-
-
Notifications
You must be signed in to change notification settings - Fork 3.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Meshlet single pass depth downsampling (SPD) #13003
Conversation
Which parts do you want me to review? Presumably |
The downsample shader, yeah. I can't post a link ATM, but you can find the full PR diff by changing the GitHub diff to compare against my meshlet-previous-frame-depth-pyramid branch. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Overall a straightforward port of FidelityFX, but I'm concerned about the lack of texture barrier and some more comments explaining the approach would be nice. Seems fine other than that.
tex = vec2(workgroup_id * 64u) + vec2(x * 2u + 32u, y * 2u + 32u); | ||
pix = vec2(workgroup_id * 32u) + vec2(x + 16u, y + 16u); | ||
v[3] = reduce_load_mip_0(tex); | ||
textureStore(mip_1, pix, vec4(v[3])); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm not a big fan of the way there's so much repetition here, but I see that you're copying from FidelityFX, so I'm OK with it.
for (var i = 0u; i < 4u; i++) { | ||
intermediate_memory[x][y] = v[i]; | ||
workgroupBarrier(); | ||
if local_invocation_index < 64u { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ok, I had to stare at this a while. I assume what you're doing here is to work around the lack of subgroupQuadSwapX
intrinsics. That's why you do local_invocation_index < 64
instead of local_invocation_index % 16 == 0
like FidelityFX does.
Can you add a comment saying that this is a workaround for lack of subgroup quad swap operations?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh wait, I guess this is similar to SpdDownsampleMips_0_1_LDS
. OK, I see. Maybe add a TODO here so we can remember to update it when we have subgroup ops.
@group(0) @binding(0) var input_depth: texture_2d<f32>; | ||
@group(0) @binding(1) var samplr: sampler; | ||
/// Generates a hierarchal depth buffer. | ||
/// Based on FidelityFX SPD https://github.com/GPUOpen-LibrariesAndSDKs/FidelityFX-SDK/blob/d7531ae47d8b36a5d4025663e731a47a38be882f/sdk/include/FidelityFX/gpu/spd/ffx_spd.h#L528 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'd add a brief comment here explaining the overall approach. Perhaps:
Every thread we dispatch is responsible for four 2x2 quads in the original depth buffer. Each workgroup has a shared 16x16 texel tile to work with. Different approaches are used at each mip level:
Mip level 0 (64x64): Use texture gather instructions to sample at the center of each 2x2 quad. Do this 4 times for the 4 quads we're responsible for. Save each downsampled depth value, effectively slicing each 32x32 piece into four 16x16 pieces.
Mip level 1 (32x32): For each of the four 16x16 pieces we have, copy it into the tile memory, then downsample it and barrier. This results in a 16x16 tile, which we store in our tile memory.
Mip levels 2-5 (16x16 to 2x2): The entire tile fits in memory, so we just downsample it in place.
Mip levels 6+: Load from the global texture and downsample 4 pixels at a time. Only one workgroup remains at this point.
Downscale is broken, will need to debug. Might be the barrier issue. |
…er, but this time from a shader
…instance-only-data-upload-squashed
…instance-only-data-upload-squashed
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good now. Looking forward to this!
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight` or `PointLight`. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, non-temporal operation, and temporal operation. The assets are my original work. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight` or `PointLight`. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, non-temporal operation, and temporal operation. The assets are my original work. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight` or `PointLight`. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, non-temporal operation, and temporal operation. The assets are my original work. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight` or `PointLight`. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, non-temporal operation, and temporal operation. The assets are my original work. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight` or `PointLight`. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, spot lights, non-temporal operation, and temporal operation. The assets are my original work. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight` or `PointLight`. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). In addition to PCSS, this commit makes the near Z plane for the shadow map configurable on a per-light basis. Previously, it had been hardcoded to 0.1. This was necessary to make the point light shadow map in the example look reasonable, as otherwise the shadows appeared far too aliased. A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, spot lights, non-temporal operation, and temporal operation. The assets are my original work. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight`, `PointLight`, or `SpotLight`, as appropriate. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). In addition to PCSS, this commit makes the near Z plane for the shadow map configurable on a per-light basis. Previously, it had been hardcoded to 0.1 meters. This change was necessary to make the point light shadow map in the example look reasonable, as otherwise the shadows appeared far too aliased. A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, spot lights, non-temporal operation, and temporal operation. The assets are my original work. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight`, `PointLight`, or `SpotLight`, as appropriate. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). In addition to PCSS, this commit makes the near Z plane for the shadow map configurable on a per-light basis. Previously, it had been hardcoded to 0.1 meters. This change was necessary to make the point light shadow map in the example look reasonable, as otherwise the shadows appeared far too aliased. A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, spot lights, non-temporal operation, and temporal operation. The assets are my original work. Both temporal and non-temporal shadows are rather noisy in the example, and, as mentioned before, this is unavoidable without downsampling the depth buffer, which we can't do yet. Note also that the shadows don't look particularly great for point lights; the example simply isn't an ideal scene for them. Nevertheless, I felt that the benefits of the ability to do a side-by-side comparison of directional and point lights outweighed the unsightliness of the point light shadows in that example, so I kept the point light feature in. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight`, `PointLight`, or `SpotLight`, as appropriate. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). In addition to PCSS, this commit makes the near Z plane for the shadow map configurable on a per-light basis. Previously, it had been hardcoded to 0.1 meters. This change was necessary to make the point light shadow map in the example look reasonable, as otherwise the shadows appeared far too aliased. A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, spot lights, non-temporal operation, and temporal operation. The assets are my original work. Both temporal and non-temporal shadows are rather noisy in the example, and, as mentioned before, this is unavoidable without downsampling the depth buffer, which we can't do yet. Note also that the shadows don't look particularly great for point lights; the example simply isn't an ideal scene for them. Nevertheless, I felt that the benefits of the ability to do a side-by-side comparison of directional and point lights outweighed the unsightliness of the point light shadows in that example, so I kept the point light feature in. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight`, `PointLight`, or `SpotLight`, as appropriate. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). In addition to PCSS, this commit makes the near Z plane for the shadow map configurable on a per-light basis. Previously, it had been hardcoded to 0.1 meters. This change was necessary to make the point light shadow map in the example look reasonable, as otherwise the shadows appeared far too aliased. A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, spot lights, non-temporal operation, and temporal operation. The assets are my original work. Both temporal and non-temporal shadows are rather noisy in the example, and, as mentioned before, this is unavoidable without downsampling the depth buffer, which we can't do yet. Note also that the shadows don't look particularly great for point lights; the example simply isn't an ideal scene for them. Nevertheless, I felt that the benefits of the ability to do a side-by-side comparison of directional and point lights outweighed the unsightliness of the point light shadows in that example, so I kept the point light feature in. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight`, `PointLight`, or `SpotLight`, as appropriate. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). In addition to PCSS, this commit makes the near Z plane for the shadow map configurable on a per-light basis. Previously, it had been hardcoded to 0.1 meters. This change was necessary to make the point light shadow map in the example look reasonable, as otherwise the shadows appeared far too aliased. A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, spot lights, non-temporal operation, and temporal operation. The assets are my original work. Both temporal and non-temporal shadows are rather noisy in the example, and, as mentioned before, this is unavoidable without downsampling the depth buffer, which we can't do yet. Note also that the shadows don't look particularly great for point lights; the example simply isn't an ideal scene for them. Nevertheless, I felt that the benefits of the ability to do a side-by-side comparison of directional and point lights outweighed the unsightliness of the point light shadows in that example, so I kept the point light feature in. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight`, `PointLight`, or `SpotLight`, as appropriate. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). In addition to PCSS, this commit makes the near Z plane for the shadow map configurable on a per-light basis. Previously, it had been hardcoded to 0.1 meters. This change was necessary to make the point light shadow map in the example look reasonable, as otherwise the shadows appeared far too aliased. A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, spot lights, non-temporal operation, and temporal operation. The assets are my original work. Both temporal and non-temporal shadows are rather noisy in the example, and, as mentioned before, this is unavoidable without downsampling the depth buffer, which we can't do yet. Note also that the shadows don't look particularly great for point lights; the example simply isn't an ideal scene for them. Nevertheless, I felt that the benefits of the ability to do a side-by-side comparison of directional and point lights outweighed the unsightliness of the point light shadows in that example, so I kept the point light feature in. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight`, `PointLight`, or `SpotLight`, as appropriate. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR bevyengine#13003 to land first). In addition to PCSS, this commit makes the near Z plane for the shadow map configurable on a per-light basis. Previously, it had been hardcoded to 0.1 meters. This change was necessary to make the point light shadow map in the example look reasonable, as otherwise the shadows appeared far too aliased. A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, spot lights, non-temporal operation, and temporal operation. The assets are my original work. Both temporal and non-temporal shadows are rather noisy in the example, and, as mentioned before, this is unavoidable without downsampling the depth buffer, which we can't do yet. Note also that the shadows don't look particularly great for point lights; the example simply isn't an ideal scene for them. Nevertheless, I felt that the benefits of the ability to do a side-by-side comparison of directional and point lights outweighed the unsightliness of the point light shadows in that example, so I kept the point light feature in. Fixes bevyengine#3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
looks good. i've already looked through this code a few times before, just never left an approve
texture_depth_2d(), | ||
write_only_r32float(), | ||
write_only_r32float(), | ||
write_only_r32float(), | ||
write_only_r32float(), | ||
write_only_r32float(), | ||
texture_storage_2d( | ||
TextureFormat::R32Float, | ||
StorageTextureAccess::ReadWrite, | ||
), | ||
write_only_r32float(), | ||
write_only_r32float(), | ||
write_only_r32float(), | ||
write_only_r32float(), | ||
write_only_r32float(), | ||
write_only_r32float(), |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I guess this could help a bit
texture_depth_2d(), | |
write_only_r32float(), | |
write_only_r32float(), | |
write_only_r32float(), | |
write_only_r32float(), | |
write_only_r32float(), | |
texture_storage_2d( | |
TextureFormat::R32Float, | |
StorageTextureAccess::ReadWrite, | |
), | |
write_only_r32float(), | |
write_only_r32float(), | |
write_only_r32float(), | |
write_only_r32float(), | |
write_only_r32float(), | |
write_only_r32float(), | |
// view depth | |
texture_depth_2d(), | |
// mip 1 | |
write_only_r32float(), | |
// mip 2 | |
write_only_r32float(), | |
// mip 3 | |
write_only_r32float(), | |
// mip 4 | |
write_only_r32float(), | |
// mip 5 | |
write_only_r32float(), | |
// mip 6 | |
texture_storage_2d( | |
TextureFormat::R32Float, | |
StorageTextureAccess::ReadWrite, | |
), | |
// mip 7 | |
write_only_r32float(), | |
// mip 8 | |
write_only_r32float(), | |
// mip 9 | |
write_only_r32float(), | |
// mip 10 | |
write_only_r32float(), | |
// mip 11 | |
write_only_r32float(), | |
// mip 12 | |
write_only_r32float(), |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I didn't go over the shader code, but I trust the other reviewers on that.
I believe that there might be an improvement that can be made here:
From what I understand from the spec, textureGather already computes the four component minimum and stores it in the
It is very possible that the compiler spots this optimization already. Frankly I'm not sure how |
textureGather does not return the minimum of the 4 values. The w component is the (u_min, v_min) value of the sample footprint. I.e. given 4 texels (the sample footprint) arranged in a 2x2 quad, the w component is the value at location (u_min, v_min). Yes, ideally I would be able to use VK_SAMPLER_REDUCTION_MODE_MIN, but unfortunately wgpu does not support it. |
[*Percentage-closer soft shadows*] are a technique from 2004 that allow shadows to become blurrier farther from the objects that cast them. It works by introducing a *blocker search* step that runs before the normal shadow map sampling. The blocker search step detects the difference between the depth of the fragment being rasterized and the depth of the nearby samples in the depth buffer. Larger depth differences result in a larger penumbra and therefore a blurrier shadow. To enable PCSS, fill in the `soft_shadow_size` value in `DirectionalLight`, `PointLight`, or `SpotLight`, as appropriate. This shadow size value represents the size of the light and should be tuned as appropriate for your scene. Higher values result in a wider penumbra (i.e. blurrier shadows). When using PCSS, temporal shadow maps (`ShadowFilteringMethod::Temporal`) are recommended. If you don't use `ShadowFilteringMethod::Temporal` and instead use `ShadowFilteringMethod::Gaussian`, Bevy will use the same technique as `Temporal`, but the result won't vary over time. This produces a rather noisy result. Doing better would likely require downsampling the shadow map, which would be complex and slower (and would require PR #13003 to land first). In addition to PCSS, this commit makes the near Z plane for the shadow map configurable on a per-light basis. Previously, it had been hardcoded to 0.1 meters. This change was necessary to make the point light shadow map in the example look reasonable, as otherwise the shadows appeared far too aliased. A new example, `pcss`, has been added. It demonstrates the percentage-closer soft shadow technique with directional lights, point lights, spot lights, non-temporal operation, and temporal operation. The assets are my original work. Both temporal and non-temporal shadows are rather noisy in the example, and, as mentioned before, this is unavoidable without downsampling the depth buffer, which we can't do yet. Note also that the shadows don't look particularly great for point lights; the example simply isn't an ideal scene for them. Nevertheless, I felt that the benefits of the ability to do a side-by-side comparison of directional and point lights outweighed the unsightliness of the point light shadows in that example, so I kept the point light feature in. Fixes #3631. [*Percentage-closer soft shadows*]: https://developer.download.nvidia.com/shaderlibrary/docs/shadow_PCSS.pdf ## Changelog ### Added * Percentage-closer soft shadows (PCSS) are now supported, allowing shadows to become blurrier as they stretch away from objects. To use them, set the `soft_shadow_size` field in `DirectionalLight`, `PointLight`, or `SpotLight`, as applicable. * The near Z value for shadow maps is now customizable via the `shadow_map_near_z` field in `DirectionalLight`, `PointLight`, and `SpotLight`. ## Screenshots PCSS off: ![Screenshot 2024-05-24 120012](https://github.com/bevyengine/bevy/assets/157897/0d35fe98-245b-44fb-8a43-8d0272a73b86) PCSS on: ![Screenshot 2024-05-24 115959](https://github.com/bevyengine/bevy/assets/157897/83397ef8-1317-49dd-bfb3-f8286d7610cd) --------- Co-authored-by: Alice Cecile <[email protected]> Co-authored-by: Torstein Grindvik <[email protected]>
Objective
Solution
singletwo compute dispatches. Lack of coherent buffers means we have to do the the last 64x64 tile from mip 7+ in a separate dispatch to ensure the mip 6 writes were flushed :(