From: Henri Verbeet Subject: Re: [PATCH 1/4] d3d9/tests: Test more formats in srgbtexture_test(). Message-Id: Date: Tue, 21 Jan 2020 19:57:48 +0330 In-Reply-To: <4173d837-7e4a-da63-30b3-eea1ced8df10@gmail.com> References: <20200113153121.529336-1-gofmanp@gmail.com> <4173d837-7e4a-da63-30b3-eea1ced8df10@gmail.com> On Tue, 21 Jan 2020 at 18:46, Paul Gofman wrote: > I've tested the game with AMD graphics card override under Wine > (VideoPciVendorID 0x1002, VideoPciDeviceId 0x67df) and it seems to fix > the rendering on its own. So the game expects _R5G6B5 to have SRGB reads > supported only if detects the GPU as Nvidia (at least), otherwise it > works some other way. Yet I suppose adding the SRGB read support as on > Nvidia might be a valid fix unless we want to make the support depend on > the reported video card. > It's a little unfortunate to use GL_SRGB8 for R5G6B5 because it implies conversions on uploads and downloads, as well as doubling GPU memory usage for the affected formats. I don't think that's necessarily prohibitive, but if there's a reasonable way to avoid supporting sRGB reads on these formats, I think that would be preferred.