Silicate crashes on files with many layers

[wilywyrm]

Very cool project! I'm trying to use your compositor to automate exports of layer groups for my weird workflow, and finding that it crashes for the largest of my files (~1GB, 102 layers at 2732 x 2048px each) with the following error:

[src\compositor\dev.rs:52] adapter.get_info() = AdapterInfo {
    name: "AMD Radeon RX 6800 XT",
    vendor: 4098,
    device: 29631,
    device_type: DiscreteGpu,
    driver: "AMD proprietary driver",
    driver_info: "23.5.2 (AMD proprietary shader compiler)",
    backend: Vulkan,
}
[src\compositor\dev.rs:53] adapter.limits() = Limits {
    max_texture_dimension_1d: 16384,
    max_texture_dimension_2d: 16384,
    max_texture_dimension_3d: 8192,
    max_texture_array_layers: 8192,
    max_bind_groups: 8,
    max_bindings_per_bind_group: 640,
    max_dynamic_uniform_buffers_per_pipeline_layout: 8,
    max_dynamic_storage_buffers_per_pipeline_layout: 8,
    max_sampled_textures_per_shader_stage: 4294967295,
    max_samplers_per_shader_stage: 4294967295,
    max_storage_buffers_per_shader_stage: 4294967295,
    max_storage_textures_per_shader_stage: 4294967295,
    max_uniform_buffers_per_shader_stage: 4294967295,
    max_uniform_buffer_binding_size: 2147483648,
    max_storage_buffer_binding_size: 2147483648,
    max_vertex_buffers: 16,
    max_buffer_size: 18446744073709551615,
    max_vertex_attributes: 64,
    max_vertex_buffer_array_stride: 16383,
    min_uniform_buffer_offset_alignment: 32,
    min_storage_buffer_offset_alignment: 32,
    max_inter_stage_shader_components: 128,
    max_compute_workgroup_storage_size: 32768,
    max_compute_invocations_per_workgroup: 1024,
    max_compute_workgroup_size_x: 1024,
    max_compute_workgroup_size_y: 1024,
    max_compute_workgroup_size_z: 1024,
    max_compute_workgroups_per_dimension: 65535,
    max_push_constant_size: 128,
}
thread 'tokio-runtime-worker' panicked at 'wgpu error: Validation Error

Caused by:
    In Device::create_texture
    Not enough memory left

', C:\Users\runneradmin\.cargo\registry\src\index.crates.io-6f17d22bba15001f\wgpu-0.16.3\src\backend\direct.rs:3019:5
stack backtrace:
   0:     0x7ff772d93bc1 - <unknown>
   1:     0x7ff772baf46b - <unknown>
   2:     0x7ff772d87639 - <unknown>
   3:     0x7ff772d95acb - <unknown>
   4:     0x7ff772d95899 - <unknown>
   5:     0x7ff772d9657b - <unknown>
   6:     0x7ff772d96025 - <unknown>
   7:     0x7ff772d95f79 - <unknown>
   8:     0x7ff772d95f64 - <unknown>
   9:     0x7ff77300b735 - <unknown>
  10:     0x7ff772dcbcec - <unknown>
  11:     0x7ff772dfd2de - <unknown>
  12:     0x7ff772e3c342 - <unknown>
  13:     0x7ff772ed6e29 - <unknown>
  14:     0x7ff772f4490d - <unknown>
  15:     0x7ff772b7086a - <unknown>
  16:     0x7ff772b23a40 - <unknown>
  17:     0x7ff772da0797 - <unknown>
  18:     0x7ff772da054d - <unknown>
  19:     0x7ff772d9333b - <unknown>
  20:     0x7ffa755f7614 - BaseThreadInitThunk
  21:     0x7ffa76ea26b1 - RtlUserThreadStart

I'm running the exe from the latest master build for windows with RUST_BACKTRACE=full. It sounds like the canvas data is being loaded into vmem and the representation used is punching through the available memory. Am new to rust but happy to dig into it if you've got pointers where to start.