750862 - Implement WEBGL_compressed_texture_pvrtc

Status: RESOLVED FIXED

(Core :: Graphics: CanvasWebGL, defect)

Description

[Benoit Jacob [:bjacob] (mostly away)]

We can't start implementing them yet as they don't have a draft-status spec . We need to make a spec proposal, submit it to the public_webgl list, and get it accepted as a draft before we can start implementing with MOZ_ prefix.

Comment 1

[Benoit Jacob [:bjacob] (mostly away)]

PVRTC proposal has been sent to the list.
https://www.khronos.org/registry/webgl/extensions/proposals/WEBGL_compressed_texture_pvrtc/

Regarding ATC we have agreed on the public_webgl list that it being officially undocumented (see http://www.khronos.org/registry/gles/extensions/AMD/AMD_compressed_ATC_texture.txt) makes it a non-starter for WebGL.

Therefore it makes sense to repurpose this bug as the PVRTC implementation bug.

Comment 2

[Martin Best (:mbest)]

Do we have any information about the compressed texture formats currently supported by Android devices in the wild?

Comment 3

[Benoit Jacob [:bjacob] (mostly away)]

Got Qualcomm to publish the ATC buffer size formula -- see public_webgl archive for September. Implementation in bug 790712.

Comment 4

[Benoit Jacob [:bjacob] (mostly away)]

(In reply to Martin Best (:mbest) from comment #2)
> Do we have any information about the compressed texture formats currently
> supported by Android devices in the wild?

No precise numbers, but we know that S3TC+PVRTC+ATC allow us to cover most devices. Qualcomm GPUs are ATC-only, while some other devices don't support ATC.

Comment 5

[Benoit Jacob [:bjacob] (mostly away)]

The extension has been promoted to draft:
http://www.khronos.org/registry/webgl/extensions/WEBGL_compressed_texture_pvrtc/

Can now implement -- writing the patch now.

Comment 6

[Benoit Jacob [:bjacob] (mostly away)]

Attachment: pvrtc

This applies on top of the ATC patch, which refactored ValidateCompressedTextureSize a bit.

This patch further refactors this by having ValidateCompressedTextureSize call ValidateLevelWidthHeightForTarget, rather than relying on the caller to call both. This fixes a bug whereby we weren't checking for negative / out-of-range texture dimensions in compressedTexSubImage2D, afaict.

Comment 7

[Kelsey Gilbert [:jgilbert]]

Comment on attachment 664105 [details] [diff] [review]
pvrtc

Review of attachment 664105 [details] [diff] [review]:
-----------------------------------------------------------------

An optional nit, but otherwise good.

::: gfx/gl/GLDefs.h
@@ +3101,5 @@
> +// IMG_texture_compression_pvrtc
> +#define LOCAL_GL_COMPRESSED_RGB_PVRTC_4BPPV1_IMG            0x8C00
> +#define LOCAL_GL_COMPRESSED_RGB_PVRTC_2BPPV1_IMG            0x8C01
> +#define LOCAL_GL_COMPRESSED_RGBA_PVRTC_4BPPV1_IMG           0x8C02
> +#define LOCAL_GL_COMPRESSED_RGBA_PVRTC_2BPPV1_IMG           0x8C03

Prefer to drop vendor suffixes for GL constants for internal use.

Comment 8

[Benoit Jacob [:bjacob] (mostly away)]

http://hg.mozilla.org/integration/mozilla-inbound/rev/33f80363eab3

Comment 9

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/33f80363eab3

Comment 10

[Stefan Sitter [:ssitter]]

I think this and Bug 790712 have broken my Thunderbird build using ac_add_options --disable-webgl. Do you need to update content/canvas/src/WebGLContextNotSupported.cpp?

Comment 11

[Benoit Jacob [:bjacob] (mostly away)]

Yes we do! Thanks for the heads-up.

Comment 12

[Benoit Jacob [:bjacob] (mostly away)]

http://hg.mozilla.org/integration/mozilla-inbound/rev/fc7098623145

Comment 13

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/fc7098623145

Comment 14

[Florian Scholz (Open Web Docs)]

https://developer.mozilla.org/en-US/docs/Web/API/WEBGL_compressed_texture_pvrtc
https://developer.mozilla.org/en-US/docs/Web/API/WebGLRenderingContext/compressedTexImage2D
https://developer.mozilla.org/en-US/docs/Web/API/WebGLRenderingContext/compressedTexSubImage2D