mctaylors/suyu
1
0
Fork 0
forked from suyu/suyu
Code Pull requests Activity

glsl: Add stubs for sparse queries and variable aoffi when not supported

Browse source
This commit is contained in:
ameerj 2021-06-13 00:05:19 -04:00
parent 6aa1bf7b6f
commit 5e7b2b9661
7 changed files with 47 additions and 13 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/shader_recompiler/backend/glsl/emit_context.cpp
View file

@ -384,7 +384,7 @@ void EmitContext::SetupExtensions() {
profile.support_viewport_index_layer_non_geometry && stage != Stage::Geometry) {
header += "#extension GL_ARB_shader_viewport_layer_array : enable\n";
}
if (info.uses_sparse_residency) {
if (info.uses_sparse_residency && profile.support_gl_sparse_textures) {
header += "#extension GL_ARB_sparse_texture2 : enable\n";
}
if (info.stores_viewport_mask && profile.support_viewport_mask) {

2
src/shader_recompiler/backend/glsl/emit_glsl.cpp
View file

@ -215,7 +215,7 @@ std::string EmitGLSL(const Profile& profile, const RuntimeInfo& runtime_info, IR
ctx.header +=
fmt::format("shared uint smem[{}];", Common::AlignUp(program.shared_memory_size, 4));
}
ctx.header += "\nvoid main(){\n";
ctx.header += "void main(){\n";
if (program.stage == Stage::VertexA || program.stage == Stage::VertexB) {
ctx.header += "gl_Position = vec4(0.0f, 0.0f, 0.0f, 1.0f);";
// TODO: Properly resolve attribute issues

46
src/shader_recompiler/backend/glsl/emit_glsl_image.cpp
View file

@ -94,7 +94,11 @@ std::string GetOffsetVec(EmitContext& ctx, const IR::Value& offset) {
break;
}
}
const auto offset_str{ctx.var_alloc.Consume(offset)};
const bool has_var_aoffi{ctx.profile.support_gl_variable_aoffi};
if (!has_var_aoffi) {
// LOG_WARNING("Device does not support variable texture offsets, STUBBING");
}
const auto offset_str{has_var_aoffi ? ctx.var_alloc.Consume(offset) : "0"};
switch (offset.Type()) {
case IR::Type::U32:
return fmt::format("int({})", offset_str);
@ -146,7 +150,12 @@ void EmitImageSampleImplicitLod(EmitContext& ctx, IR::Inst& inst, const IR::Valu
const auto bias{info.has_bias ? fmt::format(",{}", bias_lc) : ""};
const auto texel{ctx.var_alloc.Define(inst, GlslVarType::F32x4)};
const auto sparse_inst{PrepareSparse(inst)};
if (!sparse_inst) {
const bool supports_sparse{ctx.profile.support_gl_sparse_textures};
if (sparse_inst && !supports_sparse) {
// LOG_WARNING(..., "Device does not support sparse texture queries. STUBBING");
ctx.AddU1("{}=true;", *sparse_inst);
}
if (!sparse_inst || !supports_sparse) {
if (!offset.IsEmpty()) {
const auto offset_str{GetOffsetVec(ctx, offset)};
if (ctx.stage == Stage::Fragment) {
@ -163,7 +172,6 @@ void EmitImageSampleImplicitLod(EmitContext& ctx, IR::Inst& inst, const IR::Valu
}
return;
}
// TODO: Query sparseTexels extension support
if (!offset.IsEmpty()) {
ctx.AddU1("{}=sparseTexelsResidentARB(sparseTextureOffsetARB({},{},{},{}{}));",
*sparse_inst, texture, coords, GetOffsetVec(ctx, offset), texel, bias);
@ -186,7 +194,12 @@ void EmitImageSampleExplicitLod(EmitContext& ctx, IR::Inst& inst, const IR::Valu
const auto texture{Texture(ctx, info, index)};
const auto texel{ctx.var_alloc.Define(inst, GlslVarType::F32x4)};
const auto sparse_inst{PrepareSparse(inst)};
if (!sparse_inst) {
const bool supports_sparse{ctx.profile.support_gl_sparse_textures};
if (sparse_inst && !supports_sparse) {
// LOG_WARNING(..., "Device does not support sparse texture queries. STUBBING");
ctx.AddU1("{}=true;", *sparse_inst);
}
if (!sparse_inst || !supports_sparse) {
if (!offset.IsEmpty()) {
ctx.Add("{}=textureLodOffset({},{},{},{});", texel, texture, coords, lod_lc,
GetOffsetVec(ctx, offset));
@ -195,7 +208,6 @@ void EmitImageSampleExplicitLod(EmitContext& ctx, IR::Inst& inst, const IR::Valu
}
return;
}
// TODO: Query sparseTexels extension support
if (!offset.IsEmpty()) {
ctx.AddU1("{}=sparseTexelsResidentARB(sparseTexelFetchOffsetARB({},{},int({}),{},{}));",
*sparse_inst, texture, CastToIntVec(coords, info), lod_lc,
@ -315,7 +327,12 @@ void EmitImageGather(EmitContext& ctx, IR::Inst& inst, const IR::Value& index,
const auto texture{Texture(ctx, info, index)};
const auto texel{ctx.var_alloc.Define(inst, GlslVarType::F32x4)};
const auto sparse_inst{PrepareSparse(inst)};
if (!sparse_inst) {
const bool supports_sparse{ctx.profile.support_gl_sparse_textures};
if (sparse_inst && !supports_sparse) {
// LOG_WARNING(..., "Device does not support sparse texture queries. STUBBING");
ctx.AddU1("{}=true;", *sparse_inst);
}
if (!sparse_inst || !supports_sparse) {
if (offset.IsEmpty()) {
ctx.Add("{}=textureGather({},{},int({}));", texel, texture, coords,
info.gather_component);
@ -332,7 +349,6 @@ void EmitImageGather(EmitContext& ctx, IR::Inst& inst, const IR::Value& index,
info.gather_component);
return;
}
// TODO: Query sparseTexels extension support
if (offset.IsEmpty()) {
ctx.AddU1("{}=sparseTexelsResidentARB(sparseTextureGatherARB({},{},{},int({})));",
*sparse_inst, texture, coords, texel, info.gather_component);
@ -358,7 +374,12 @@ void EmitImageGatherDref(EmitContext& ctx, IR::Inst& inst, const IR::Value& inde
const auto texture{Texture(ctx, info, index)};
const auto texel{ctx.var_alloc.Define(inst, GlslVarType::F32x4)};
const auto sparse_inst{PrepareSparse(inst)};
if (!sparse_inst) {
const bool supports_sparse{ctx.profile.support_gl_sparse_textures};
if (sparse_inst && !supports_sparse) {
// LOG_WARNING(..., "Device does not support sparse texture queries. STUBBING");
ctx.AddU1("{}=true;", *sparse_inst);
}
if (!sparse_inst || !supports_sparse) {
if (offset.IsEmpty()) {
ctx.Add("{}=textureGather({},{},{});", texel, texture, coords, dref);
return;
@ -373,7 +394,6 @@ void EmitImageGatherDref(EmitContext& ctx, IR::Inst& inst, const IR::Value& inde
ctx.Add("{}=textureGatherOffsets({},{},{},{});", texel, texture, coords, dref, offsets);
return;
}
// TODO: Query sparseTexels extension support
if (offset.IsEmpty()) {
ctx.AddU1("{}=sparseTexelsResidentARB(sparseTextureGatherARB({},{},{},{}));", *sparse_inst,
texture, coords, dref, texel);
@ -404,7 +424,12 @@ void EmitImageFetch(EmitContext& ctx, IR::Inst& inst, const IR::Value& index,
const auto texture{Texture(ctx, info, index)};
const auto sparse_inst{PrepareSparse(inst)};
const auto texel{ctx.var_alloc.Define(inst, GlslVarType::F32x4)};
if (!sparse_inst) {
const bool supports_sparse{ctx.profile.support_gl_sparse_textures};
if (sparse_inst && !supports_sparse) {
// LOG_WARNING(..., "Device does not support sparse texture queries. STUBBING");
ctx.AddU1("{}=true;", *sparse_inst);
}
if (!sparse_inst || !supports_sparse) {
if (!offset.empty()) {
ctx.Add("{}=texelFetchOffset({},{},int({}),{});", texel, texture,
CoordsCastToInt(coords, info), lod, CoordsCastToInt(offset, info));
@ -418,7 +443,6 @@ void EmitImageFetch(EmitContext& ctx, IR::Inst& inst, const IR::Value& index,
}
return;
}
// TODO: Query sparseTexels extension support
if (!offset.empty()) {
ctx.AddU1("{}=sparseTexelsResidentARB(sparseTexelFetchOffsetARB({},{},int({}),{},{}));",
*sparse_inst, texture, CastToIntVec(coords, info), lod,

2
src/shader_recompiler/profile.h
View file

@ -87,6 +87,8 @@ struct Profile {
bool support_gl_amd_gpu_shader_half_float{};
bool support_gl_texture_shadow_lod{};
bool support_gl_warp_intrinsics{};
bool support_gl_variable_aoffi{};
bool support_gl_sparse_textures{};
bool warp_size_potentially_larger_than_guest{};

1
src/video_core/renderer_opengl/gl_device.cpp
View file

@ -160,6 +160,7 @@ Device::Device() {
has_depth_buffer_float = HasExtension(extensions, "GL_NV_depth_buffer_float");
has_nv_gpu_shader_5 = GLAD_GL_NV_gpu_shader5;
has_amd_shader_half_float = GLAD_GL_AMD_gpu_shader_half_float;
has_sparse_texture_2 = GLAD_GL_ARB_sparse_texture2;
warp_size_potentially_larger_than_guest = !is_nvidia && !is_intel;
// At the moment of writing this, only Nvidia's driver optimizes BufferSubData on exclusive

5
src/video_core/renderer_opengl/gl_device.h
View file

@ -128,6 +128,10 @@ public:
return has_amd_shader_half_float;
}
bool HasSparseTexture2() const {
return has_sparse_texture_2;
}
bool IsWarpSizePotentiallyLargerThanGuest() const {
return warp_size_potentially_larger_than_guest;
}
@ -165,6 +169,7 @@ private:
bool has_depth_buffer_float{};
bool has_nv_gpu_shader_5{};
bool has_amd_shader_half_float{};
bool has_sparse_texture_2{};
bool warp_size_potentially_larger_than_guest{};
};

2
src/video_core/renderer_opengl/gl_shader_cache.cpp
View file

@ -193,6 +193,8 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
.support_gl_amd_gpu_shader_half_float = device.HasAmdShaderHalfFloat(),
.support_gl_texture_shadow_lod = device.HasTextureShadowLod(),
.support_gl_warp_intrinsics = false,
.support_gl_variable_aoffi = device.HasVariableAoffi(),
.support_gl_sparse_textures = device.HasSparseTexture2(),
.warp_size_potentially_larger_than_guest = device.IsWarpSizePotentiallyLargerThanGuest(),