gl_device: Reserve at least 4 image bindings for fragment stage

Due to the limitation of GL_MAX_IMAGE_UNITS being low (8) on Intel's and Nvidia's proprietary drivers, we have to reserve an appropriate amount of image bindings for each of the stages. So far games have been observed to use 4 image bindings on the fragment stage (Kirby Star Allies) and 1 on the vertex stage (TWD series).
No games thus far in my limited testing used more than 4 images concurrently and across all currently active programs.
This fixes shader compilation errors on Kirby Star Allies on OpenGL (GLSL/GLASM)
This commit is contained in:
Morph 2020-06-16 01:30:01 -04:00
parent ed2cd9d8f3
commit e2f5d16540
1 changed files with 14 additions and 6 deletions

View File

@ -123,16 +123,24 @@ std::array<Device::BaseBindings, Tegra::Engines::MaxShaderTypes> BuildBaseBindin
u32 num_images = GetInteger<u32>(GL_MAX_IMAGE_UNITS);
u32 base_images = 0;
// Reserve more image bindings on fragment and vertex stages.
// GL_MAX_IMAGE_UNITS is guaranteed by the spec to have a minimum value of 8.
// Due to the limitation of GL_MAX_IMAGE_UNITS, reserve at least 4 image bindings on the
// fragment stage, and at least 1 for the rest of the stages.
// So far games are observed to use 1 image binding on vertex and 4 on fragment stages.
// Reserve at least 4 image bindings on the fragment stage.
bindings[4].image =
Extract(base_images, num_images, num_images / NumStages + 2, LimitImages[4]);
bindings[0].image =
Extract(base_images, num_images, num_images / NumStages + 1, LimitImages[0]);
Extract(base_images, num_images, std::max(4U, num_images / NumStages), LimitImages[4]);
// This is guaranteed to be at least 1.
const u32 total_extracted_images = num_images / (NumStages - 1);
// Reserve the other image bindings.
const u32 total_extracted_images = num_images / (NumStages - 2);
for (std::size_t i = 2; i < NumStages; ++i) {
for (std::size_t i = 0; i < NumStages; ++i) {
const std::size_t stage = stage_swizzle[i];
if (stage == 4) {
continue;
}
bindings[stage].image =
Extract(base_images, num_images, total_extracted_images, LimitImages[stage]);
}