diff options
29 files changed, 203 insertions, 72 deletions
diff --git a/externals/dynarmic b/externals/dynarmic -Subproject cce7e4ee5d7b07a4609c73c053fbf57dc8c7845 +Subproject 28714ee75aa079cbb706e38bdabc8ee1f6c6951 diff --git a/src/core/frontend/applets/controller.cpp b/src/core/frontend/applets/controller.cpp index 6dbd38ffa..e1033b634 100644 --- a/src/core/frontend/applets/controller.cpp +++ b/src/core/frontend/applets/controller.cpp @@ -45,26 +45,26 @@ void DefaultControllerApplet::ReconfigureControllers(std::function<void()> callb // Pro Controller -> Dual Joycons -> Left Joycon/Right Joycon -> Handheld if (parameters.allow_pro_controller) { controller->SetNpadStyleIndex(Core::HID::NpadStyleIndex::ProController); - controller->Connect(); + controller->Connect(true); } else if (parameters.allow_dual_joycons) { controller->SetNpadStyleIndex(Core::HID::NpadStyleIndex::JoyconDual); - controller->Connect(); + controller->Connect(true); } else if (parameters.allow_left_joycon && parameters.allow_right_joycon) { // Assign left joycons to even player indices and right joycons to odd player indices. // We do this since Captain Toad Treasure Tracker expects a left joycon for Player 1 and // a right Joycon for Player 2 in 2 Player Assist mode. if (index % 2 == 0) { controller->SetNpadStyleIndex(Core::HID::NpadStyleIndex::JoyconLeft); - controller->Connect(); + controller->Connect(true); } else { controller->SetNpadStyleIndex(Core::HID::NpadStyleIndex::JoyconRight); - controller->Connect(); + controller->Connect(true); } } else if (index == 0 && parameters.enable_single_mode && parameters.allow_handheld && !Settings::values.use_docked_mode.GetValue()) { // We should *never* reach here under any normal circumstances. controller->SetNpadStyleIndex(Core::HID::NpadStyleIndex::Handheld); - controller->Connect(); + controller->Connect(true); } else { UNREACHABLE_MSG("Unable to add a new controller based on the given parameters!"); } diff --git a/src/core/hid/emulated_controller.cpp b/src/core/hid/emulated_controller.cpp index ff9d7a7e3..2d3fce276 100644 --- a/src/core/hid/emulated_controller.cpp +++ b/src/core/hid/emulated_controller.cpp @@ -886,8 +886,9 @@ void EmulatedController::SetSupportedNpadStyleTag(NpadStyleTag supported_styles) } } -bool EmulatedController::IsControllerSupported() const { - switch (npad_type) { +bool EmulatedController::IsControllerSupported(bool use_temporary_value) const { + const auto type = is_configuring && use_temporary_value ? tmp_npad_type : npad_type; + switch (type) { case NpadStyleIndex::ProController: return supported_style_tag.fullkey; case NpadStyleIndex::Handheld: @@ -915,9 +916,10 @@ bool EmulatedController::IsControllerSupported() const { } } -void EmulatedController::Connect() { - if (!IsControllerSupported()) { - LOG_ERROR(Service_HID, "Controller type {} is not supported", npad_type); +void EmulatedController::Connect(bool use_temporary_value) { + if (!IsControllerSupported(use_temporary_value)) { + const auto type = is_configuring && use_temporary_value ? tmp_npad_type : npad_type; + LOG_ERROR(Service_HID, "Controller type {} is not supported", type); return; } { diff --git a/src/core/hid/emulated_controller.h b/src/core/hid/emulated_controller.h index e42aafebc..d887eca87 100644 --- a/src/core/hid/emulated_controller.h +++ b/src/core/hid/emulated_controller.h @@ -167,8 +167,11 @@ public: */ void SetSupportedNpadStyleTag(NpadStyleTag supported_styles); - /// Sets the connected status to true - void Connect(); + /** + * Sets the connected status to true + * @param use_temporary_value If true tmp_npad_type will be used + */ + void Connect(bool use_temporary_value = false); /// Sets the connected status to false void Disconnect(); @@ -319,9 +322,10 @@ private: /** * Checks the current controller type against the supported_style_tag + * @param use_temporary_value If true tmp_npad_type will be used * @return true if the controller is supported */ - bool IsControllerSupported() const; + bool IsControllerSupported(bool use_temporary_value = false) const; /** * Updates the button status of the controller diff --git a/src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp b/src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp index 6f98d0998..7434a1f92 100644 --- a/src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp +++ b/src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp @@ -126,6 +126,22 @@ void EmitGetAttribute(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, Scal } } +void EmitGetAttributeU32(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, ScalarU32) { + switch (attr) { + case IR::Attribute::PrimitiveId: + ctx.Add("MOV.S {}.x,primitive.id;", inst); + break; + case IR::Attribute::InstanceId: + ctx.Add("MOV.S {}.x,{}.instance;", inst, ctx.attrib_name); + break; + case IR::Attribute::VertexId: + ctx.Add("MOV.S {}.x,{}.id;", inst, ctx.attrib_name); + break; + default: + throw NotImplementedException("Get U32 attribute {}", attr); + } +} + void EmitSetAttribute(EmitContext& ctx, IR::Attribute attr, ScalarF32 value, [[maybe_unused]] ScalarU32 vertex) { const u32 element{static_cast<u32>(attr) % 4}; diff --git a/src/shader_recompiler/backend/glasm/emit_glasm_instructions.h b/src/shader_recompiler/backend/glasm/emit_glasm_instructions.h index 1f343bff5..b48007856 100644 --- a/src/shader_recompiler/backend/glasm/emit_glasm_instructions.h +++ b/src/shader_recompiler/backend/glasm/emit_glasm_instructions.h @@ -50,6 +50,7 @@ void EmitGetCbufU32(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, void EmitGetCbufF32(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, ScalarU32 offset); void EmitGetCbufU32x2(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, ScalarU32 offset); void EmitGetAttribute(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, ScalarU32 vertex); +void EmitGetAttributeU32(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, ScalarU32 vertex); void EmitSetAttribute(EmitContext& ctx, IR::Attribute attr, ScalarF32 value, ScalarU32 vertex); void EmitGetAttributeIndexed(EmitContext& ctx, IR::Inst& inst, ScalarS32 offset, ScalarU32 vertex); void EmitSetAttributeIndexed(EmitContext& ctx, ScalarU32 offset, ScalarF32 value, ScalarU32 vertex); diff --git a/src/shader_recompiler/backend/glsl/emit_glsl_bitwise_conversion.cpp b/src/shader_recompiler/backend/glsl/emit_glsl_bitwise_conversion.cpp index 0f2668d9e..e0ead7a53 100644 --- a/src/shader_recompiler/backend/glsl/emit_glsl_bitwise_conversion.cpp +++ b/src/shader_recompiler/backend/glsl/emit_glsl_bitwise_conversion.cpp @@ -7,6 +7,7 @@ #include "shader_recompiler/backend/glsl/emit_glsl_instructions.h" #include "shader_recompiler/backend/glsl/glsl_emit_context.h" #include "shader_recompiler/frontend/ir/value.h" +#include "shader_recompiler/profile.h" namespace Shader::Backend::GLSL { namespace { @@ -30,8 +31,9 @@ void EmitConditionRef(EmitContext& ctx, IR::Inst& inst, const IR::Value& value) inst.DestructiveAddUsage(1); const auto ret{ctx.var_alloc.Define(inst, GlslVarType::U1)}; const auto input{ctx.var_alloc.Consume(value)}; + const auto suffix{ctx.profile.has_gl_bool_ref_bug ? "?true:false" : ""}; if (ret != input) { - ctx.Add("{}={};", ret, input); + ctx.Add("{}={}{};", ret, input, suffix); } } diff --git a/src/shader_recompiler/backend/glsl/emit_glsl_context_get_set.cpp b/src/shader_recompiler/backend/glsl/emit_glsl_context_get_set.cpp index 6477bd192..0c1fbc7b1 100644 --- a/src/shader_recompiler/backend/glsl/emit_glsl_context_get_set.cpp +++ b/src/shader_recompiler/backend/glsl/emit_glsl_context_get_set.cpp @@ -102,39 +102,46 @@ void GetCbuf16(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, const void EmitGetCbufU8(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, const IR::Value& offset) { - GetCbuf8(ctx, inst, binding, offset, "ftou"); + const auto cast{ctx.profile.has_gl_cbuf_ftou_bug ? "" : "ftou"}; + GetCbuf8(ctx, inst, binding, offset, cast); } void EmitGetCbufS8(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, const IR::Value& offset) { - GetCbuf8(ctx, inst, binding, offset, "ftoi"); + const auto cast{ctx.profile.has_gl_cbuf_ftou_bug ? "int" : "ftoi"}; + GetCbuf8(ctx, inst, binding, offset, cast); } void EmitGetCbufU16(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, const IR::Value& offset) { - GetCbuf16(ctx, inst, binding, offset, "ftou"); + const auto cast{ctx.profile.has_gl_cbuf_ftou_bug ? "" : "ftou"}; + GetCbuf16(ctx, inst, binding, offset, cast); } void EmitGetCbufS16(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, const IR::Value& offset) { - GetCbuf16(ctx, inst, binding, offset, "ftoi"); + const auto cast{ctx.profile.has_gl_cbuf_ftou_bug ? "int" : "ftoi"}; + GetCbuf16(ctx, inst, binding, offset, cast); } void EmitGetCbufU32(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, const IR::Value& offset) { const auto ret{ctx.var_alloc.Define(inst, GlslVarType::U32)}; - GetCbuf(ctx, ret, binding, offset, 32, "ftou"); + const auto cast{ctx.profile.has_gl_cbuf_ftou_bug ? "" : "ftou"}; + GetCbuf(ctx, ret, binding, offset, 32, cast); } void EmitGetCbufF32(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, const IR::Value& offset) { const auto ret{ctx.var_alloc.Define(inst, GlslVarType::F32)}; - GetCbuf(ctx, ret, binding, offset, 32); + const auto cast{ctx.profile.has_gl_cbuf_ftou_bug ? "utof" : ""}; + GetCbuf(ctx, ret, binding, offset, 32, cast); } void EmitGetCbufU32x2(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding, const IR::Value& offset) { const auto cbuf{fmt::format("{}_cbuf{}", ctx.stage_name, binding.U32())}; + const auto cast{ctx.profile.has_gl_cbuf_ftou_bug ? "" : "ftou"}; if (offset.IsImmediate()) { static constexpr u32 cbuf_size{0x10000}; const u32 u32_offset{offset.U32()}; @@ -145,26 +152,26 @@ void EmitGetCbufU32x2(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding return; } if (u32_offset % 2 == 0) { - ctx.AddU32x2("{}=ftou({}[{}].{}{});", inst, cbuf, u32_offset / 16, + ctx.AddU32x2("{}={}({}[{}].{}{});", inst, cast, cbuf, u32_offset / 16, OffsetSwizzle(u32_offset), OffsetSwizzle(u32_offset + 4)); } else { - ctx.AddU32x2("{}=uvec2(ftou({}[{}].{}),ftou({}[{}].{}));", inst, cbuf, u32_offset / 16, - OffsetSwizzle(u32_offset), cbuf, (u32_offset + 4) / 16, - OffsetSwizzle(u32_offset + 4)); + ctx.AddU32x2("{}=uvec2({}({}[{}].{}),{}({}[{}].{}));", inst, cast, cbuf, + u32_offset / 16, OffsetSwizzle(u32_offset), cast, cbuf, + (u32_offset + 4) / 16, OffsetSwizzle(u32_offset + 4)); } return; } const auto offset_var{ctx.var_alloc.Consume(offset)}; if (!ctx.profile.has_gl_component_indexing_bug) { - ctx.AddU32x2("{}=uvec2(ftou({}[{}>>4][({}>>2)%4]),ftou({}[({}+4)>>4][(({}+4)>>2)%4]));", - inst, cbuf, offset_var, offset_var, cbuf, offset_var, offset_var); + ctx.AddU32x2("{}=uvec2({}({}[{}>>4][({}>>2)%4]),{}({}[({}+4)>>4][(({}+4)>>2)%4]));", inst, + cast, cbuf, offset_var, offset_var, cast, cbuf, offset_var, offset_var); return; } const auto ret{ctx.var_alloc.Define(inst, GlslVarType::U32x2)}; const auto cbuf_offset{fmt::format("{}>>2", offset_var)}; for (u32 swizzle = 0; swizzle < 4; ++swizzle) { - ctx.Add("if(({}&3)=={}){}=uvec2(ftou({}[{}>>4].{}),ftou({}[({}+4)>>4].{}));", cbuf_offset, - swizzle, ret, cbuf, offset_var, "xyzw"[swizzle], cbuf, offset_var, + ctx.Add("if(({}&3)=={}){}=uvec2({}({}[{}>>4].{}),{}({}[({}+4)>>4].{}));", cbuf_offset, + swizzle, ret, cast, cbuf, offset_var, "xyzw"[swizzle], cast, cbuf, offset_var, "xyzw"[(swizzle + 1) % 4]); } } @@ -221,6 +228,22 @@ void EmitGetAttribute(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, } } +void EmitGetAttributeU32(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, std::string_view) { + switch (attr) { + case IR::Attribute::PrimitiveId: + ctx.AddU32("{}=uint(gl_PrimitiveID);", inst); + break; + case IR::Attribute::InstanceId: + ctx.AddU32("{}=uint(gl_InstanceID);", inst); + break; + case IR::Attribute::VertexId: + ctx.AddU32("{}=uint(gl_VertexID);", inst); + break; + default: + throw NotImplementedException("Get U32 attribute {}", attr); + } +} + void EmitSetAttribute(EmitContext& ctx, IR::Attribute attr, std::string_view value, [[maybe_unused]] std::string_view vertex) { if (IR::IsGeneric(attr)) { diff --git a/src/shader_recompiler/backend/glsl/emit_glsl_floating_point.cpp b/src/shader_recompiler/backend/glsl/emit_glsl_floating_point.cpp index b765a251b..474189d87 100644 --- a/src/shader_recompiler/backend/glsl/emit_glsl_floating_point.cpp +++ b/src/shader_recompiler/backend/glsl/emit_glsl_floating_point.cpp @@ -125,11 +125,11 @@ void EmitFPNeg16([[maybe_unused]] EmitContext& ctx, [[maybe_unused]] IR::Inst& i } void EmitFPNeg32(EmitContext& ctx, IR::Inst& inst, std::string_view value) { - ctx.AddF32("{}=-({});", inst, value); + ctx.AddF32("{}=0.f-({});", inst, value); } void EmitFPNeg64(EmitContext& ctx, IR::Inst& inst, std::string_view value) { - ctx.AddF64("{}=-({});", inst, value); + ctx.AddF64("{}=double(0.)-({});", inst, value); } void EmitFPSin(EmitContext& ctx, IR::Inst& inst, std::string_view value) { diff --git a/src/shader_recompiler/backend/glsl/emit_glsl_instructions.h b/src/shader_recompiler/backend/glsl/emit_glsl_instructions.h index f86502e4c..6cabbc717 100644 --- a/src/shader_recompiler/backend/glsl/emit_glsl_instructions.h +++ b/src/shader_recompiler/backend/glsl/emit_glsl_instructions.h @@ -60,6 +60,8 @@ void EmitGetCbufU32x2(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding const IR::Value& offset); void EmitGetAttribute(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, std::string_view vertex); +void EmitGetAttributeU32(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, + std::string_view vertex); void EmitSetAttribute(EmitContext& ctx, IR::Attribute attr, std::string_view value, std::string_view vertex); void EmitGetAttributeIndexed(EmitContext& ctx, IR::Inst& inst, std::string_view offset, diff --git a/src/shader_recompiler/backend/glsl/emit_glsl_integer.cpp b/src/shader_recompiler/backend/glsl/emit_glsl_integer.cpp index 44060df33..b0d85be99 100644 --- a/src/shader_recompiler/backend/glsl/emit_glsl_integer.cpp +++ b/src/shader_recompiler/backend/glsl/emit_glsl_integer.cpp @@ -87,11 +87,11 @@ void EmitUDiv32(EmitContext& ctx, IR::Inst& inst, std::string_view a, std::strin } void EmitINeg32(EmitContext& ctx, IR::Inst& inst, std::string_view value) { - ctx.AddU32("{}=uint(-({}));", inst, value); + ctx.AddU32("{}=uint(int(0)-int({}));", inst, value); } void EmitINeg64(EmitContext& ctx, IR::Inst& inst, std::string_view value) { - ctx.AddU64("{}=-({});", inst, value); + ctx.AddU64("{}=uint64_t(int64_t(0)-int64_t({}));", inst, value); } void EmitIAbs32(EmitContext& ctx, IR::Inst& inst, std::string_view value) { diff --git a/src/shader_recompiler/backend/glsl/emit_glsl_special.cpp b/src/shader_recompiler/backend/glsl/emit_glsl_special.cpp index b8ddafe48..fcf620b79 100644 --- a/src/shader_recompiler/backend/glsl/emit_glsl_special.cpp +++ b/src/shader_recompiler/backend/glsl/emit_glsl_special.cpp @@ -90,7 +90,9 @@ void EmitPhiMove(EmitContext& ctx, const IR::Value& phi_value, const IR::Value& if (phi_reg == val_reg) { return; } - ctx.Add("{}={};", phi_reg, val_reg); + const bool needs_workaround{ctx.profile.has_gl_bool_ref_bug && phi_type == IR::Type::U1}; + const auto suffix{needs_workaround ? "?true:false" : ""}; + ctx.Add("{}={}{};", phi_reg, val_reg, suffix); } void EmitPrologue(EmitContext& ctx) { diff --git a/src/shader_recompiler/backend/glsl/glsl_emit_context.cpp b/src/shader_recompiler/backend/glsl/glsl_emit_context.cpp index bc9d2a904..bb7f1a0fd 100644 --- a/src/shader_recompiler/backend/glsl/glsl_emit_context.cpp +++ b/src/shader_recompiler/backend/glsl/glsl_emit_context.cpp @@ -428,9 +428,10 @@ void EmitContext::DefineConstantBuffers(Bindings& bindings) { return; } for (const auto& desc : info.constant_buffer_descriptors) { - header += fmt::format( - "layout(std140,binding={}) uniform {}_cbuf_{}{{vec4 {}_cbuf{}[{}];}};", - bindings.uniform_buffer, stage_name, desc.index, stage_name, desc.index, 4 * 1024); + const auto cbuf_type{profile.has_gl_cbuf_ftou_bug ? "uvec4" : "vec4"}; + header += fmt::format("layout(std140,binding={}) uniform {}_cbuf_{}{{{} {}_cbuf{}[{}];}};", + bindings.uniform_buffer, stage_name, desc.index, cbuf_type, + stage_name, desc.index, 4 * 1024); bindings.uniform_buffer += desc.count; } } diff --git a/src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp b/src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp index 14f470812..8ea730c80 100644 --- a/src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp +++ b/src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp @@ -355,6 +355,31 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) { } } +Id EmitGetAttributeU32(EmitContext& ctx, IR::Attribute attr, Id) { + switch (attr) { + case IR::Attribute::PrimitiveId: + return ctx.OpLoad(ctx.U32[1], ctx.primitive_id); + case IR::Attribute::InstanceId: + if (ctx.profile.support_vertex_instance_id) { + return ctx.OpLoad(ctx.U32[1], ctx.instance_id); + } else { + const Id index{ctx.OpLoad(ctx.U32[1], ctx.instance_index)}; + const Id base{ctx.OpLoad(ctx.U32[1], ctx.base_instance)}; + return ctx.OpISub(ctx.U32[1], index, base); + } + case IR::Attribute::VertexId: + if (ctx.profile.support_vertex_instance_id) { + return ctx.OpLoad(ctx.U32[1], ctx.vertex_id); + } else { + const Id index{ctx.OpLoad(ctx.U32[1], ctx.vertex_index)}; + const Id base{ctx.OpLoad(ctx.U32[1], ctx.base_vertex)}; + return ctx.OpISub(ctx.U32[1], index, base); + } + default: + throw NotImplementedException("Read U32 attribute {}", attr); + } +} + void EmitSetAttribute(EmitContext& ctx, IR::Attribute attr, Id value, [[maybe_unused]] Id vertex) { const std::optional<OutAttr> output{OutputAttrPointer(ctx, attr)}; if (!output) { diff --git a/src/shader_recompiler/backend/spirv/emit_spirv_instructions.h b/src/shader_recompiler/backend/spirv/emit_spirv_instructions.h index 6cd22dd3e..887112deb 100644 --- a/src/shader_recompiler/backend/spirv/emit_spirv_instructions.h +++ b/src/shader_recompiler/backend/spirv/emit_spirv_instructions.h @@ -53,6 +53,7 @@ Id EmitGetCbufU32(EmitContext& ctx, const IR::Value& binding, const IR::Value& o Id EmitGetCbufF32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset); Id EmitGetCbufU32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset); Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex); +Id EmitGetAttributeU32(EmitContext& ctx, IR::Attribute attr, Id vertex); void EmitSetAttribute(EmitContext& ctx, IR::Attribute attr, Id value, Id vertex); Id EmitGetAttributeIndexed(EmitContext& ctx, Id offset, Id vertex); void EmitSetAttributeIndexed(EmitContext& ctx, Id offset, Id value, Id vertex); diff --git a/src/shader_recompiler/frontend/ir/opcodes.inc b/src/shader_recompiler/frontend/ir/opcodes.inc index 6929919df..b94ce7406 100644 --- a/src/shader_recompiler/frontend/ir/opcodes.inc +++ b/src/shader_recompiler/frontend/ir/opcodes.inc @@ -40,6 +40,7 @@ OPCODE(GetCbufU32, U32, U32, OPCODE(GetCbufF32, F32, U32, U32, ) OPCODE(GetCbufU32x2, U32x2, U32, U32, ) OPCODE(GetAttribute, F32, Attribute, U32, ) +OPCODE(GetAttributeU32, U32, Attribute, U32, ) OPCODE(SetAttribute, Void, Attribute, F32, U32, ) OPCODE(GetAttributeIndexed, F32, U32, U32, ) OPCODE(SetAttributeIndexed, Void, U32, F32, U32, ) diff --git a/src/shader_recompiler/ir_opt/collect_shader_info_pass.cpp b/src/shader_recompiler/ir_opt/collect_shader_info_pass.cpp index 1e476d83d..a78c469be 100644 --- a/src/shader_recompiler/ir_opt/collect_shader_info_pass.cpp +++ b/src/shader_recompiler/ir_opt/collect_shader_info_pass.cpp @@ -389,6 +389,7 @@ void VisitUsages(Info& info, IR::Inst& inst) { info.uses_demote_to_helper_invocation = true; break; case IR::Opcode::GetAttribute: + case IR::Opcode::GetAttributeU32: info.loads.mask[static_cast<size_t>(inst.Arg(0).Attribute())] = true; break; case IR::Opcode::SetAttribute: diff --git a/src/shader_recompiler/ir_opt/constant_propagation_pass.cpp b/src/shader_recompiler/ir_opt/constant_propagation_pass.cpp index d089fdd12..c134a12bc 100644 --- a/src/shader_recompiler/ir_opt/constant_propagation_pass.cpp +++ b/src/shader_recompiler/ir_opt/constant_propagation_pass.cpp @@ -505,6 +505,29 @@ void FoldBitCast(IR::Inst& inst, IR::Opcode reverse) { return; } } + if constexpr (op == IR::Opcode::BitCastU32F32) { + // Workaround for new NVIDIA driver bug, where: + // uint attr = ftou(itof(gl_InstanceID)); + // always returned 0. + // We can instead manually optimize this and work around the driver bug: + // uint attr = uint(gl_InstanceID); + if (arg_inst->GetOpcode() == IR::Opcode::GetAttribute) { + const IR::Attribute attr{arg_inst->Arg(0).Attribute()}; + switch (attr) { + case IR::Attribute::PrimitiveId: + case IR::Attribute::InstanceId: + case IR::Attribute::VertexId: + break; + default: + return; + } + // Replace the bitcasts with an integer attribute get + inst.ReplaceOpcode(IR::Opcode::GetAttributeU32); + inst.SetArg(0, arg_inst->Arg(0)); + inst.SetArg(1, arg_inst->Arg(1)); + return; + } + } } void FoldInverseFunc(IR::Inst& inst, IR::Opcode reverse) { diff --git a/src/shader_recompiler/profile.h b/src/shader_recompiler/profile.h index f0c3b3b17..dc4c806ff 100644 --- a/src/shader_recompiler/profile.h +++ b/src/shader_recompiler/profile.h @@ -65,6 +65,10 @@ struct Profile { bool has_gl_component_indexing_bug{}; /// The precise type qualifier is broken in the fragment stage of some drivers bool has_gl_precise_bug{}; + /// Some drivers do not properly support floatBitsToUint when used on cbufs + bool has_gl_cbuf_ftou_bug{}; + /// Some drivers poorly optimize boolean variable references + bool has_gl_bool_ref_bug{}; /// Ignores SPIR-V ordered vs unordered using GLSL semantics bool ignore_nan_fp_comparisons{}; diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp index 0764ea6e0..e62912a22 100644 --- a/src/video_core/renderer_opengl/gl_device.cpp +++ b/src/video_core/renderer_opengl/gl_device.cpp @@ -182,17 +182,13 @@ Device::Device() { shader_backend = Settings::ShaderBackend::GLSL; } - if (shader_backend == Settings::ShaderBackend::GLSL && is_nvidia && - !Settings::values.renderer_debug) { + if (shader_backend == Settings::ShaderBackend::GLSL && is_nvidia) { const std::string_view driver_version = version.substr(13); const int version_major = std::atoi(driver_version.substr(0, driver_version.find(".")).data()); - if (version_major >= 495) { - LOG_WARNING(Render_OpenGL, "NVIDIA drivers 495 and later causes significant problems " - "with yuzu. Forcing GLASM as a mitigation."); - shader_backend = Settings::ShaderBackend::GLASM; - use_assembly_shaders = true; + has_cbuf_ftou_bug = true; + has_bool_ref_bug = true; } } diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h index de9e41659..95c2e8d38 100644 --- a/src/video_core/renderer_opengl/gl_device.h +++ b/src/video_core/renderer_opengl/gl_device.h @@ -152,6 +152,14 @@ public: return need_fastmath_off; } + bool HasCbufFtouBug() const { + return has_cbuf_ftou_bug; + } + + bool HasBoolRefBug() const { + return has_bool_ref_bug; + } + Settings::ShaderBackend GetShaderBackend() const { return shader_backend; } @@ -200,6 +208,8 @@ private: bool has_sparse_texture_2{}; bool warp_size_potentially_larger_than_guest{}; bool need_fastmath_off{}; + bool has_cbuf_ftou_bug{}; + bool has_bool_ref_bug{}; std::string vendor_name; }; diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index 29c6e1a5f..ec558a9af 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -214,6 +214,8 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo .has_broken_fp16_float_controls = false, .has_gl_component_indexing_bug = device.HasComponentIndexingBug(), .has_gl_precise_bug = device.HasPreciseBug(), + .has_gl_cbuf_ftou_bug = device.HasCbufFtouBug(), + .has_gl_bool_ref_bug = device.HasBoolRefBug(), .ignore_nan_fp_comparisons = true, .gl_max_compute_smem_size = device.GetMaxComputeSharedMemorySize(), }, diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index c3050887c..0ba56ff1e 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -1344,7 +1344,6 @@ bool Image::ScaleUp(bool ignore) { return false; } has_scaled = true; - const auto& device = runtime->device; if (!scaled_image) { const bool is_2d = info.type == ImageType::e2D; const u32 scaled_width = resolution.ScaleUp(info.size.width); @@ -1352,7 +1351,7 @@ bool Image::ScaleUp(bool ignore) { auto scaled_info = info; scaled_info.size.width = scaled_width; scaled_info.size.height = scaled_height; - scaled_image = MakeImage(device, scaled_info); + scaled_image = MakeImage(runtime->device, scaled_info); auto& allocator = runtime->memory_allocator; scaled_commit = MemoryCommit(allocator.Commit(scaled_image, MemoryUsage::DeviceLocal)); ignore = false; @@ -1361,18 +1360,13 @@ bool Image::ScaleUp(bool ignore) { if (ignore) { return true; } - if (aspect_mask == 0) { aspect_mask = ImageAspectMask(info.format); } - static constexpr auto OPTIMAL_FORMAT = FormatType::Optimal; - const PixelFormat format = StorageFormat(info.format); - const auto vk_format = MaxwellToVK::SurfaceFormat(device, OPTIMAL_FORMAT, false, format).format; - const auto blit_usage = VK_FORMAT_FEATURE_BLIT_SRC_BIT | VK_FORMAT_FEATURE_BLIT_DST_BIT; - if (device.IsFormatSupported(vk_format, blit_usage, OPTIMAL_FORMAT)) { - BlitScale(*scheduler, *original_image, *scaled_image, info, aspect_mask, resolution); - } else { + if (NeedsScaleHelper()) { return BlitScaleHelper(true); + } else { + BlitScale(*scheduler, *original_image, *scaled_image, info, aspect_mask, resolution); } return true; } @@ -1394,15 +1388,10 @@ bool Image::ScaleDown(bool ignore) { if (aspect_mask == 0) { aspect_mask = ImageAspectMask(info.format); } - static constexpr auto OPTIMAL_FORMAT = FormatType::Optimal; - const PixelFormat format = StorageFormat(info.format); - const auto& device = runtime->device; - const auto vk_format = MaxwellToVK::SurfaceFormat(device, OPTIMAL_FORMAT, false, format).format; - const auto blit_usage = VK_FORMAT_FEATURE_BLIT_SRC_BIT | VK_FORMAT_FEATURE_BLIT_DST_BIT; - if (device.IsFormatSupported(vk_format, blit_usage, OPTIMAL_FORMAT)) { - BlitScale(*scheduler, *scaled_image, *original_image, info, aspect_mask, resolution, false); - } else { + if (NeedsScaleHelper()) { return BlitScaleHelper(false); + } else { + BlitScale(*scheduler, *scaled_image, *original_image, info, aspect_mask, resolution, false); } return true; } @@ -1470,6 +1459,20 @@ bool Image::BlitScaleHelper(bool scale_up) { return true; } +bool Image::NeedsScaleHelper() const { + const auto& device = runtime->device; + const bool needs_msaa_helper = info.num_samples > 1 && device.CantBlitMSAA(); + if (needs_msaa_helper) { + return true; + } + static constexpr auto OPTIMAL_FORMAT = FormatType::Optimal; + const PixelFormat format = StorageFormat(info.format); + const auto vk_format = MaxwellToVK::SurfaceFormat(device, OPTIMAL_FORMAT, false, format).format; + const auto blit_usage = VK_FORMAT_FEATURE_BLIT_SRC_BIT | VK_FORMAT_FEATURE_BLIT_DST_BIT; + const bool needs_blit_helper = !device.IsFormatSupported(vk_format, blit_usage, OPTIMAL_FORMAT); + return needs_blit_helper; +} + ImageView::ImageView(TextureCacheRuntime& runtime, const VideoCommon::ImageViewInfo& info, ImageId image_id_, Image& image) : VideoCommon::ImageViewBase{info, image.info, image_id_}, device{&runtime.device}, diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h index 2f12be78b..c81130dd2 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.h +++ b/src/video_core/renderer_vulkan/vk_texture_cache.h @@ -149,6 +149,8 @@ public: private: bool BlitScaleHelper(bool scale_up); + bool NeedsScaleHelper() const; + VKScheduler* scheduler{}; TextureCacheRuntime* runtime{}; diff --git a/src/video_core/texture_cache/util.cpp b/src/video_core/texture_cache/util.cpp index 7bd31b211..d8e19cb2f 100644 --- a/src/video_core/texture_cache/util.cpp +++ b/src/video_core/texture_cache/util.cpp @@ -364,14 +364,14 @@ template <u32 GOB_EXTENT> [[nodiscard]] std::optional<SubresourceExtent> ResolveOverlapRightAddress2D( const ImageInfo& new_info, GPUVAddr gpu_addr, const ImageBase& overlap, bool strict_size) { - const u32 layer_stride = new_info.layer_stride; - const s32 new_size = layer_stride * new_info.resources.layers; - const s32 diff = static_cast<s32>(overlap.gpu_addr - gpu_addr); + const u64 layer_stride = new_info.layer_stride; + const u64 new_size = layer_stride * new_info.resources.layers; + const u64 diff = overlap.gpu_addr - gpu_addr; if (diff > new_size) { return std::nullopt; } - const s32 base_layer = diff / layer_stride; - const s32 mip_offset = diff % layer_stride; + const s32 base_layer = static_cast<s32>(diff / layer_stride); + const s32 mip_offset = static_cast<s32>(diff % layer_stride); const std::array offsets = CalculateMipLevelOffsets(new_info); const auto end = offsets.begin() + new_info.resources.levels; const auto it = std::find(offsets.begin(), end, static_cast<u32>(mip_offset)); diff --git a/src/video_core/vulkan_common/vulkan_device.cpp b/src/video_core/vulkan_common/vulkan_device.cpp index 9862b815b..3d78efddc 100644 --- a/src/video_core/vulkan_common/vulkan_device.cpp +++ b/src/video_core/vulkan_common/vulkan_device.cpp @@ -638,15 +638,20 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR } } - if (ext_vertex_input_dynamic_state && driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS) { + const bool is_intel_windows = driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS; + if (ext_vertex_input_dynamic_state && is_intel_windows) { LOG_WARNING(Render_Vulkan, "Blacklisting Intel for VK_EXT_vertex_input_dynamic_state"); ext_vertex_input_dynamic_state = false; } - if (is_float16_supported && driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS) { + if (is_float16_supported && is_intel_windows) { // Intel's compiler crashes when using fp16 on Astral Chain, disable it for the time being. LOG_WARNING(Render_Vulkan, "Blacklisting Intel proprietary from float16 math"); is_float16_supported = false; } + if (is_intel_windows) { + LOG_WARNING(Render_Vulkan, "Intel proprietary drivers do not support MSAA image blits"); + cant_blit_msaa = true; + } supports_d24_depth = IsFormatSupported(VK_FORMAT_D24_UNORM_S8_UINT, diff --git a/src/video_core/vulkan_common/vulkan_device.h b/src/video_core/vulkan_common/vulkan_device.h index 4c9d86aad..37d140ebd 100644 --- a/src/video_core/vulkan_common/vulkan_device.h +++ b/src/video_core/vulkan_common/vulkan_device.h @@ -350,6 +350,10 @@ public: return supports_d24_depth; } + bool CantBlitMSAA() const { + return cant_blit_msaa; + } + private: /// Checks if the physical device is suitable. void CheckSuitability(bool requires_swapchain) const; @@ -443,6 +447,7 @@ private: bool has_renderdoc{}; ///< Has RenderDoc attached bool has_nsight_graphics{}; ///< Has Nsight Graphics attached bool supports_d24_depth{}; ///< Supports D24 depth buffers. + bool cant_blit_msaa{}; ///< Does not support MSAA<->MSAA blitting. // Telemetry parameters std::string vendor_name; ///< Device's driver name. diff --git a/src/yuzu/applets/qt_controller.cpp b/src/yuzu/applets/qt_controller.cpp index c6222b571..d63193131 100644 --- a/src/yuzu/applets/qt_controller.cpp +++ b/src/yuzu/applets/qt_controller.cpp @@ -33,7 +33,7 @@ void UpdateController(Core::HID::EmulatedController* controller, } controller->SetNpadStyleIndex(controller_type); if (connected) { - controller->Connect(); + controller->Connect(true); } } diff --git a/src/yuzu/configuration/configure_input_player.cpp b/src/yuzu/configuration/configure_input_player.cpp index 8a8be8e40..cb6163702 100644 --- a/src/yuzu/configuration/configure_input_player.cpp +++ b/src/yuzu/configuration/configure_input_player.cpp @@ -599,11 +599,11 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i if (is_connected) { if (type == Core::HID::NpadStyleIndex::Handheld) { emulated_controller_p1->Disconnect(); - emulated_controller_handheld->Connect(); + emulated_controller_handheld->Connect(true); emulated_controller = emulated_controller_handheld; } else { emulated_controller_handheld->Disconnect(); - emulated_controller_p1->Connect(); + emulated_controller_p1->Connect(true); emulated_controller = emulated_controller_p1; } } @@ -718,7 +718,7 @@ void ConfigureInputPlayer::LoadConfiguration() { void ConfigureInputPlayer::ConnectPlayer(bool connected) { ui->groupConnectedController->setChecked(connected); if (connected) { - emulated_controller->Connect(); + emulated_controller->Connect(true); } else { emulated_controller->Disconnect(); } |