dmeans3D, dL_dcov3D, dL_dsh, dL_dscales, dL_drotations);
-}
-
-torch::Tensor markVisible(
-		torch::Tensor& means3D,
-		torch::Tensor& viewmatrix,
-		torch::Tensor& projmatrix)
-{ 
-  const int P = means3D.size(0);
-  
-  torch::Tensor present = torch::full({P}, false, means3D.options().dtype(at::kBool));
- 
-  if(P != 0)
-  {
-	CudaRasterizer::Rasterizer::markVisible(P,
-		means3D.contiguous().data<float>(),
-		viewmatrix.contiguous().data<float>(),
-		projmatrix.contiguous().data<float>(),
-		present.contiguous().data<bool>());
-  }
-  
-  return present;
-}

diff-gaussian-rasterization/rasterize_points.h

@@ -1,70 +0,0 @@
-/*
- * Copyright (C) 2023, Inria
- * GRAPHDECO research group, https://team.inria.fr/graphdeco
- * All rights reserved.
- *
- * This software is free for non-commercial, research and evaluation use 
- * under the terms of the LICENSE.md file.
- *
- * For inquiries contact  george.drettakis@inria.fr
- */
-
-#pragma once
-#include <torch/extension.h>
-#include <cstdio>
-#include <tuple>
-#include <string>
-	
-std::tuple<int, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor>
-RasterizeGaussiansCUDA(
-	const torch::Tensor& background,
-	const torch::Tensor& means3D,
-    const torch::Tensor& colors,
-    const torch::Tensor& opacity,
-	const torch::Tensor& scales,
-	const torch::Tensor& rotations,
-	const float scale_modifier,
-	const torch::Tensor& cov3D_precomp,
-	const torch::Tensor& viewmatrix,
-	const torch::Tensor& projmatrix,
-	const float tan_fovx, 
-	const float tan_fovy,
-    const int image_height,
-    const int image_width,
-	const torch::Tensor& sh,
-	const int degree,
-	const torch::Tensor& campos,
-	const bool prefiltered,
-	const bool debug);
-
-std::tuple<torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor, torch::Tensor>
- RasterizeGaussiansBackwardCUDA(
- 	const torch::Tensor& background,
-	const torch::Tensor& means3D,
-	const torch::Tensor& radii,
-    const torch::Tensor& colors,
-	const torch::Tensor& scales,
-	const torch::Tensor& rotations,
-	const float scale_modifier,
-	const torch::Tensor& cov3D_precomp,
-	const torch::Tensor& viewmatrix,
-    const torch::Tensor& projmatrix,
-	const float tan_fovx, 
-	const float tan_fovy,
-    const torch::Tensor& dL_dout_color,
-	const torch::Tensor& dL_dout_depth,
-	const torch::Tensor& dL_dout_alpha,
-	const torch::Tensor& sh,
-	const int degree,
-	const torch::Tensor& campos,
-	const torch::Tensor& geomBuffer,
-	const int R,
-	const torch::Tensor& binningBuffer,
-	const torch::Tensor& imageBuffer,
-	const torch::Tensor& alpha,
-	const bool debug);
-		
-torch::Tensor markVisible(
-		torch::Tensor& means3D,
-		torch::Tensor& viewmatrix,
-		torch::Tensor& projmatrix);

diff-gaussian-rasterization/setup.py

@@ -1,34 +0,0 @@
-#
-# Copyright (C) 2023, Inria
-# GRAPHDECO research group, https://team.inria.fr/graphdeco
-# All rights reserved.
-#
-# This software is free for non-commercial, research and evaluation use 
-# under the terms of the LICENSE.md file.
-#
-# For inquiries contact  george.drettakis@inria.fr
-#
-
-from setuptools import setup
-from torch.utils.cpp_extension import CUDAExtension, BuildExtension
-import os
-os.path.dirname(os.path.abspath(__file__))
-
-setup(
-    name="diff_gaussian_rasterization",
-    packages=['diff_gaussian_rasterization'],
-    ext_modules=[
-        CUDAExtension(
-            name="diff_gaussian_rasterization._C",
-            sources=[
-            "cuda_rasterizer/rasterizer_impl.cu",
-            "cuda_rasterizer/forward.cu",
-            "cuda_rasterizer/backward.cu",
-            "rasterize_points.cu",
-            "ext.cpp"],
-            extra_compile_args={"nvcc": ["-I" + os.path.join(os.path.dirname(os.path.abspath(__file__)), "third_party/glm/")]})
-        ],
-    cmdclass={
-        'build_ext': BuildExtension
-    }
-)

diff-gaussian-rasterization/third_party/glm/.appveyor.yml

@@ -1,92 +0,0 @@
-shallow_clone: true
-
-platform:
-  - x86
-  - x64
-
-configuration:
-  - Debug
-  - Release
-
-image:
-  - Visual Studio 2013
-  - Visual Studio 2015
-  - Visual Studio 2017
-  - Visual Studio 2019
-
-environment:
-  matrix:
-    - GLM_ARGUMENTS: -DGLM_TEST_FORCE_PURE=ON
-    - GLM_ARGUMENTS: -DGLM_TEST_ENABLE_SIMD_SSE2=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON
-    - GLM_ARGUMENTS: -DGLM_TEST_ENABLE_SIMD_AVX=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON
-    - GLM_ARGUMENTS: -DGLM_TEST_ENABLE_SIMD_AVX=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_CXX_14=ON
-    - GLM_ARGUMENTS: -DGLM_TEST_ENABLE_SIMD_AVX=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_CXX_17=ON
-
-matrix:
-    exclude:
-    - image: Visual Studio 2013
-      GLM_ARGUMENTS: -DGLM_TEST_ENABLE_SIMD_AVX=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON
-    - image: Visual Studio 2013
-      GLM_ARGUMENTS: -DGLM_TEST_ENABLE_SIMD_AVX=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_CXX_14=ON
-    - image: Visual Studio 2013
-      GLM_ARGUMENTS: -DGLM_TEST_ENABLE_SIMD_AVX=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_CXX_17=ON
-    - image: Visual Studio 2013
-      configuration: Debug
-    - image: Visual Studio 2015
-      GLM_ARGUMENTS: -DGLM_TEST_ENABLE_SIMD_SSE2=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON
-    - image: Visual Studio 2015
-      GLM_ARGUMENTS: -DGLM_TEST_ENABLE_SIMD_AVX=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_CXX_14=ON
-    - image: Visual Studio 2015
-      GLM_ARGUMENTS: -DGLM_TEST_ENABLE_SIMD_AVX=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_CXX_17=ON
-    - image: Visual Studio 2015
-      platform: x86
-    - image: Visual Studio 2015
-      configuration: Debug
-    - image: Visual Studio 2017
-      platform: x86
-    - image: Visual Studio 2017
-      configuration: Debug
-    - image: Visual Studio 2019
-      platform: x64
-
-branches:
-  only:
-    - master
-
-before_build:
-  - ps: |
-      mkdir build
-      cd build
-
-      if ("$env:APPVEYOR_JOB_NAME" -match "Image: Visual Studio 2013") {
-          $env:generator="Visual Studio 12 2013"
-      } 
-      if ("$env:APPVEYOR_JOB_NAME" -match "Image: Visual Studio 2015") {
-          $env:generator="Visual Studio 14 2015"
-      } 
-      if ("$env:APPVEYOR_JOB_NAME" -match "Image: Visual Studio 2017") {
-          $env:generator="Visual Studio 15 2017"
-      }
-      if ("$env:APPVEYOR_JOB_NAME" -match "Image: Visual Studio 2019") {
-          $env:generator="Visual Studio 16 2019"
-      }
-      if ($env:PLATFORM -eq "x64") {
-          $env:generator="$env:generator Win64"
-      }
-      echo generator="$env:generator"
-      cmake .. -G "$env:generator" -DCMAKE_INSTALL_PREFIX="$env:APPVEYOR_BUILD_FOLDER/install" -DGLM_QUIET=ON -DGLM_TEST_ENABLE=ON "$env:GLM_ARGUMENTS"
-
-build_script:
-  - cmake --build . --parallel --config %CONFIGURATION% -- /m /v:minimal
-  - cmake --build . --target install --parallel --config %CONFIGURATION% -- /m /v:minimal
-
-test_script:
-  - ctest --parallel 4 --verbose -C %CONFIGURATION%
-  - cd ..
-  - ps: |
-      mkdir build_test_cmake
-      cd build_test_cmake
-      cmake ..\test\cmake\ -G "$env:generator" -DCMAKE_PREFIX_PATH="$env:APPVEYOR_BUILD_FOLDER/install"
-  - cmake --build . --parallel --config %CONFIGURATION% -- /m /v:minimal
-
-deploy: off

diff-gaussian-rasterization/third_party/glm/.gitignore

@@ -1,61 +0,0 @@
-# Compiled Object files
-*.slo
-*.lo
-*.o
-*.obj
-
-# Precompiled Headers
-*.gch
-*.pch
-
-# Compiled Dynamic libraries
-*.so
-*.dylib
-*.dll
-
-# Fortran module files
-*.mod
-
-# Compiled Static libraries
-*.lai
-*.la
-*.a
-*.lib
-
-# Executables
-*.exe
-*.out
-*.app
-
-# CMake
-CMakeCache.txt
-CMakeFiles
-cmake_install.cmake
-install_manifest.txt
-*.cmake
-!glmConfig.cmake
-!glmConfig-version.cmake
-# ^ May need to add future .cmake files as exceptions
-
-# Test logs
-Testing/*
-
-# Test input
-test/gtc/*.dds
-
-# Project Files
-Makefile
-*.cbp
-*.user
-
-# Misc.
-*.log
-
-# local build(s)
-build*
-
-/.vs
-/.vscode
-/CMakeSettings.json
-.DS_Store
-*.swp

diff-gaussian-rasterization/third_party/glm/.travis.yml

@@ -1,388 +0,0 @@
-language: cpp
-
-branches:
-  only:
-    - master
-    - stable
-
-jobs:
-  include:
-    - name: "Xcode 7.3 C++98 pure release"
-      os: osx
-      osx_image: xcode7.3
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_98=ON -DGLM_TEST_FORCE_PURE=ON"
-
-    - name: "Xcode 7.3 C++98 sse2 release"
-      os: osx
-      osx_image: xcode7.3
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_98=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_SSE2=ON"
-
-    - name: "Xcode 7.3 C++98 ms release"
-      os: osx
-      osx_image: xcode7.3
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_98=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON"
-
-    - name: "XCode 7.3 C++11 pure release"
-      os: osx
-      osx_image: xcode7.3
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_11=ON -DGLM_TEST_FORCE_PURE=ON"
-
-    - name: "XCode 7.3 C++11 sse2 release"
-      os: osx
-      osx_image: xcode7.3
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_11=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_SSE3=ON"
-
-    - name: "XCode 10.3 C++11 sse2 release"
-      os: osx
-      osx_image: xcode10.3
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_11=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_SSE3=ON"
-
-    - name: "XCode 12.2 C++11 sse2 release"
-      os: osx
-      osx_image: xcode12.2
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_11=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_SSE3=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "XCode 12.2 C++11 sse2 debug"
-      os: osx
-      osx_image: xcode12.2
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_11=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_SSE3=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "XCode 12.2 C++11 avx debug"
-      os: osx
-      osx_image: xcode12.2
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_11=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_AVX=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "XCode 12.2 C++14 avx debug"
-      os: osx
-      osx_image: xcode12.2
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_14=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_AVX=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "XCode 12.2 C++14 pure debug"
-      os: osx
-      osx_image: xcode12.2
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_14=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "XCode 12.2 C++17 pure debug"
-      os: osx
-      osx_image: xcode12.2
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "XCode 12.2 C++17 sse2 debug"
-      os: osx
-      osx_image: xcode12.2
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_SSE2=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "XCode 12.2 C++17 sse2 release"
-      os: osx
-      osx_image: xcode12.2
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_SSE2=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "XCode 12.2 C++17 avx release"
-      os: osx
-      osx_image: xcode12.2
-      env:
-        - MATRIX_EVAL=""
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_AVX=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "GCC 4.9 C++98 pure release"
-      os: linux
-      dist: Xenial
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.9
-      env:
-        - MATRIX_EVAL="CC=gcc-4.9 && CXX=g++-4.9"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_98=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "GCC 4.9 C++98 pure debug"
-      os: linux
-      dist: Xenial
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.9
-      env:
-        - MATRIX_EVAL="CC=gcc-4.9 && CXX=g++-4.9"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_98=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "GCC 4.9 C++98 ms debug"
-      os: linux
-      dist: Xenial
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.9
-      env:
-        - MATRIX_EVAL="CC=gcc-4.9 && CXX=g++-4.9"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_98=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "GCC 4.9 C++11 ms debug"
-      os: linux
-      dist: Xenial
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.9
-      env:
-        - MATRIX_EVAL="CC=gcc-4.9 && CXX=g++-4.9"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_11=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "GCC 4.9 C++11 pure debug"
-      os: linux
-      dist: Xenial
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.9
-      env:
-        - MATRIX_EVAL="CC=gcc-4.9 && CXX=g++-4.9"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_11=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "GCC 6 C++14 pure debug"
-      os: linux
-      dist: bionic
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-6
-      env:
-        - MATRIX_EVAL="CC=gcc-6 && CXX=g++-6"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_14=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "GCC 6 C++14 ms debug"
-      os: linux
-      dist: bionic
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-6
-      env:
-        - MATRIX_EVAL="CC=gcc-6 && CXX=g++-6"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_14=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "GCC 7 C++17 ms debug"
-      os: linux
-      dist: bionic
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-7
-      env:
-        - MATRIX_EVAL="CC=gcc-7 && CXX=g++-7"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "GCC 7 C++17 pure debug"
-      os: linux
-      dist: bionic
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-7
-      env:
-        - MATRIX_EVAL="CC=gcc-7 && CXX=g++-7"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "GCC 10 C++17 pure debug"
-      os: linux
-      dist: bionic
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-10
-      env:
-        - MATRIX_EVAL="CC=gcc-10 && CXX=g++-10"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "GCC 10 C++17 pure release"
-      os: linux
-      dist: bionic
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-10
-      env:
-        - MATRIX_EVAL="CC=gcc-10 && CXX=g++-10"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "Clang C++14 pure release"
-      os: linux
-      dist: Xenial
-      env:
-        - MATRIX_EVAL="CC=clang && CXX=clang++"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_14=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "Clang C++14 pure debug"
-      os: linux
-      dist: Xenial
-      env:
-        - MATRIX_EVAL="CC=clang && CXX=clang++"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_14=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "Clang C++14 sse2 debug"
-      os: linux
-      dist: Xenial
-      env:
-        - MATRIX_EVAL="CC=clang && CXX=clang++"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_14=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_SSE2=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "Clang C++14 sse2 debug"
-      os: linux
-      dist: focal
-      env:
-        - MATRIX_EVAL="CC=clang && CXX=clang++"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_14=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_SSE2=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "Clang C++17 sse2 debug"
-      os: linux
-      dist: focal
-      env:
-        - MATRIX_EVAL="CC=clang && CXX=clang++"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_SSE2=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "Clang C++17 avx2 debug"
-      os: linux
-      dist: focal
-      env:
-        - MATRIX_EVAL="CC=clang && CXX=clang++"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_ENABLE_LANG_EXTENSIONS=ON -DGLM_TEST_ENABLE_SIMD_AVX2=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "Clang C++17 pure debug"
-      os: linux
-      dist: focal
-      env:
-        - MATRIX_EVAL="CC=clang && CXX=clang++"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Debug -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-    - name: "Clang C++17 pure release"
-      os: linux
-      dist: focal
-      env:
-        - MATRIX_EVAL="CC=clang && CXX=clang++"
-        - CMAKE_BUILD_ENV="-DCMAKE_BUILD_TYPE=Release -DGLM_TEST_ENABLE=ON -DGLM_TEST_ENABLE_CXX_17=ON -DGLM_TEST_FORCE_PURE=ON"
-        - CTEST_ENV="--parallel 4 --output-on-failure"
-        - CMAKE_ENV="--parallel"
-
-before_script:
-    - cmake --version
-    - eval "${MATRIX_EVAL}"
-
-script:
-  - ${CC} --version
-  - mkdir ./build
-  - cd ./build
-  - cmake -DCMAKE_INSTALL_PREFIX=$TRAVIS_BUILD_DIR/install -DCMAKE_CXX_COMPILER=$COMPILER ${CMAKE_BUILD_ENV} ..
-  - cmake --build . ${CMAKE_ENV}
-  - ctest ${CTEST_ENV}
-  - cmake --build . --target install ${CMAKE_ENV}
-  - cd $TRAVIS_BUILD_DIR
-  - mkdir ./build_test_cmake
-  - cd ./build_test_cmake
-  - cmake -DCMAKE_CXX_COMPILER=$COMPILER $TRAVIS_BUILD_DIR/test/cmake/ -DCMAKE_PREFIX_PATH=$TRAVIS_BUILD_DIR/install
-  - cmake --build .
-
-

diff-gaussian-rasterization/third_party/glm/CMakeLists.txt

@@ -1,45 +0,0 @@
-cmake_minimum_required(VERSION 3.2 FATAL_ERROR)
-cmake_policy(VERSION 3.2)
-
-
-file(READ "glm/detail/setup.hpp" GLM_SETUP_FILE)
-string(REGEX MATCH "#define[ ]+GLM_VERSION_MAJOR[ ]+([0-9]+)" _ ${GLM_SETUP_FILE})
-set(GLM_VERSION_MAJOR "${CMAKE_MATCH_1}")
-string(REGEX MATCH "#define[ ]+GLM_VERSION_MINOR[ ]+([0-9]+)" _ ${GLM_SETUP_FILE})
-set(GLM_VERSION_MINOR "${CMAKE_MATCH_1}")
-string(REGEX MATCH "#define[ ]+GLM_VERSION_PATCH[ ]+([0-9]+)" _ ${GLM_SETUP_FILE})
-set(GLM_VERSION_PATCH "${CMAKE_MATCH_1}")
-string(REGEX MATCH "#define[ ]+GLM_VERSION_REVISION[ ]+([0-9]+)" _ ${GLM_SETUP_FILE})
-set(GLM_VERSION_REVISION "${CMAKE_MATCH_1}")
-
-set(GLM_VERSION ${GLM_VERSION_MAJOR}.${GLM_VERSION_MINOR}.${GLM_VERSION_PATCH}.${GLM_VERSION_REVISION})
-project(glm VERSION ${GLM_VERSION} LANGUAGES CXX)
-message(STATUS "GLM: Version " ${GLM_VERSION})
-
-add_subdirectory(glm)
-add_library(glm::glm ALIAS glm)
-
-if(${CMAKE_SOURCE_DIR} STREQUAL ${CMAKE_CURRENT_SOURCE_DIR})
-
-	include(CPack)
-	install(DIRECTORY glm DESTINATION ${CMAKE_INSTALL_INCLUDEDIR} PATTERN "CMakeLists.txt" EXCLUDE)
-	install(EXPORT glm FILE glmConfig.cmake DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/glm NAMESPACE glm::)
-	include(CMakePackageConfigHelpers)
-	write_basic_package_version_file("glmConfigVersion.cmake" COMPATIBILITY AnyNewerVersion)
-	install(FILES ${CMAKE_CURRENT_BINARY_DIR}/glmConfigVersion.cmake DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/glm)
-
-	include(CTest)
-	if(BUILD_TESTING)
-		add_subdirectory(test)
-	endif()
-
-endif(${CMAKE_SOURCE_DIR} STREQUAL ${CMAKE_CURRENT_SOURCE_DIR})
-
-if (NOT TARGET uninstall)
-configure_file(cmake/cmake_uninstall.cmake.in
-               cmake_uninstall.cmake IMMEDIATE @ONLY)
-
-add_custom_target(uninstall
-                  "${CMAKE_COMMAND}" -P
-                  "${CMAKE_BINARY_DIR}/cmake_uninstall.cmake")
-endif()

diff-gaussian-rasterization/third_party/glm/cmake/cmake_uninstall.cmake.in

@@ -1,21 +0,0 @@
-if(NOT EXISTS "@CMAKE_BINARY_DIR@/install_manifest.txt")
-  message(FATAL_ERROR "Cannot find install manifest: @CMAKE_BINARY_DIR@/install_manifest.txt")
-endif()
-
-file(READ "@CMAKE_BINARY_DIR@/install_manifest.txt" files)
-string(REGEX REPLACE "\n" ";" files "${files}")
-foreach(file ${files})
-  message(STATUS "Uninstalling $ENV{DESTDIR}${file}")
-  if(IS_SYMLINK "$ENV{DESTDIR}${file}" OR EXISTS "$ENV{DESTDIR}${file}")
-    exec_program(
-      "@CMAKE_COMMAND@" ARGS "-E remove \"$ENV{DESTDIR}${file}\""
-      OUTPUT_VARIABLE rm_out
-      RETURN_VALUE rm_retval
-      )
-    if(NOT "${rm_retval}" STREQUAL 0)
-      message(FATAL_ERROR "Problem when removing $ENV{DESTDIR}${file}")
-    endif()
-  else(IS_SYMLINK "$ENV{DESTDIR}${file}" OR EXISTS "$ENV{DESTDIR}${file}")
-    message(STATUS "File $ENV{DESTDIR}${file} does not exist.")
-  endif()
-endforeach()

diff-gaussian-rasterization/third_party/glm/copying.txt

@@ -1,54 +0,0 @@
-================================================================================
-OpenGL Mathematics (GLM)
---------------------------------------------------------------------------------
-GLM is licensed under The Happy Bunny License or MIT License
-
-================================================================================
-The Happy Bunny License (Modified MIT License)
---------------------------------------------------------------------------------
-Copyright (c) 2005 - G-Truc Creation
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in
-all copies or substantial portions of the Software.
-
-Restrictions:
- By making use of the Software for military purposes, you choose to make a
- Bunny unhappy.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
-THE SOFTWARE.
-
-================================================================================
-The MIT License
---------------------------------------------------------------------------------
-Copyright (c) 2005 - G-Truc Creation
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in
-all copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
-THE SOFTWARE.

diff-gaussian-rasterization/third_party/glm/doc/api/a00001_source.html

@@ -1,493 +0,0 @@
-<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
-<html xmlns="http://www.w3.org/1999/xhtml">
-<head>
-<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
-<meta http-equiv="X-UA-Compatible" content="IE=9"/>
-<meta name="generator" content="Doxygen 1.8.10"/>
-<title>0.9.9 API documentation: _features.hpp Source File</title>
-<link href="tabs.css" rel="stylesheet" type="text/css"/>
-<script type="text/javascript" src="jquery.js"></script>
-<script type="text/javascript" src="dynsections.js"></script>
-<link href="search/search.css" rel="stylesheet" type="text/css"/>
-<script type="text/javascript" src="search/searchdata.js"></script>
-<script type="text/javascript" src="search/search.js"></script>
-<script type="text/javascript">
-  $(document).ready(function() { init_search(); });
-</script>
-<link href="doxygen.css" rel="stylesheet" type="text/css" />
-</head>
-<body>
-<div id="top"><!-- do not remove this div, it is closed by doxygen! -->
-<div id="titlearea">
-<table cellspacing="0" cellpadding="0">
- <tbody>
- <tr style="height: 56px;">
-  <td id="projectlogo"><img alt="Logo" src="logo-mini.png"/></td>
-  <td id="projectalign" style="padding-left: 0.5em;">
-   <div id="projectname">0.9.9 API documentation
-   </div>
-  </td>
- </tr>
- </tbody>
-</table>
-</div>
-<!-- end header part -->
-<!-- Generated by Doxygen 1.8.10 -->
-<script type="text/javascript">
-var searchBox = new SearchBox("searchBox", "search",false,'Search');
-</script>
-  <div id="navrow1" class="tabs">
-    <ul class="tablist">
-      <li><a href="index.html"><span>Main&#160;Page</span></a></li>
-      <li><a href="modules.html"><span>Modules</span></a></li>
-      <li class="current"><a href="files.html"><span>Files</span></a></li>
-      <li>
-        <div id="MSearchBox" class="MSearchBoxInactive">
-        <span class="left">
-          <img id="MSearchSelect" src="search/mag_sel.png"
-               onmouseover="return searchBox.OnSearchSelectShow()"
-               onmouseout="return searchBox.OnSearchSelectHide()"
-               alt=""/>
-          <input type="text" id="MSearchField" value="Search" accesskey="S"
-               onfocus="searchBox.OnSearchFieldFocus(true)" 
-               onblur="searchBox.OnSearchFieldFocus(false)" 
-               onkeyup="searchBox.OnSearchFieldChange(event)"/>
-          </span><span class="right">
-            <a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="search/close.png" alt=""/></a>
-          </span>
-        </div>
-      </li>
-    </ul>
-  </div>
-  <div id="navrow2" class="tabs2">
-    <ul class="tablist">
-      <li><a href="files.html"><span>File&#160;List</span></a></li>
-    </ul>
-  </div>
-<!-- window showing the filter options -->
-<div id="MSearchSelectWindow"
-     onmouseover="return searchBox.OnSearchSelectShow()"
-     onmouseout="return searchBox.OnSearchSelectHide()"
-     onkeydown="return searchBox.OnSearchSelectKey(event)">
-</div>
-
-<!-- iframe showing the search results (closed by default) -->
-<div id="MSearchResultsWindow">
-<iframe src="javascript:void(0)" frameborder="0" 
-        name="MSearchResults" id="MSearchResults">
-</iframe>
-</div>
-
-<div id="nav-path" class="navpath">
-  <ul>
-<li class="navelem"><a class="el" href="dir_3a581ba30d25676e4b797b1f96d53b45.html">F:</a></li><li class="navelem"><a class="el" href="dir_9e5fe034a00e89334fd5186c3e7db156.html">G-Truc</a></li><li class="navelem"><a class="el" href="dir_d9496f0844b48bc7e53b5af8c99b9ab2.html">Source</a></li><li class="navelem"><a class="el" href="dir_a8bee7be44182a33f3820393ae0b105d.html">G-Truc</a></li><li class="navelem"><a class="el" href="dir_44e5e654415abd9ca6fdeaddaff8565e.html">glm</a></li><li class="navelem"><a class="el" href="dir_cef2d71d502cb69a9252bca2297d9549.html">glm</a></li><li class="navelem"><a class="el" href="dir_033f5edb0915b828d2c46ed4804e5503.html">detail</a></li>  </ul>
-</div>
-</div><!-- top -->
-<div class="header">
-  <div class="headertitle">
-<div class="title">_features.hpp</div>  </div>
-</div><!--header-->
-<div class="contents">
-<div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno">    1</span>&#160;<span class="preprocessor">#pragma once</span></div>
-<div class="line"><a name="l00002"></a><span class="lineno">    2</span>&#160;</div>
-<div class="line"><a name="l00003"></a><span class="lineno">    3</span>&#160;<span class="comment">// #define GLM_CXX98_EXCEPTIONS</span></div>
-<div class="line"><a name="l00004"></a><span class="lineno">    4</span>&#160;<span class="comment">// #define GLM_CXX98_RTTI</span></div>
-<div class="line"><a name="l00005"></a><span class="lineno">    5</span>&#160;</div>
-<div class="line"><a name="l00006"></a><span class="lineno">    6</span>&#160;<span class="comment">// #define GLM_CXX11_RVALUE_REFERENCES</span></div>
-<div class="line"><a name="l00007"></a><span class="lineno">    7</span>&#160;<span class="comment">// Rvalue references - GCC 4.3</span></div>
-<div class="line"><a name="l00008"></a><span class="lineno">    8</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2118.html</span></div>
-<div class="line"><a name="l00009"></a><span class="lineno">    9</span>&#160;</div>
-<div class="line"><a name="l00010"></a><span class="lineno">   10</span>&#160;<span class="comment">// GLM_CXX11_TRAILING_RETURN</span></div>
-<div class="line"><a name="l00011"></a><span class="lineno">   11</span>&#160;<span class="comment">// Rvalue references for *this - GCC not supported</span></div>
-<div class="line"><a name="l00012"></a><span class="lineno">   12</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2439.htm</span></div>
-<div class="line"><a name="l00013"></a><span class="lineno">   13</span>&#160;</div>
-<div class="line"><a name="l00014"></a><span class="lineno">   14</span>&#160;<span class="comment">// GLM_CXX11_NONSTATIC_MEMBER_INIT</span></div>
-<div class="line"><a name="l00015"></a><span class="lineno">   15</span>&#160;<span class="comment">// Initialization of class objects by rvalues - GCC any</span></div>
-<div class="line"><a name="l00016"></a><span class="lineno">   16</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1610.html</span></div>
-<div class="line"><a name="l00017"></a><span class="lineno">   17</span>&#160;</div>
-<div class="line"><a name="l00018"></a><span class="lineno">   18</span>&#160;<span class="comment">// GLM_CXX11_NONSTATIC_MEMBER_INIT</span></div>
-<div class="line"><a name="l00019"></a><span class="lineno">   19</span>&#160;<span class="comment">// Non-static data member initializers - GCC 4.7</span></div>
-<div class="line"><a name="l00020"></a><span class="lineno">   20</span>&#160;<span class="comment">// http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2008/n2756.htm</span></div>
-<div class="line"><a name="l00021"></a><span class="lineno">   21</span>&#160;</div>
-<div class="line"><a name="l00022"></a><span class="lineno">   22</span>&#160;<span class="comment">// #define GLM_CXX11_VARIADIC_TEMPLATE</span></div>
-<div class="line"><a name="l00023"></a><span class="lineno">   23</span>&#160;<span class="comment">// Variadic templates - GCC 4.3</span></div>
-<div class="line"><a name="l00024"></a><span class="lineno">   24</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2242.pdf</span></div>
-<div class="line"><a name="l00025"></a><span class="lineno">   25</span>&#160;</div>
-<div class="line"><a name="l00026"></a><span class="lineno">   26</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00027"></a><span class="lineno">   27</span>&#160;<span class="comment">// Extending variadic template template parameters - GCC 4.4</span></div>
-<div class="line"><a name="l00028"></a><span class="lineno">   28</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2555.pdf</span></div>
-<div class="line"><a name="l00029"></a><span class="lineno">   29</span>&#160;</div>
-<div class="line"><a name="l00030"></a><span class="lineno">   30</span>&#160;<span class="comment">// #define GLM_CXX11_GENERALIZED_INITIALIZERS</span></div>
-<div class="line"><a name="l00031"></a><span class="lineno">   31</span>&#160;<span class="comment">// Initializer lists - GCC 4.4</span></div>
-<div class="line"><a name="l00032"></a><span class="lineno">   32</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2672.htm</span></div>
-<div class="line"><a name="l00033"></a><span class="lineno">   33</span>&#160;</div>
-<div class="line"><a name="l00034"></a><span class="lineno">   34</span>&#160;<span class="comment">// #define GLM_CXX11_STATIC_ASSERT</span></div>
-<div class="line"><a name="l00035"></a><span class="lineno">   35</span>&#160;<span class="comment">// Static assertions - GCC 4.3</span></div>
-<div class="line"><a name="l00036"></a><span class="lineno">   36</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1720.html</span></div>
-<div class="line"><a name="l00037"></a><span class="lineno">   37</span>&#160;</div>
-<div class="line"><a name="l00038"></a><span class="lineno">   38</span>&#160;<span class="comment">// #define GLM_CXX11_AUTO_TYPE</span></div>
-<div class="line"><a name="l00039"></a><span class="lineno">   39</span>&#160;<span class="comment">// auto-typed variables - GCC 4.4</span></div>
-<div class="line"><a name="l00040"></a><span class="lineno">   40</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1984.pdf</span></div>
-<div class="line"><a name="l00041"></a><span class="lineno">   41</span>&#160;</div>
-<div class="line"><a name="l00042"></a><span class="lineno">   42</span>&#160;<span class="comment">// #define GLM_CXX11_AUTO_TYPE</span></div>
-<div class="line"><a name="l00043"></a><span class="lineno">   43</span>&#160;<span class="comment">// Multi-declarator auto - GCC 4.4</span></div>
-<div class="line"><a name="l00044"></a><span class="lineno">   44</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1737.pdf</span></div>
-<div class="line"><a name="l00045"></a><span class="lineno">   45</span>&#160;</div>
-<div class="line"><a name="l00046"></a><span class="lineno">   46</span>&#160;<span class="comment">// #define GLM_CXX11_AUTO_TYPE</span></div>
-<div class="line"><a name="l00047"></a><span class="lineno">   47</span>&#160;<span class="comment">// Removal of auto as a storage-class specifier - GCC 4.4</span></div>
-<div class="line"><a name="l00048"></a><span class="lineno">   48</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2546.htm</span></div>
-<div class="line"><a name="l00049"></a><span class="lineno">   49</span>&#160;</div>
-<div class="line"><a name="l00050"></a><span class="lineno">   50</span>&#160;<span class="comment">// #define GLM_CXX11_AUTO_TYPE</span></div>
-<div class="line"><a name="l00051"></a><span class="lineno">   51</span>&#160;<span class="comment">// New function declarator syntax - GCC 4.4</span></div>
-<div class="line"><a name="l00052"></a><span class="lineno">   52</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2541.htm</span></div>
-<div class="line"><a name="l00053"></a><span class="lineno">   53</span>&#160;</div>
-<div class="line"><a name="l00054"></a><span class="lineno">   54</span>&#160;<span class="comment">// #define GLM_CXX11_LAMBDAS</span></div>
-<div class="line"><a name="l00055"></a><span class="lineno">   55</span>&#160;<span class="comment">// New wording for C++0x lambdas - GCC 4.5</span></div>
-<div class="line"><a name="l00056"></a><span class="lineno">   56</span>&#160;<span class="comment">// http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2009/n2927.pdf</span></div>
-<div class="line"><a name="l00057"></a><span class="lineno">   57</span>&#160;</div>
-<div class="line"><a name="l00058"></a><span class="lineno">   58</span>&#160;<span class="comment">// #define GLM_CXX11_DECLTYPE</span></div>
-<div class="line"><a name="l00059"></a><span class="lineno">   59</span>&#160;<span class="comment">// Declared type of an expression - GCC 4.3</span></div>
-<div class="line"><a name="l00060"></a><span class="lineno">   60</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2343.pdf</span></div>
-<div class="line"><a name="l00061"></a><span class="lineno">   61</span>&#160;</div>
-<div class="line"><a name="l00062"></a><span class="lineno">   62</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00063"></a><span class="lineno">   63</span>&#160;<span class="comment">// Right angle brackets - GCC 4.3</span></div>
-<div class="line"><a name="l00064"></a><span class="lineno">   64</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1757.html</span></div>
-<div class="line"><a name="l00065"></a><span class="lineno">   65</span>&#160;</div>
-<div class="line"><a name="l00066"></a><span class="lineno">   66</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00067"></a><span class="lineno">   67</span>&#160;<span class="comment">// Default template arguments for function templates    DR226   GCC 4.3</span></div>
-<div class="line"><a name="l00068"></a><span class="lineno">   68</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#226</span></div>
-<div class="line"><a name="l00069"></a><span class="lineno">   69</span>&#160;</div>
-<div class="line"><a name="l00070"></a><span class="lineno">   70</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00071"></a><span class="lineno">   71</span>&#160;<span class="comment">// Solving the SFINAE problem for expressions   DR339   GCC 4.4</span></div>
-<div class="line"><a name="l00072"></a><span class="lineno">   72</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2634.html</span></div>
-<div class="line"><a name="l00073"></a><span class="lineno">   73</span>&#160;</div>
-<div class="line"><a name="l00074"></a><span class="lineno">   74</span>&#160;<span class="comment">// #define GLM_CXX11_ALIAS_TEMPLATE</span></div>
-<div class="line"><a name="l00075"></a><span class="lineno">   75</span>&#160;<span class="comment">// Template aliases     N2258   GCC 4.7</span></div>
-<div class="line"><a name="l00076"></a><span class="lineno">   76</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2258.pdf</span></div>
-<div class="line"><a name="l00077"></a><span class="lineno">   77</span>&#160;</div>
-<div class="line"><a name="l00078"></a><span class="lineno">   78</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00079"></a><span class="lineno">   79</span>&#160;<span class="comment">// Extern templates     N1987   Yes</span></div>
-<div class="line"><a name="l00080"></a><span class="lineno">   80</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1987.htm</span></div>
-<div class="line"><a name="l00081"></a><span class="lineno">   81</span>&#160;</div>
-<div class="line"><a name="l00082"></a><span class="lineno">   82</span>&#160;<span class="comment">// #define GLM_CXX11_NULLPTR</span></div>
-<div class="line"><a name="l00083"></a><span class="lineno">   83</span>&#160;<span class="comment">// Null pointer constant        N2431   GCC 4.6</span></div>
-<div class="line"><a name="l00084"></a><span class="lineno">   84</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2431.pdf</span></div>
-<div class="line"><a name="l00085"></a><span class="lineno">   85</span>&#160;</div>
-<div class="line"><a name="l00086"></a><span class="lineno">   86</span>&#160;<span class="comment">// #define GLM_CXX11_STRONG_ENUMS</span></div>
-<div class="line"><a name="l00087"></a><span class="lineno">   87</span>&#160;<span class="comment">// Strongly-typed enums N2347   GCC 4.4</span></div>
-<div class="line"><a name="l00088"></a><span class="lineno">   88</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2347.pdf</span></div>
-<div class="line"><a name="l00089"></a><span class="lineno">   89</span>&#160;</div>
-<div class="line"><a name="l00090"></a><span class="lineno">   90</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00091"></a><span class="lineno">   91</span>&#160;<span class="comment">// Forward declarations for enums       N2764   GCC 4.6</span></div>
-<div class="line"><a name="l00092"></a><span class="lineno">   92</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2764.pdf</span></div>
-<div class="line"><a name="l00093"></a><span class="lineno">   93</span>&#160;</div>
-<div class="line"><a name="l00094"></a><span class="lineno">   94</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00095"></a><span class="lineno">   95</span>&#160;<span class="comment">// Generalized attributes       N2761   GCC 4.8</span></div>
-<div class="line"><a name="l00096"></a><span class="lineno">   96</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2761.pdf</span></div>
-<div class="line"><a name="l00097"></a><span class="lineno">   97</span>&#160;</div>
-<div class="line"><a name="l00098"></a><span class="lineno">   98</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00099"></a><span class="lineno">   99</span>&#160;<span class="comment">// Generalized constant expressions     N2235   GCC 4.6</span></div>
-<div class="line"><a name="l00100"></a><span class="lineno">  100</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2235.pdf</span></div>
-<div class="line"><a name="l00101"></a><span class="lineno">  101</span>&#160;</div>
-<div class="line"><a name="l00102"></a><span class="lineno">  102</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00103"></a><span class="lineno">  103</span>&#160;<span class="comment">// Alignment support    N2341   GCC 4.8</span></div>
-<div class="line"><a name="l00104"></a><span class="lineno">  104</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2341.pdf</span></div>
-<div class="line"><a name="l00105"></a><span class="lineno">  105</span>&#160;</div>
-<div class="line"><a name="l00106"></a><span class="lineno">  106</span>&#160;<span class="comment">// #define GLM_CXX11_DELEGATING_CONSTRUCTORS</span></div>
-<div class="line"><a name="l00107"></a><span class="lineno">  107</span>&#160;<span class="comment">// Delegating constructors      N1986   GCC 4.7</span></div>
-<div class="line"><a name="l00108"></a><span class="lineno">  108</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1986.pdf</span></div>
-<div class="line"><a name="l00109"></a><span class="lineno">  109</span>&#160;</div>
-<div class="line"><a name="l00110"></a><span class="lineno">  110</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00111"></a><span class="lineno">  111</span>&#160;<span class="comment">// Inheriting constructors      N2540   GCC 4.8</span></div>
-<div class="line"><a name="l00112"></a><span class="lineno">  112</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2540.htm</span></div>
-<div class="line"><a name="l00113"></a><span class="lineno">  113</span>&#160;</div>
-<div class="line"><a name="l00114"></a><span class="lineno">  114</span>&#160;<span class="comment">// #define GLM_CXX11_EXPLICIT_CONVERSIONS</span></div>
-<div class="line"><a name="l00115"></a><span class="lineno">  115</span>&#160;<span class="comment">// Explicit conversion operators        N2437   GCC 4.5</span></div>
-<div class="line"><a name="l00116"></a><span class="lineno">  116</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2437.pdf</span></div>
-<div class="line"><a name="l00117"></a><span class="lineno">  117</span>&#160;</div>
-<div class="line"><a name="l00118"></a><span class="lineno">  118</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00119"></a><span class="lineno">  119</span>&#160;<span class="comment">// New character types  N2249   GCC 4.4</span></div>
-<div class="line"><a name="l00120"></a><span class="lineno">  120</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2249.html</span></div>
-<div class="line"><a name="l00121"></a><span class="lineno">  121</span>&#160;</div>
-<div class="line"><a name="l00122"></a><span class="lineno">  122</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00123"></a><span class="lineno">  123</span>&#160;<span class="comment">// Unicode string literals      N2442   GCC 4.5</span></div>
-<div class="line"><a name="l00124"></a><span class="lineno">  124</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2442.htm</span></div>
-<div class="line"><a name="l00125"></a><span class="lineno">  125</span>&#160;</div>
-<div class="line"><a name="l00126"></a><span class="lineno">  126</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00127"></a><span class="lineno">  127</span>&#160;<span class="comment">// Raw string literals  N2442   GCC 4.5</span></div>
-<div class="line"><a name="l00128"></a><span class="lineno">  128</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2442.htm</span></div>
-<div class="line"><a name="l00129"></a><span class="lineno">  129</span>&#160;</div>
-<div class="line"><a name="l00130"></a><span class="lineno">  130</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00131"></a><span class="lineno">  131</span>&#160;<span class="comment">// Universal character name literals    N2170   GCC 4.5</span></div>
-<div class="line"><a name="l00132"></a><span class="lineno">  132</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2170.html</span></div>
-<div class="line"><a name="l00133"></a><span class="lineno">  133</span>&#160;</div>
-<div class="line"><a name="l00134"></a><span class="lineno">  134</span>&#160;<span class="comment">// #define GLM_CXX11_USER_LITERALS</span></div>
-<div class="line"><a name="l00135"></a><span class="lineno">  135</span>&#160;<span class="comment">// User-defined literals                N2765   GCC 4.7</span></div>
-<div class="line"><a name="l00136"></a><span class="lineno">  136</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2765.pdf</span></div>
-<div class="line"><a name="l00137"></a><span class="lineno">  137</span>&#160;</div>
-<div class="line"><a name="l00138"></a><span class="lineno">  138</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00139"></a><span class="lineno">  139</span>&#160;<span class="comment">// Standard Layout Types        N2342   GCC 4.5</span></div>
-<div class="line"><a name="l00140"></a><span class="lineno">  140</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2342.htm</span></div>
-<div class="line"><a name="l00141"></a><span class="lineno">  141</span>&#160;</div>
-<div class="line"><a name="l00142"></a><span class="lineno">  142</span>&#160;<span class="comment">// #define GLM_CXX11_DEFAULTED_FUNCTIONS</span></div>
-<div class="line"><a name="l00143"></a><span class="lineno">  143</span>&#160;<span class="comment">// #define GLM_CXX11_DELETED_FUNCTIONS</span></div>
-<div class="line"><a name="l00144"></a><span class="lineno">  144</span>&#160;<span class="comment">// Defaulted and deleted functions      N2346   GCC 4.4</span></div>
-<div class="line"><a name="l00145"></a><span class="lineno">  145</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2346.htm</span></div>
-<div class="line"><a name="l00146"></a><span class="lineno">  146</span>&#160;</div>
-<div class="line"><a name="l00147"></a><span class="lineno">  147</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00148"></a><span class="lineno">  148</span>&#160;<span class="comment">// Extended friend declarations N1791   GCC 4.7</span></div>
-<div class="line"><a name="l00149"></a><span class="lineno">  149</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1791.pdf</span></div>
-<div class="line"><a name="l00150"></a><span class="lineno">  150</span>&#160;</div>
-<div class="line"><a name="l00151"></a><span class="lineno">  151</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00152"></a><span class="lineno">  152</span>&#160;<span class="comment">// Extending sizeof     N2253   GCC 4.4</span></div>
-<div class="line"><a name="l00153"></a><span class="lineno">  153</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2253.html</span></div>
-<div class="line"><a name="l00154"></a><span class="lineno">  154</span>&#160;</div>
-<div class="line"><a name="l00155"></a><span class="lineno">  155</span>&#160;<span class="comment">// #define GLM_CXX11_INLINE_NAMESPACES</span></div>
-<div class="line"><a name="l00156"></a><span class="lineno">  156</span>&#160;<span class="comment">// Inline namespaces    N2535   GCC 4.4</span></div>
-<div class="line"><a name="l00157"></a><span class="lineno">  157</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2535.htm</span></div>
-<div class="line"><a name="l00158"></a><span class="lineno">  158</span>&#160;</div>
-<div class="line"><a name="l00159"></a><span class="lineno">  159</span>&#160;<span class="comment">// #define GLM_CXX11_UNRESTRICTED_UNIONS</span></div>
-<div class="line"><a name="l00160"></a><span class="lineno">  160</span>&#160;<span class="comment">// Unrestricted unions  N2544   GCC 4.6</span></div>
-<div class="line"><a name="l00161"></a><span class="lineno">  161</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2544.pdf</span></div>
-<div class="line"><a name="l00162"></a><span class="lineno">  162</span>&#160;</div>
-<div class="line"><a name="l00163"></a><span class="lineno">  163</span>&#160;<span class="comment">// #define GLM_CXX11_LOCAL_TYPE_TEMPLATE_ARGS</span></div>
-<div class="line"><a name="l00164"></a><span class="lineno">  164</span>&#160;<span class="comment">// Local and unnamed types as template arguments        N2657   GCC 4.5</span></div>
-<div class="line"><a name="l00165"></a><span class="lineno">  165</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm</span></div>
-<div class="line"><a name="l00166"></a><span class="lineno">  166</span>&#160;</div>
-<div class="line"><a name="l00167"></a><span class="lineno">  167</span>&#160;<span class="comment">// #define GLM_CXX11_RANGE_FOR</span></div>
-<div class="line"><a name="l00168"></a><span class="lineno">  168</span>&#160;<span class="comment">// Range-based for      N2930   GCC 4.6</span></div>
-<div class="line"><a name="l00169"></a><span class="lineno">  169</span>&#160;<span class="comment">// http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2009/n2930.html</span></div>
-<div class="line"><a name="l00170"></a><span class="lineno">  170</span>&#160;</div>
-<div class="line"><a name="l00171"></a><span class="lineno">  171</span>&#160;<span class="comment">// #define GLM_CXX11_OVERRIDE_CONTROL</span></div>
-<div class="line"><a name="l00172"></a><span class="lineno">  172</span>&#160;<span class="comment">// Explicit virtual overrides   N2928 N3206 N3272       GCC 4.7</span></div>
-<div class="line"><a name="l00173"></a><span class="lineno">  173</span>&#160;<span class="comment">// http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2009/n2928.htm</span></div>
-<div class="line"><a name="l00174"></a><span class="lineno">  174</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3206.htm</span></div>
-<div class="line"><a name="l00175"></a><span class="lineno">  175</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3272.htm</span></div>
-<div class="line"><a name="l00176"></a><span class="lineno">  176</span>&#160;</div>
-<div class="line"><a name="l00177"></a><span class="lineno">  177</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00178"></a><span class="lineno">  178</span>&#160;<span class="comment">// Minimal support for garbage collection and reachability-based leak detection N2670   No</span></div>
-<div class="line"><a name="l00179"></a><span class="lineno">  179</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2670.htm</span></div>
-<div class="line"><a name="l00180"></a><span class="lineno">  180</span>&#160;</div>
-<div class="line"><a name="l00181"></a><span class="lineno">  181</span>&#160;<span class="comment">// #define GLM_CXX11_NOEXCEPT</span></div>
-<div class="line"><a name="l00182"></a><span class="lineno">  182</span>&#160;<span class="comment">// Allowing move constructors to throw [noexcept]       N3050   GCC 4.6 (core language only)</span></div>
-<div class="line"><a name="l00183"></a><span class="lineno">  183</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3050.html</span></div>
-<div class="line"><a name="l00184"></a><span class="lineno">  184</span>&#160;</div>
-<div class="line"><a name="l00185"></a><span class="lineno">  185</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00186"></a><span class="lineno">  186</span>&#160;<span class="comment">// Defining move special member functions       N3053   GCC 4.6</span></div>
-<div class="line"><a name="l00187"></a><span class="lineno">  187</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3053.html</span></div>
-<div class="line"><a name="l00188"></a><span class="lineno">  188</span>&#160;</div>
-<div class="line"><a name="l00189"></a><span class="lineno">  189</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00190"></a><span class="lineno">  190</span>&#160;<span class="comment">// Sequence points      N2239   Yes</span></div>
-<div class="line"><a name="l00191"></a><span class="lineno">  191</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2239.html</span></div>
-<div class="line"><a name="l00192"></a><span class="lineno">  192</span>&#160;</div>
-<div class="line"><a name="l00193"></a><span class="lineno">  193</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00194"></a><span class="lineno">  194</span>&#160;<span class="comment">// Atomic operations    N2427   GCC 4.4</span></div>
-<div class="line"><a name="l00195"></a><span class="lineno">  195</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2239.html</span></div>
-<div class="line"><a name="l00196"></a><span class="lineno">  196</span>&#160;</div>
-<div class="line"><a name="l00197"></a><span class="lineno">  197</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00198"></a><span class="lineno">  198</span>&#160;<span class="comment">// Strong Compare and Exchange  N2748   GCC 4.5</span></div>
-<div class="line"><a name="l00199"></a><span class="lineno">  199</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2427.html</span></div>
-<div class="line"><a name="l00200"></a><span class="lineno">  200</span>&#160;</div>
-<div class="line"><a name="l00201"></a><span class="lineno">  201</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00202"></a><span class="lineno">  202</span>&#160;<span class="comment">// Bidirectional Fences N2752   GCC 4.8</span></div>
-<div class="line"><a name="l00203"></a><span class="lineno">  203</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2752.htm</span></div>
-<div class="line"><a name="l00204"></a><span class="lineno">  204</span>&#160;</div>
-<div class="line"><a name="l00205"></a><span class="lineno">  205</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00206"></a><span class="lineno">  206</span>&#160;<span class="comment">// Memory model N2429   GCC 4.8</span></div>
-<div class="line"><a name="l00207"></a><span class="lineno">  207</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2429.htm</span></div>
-<div class="line"><a name="l00208"></a><span class="lineno">  208</span>&#160;</div>
-<div class="line"><a name="l00209"></a><span class="lineno">  209</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00210"></a><span class="lineno">  210</span>&#160;<span class="comment">// Data-dependency ordering: atomics and memory model   N2664   GCC 4.4</span></div>
-<div class="line"><a name="l00211"></a><span class="lineno">  211</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2664.htm</span></div>
-<div class="line"><a name="l00212"></a><span class="lineno">  212</span>&#160;</div>
-<div class="line"><a name="l00213"></a><span class="lineno">  213</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00214"></a><span class="lineno">  214</span>&#160;<span class="comment">// Propagating exceptions       N2179   GCC 4.4</span></div>
-<div class="line"><a name="l00215"></a><span class="lineno">  215</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2179.html</span></div>
-<div class="line"><a name="l00216"></a><span class="lineno">  216</span>&#160;</div>
-<div class="line"><a name="l00217"></a><span class="lineno">  217</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00218"></a><span class="lineno">  218</span>&#160;<span class="comment">// Abandoning a process and at_quick_exit       N2440   GCC 4.8</span></div>
-<div class="line"><a name="l00219"></a><span class="lineno">  219</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2440.htm</span></div>
-<div class="line"><a name="l00220"></a><span class="lineno">  220</span>&#160;</div>
-<div class="line"><a name="l00221"></a><span class="lineno">  221</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00222"></a><span class="lineno">  222</span>&#160;<span class="comment">// Allow atomics use in signal handlers N2547   Yes</span></div>
-<div class="line"><a name="l00223"></a><span class="lineno">  223</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2547.htm</span></div>
-<div class="line"><a name="l00224"></a><span class="lineno">  224</span>&#160;</div>
-<div class="line"><a name="l00225"></a><span class="lineno">  225</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00226"></a><span class="lineno">  226</span>&#160;<span class="comment">// Thread-local storage N2659   GCC 4.8</span></div>
-<div class="line"><a name="l00227"></a><span class="lineno">  227</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2659.htm</span></div>
-<div class="line"><a name="l00228"></a><span class="lineno">  228</span>&#160;</div>
-<div class="line"><a name="l00229"></a><span class="lineno">  229</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00230"></a><span class="lineno">  230</span>&#160;<span class="comment">// Dynamic initialization and destruction with concurrency      N2660   GCC 4.3</span></div>
-<div class="line"><a name="l00231"></a><span class="lineno">  231</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2660.htm</span></div>
-<div class="line"><a name="l00232"></a><span class="lineno">  232</span>&#160;</div>
-<div class="line"><a name="l00233"></a><span class="lineno">  233</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00234"></a><span class="lineno">  234</span>&#160;<span class="comment">// __func__ predefined identifier       N2340   GCC 4.3</span></div>
-<div class="line"><a name="l00235"></a><span class="lineno">  235</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2340.htm</span></div>
-<div class="line"><a name="l00236"></a><span class="lineno">  236</span>&#160;</div>
-<div class="line"><a name="l00237"></a><span class="lineno">  237</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00238"></a><span class="lineno">  238</span>&#160;<span class="comment">// C99 preprocessor     N1653   GCC 4.3</span></div>
-<div class="line"><a name="l00239"></a><span class="lineno">  239</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1653.htm</span></div>
-<div class="line"><a name="l00240"></a><span class="lineno">  240</span>&#160;</div>
-<div class="line"><a name="l00241"></a><span class="lineno">  241</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00242"></a><span class="lineno">  242</span>&#160;<span class="comment">// long long    N1811   GCC 4.3</span></div>
-<div class="line"><a name="l00243"></a><span class="lineno">  243</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1811.pdf</span></div>
-<div class="line"><a name="l00244"></a><span class="lineno">  244</span>&#160;</div>
-<div class="line"><a name="l00245"></a><span class="lineno">  245</span>&#160;<span class="comment">//</span></div>
-<div class="line"><a name="l00246"></a><span class="lineno">  246</span>&#160;<span class="comment">// Extended integral types      N1988   Yes</span></div>
-<div class="line"><a name="l00247"></a><span class="lineno">  247</span>&#160;<span class="comment">// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1988.pdf</span></div>
-<div class="line"><a name="l00248"></a><span class="lineno">  248</span>&#160;</div>
-<div class="line"><a name="l00249"></a><span class="lineno">  249</span>&#160;<span class="preprocessor">#if(GLM_COMPILER &amp; GLM_COMPILER_GCC)</span></div>
-<div class="line"><a name="l00250"></a><span class="lineno">  250</span>&#160;</div>
-<div class="line"><a name="l00251"></a><span class="lineno">  251</span>&#160;<span class="preprocessor">#       define GLM_CXX11_STATIC_ASSERT</span></div>
-<div class="line"><a name="l00252"></a><span class="lineno">  252</span>&#160;</div>
-<div class="line"><a name="l00253"></a><span class="lineno">  253</span>&#160;<span class="preprocessor">#elif(GLM_COMPILER &amp; GLM_COMPILER_CLANG)</span></div>
-<div class="line"><a name="l00254"></a><span class="lineno">  254</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_exceptions))</span></div>
-<div class="line"><a name="l00255"></a><span class="lineno">  255</span>&#160;<span class="preprocessor">#               define GLM_CXX98_EXCEPTIONS</span></div>
-<div class="line"><a name="l00256"></a><span class="lineno">  256</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00257"></a><span class="lineno">  257</span>&#160;</div>
-<div class="line"><a name="l00258"></a><span class="lineno">  258</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_rtti))</span></div>
-<div class="line"><a name="l00259"></a><span class="lineno">  259</span>&#160;<span class="preprocessor">#               define GLM_CXX98_RTTI</span></div>
-<div class="line"><a name="l00260"></a><span class="lineno">  260</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00261"></a><span class="lineno">  261</span>&#160;</div>
-<div class="line"><a name="l00262"></a><span class="lineno">  262</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_access_control_sfinae))</span></div>
-<div class="line"><a name="l00263"></a><span class="lineno">  263</span>&#160;<span class="preprocessor">#               define GLM_CXX11_ACCESS_CONTROL_SFINAE</span></div>
-<div class="line"><a name="l00264"></a><span class="lineno">  264</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00265"></a><span class="lineno">  265</span>&#160;</div>
-<div class="line"><a name="l00266"></a><span class="lineno">  266</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_alias_templates))</span></div>
-<div class="line"><a name="l00267"></a><span class="lineno">  267</span>&#160;<span class="preprocessor">#               define GLM_CXX11_ALIAS_TEMPLATE</span></div>
-<div class="line"><a name="l00268"></a><span class="lineno">  268</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00269"></a><span class="lineno">  269</span>&#160;</div>
-<div class="line"><a name="l00270"></a><span class="lineno">  270</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_alignas))</span></div>
-<div class="line"><a name="l00271"></a><span class="lineno">  271</span>&#160;<span class="preprocessor">#               define GLM_CXX11_ALIGNAS</span></div>
-<div class="line"><a name="l00272"></a><span class="lineno">  272</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00273"></a><span class="lineno">  273</span>&#160;</div>
-<div class="line"><a name="l00274"></a><span class="lineno">  274</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_attributes))</span></div>
-<div class="line"><a name="l00275"></a><span class="lineno">  275</span>&#160;<span class="preprocessor">#               define GLM_CXX11_ATTRIBUTES</span></div>
-<div class="line"><a name="l00276"></a><span class="lineno">  276</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00277"></a><span class="lineno">  277</span>&#160;</div>
-<div class="line"><a name="l00278"></a><span class="lineno">  278</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_constexpr))</span></div>
-<div class="line"><a name="l00279"></a><span class="lineno">  279</span>&#160;<span class="preprocessor">#               define GLM_CXX11_CONSTEXPR</span></div>
-<div class="line"><a name="l00280"></a><span class="lineno">  280</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00281"></a><span class="lineno">  281</span>&#160;</div>
-<div class="line"><a name="l00282"></a><span class="lineno">  282</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_decltype))</span></div>
-<div class="line"><a name="l00283"></a><span class="lineno">  283</span>&#160;<span class="preprocessor">#               define GLM_CXX11_DECLTYPE</span></div>
-<div class="line"><a name="l00284"></a><span class="lineno">  284</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00285"></a><span class="lineno">  285</span>&#160;</div>
-<div class="line"><a name="l00286"></a><span class="lineno">  286</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_default_function_template_args))</span></div>
-<div class="line"><a name="l00287"></a><span class="lineno">  287</span>&#160;<span class="preprocessor">#               define GLM_CXX11_DEFAULT_FUNCTION_TEMPLATE_ARGS</span></div>
-<div class="line"><a name="l00288"></a><span class="lineno">  288</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00289"></a><span class="lineno">  289</span>&#160;</div>
-<div class="line"><a name="l00290"></a><span class="lineno">  290</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_defaulted_functions))</span></div>
-<div class="line"><a name="l00291"></a><span class="lineno">  291</span>&#160;<span class="preprocessor">#               define GLM_CXX11_DEFAULTED_FUNCTIONS</span></div>
-<div class="line"><a name="l00292"></a><span class="lineno">  292</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00293"></a><span class="lineno">  293</span>&#160;</div>
-<div class="line"><a name="l00294"></a><span class="lineno">  294</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_delegating_constructors))</span></div>
-<div class="line"><a name="l00295"></a><span class="lineno">  295</span>&#160;<span class="preprocessor">#               define GLM_CXX11_DELEGATING_CONSTRUCTORS</span></div>
-<div class="line"><a name="l00296"></a><span class="lineno">  296</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00297"></a><span class="lineno">  297</span>&#160;</div>
-<div class="line"><a name="l00298"></a><span class="lineno">  298</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_deleted_functions))</span></div>
-<div class="line"><a name="l00299"></a><span class="lineno">  299</span>&#160;<span class="preprocessor">#               define GLM_CXX11_DELETED_FUNCTIONS</span></div>
-<div class="line"><a name="l00300"></a><span class="lineno">  300</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00301"></a><span class="lineno">  301</span>&#160;</div>
-<div class="line"><a name="l00302"></a><span class="lineno">  302</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_explicit_conversions))</span></div>
-<div class="line"><a name="l00303"></a><span class="lineno">  303</span>&#160;<span class="preprocessor">#               define GLM_CXX11_EXPLICIT_CONVERSIONS</span></div>
-<div class="line"><a name="l00304"></a><span class="lineno">  304</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00305"></a><span class="lineno">  305</span>&#160;</div>
-<div class="line"><a name="l00306"></a><span class="lineno">  306</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_generalized_initializers))</span></div>
-<div class="line"><a name="l00307"></a><span class="lineno">  307</span>&#160;<span class="preprocessor">#               define GLM_CXX11_GENERALIZED_INITIALIZERS</span></div>
-<div class="line"><a name="l00308"></a><span class="lineno">  308</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00309"></a><span class="lineno">  309</span>&#160;</div>
-<div class="line"><a name="l00310"></a><span class="lineno">  310</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_implicit_moves))</span></div>
-<div class="line"><a name="l00311"></a><span class="lineno">  311</span>&#160;<span class="preprocessor">#               define GLM_CXX11_IMPLICIT_MOVES</span></div>
-<div class="line"><a name="l00312"></a><span class="lineno">  312</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00313"></a><span class="lineno">  313</span>&#160;</div>
-<div class="line"><a name="l00314"></a><span class="lineno">  314</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_inheriting_constructors))</span></div>
-<div class="line"><a name="l00315"></a><span class="lineno">  315</span>&#160;<span class="preprocessor">#               define GLM_CXX11_INHERITING_CONSTRUCTORS</span></div>
-<div class="line"><a name="l00316"></a><span class="lineno">  316</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00317"></a><span class="lineno">  317</span>&#160;</div>
-<div class="line"><a name="l00318"></a><span class="lineno">  318</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_inline_namespaces))</span></div>
-<div class="line"><a name="l00319"></a><span class="lineno">  319</span>&#160;<span class="preprocessor">#               define GLM_CXX11_INLINE_NAMESPACES</span></div>
-<div class="line"><a name="l00320"></a><span class="lineno">  320</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00321"></a><span class="lineno">  321</span>&#160;</div>
-<div class="line"><a name="l00322"></a><span class="lineno">  322</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_lambdas))</span></div>
-<div class="line"><a name="l00323"></a><span class="lineno">  323</span>&#160;<span class="preprocessor">#               define GLM_CXX11_LAMBDAS</span></div>
-<div class="line"><a name="l00324"></a><span class="lineno">  324</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00325"></a><span class="lineno">  325</span>&#160;</div>
-<div class="line"><a name="l00326"></a><span class="lineno">  326</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_local_type_template_args))</span></div>
-<div class="line"><a name="l00327"></a><span class="lineno">  327</span>&#160;<span class="preprocessor">#               define GLM_CXX11_LOCAL_TYPE_TEMPLATE_ARGS</span></div>
-<div class="line"><a name="l00328"></a><span class="lineno">  328</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00329"></a><span class="lineno">  329</span>&#160;</div>
-<div class="line"><a name="l00330"></a><span class="lineno">  330</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_noexcept))</span></div>
-<div class="line"><a name="l00331"></a><span class="lineno">  331</span>&#160;<span class="preprocessor">#               define GLM_CXX11_NOEXCEPT</span></div>
-<div class="line"><a name="l00332"></a><span class="lineno">  332</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00333"></a><span class="lineno">  333</span>&#160;</div>
-<div class="line"><a name="l00334"></a><span class="lineno">  334</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_nonstatic_member_init))</span></div>
-<div class="line"><a name="l00335"></a><span class="lineno">  335</span>&#160;<span class="preprocessor">#               define GLM_CXX11_NONSTATIC_MEMBER_INIT</span></div>
-<div class="line"><a name="l00336"></a><span class="lineno">  336</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00337"></a><span class="lineno">  337</span>&#160;</div>
-<div class="line"><a name="l00338"></a><span class="lineno">  338</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_nullptr))</span></div>
-<div class="line"><a name="l00339"></a><span class="lineno">  339</span>&#160;<span class="preprocessor">#               define GLM_CXX11_NULLPTR</span></div>
-<div class="line"><a name="l00340"></a><span class="lineno">  340</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00341"></a><span class="lineno">  341</span>&#160;</div>
-<div class="line"><a name="l00342"></a><span class="lineno">  342</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_override_control))</span></div>
-<div class="line"><a name="l00343"></a><span class="lineno">  343</span>&#160;<span class="preprocessor">#               define GLM_CXX11_OVERRIDE_CONTROL</span></div>
-<div class="line"><a name="l00344"></a><span class="lineno">  344</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00345"></a><span class="lineno">  345</span>&#160;</div>
-<div class="line"><a name="l00346"></a><span class="lineno">  346</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_reference_qualified_functions))</span></div>
-<div class="line"><a name="l00347"></a><span class="lineno">  347</span>&#160;<span class="preprocessor">#               define GLM_CXX11_REFERENCE_QUALIFIED_FUNCTIONS</span></div>
-<div class="line"><a name="l00348"></a><span class="lineno">  348</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00349"></a><span class="lineno">  349</span>&#160;</div>
-<div class="line"><a name="l00350"></a><span class="lineno">  350</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_range_for))</span></div>
-<div class="line"><a name="l00351"></a><span class="lineno">  351</span>&#160;<span class="preprocessor">#               define GLM_CXX11_RANGE_FOR</span></div>
-<div class="line"><a name="l00352"></a><span class="lineno">  352</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00353"></a><span class="lineno">  353</span>&#160;</div>
-<div class="line"><a name="l00354"></a><span class="lineno">  354</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_raw_string_literals))</span></div>
-<div class="line"><a name="l00355"></a><span class="lineno">  355</span>&#160;<span class="preprocessor">#               define GLM_CXX11_RAW_STRING_LITERALS</span></div>
-<div class="line"><a name="l00356"></a><span class="lineno">  356</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00357"></a><span class="lineno">  357</span>&#160;</div>
-<div class="line"><a name="l00358"></a><span class="lineno">  358</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_rvalue_references))</span></div>
-<div class="line"><a name="l00359"></a><span class="lineno">  359</span>&#160;<span class="preprocessor">#               define GLM_CXX11_RVALUE_REFERENCES</span></div>
-<div class="line"><a name="l00360"></a><span class="lineno">  360</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00361"></a><span class="lineno">  361</span>&#160;</div>
-<div class="line"><a name="l00362"></a><span class="lineno">  362</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_static_assert))</span></div>
-<div class="line"><a name="l00363"></a><span class="lineno">  363</span>&#160;<span class="preprocessor">#               define GLM_CXX11_STATIC_ASSERT</span></div>
-<div class="line"><a name="l00364"></a><span class="lineno">  364</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00365"></a><span class="lineno">  365</span>&#160;</div>
-<div class="line"><a name="l00366"></a><span class="lineno">  366</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_auto_type))</span></div>
-<div class="line"><a name="l00367"></a><span class="lineno">  367</span>&#160;<span class="preprocessor">#               define GLM_CXX11_AUTO_TYPE</span></div>
-<div class="line"><a name="l00368"></a><span class="lineno">  368</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00369"></a><span class="lineno">  369</span>&#160;</div>
-<div class="line"><a name="l00370"></a><span class="lineno">  370</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_strong_enums))</span></div>
-<div class="line"><a name="l00371"></a><span class="lineno">  371</span>&#160;<span class="preprocessor">#               define GLM_CXX11_STRONG_ENUMS</span></div>
-<div class="line"><a name="l00372"></a><span class="lineno">  372</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00373"></a><span class="lineno">  373</span>&#160;</div>
-<div class="line"><a name="l00374"></a><span class="lineno">  374</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_trailing_return))</span></div>
-<div class="line"><a name="l00375"></a><span class="lineno">  375</span>&#160;<span class="preprocessor">#               define GLM_CXX11_TRAILING_RETURN</span></div>
-<div class="line"><a name="l00376"></a><span class="lineno">  376</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00377"></a><span class="lineno">  377</span>&#160;</div>
-<div class="line"><a name="l00378"></a><span class="lineno">  378</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_unicode_literals))</span></div>
-<div class="line"><a name="l00379"></a><span class="lineno">  379</span>&#160;<span class="preprocessor">#               define GLM_CXX11_UNICODE_LITERALS</span></div>
-<div class="line"><a name="l00380"></a><span class="lineno">  380</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00381"></a><span class="lineno">  381</span>&#160;</div>
-<div class="line"><a name="l00382"></a><span class="lineno">  382</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_unrestricted_unions))</span></div>
-<div class="line"><a name="l00383"></a><span class="lineno">  383</span>&#160;<span class="preprocessor">#               define GLM_CXX11_UNRESTRICTED_UNIONS</span></div>
-<div class="line"><a name="l00384"></a><span class="lineno">  384</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00385"></a><span class="lineno">  385</span>&#160;</div>
-<div class="line"><a name="l00386"></a><span class="lineno">  386</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_user_literals))</span></div>
-<div class="line"><a name="l00387"></a><span class="lineno">  387</span>&#160;<span class="preprocessor">#               define GLM_CXX11_USER_LITERALS</span></div>
-<div class="line"><a name="l00388"></a><span class="lineno">  388</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00389"></a><span class="lineno">  389</span>&#160;</div>
-<div class="line"><a name="l00390"></a><span class="lineno">  390</span>&#160;<span class="preprocessor">#       if(__has_feature(cxx_variadic_templates))</span></div>
-<div class="line"><a name="l00391"></a><span class="lineno">  391</span>&#160;<span class="preprocessor">#               define GLM_CXX11_VARIADIC_TEMPLATES</span></div>
-<div class="line"><a name="l00392"></a><span class="lineno">  392</span>&#160;<span class="preprocessor">#       endif</span></div>
-<div class="line"><a name="l00393"></a><span class="lineno">  393</span>&#160;</div>
-<div class="line"><a name="l00394"></a><span class="lineno">  394</span>&#160;<span class="preprocessor">#endif//(GLM_COMPILER &amp; GLM_COMPILER_CLANG)</span></div>
-</div><!-- fragment --></div><!-- contents -->
-<!-- start footer part -->
-<hr class="footer"/><address class="footer"><small>
-Generated by &#160;<a href="http://www.doxygen.org/index.html">
-<img class="footer" src="doxygen.png" alt="doxygen"/>
-</a> 1.8.10
-</small></address>
-</body>
-</html>

diff-gaussian-rasterization/third_party/glm/doc/api/a00002_source.html

@@ -1,121 +0,0 @@
-<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
-<html xmlns="http://www.w3.org/1999/xhtml">
-<head>
-<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
-<meta http-equiv="X-UA-Compatible" content="IE=9"/>
-<meta name="generator" content="Doxygen 1.8.10"/>
-<title>0.9.9 API documentation: _fixes.hpp Source File</title>
-<link href="tabs.css" rel="stylesheet" type="text/css"/>
-<script type="text/javascript" src="jquery.js"></script>
-<script type="text/javascript" src="dynsections.js"></script>
-<link href="search/search.css" rel="stylesheet" type="text/css"/>
-<script type="text/javascript" src="search/searchdata.js"></script>
-<script type="text/javascript" src="search/search.js"></script>
-<script type="text/javascript">
-  $(document).ready(function() { init_search(); });
-</script>
-<link href="doxygen.css" rel="stylesheet" type="text/css" />
-</head>
-<body>
-<div id="top"><!-- do not remove this div, it is closed by doxygen! -->
-<div id="titlearea">
-<table cellspacing="0" cellpadding="0">
- <tbody>
- <tr style="height: 56px;">
-  <td id="projectlogo"><img alt="Logo" src="logo-mini.png"/></td>
-  <td id="projectalign" style="padding-left: 0.5em;">
-   <div id="projectname">0.9.9 API documentation
-   </div>
-  </td>
- </tr>
- </tbody>
-</table>
-</div>
-<!-- end header part -->
-<!-- Generated by Doxygen 1.8.10 -->
-<script type="text/javascript">
-var searchBox = new SearchBox("searchBox", "search",false,'Search');
-</script>
-  <div id="navrow1" class="tabs">
-    <ul class="tablist">
-      <li><a href="index.html"><span>Main&#160;Page</span></a></li>
-      <li><a href="modules.html"><span>Modules</span></a></li>
-      <li class="current"><a href="files.html"><span>Files</span></a></li>
-      <li>
-        <div id="MSearchBox" class="MSearchBoxInactive">
-        <span class="left">
-          <img id="MSearchSelect" src="search/mag_sel.png"
-               onmouseover="return searchBox.OnSearchSelectShow()"
-               onmouseout="return searchBox.OnSearchSelectHide()"
-               alt=""/>
-          <input type="text" id="MSearchField" value="Search" accesskey="S"
-               onfocus="searchBox.OnSearchFieldFocus(true)" 
-               onblur="searchBox.OnSearchFieldFocus(false)" 
-               onkeyup="searchBox.OnSearchFieldChange(event)"/>
-          </span><span class="right">
-            <a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="search/close.png" alt=""/></a>
-          </span>
-        </div>
-      </li>
-    </ul>
-  </div>
-  <div id="navrow2" class="tabs2">
-    <ul class="tablist">
-      <li><a href="files.html"><span>File&#160;List</span></a></li>
-    </ul>
-  </div>
-<!-- window showing the filter options -->
-<div id="MSearchSelectWindow"
-     onmouseover="return searchBox.OnSearchSelectShow()"
-     onmouseout="return searchBox.OnSearchSelectHide()"
-     onkeydown="return searchBox.OnSearchSelectKey(event)">
-</div>
-
-<!-- iframe showing the search results (closed by default) -->
-<div id="MSearchResultsWindow">
-<iframe src="javascript:void(0)" frameborder="0" 
-        name="MSearchResults" id="MSearchResults">
-</iframe>
-</div>
-
-<div id="nav-path" class="navpath">
-  <ul>
-<li class="navelem"><a class="el" href="dir_3a581ba30d25676e4b797b1f96d53b45.html">F:</a></li><li class="navelem"><a class="el" href="dir_9e5fe034a00e89334fd5186c3e7db156.html">G-Truc</a></li><li class="navelem"><a class="el" href="dir_d9496f0844b48bc7e53b5af8c99b9ab2.html">Source</a></li><li class="navelem"><a class="el" href="dir_a8bee7be44182a33f3820393ae0b105d.html">G-Truc</a></li><li class="navelem"><a class="el" href="dir_44e5e654415abd9ca6fdeaddaff8565e.html">glm</a></li><li class="navelem"><a class="el" href="dir_cef2d71d502cb69a9252bca2297d9549.html">glm</a></li><li class="navelem"><a class="el" href="dir_033f5edb0915b828d2c46ed4804e5503.html">detail</a></li>  </ul>
-</div>
-</div><!-- top -->
-<div class="header">
-  <div class="headertitle">
-<div class="title">_fixes.hpp</div>  </div>
-</div><!--header-->
-<div class="contents">
-<div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno">    1</span>&#160;<span class="preprocessor">#include &lt;cmath&gt;</span></div>
-<div class="line"><a name="l00002"></a><span class="lineno">    2</span>&#160;</div>
-<div class="line"><a name="l00004"></a><span class="lineno">    4</span>&#160;<span class="preprocessor">#ifdef max</span></div>
-<div class="line"><a name="l00005"></a><span class="lineno">    5</span>&#160;<span class="preprocessor">#undef max</span></div>
-<div class="line"><a name="l00006"></a><span class="lineno">    6</span>&#160;<span class="preprocessor">#endif</span></div>
-<div class="line"><a name="l00007"></a><span class="lineno">    7</span>&#160;</div>
-<div class="line"><a name="l00009"></a><span class="lineno">    9</span>&#160;<span class="preprocessor">#ifdef min</span></div>
-<div class="line"><a name="l00010"></a><span class="lineno">   10</span>&#160;<span class="preprocessor">#undef min</span></div>
-<div class="line"><a name="l00011"></a><span class="lineno">   11</span>&#160;<span class="preprocessor">#endif</span></div>
-<div class="line"><a name="l00012"></a><span class="lineno">   12</span>&#160;</div>
-<div class="line"><a name="l00014"></a><span class="lineno">   14</span>&#160;<span class="preprocessor">#ifdef isnan</span></div>
-<div class="line"><a name="l00015"></a><span class="lineno">   15</span>&#160;<span class="preprocessor">#undef isnan</span></div>
-<div class="line"><a name="l00016"></a><span class="lineno">   16</span>&#160;<span class="preprocessor">#endif</span></div>
-<div class="line"><a name="l00017"></a><span class="lineno">   17</span>&#160;</div>
-<div class="line"><a name="l00019"></a><span class="lineno">   19</span>&#160;<span class="preprocessor">#ifdef isinf</span></div>
-<div class="line"><a name="l00020"></a><span class="lineno">   20</span>&#160;<span class="preprocessor">#undef isinf</span></div>
-<div class="line"><a name="l00021"></a><span class="lineno">   21</span>&#160;<span class="preprocessor">#endif</span></div>
-<div class="line"><a name="l00022"></a><span class="lineno">   22</span>&#160;</div>
-<div class="line"><a name="l00024"></a><span class="lineno">   24</span>&#160;<span class="preprocessor">#ifdef log2</span></div>
-<div class="line"><a name="l00025"></a><span class="lineno">   25</span>&#160;<span class="preprocessor">#undef log2</span></div>
-<div class="line"><a name="l00026"></a><span class="lineno">   26</span>&#160;<span class="preprocessor">#endif</span></div>
-<div class="line"><a name="l00027"></a><span class="lineno">   27</span>&#160;</div>
-</div><!-- fragment --></div><!-- contents -->
-<!-- start footer part -->
-<hr class="footer"/><address class="footer"><small>
-Generated by &#160;<a href="http://www.doxygen.org/index.html">
-<img class="footer" src="doxygen.png" alt="doxygen"/>
-</a> 1.8.10
-</small></address>
-</body>
-</html>

diff-gaussian-rasterization/third_party/glm/doc/api/a00003_source.html

@@ -1,182 +0,0 @@
-<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
-<html xmlns="http://www.w3.org/1999/xhtml">
-<head>
-<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
-<meta http-equiv="X-UA-Compatible" content="IE=9"/>
-<meta name="generator" content="Doxygen 1.8.10"/>
-<title>0.9.9 API documentation: _noise.hpp Source File</title>
-<link href="tabs.css" rel="stylesheet" type="text/css"/>
-<script type="text/javascript" src="jquery.js"></script>
-<script type="text/javascript" src="dynsections.js"></script>
-<link href="search/search.css" rel="stylesheet" type="text/css"/>
-<script type="text/javascript" src="search/searchdata.js"></script>
-<script type="text/javascript" src="search/search.js"></script>
-<script type="text/javascript">
-  $(document).ready(function() { init_search(); });
-</script>
-<link href="doxygen.css" rel="stylesheet" type="text/css" />
-</head>
-<body>
-<div id="top"><!-- do not remove this div, it is closed by doxygen! -->
-<div id="titlearea">
-<table cellspacing="0" cellpadding="0">
- <tbody>
- <tr style="height: 56px;">
-  <td id="projectlogo"><img alt="Logo" src="logo-mini.png"/></td>
-  <td id="projectalign" style="padding-left: 0.5em;">
-   <div id="projectname">0.9.9 API documentation
-   </div>
-  </td>
- </tr>
- </tbody>
-</table>
-</div>
-<!-- end header part -->
-<!-- Generated by Doxygen 1.8.10 -->
-<script type="text/javascript">
-var searchBox = new SearchBox("searchBox", "search",false,'Search');
-</script>
-  <div id="navrow1" class="tabs">
-    <ul class="tablist">
-      <li><a href="index.html"><span>Main&#160;Page</span></a></li>
-      <li><a href="modules.html"><span>Modules</span></a></li>
-      <li class="current"><a href="files.html"><span>Files</span></a></li>
-      <li>
-        <div id="MSearchBox" class="MSearchBoxInactive">
-        <span class="left">
-          <img id="MSearchSelect" src="search/mag_sel.png"
-               onmouseover="return searchBox.OnSearchSelectShow()"
-               onmouseout="return searchBox.OnSearchSelectHide()"
-               alt=""/>
-          <input type="text" id="MSearchField" value="Search" accesskey="S"
-               onfocus="searchBox.OnSearchFieldFocus(true)" 
-               onblur="searchBox.OnSearchFieldFocus(false)" 
-               onkeyup="searchBox.OnSearchFieldChange(event)"/>
-          </span><span class="right">
-            <a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="search/close.png" alt=""/></a>
-          </span>
-        </div>
-      </li>
-    </ul>
-  </div>
-  <div id="navrow2" class="tabs2">
-    <ul class="tablist">
-      <li><a href="files.html"><span>File&#160;List</span></a></li>
-    </ul>
-  </div>
-<!-- window showing the filter options -->
-<div id="MSearchSelectWindow"
-     onmouseover="return searchBox.OnSearchSelectShow()"
-     onmouseout="return searchBox.OnSearchSelectHide()"
-     onkeydown="return searchBox.OnSearchSelectKey(event)">
-</div>
-
-<!-- iframe showing the search results (closed by default) -->
-<div id="MSearchResultsWindow">
-<iframe src="javascript:void(0)" frameborder="0" 
-        name="MSearchResults" id="MSearchResults">
-</iframe>
-</div>
-
-<div id="nav-path" class="navpath">
-  <ul>
-<li class="navelem"><a class="el" href="dir_3a581ba30d25676e4b797b1f96d53b45.html">F:</a></li><li class="navelem"><a class="el" href="dir_9e5fe034a00e89334fd5186c3e7db156.html">G-Truc</a></li><li class="navelem"><a class="el" href="dir_d9496f0844b48bc7e53b5af8c99b9ab2.html">Source</a></li><li class="navelem"><a class="el" href="dir_a8bee7be44182a33f3820393ae0b105d.html">G-Truc</a></li><li class="navelem"><a class="el" href="dir_44e5e654415abd9ca6fdeaddaff8565e.html">glm</a></li><li class="navelem"><a class="el" href="dir_cef2d71d502cb69a9252bca2297d9549.html">glm</a></li><li class="navelem"><a class="el" href="dir_033f5edb0915b828d2c46ed4804e5503.html">detail</a></li>  </ul>
-</div>
-</div><!-- top -->
-<div class="header">
-  <div class="headertitle">
-<div class="title">_noise.hpp</div>  </div>
-</div><!--header-->
-<div class="contents">
-<div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno">    1</span>&#160;<span class="preprocessor">#pragma once</span></div>
-<div class="line"><a name="l00002"></a><span class="lineno">    2</span>&#160;</div>
-<div class="line"><a name="l00003"></a><span class="lineno">    3</span>&#160;<span class="preprocessor">#include &quot;../common.hpp&quot;</span></div>
-<div class="line"><a name="l00004"></a><span class="lineno">    4</span>&#160;</div>
-<div class="line"><a name="l00005"></a><span class="lineno">    5</span>&#160;<span class="keyword">namespace </span><a class="code" href="a00236.html">glm</a>{</div>
-<div class="line"><a name="l00006"></a><span class="lineno">    6</span>&#160;<span class="keyword">namespace </span>detail</div>
-<div class="line"><a name="l00007"></a><span class="lineno">    7</span>&#160;{</div>
-<div class="line"><a name="l00008"></a><span class="lineno">    8</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T&gt;</div>
-<div class="line"><a name="l00009"></a><span class="lineno">    9</span>&#160;        GLM_FUNC_QUALIFIER T mod289(T <span class="keyword">const</span>&amp; x)</div>
-<div class="line"><a name="l00010"></a><span class="lineno">   10</span>&#160;        {</div>
-<div class="line"><a name="l00011"></a><span class="lineno">   11</span>&#160;                <span class="keywordflow">return</span> x - <a class="code" href="a00241.html#gaa9d0742639e85b29c7c5de11cfd6840d">floor</a>(x * (static_cast&lt;T&gt;(1.0) / static_cast&lt;T&gt;(289.0))) * <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(289.0);</div>
-<div class="line"><a name="l00012"></a><span class="lineno">   12</span>&#160;        }</div>
-<div class="line"><a name="l00013"></a><span class="lineno">   13</span>&#160;</div>
-<div class="line"><a name="l00014"></a><span class="lineno">   14</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T&gt;</div>
-<div class="line"><a name="l00015"></a><span class="lineno">   15</span>&#160;        GLM_FUNC_QUALIFIER T permute(T <span class="keyword">const</span>&amp; x)</div>
-<div class="line"><a name="l00016"></a><span class="lineno">   16</span>&#160;        {</div>
-<div class="line"><a name="l00017"></a><span class="lineno">   17</span>&#160;                <span class="keywordflow">return</span> mod289(((x * static_cast&lt;T&gt;(34)) + static_cast&lt;T&gt;(1)) * x);</div>
-<div class="line"><a name="l00018"></a><span class="lineno">   18</span>&#160;        }</div>
-<div class="line"><a name="l00019"></a><span class="lineno">   19</span>&#160;</div>
-<div class="line"><a name="l00020"></a><span class="lineno">   20</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T, qualifier Q&gt;</div>
-<div class="line"><a name="l00021"></a><span class="lineno">   21</span>&#160;        GLM_FUNC_QUALIFIER vec&lt;2, T, Q&gt; permute(vec&lt;2, T, Q&gt; <span class="keyword">const</span>&amp; x)</div>
-<div class="line"><a name="l00022"></a><span class="lineno">   22</span>&#160;        {</div>
-<div class="line"><a name="l00023"></a><span class="lineno">   23</span>&#160;                <span class="keywordflow">return</span> mod289(((x * static_cast&lt;T&gt;(34)) + static_cast&lt;T&gt;(1)) * x);</div>
-<div class="line"><a name="l00024"></a><span class="lineno">   24</span>&#160;        }</div>
-<div class="line"><a name="l00025"></a><span class="lineno">   25</span>&#160;</div>
-<div class="line"><a name="l00026"></a><span class="lineno">   26</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T, qualifier Q&gt;</div>
-<div class="line"><a name="l00027"></a><span class="lineno">   27</span>&#160;        GLM_FUNC_QUALIFIER vec&lt;3, T, Q&gt; permute(vec&lt;3, T, Q&gt; <span class="keyword">const</span>&amp; x)</div>
-<div class="line"><a name="l00028"></a><span class="lineno">   28</span>&#160;        {</div>
-<div class="line"><a name="l00029"></a><span class="lineno">   29</span>&#160;                <span class="keywordflow">return</span> mod289(((x * static_cast&lt;T&gt;(34)) + static_cast&lt;T&gt;(1)) * x);</div>
-<div class="line"><a name="l00030"></a><span class="lineno">   30</span>&#160;        }</div>
-<div class="line"><a name="l00031"></a><span class="lineno">   31</span>&#160;</div>
-<div class="line"><a name="l00032"></a><span class="lineno">   32</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T, qualifier Q&gt;</div>
-<div class="line"><a name="l00033"></a><span class="lineno">   33</span>&#160;        GLM_FUNC_QUALIFIER vec&lt;4, T, Q&gt; permute(vec&lt;4, T, Q&gt; <span class="keyword">const</span>&amp; x)</div>
-<div class="line"><a name="l00034"></a><span class="lineno">   34</span>&#160;        {</div>
-<div class="line"><a name="l00035"></a><span class="lineno">   35</span>&#160;                <span class="keywordflow">return</span> mod289(((x * static_cast&lt;T&gt;(34)) + static_cast&lt;T&gt;(1)) * x);</div>
-<div class="line"><a name="l00036"></a><span class="lineno">   36</span>&#160;        }</div>
-<div class="line"><a name="l00037"></a><span class="lineno">   37</span>&#160;</div>
-<div class="line"><a name="l00038"></a><span class="lineno">   38</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T&gt;</div>
-<div class="line"><a name="l00039"></a><span class="lineno">   39</span>&#160;        GLM_FUNC_QUALIFIER T taylorInvSqrt(T <span class="keyword">const</span>&amp; r)</div>
-<div class="line"><a name="l00040"></a><span class="lineno">   40</span>&#160;        {</div>
-<div class="line"><a name="l00041"></a><span class="lineno">   41</span>&#160;                <span class="keywordflow">return</span> <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(1.79284291400159) - <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(0.85373472095314) * r;</div>
-<div class="line"><a name="l00042"></a><span class="lineno">   42</span>&#160;        }</div>
-<div class="line"><a name="l00043"></a><span class="lineno">   43</span>&#160;</div>
-<div class="line"><a name="l00044"></a><span class="lineno">   44</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T, qualifier Q&gt;</div>
-<div class="line"><a name="l00045"></a><span class="lineno">   45</span>&#160;        GLM_FUNC_QUALIFIER vec&lt;2, T, Q&gt; taylorInvSqrt(vec&lt;2, T, Q&gt; <span class="keyword">const</span>&amp; r)</div>
-<div class="line"><a name="l00046"></a><span class="lineno">   46</span>&#160;        {</div>
-<div class="line"><a name="l00047"></a><span class="lineno">   47</span>&#160;                <span class="keywordflow">return</span> <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(1.79284291400159) - <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(0.85373472095314) * r;</div>
-<div class="line"><a name="l00048"></a><span class="lineno">   48</span>&#160;        }</div>
-<div class="line"><a name="l00049"></a><span class="lineno">   49</span>&#160;</div>
-<div class="line"><a name="l00050"></a><span class="lineno">   50</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T, qualifier Q&gt;</div>
-<div class="line"><a name="l00051"></a><span class="lineno">   51</span>&#160;        GLM_FUNC_QUALIFIER vec&lt;3, T, Q&gt; taylorInvSqrt(vec&lt;3, T, Q&gt; <span class="keyword">const</span>&amp; r)</div>
-<div class="line"><a name="l00052"></a><span class="lineno">   52</span>&#160;        {</div>
-<div class="line"><a name="l00053"></a><span class="lineno">   53</span>&#160;                <span class="keywordflow">return</span> <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(1.79284291400159) - <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(0.85373472095314) * r;</div>
-<div class="line"><a name="l00054"></a><span class="lineno">   54</span>&#160;        }</div>
-<div class="line"><a name="l00055"></a><span class="lineno">   55</span>&#160;</div>
-<div class="line"><a name="l00056"></a><span class="lineno">   56</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T, qualifier Q&gt;</div>
-<div class="line"><a name="l00057"></a><span class="lineno">   57</span>&#160;        GLM_FUNC_QUALIFIER vec&lt;4, T, Q&gt; taylorInvSqrt(vec&lt;4, T, Q&gt; <span class="keyword">const</span>&amp; r)</div>
-<div class="line"><a name="l00058"></a><span class="lineno">   58</span>&#160;        {</div>
-<div class="line"><a name="l00059"></a><span class="lineno">   59</span>&#160;                <span class="keywordflow">return</span> <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(1.79284291400159) - <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(0.85373472095314) * r;</div>
-<div class="line"><a name="l00060"></a><span class="lineno">   60</span>&#160;        }</div>
-<div class="line"><a name="l00061"></a><span class="lineno">   61</span>&#160;</div>
-<div class="line"><a name="l00062"></a><span class="lineno">   62</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T, qualifier Q&gt;</div>
-<div class="line"><a name="l00063"></a><span class="lineno">   63</span>&#160;        GLM_FUNC_QUALIFIER vec&lt;2, T, Q&gt; fade(vec&lt;2, T, Q&gt; <span class="keyword">const</span>&amp; t)</div>
-<div class="line"><a name="l00064"></a><span class="lineno">   64</span>&#160;        {</div>
-<div class="line"><a name="l00065"></a><span class="lineno">   65</span>&#160;                <span class="keywordflow">return</span> (t * t * t) * (t * (t * <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(6) - static_cast&lt;T&gt;(15)) + static_cast&lt;T&gt;(10));</div>
-<div class="line"><a name="l00066"></a><span class="lineno">   66</span>&#160;        }</div>
-<div class="line"><a name="l00067"></a><span class="lineno">   67</span>&#160;</div>
-<div class="line"><a name="l00068"></a><span class="lineno">   68</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T, qualifier Q&gt;</div>
-<div class="line"><a name="l00069"></a><span class="lineno">   69</span>&#160;        GLM_FUNC_QUALIFIER vec&lt;3, T, Q&gt; fade(vec&lt;3, T, Q&gt; <span class="keyword">const</span>&amp; t)</div>
-<div class="line"><a name="l00070"></a><span class="lineno">   70</span>&#160;        {</div>
-<div class="line"><a name="l00071"></a><span class="lineno">   71</span>&#160;                <span class="keywordflow">return</span> (t * t * t) * (t * (t * <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(6) - static_cast&lt;T&gt;(15)) + static_cast&lt;T&gt;(10));</div>
-<div class="line"><a name="l00072"></a><span class="lineno">   72</span>&#160;        }</div>
-<div class="line"><a name="l00073"></a><span class="lineno">   73</span>&#160;</div>
-<div class="line"><a name="l00074"></a><span class="lineno">   74</span>&#160;        <span class="keyword">template</span>&lt;<span class="keyword">typename</span> T, qualifier Q&gt;</div>
-<div class="line"><a name="l00075"></a><span class="lineno">   75</span>&#160;        GLM_FUNC_QUALIFIER vec&lt;4, T, Q&gt; fade(vec&lt;4, T, Q&gt; <span class="keyword">const</span>&amp; t)</div>
-<div class="line"><a name="l00076"></a><span class="lineno">   76</span>&#160;        {</div>
-<div class="line"><a name="l00077"></a><span class="lineno">   77</span>&#160;                <span class="keywordflow">return</span> (t * t * t) * (t * (t * <span class="keyword">static_cast&lt;</span>T<span class="keyword">&gt;</span>(6) - static_cast&lt;T&gt;(15)) + static_cast&lt;T&gt;(10));</div>
-<div class="line"><a name="l00078"></a><span class="lineno">   78</span>&#160;        }</div>
-<div class="line"><a name="l00079"></a><span class="lineno">   79</span>&#160;}<span class="comment">//namespace detail</span></div>
-<div class="line"><a name="l00080"></a><span class="lineno">   80</span>&#160;}<span class="comment">//namespace glm</span></div>
-<div class="line"><a name="l00081"></a><span class="lineno">   81</span>&#160;</div>
-<div class="ttc" id="a00241_html_gaa9d0742639e85b29c7c5de11cfd6840d"><div class="ttname"><a href="a00241.html#gaa9d0742639e85b29c7c5de11cfd6840d">glm::floor</a></div><div class="ttdeci">GLM_FUNC_DECL vec&lt; L, T, Q &gt; floor(vec&lt; L, T, Q &gt; const &amp;x)</div><div class="ttdoc">Returns a value equal to the nearest integer that is less then or equal to x. </div></div>
-<div class="ttc" id="a00236_html"><div class="ttname"><a href="a00236.html">glm</a></div><div class="ttdef"><b>Definition:</b> <a href="a00015_source.html#l00020">common.hpp:20</a></div></div>
-</div><!-- fragment --></div><!-- contents -->
-<!-- start footer part -->
-<hr class="footer"/><address class="footer"><small>
-Generated by &#160;<a href="http://www.doxygen.org/index.html">
-<img class="footer" src="doxygen.png" alt="doxygen"/>
-</a> 1.8.10
-</small></address>
-</body>
-</html>