aboutsummaryrefslogtreecommitdiff
path: root/src/video_core/shader
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core/shader')
-rw-r--r--src/video_core/shader/decode.cpp2
-rw-r--r--src/video_core/shader/decode/half_set.cpp88
-rw-r--r--src/video_core/shader/decode/image.cpp26
-rw-r--r--src/video_core/shader/decode/memory.cpp3
-rw-r--r--src/video_core/shader/decode/other.cpp42
-rw-r--r--src/video_core/shader/decode/texture.cpp55
-rw-r--r--src/video_core/shader/decode/xmad.cpp12
-rw-r--r--src/video_core/shader/memory_util.cpp4
-rw-r--r--src/video_core/shader/node.h105
-rw-r--r--src/video_core/shader/node_helper.h2
-rw-r--r--src/video_core/shader/registry.cpp20
-rw-r--r--src/video_core/shader/registry.h35
-rw-r--r--src/video_core/shader/shader_ir.cpp109
-rw-r--r--src/video_core/shader/shader_ir.h14
-rw-r--r--src/video_core/shader/track.cpp78
15 files changed, 385 insertions, 210 deletions
diff --git a/src/video_core/shader/decode.cpp b/src/video_core/shader/decode.cpp
index a75a5cc63..eeac328a6 100644
--- a/src/video_core/shader/decode.cpp
+++ b/src/video_core/shader/decode.cpp
@@ -255,7 +255,7 @@ void ShaderIR::InsertControlFlow(NodeBlock& bb, const ShaderBlock& block) {
Node n = Operation(OperationCode::Branch, Immediate(branch_case.address));
Node op_b = Immediate(branch_case.cmp_value);
Node condition =
- GetPredicateComparisonInteger(Tegra::Shader::PredCondition::Equal, false, op_a, op_b);
+ GetPredicateComparisonInteger(Tegra::Shader::PredCondition::EQ, false, op_a, op_b);
auto result = Conditional(condition, {n});
bb.push_back(result);
global_code.push_back(result);
diff --git a/src/video_core/shader/decode/half_set.cpp b/src/video_core/shader/decode/half_set.cpp
index 848e46874..b2e88fa20 100644
--- a/src/video_core/shader/decode/half_set.cpp
+++ b/src/video_core/shader/decode/half_set.cpp
@@ -13,55 +13,101 @@
namespace VideoCommon::Shader {
+using std::move;
using Tegra::Shader::Instruction;
using Tegra::Shader::OpCode;
+using Tegra::Shader::PredCondition;
u32 ShaderIR::DecodeHalfSet(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
const auto opcode = OpCode::Decode(instr);
- if (instr.hset2.ftz == 0) {
- LOG_DEBUG(HW_GPU, "{} without FTZ is not implemented", opcode->get().GetName());
+ PredCondition cond;
+ bool bf;
+ bool ftz;
+ bool neg_a;
+ bool abs_a;
+ bool neg_b;
+ bool abs_b;
+ switch (opcode->get().GetId()) {
+ case OpCode::Id::HSET2_C:
+ case OpCode::Id::HSET2_IMM:
+ cond = instr.hsetp2.cbuf_and_imm.cond;
+ bf = instr.Bit(53);
+ ftz = instr.Bit(54);
+ neg_a = instr.Bit(43);
+ abs_a = instr.Bit(44);
+ neg_b = instr.Bit(56);
+ abs_b = instr.Bit(54);
+ break;
+ case OpCode::Id::HSET2_R:
+ cond = instr.hsetp2.reg.cond;
+ bf = instr.Bit(49);
+ ftz = instr.Bit(50);
+ neg_a = instr.Bit(43);
+ abs_a = instr.Bit(44);
+ neg_b = instr.Bit(31);
+ abs_b = instr.Bit(30);
+ break;
+ default:
+ UNREACHABLE();
}
- Node op_a = UnpackHalfFloat(GetRegister(instr.gpr8), instr.hset2.type_a);
- op_a = GetOperandAbsNegHalf(op_a, instr.hset2.abs_a, instr.hset2.negate_a);
-
- Node op_b = [&]() {
+ Node op_b = [this, instr, opcode] {
switch (opcode->get().GetId()) {
+ case OpCode::Id::HSET2_C:
+ // Inform as unimplemented as this is not tested.
+ UNIMPLEMENTED_MSG("HSET2_C is not implemented");
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
case OpCode::Id::HSET2_R:
return GetRegister(instr.gpr20);
+ case OpCode::Id::HSET2_IMM:
+ return UnpackHalfImmediate(instr, true);
default:
UNREACHABLE();
- return Immediate(0);
+ return Node{};
}
}();
- op_b = UnpackHalfFloat(op_b, instr.hset2.type_b);
- op_b = GetOperandAbsNegHalf(op_b, instr.hset2.abs_b, instr.hset2.negate_b);
- const Node second_pred = GetPredicate(instr.hset2.pred39, instr.hset2.neg_pred);
+ if (!ftz) {
+ LOG_DEBUG(HW_GPU, "{} without FTZ is not implemented", opcode->get().GetName());
+ }
+
+ Node op_a = UnpackHalfFloat(GetRegister(instr.gpr8), instr.hset2.type_a);
+ op_a = GetOperandAbsNegHalf(op_a, abs_a, neg_a);
+
+ switch (opcode->get().GetId()) {
+ case OpCode::Id::HSET2_R:
+ op_b = GetOperandAbsNegHalf(move(op_b), abs_b, neg_b);
+ [[fallthrough]];
+ case OpCode::Id::HSET2_C:
+ op_b = UnpackHalfFloat(move(op_b), instr.hset2.type_b);
+ break;
+ default:
+ break;
+ }
- const Node comparison_pair = GetPredicateComparisonHalf(instr.hset2.cond, op_a, op_b);
+ Node second_pred = GetPredicate(instr.hset2.pred39, instr.hset2.neg_pred);
+
+ Node comparison_pair = GetPredicateComparisonHalf(cond, op_a, op_b);
const OperationCode combiner = GetPredicateCombiner(instr.hset2.op);
// HSET2 operates on each half float in the pack.
std::array<Node, 2> values;
for (u32 i = 0; i < 2; ++i) {
- const u32 raw_value = instr.hset2.bf ? 0x3c00 : 0xffff;
- const Node true_value = Immediate(raw_value << (i * 16));
- const Node false_value = Immediate(0);
-
- const Node comparison =
- Operation(OperationCode::LogicalPick2, comparison_pair, Immediate(i));
- const Node predicate = Operation(combiner, comparison, second_pred);
+ const u32 raw_value = bf ? 0x3c00 : 0xffff;
+ Node true_value = Immediate(raw_value << (i * 16));
+ Node false_value = Immediate(0);
+ Node comparison = Operation(OperationCode::LogicalPick2, comparison_pair, Immediate(i));
+ Node predicate = Operation(combiner, comparison, second_pred);
values[i] =
- Operation(OperationCode::Select, NO_PRECISE, predicate, true_value, false_value);
+ Operation(OperationCode::Select, predicate, move(true_value), move(false_value));
}
- const Node value = Operation(OperationCode::UBitwiseOr, NO_PRECISE, values[0], values[1]);
- SetRegister(bb, instr.gpr0, value);
+ Node value = Operation(OperationCode::UBitwiseOr, values[0], values[1]);
+ SetRegister(bb, instr.gpr0, move(value));
return pc;
}
diff --git a/src/video_core/shader/decode/image.cpp b/src/video_core/shader/decode/image.cpp
index 60b6ad72a..07778dc3e 100644
--- a/src/video_core/shader/decode/image.cpp
+++ b/src/video_core/shader/decode/image.cpp
@@ -97,6 +97,7 @@ ComponentType GetComponentType(Tegra::Engines::SamplerDescriptor descriptor,
break;
case TextureFormat::B5G6R5:
case TextureFormat::B6G5R5:
+ case TextureFormat::BF10GF11RF11:
if (component == 0) {
return descriptor.b_type;
}
@@ -119,7 +120,7 @@ ComponentType GetComponentType(Tegra::Engines::SamplerDescriptor descriptor,
}
break;
}
- UNIMPLEMENTED_MSG("texture format not implement={}", format);
+ UNIMPLEMENTED_MSG("Texture format not implemented={}", format);
return ComponentType::FLOAT;
}
@@ -191,6 +192,14 @@ u32 GetComponentSize(TextureFormat format, std::size_t component) {
return 6;
}
return 0;
+ case TextureFormat::BF10GF11RF11:
+ if (component == 1 || component == 2) {
+ return 11;
+ }
+ if (component == 0) {
+ return 10;
+ }
+ return 0;
case TextureFormat::G8R24:
if (component == 0) {
return 8;
@@ -211,10 +220,9 @@ u32 GetComponentSize(TextureFormat format, std::size_t component) {
return (component == 0 || component == 1) ? 8 : 0;
case TextureFormat::G4R4:
return (component == 0 || component == 1) ? 4 : 0;
- default:
- UNIMPLEMENTED_MSG("texture format not implement={}", format);
- return 0;
}
+ UNIMPLEMENTED_MSG("Texture format not implemented={}", format);
+ return 0;
}
std::size_t GetImageComponentMask(TextureFormat format) {
@@ -235,6 +243,7 @@ std::size_t GetImageComponentMask(TextureFormat format) {
case TextureFormat::R32_B24G8:
case TextureFormat::B5G6R5:
case TextureFormat::B6G5R5:
+ case TextureFormat::BF10GF11RF11:
return std::size_t{R | G | B};
case TextureFormat::R32_G32:
case TextureFormat::R16_G16:
@@ -248,10 +257,9 @@ std::size_t GetImageComponentMask(TextureFormat format) {
case TextureFormat::R8:
case TextureFormat::R1:
return std::size_t{R};
- default:
- UNIMPLEMENTED_MSG("texture format not implement={}", format);
- return std::size_t{R | G | B | A};
}
+ UNIMPLEMENTED_MSG("Texture format not implemented={}", format);
+ return std::size_t{R | G | B | A};
}
std::size_t GetImageTypeNumCoordinates(Tegra::Shader::ImageType image_type) {
@@ -299,7 +307,7 @@ std::pair<Node, bool> ShaderIR::GetComponentValue(ComponentType component_type,
return {std::move(original_value), true};
}
default:
- UNIMPLEMENTED_MSG("Unimplement component type={}", component_type);
+ UNIMPLEMENTED_MSG("Unimplemented component type={}", component_type);
return {std::move(original_value), true};
}
}
@@ -459,7 +467,7 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
default:
break;
}
- UNIMPLEMENTED_MSG("Unimplemented operation={} type={}",
+ UNIMPLEMENTED_MSG("Unimplemented operation={}, type={}",
static_cast<u64>(instr.suatom_d.operation.Value()),
static_cast<u64>(instr.suatom_d.operation_type.Value()));
return OperationCode::AtomicImageAdd;
diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp
index 9392f065b..63adbc4a3 100644
--- a/src/video_core/shader/decode/memory.cpp
+++ b/src/video_core/shader/decode/memory.cpp
@@ -387,7 +387,6 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
}
case OpCode::Id::RED: {
UNIMPLEMENTED_IF_MSG(instr.red.type != GlobalAtomicType::U32);
- UNIMPLEMENTED_IF_MSG(instr.red.operation != AtomicOp::Add);
const auto [real_address, base_address, descriptor] =
TrackGlobalMemory(bb, instr, true, true);
if (!real_address || !base_address) {
@@ -396,7 +395,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
}
Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor);
Node value = GetRegister(instr.gpr0);
- bb.push_back(Operation(OperationCode::ReduceIAdd, move(gmem), move(value)));
+ bb.push_back(Operation(GetAtomOperation(instr.red.operation), move(gmem), move(value)));
break;
}
case OpCode::Id::ATOM: {
diff --git a/src/video_core/shader/decode/other.cpp b/src/video_core/shader/decode/other.cpp
index d4f95b18c..c0a8f233f 100644
--- a/src/video_core/shader/decode/other.cpp
+++ b/src/video_core/shader/decode/other.cpp
@@ -83,7 +83,7 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
return Operation(OperationCode::YNegate);
case SystemVariable::InvocationInfo:
LOG_WARNING(HW_GPU, "S2R instruction with InvocationInfo is incomplete");
- return Immediate(0U);
+ return Immediate(0x00ff'0000U);
case SystemVariable::WscaleFactorXY:
UNIMPLEMENTED_MSG("S2R WscaleFactorXY is not implemented");
return Immediate(0U);
@@ -109,6 +109,27 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
return Operation(OperationCode::WorkGroupIdY);
case SystemVariable::CtaIdZ:
return Operation(OperationCode::WorkGroupIdZ);
+ case SystemVariable::EqMask:
+ case SystemVariable::LtMask:
+ case SystemVariable::LeMask:
+ case SystemVariable::GtMask:
+ case SystemVariable::GeMask:
+ uses_warps = true;
+ switch (instr.sys20) {
+ case SystemVariable::EqMask:
+ return Operation(OperationCode::ThreadEqMask);
+ case SystemVariable::LtMask:
+ return Operation(OperationCode::ThreadLtMask);
+ case SystemVariable::LeMask:
+ return Operation(OperationCode::ThreadLeMask);
+ case SystemVariable::GtMask:
+ return Operation(OperationCode::ThreadGtMask);
+ case SystemVariable::GeMask:
+ return Operation(OperationCode::ThreadGeMask);
+ default:
+ UNREACHABLE();
+ return Immediate(0u);
+ }
default:
UNIMPLEMENTED_MSG("Unhandled system move: {}",
static_cast<u32>(instr.sys20.Value()));
@@ -272,10 +293,25 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
SetRegister(bb, instr.gpr0, GetRegister(instr.gpr8));
break;
}
+ case OpCode::Id::BAR: {
+ UNIMPLEMENTED_IF_MSG(instr.value != 0xF0A81B8000070000ULL, "BAR is not BAR.SYNC 0x0");
+ bb.push_back(Operation(OperationCode::Barrier));
+ break;
+ }
case OpCode::Id::MEMBAR: {
- UNIMPLEMENTED_IF(instr.membar.type != Tegra::Shader::MembarType::GL);
UNIMPLEMENTED_IF(instr.membar.unknown != Tegra::Shader::MembarUnknown::Default);
- bb.push_back(Operation(OperationCode::MemoryBarrierGL));
+ const OperationCode type = [instr] {
+ switch (instr.membar.type) {
+ case Tegra::Shader::MembarType::CTA:
+ return OperationCode::MemoryBarrierGroup;
+ case Tegra::Shader::MembarType::GL:
+ return OperationCode::MemoryBarrierGlobal;
+ default:
+ UNIMPLEMENTED_MSG("MEMBAR type={}", static_cast<int>(instr.membar.type.Value()));
+ return OperationCode::MemoryBarrierGlobal;
+ }
+ }();
+ bb.push_back(Operation(type));
break;
}
case OpCode::Id::DEPBAR: {
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index 8f0bb996e..29ebf65ba 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -357,13 +357,11 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
return pc;
}
-ShaderIR::SamplerInfo ShaderIR::GetSamplerInfo(SamplerInfo info, u32 offset,
- std::optional<u32> buffer) {
+ShaderIR::SamplerInfo ShaderIR::GetSamplerInfo(
+ SamplerInfo info, std::optional<Tegra::Engines::SamplerDescriptor> sampler) {
if (info.IsComplete()) {
return info;
}
- const auto sampler = buffer ? registry.ObtainBindlessSampler(*buffer, offset)
- : registry.ObtainBoundSampler(offset);
if (!sampler) {
LOG_WARNING(HW_GPU, "Unknown sampler info");
info.type = info.type.value_or(Tegra::Shader::TextureType::Texture2D);
@@ -381,8 +379,8 @@ ShaderIR::SamplerInfo ShaderIR::GetSamplerInfo(SamplerInfo info, u32 offset,
std::optional<Sampler> ShaderIR::GetSampler(Tegra::Shader::Sampler sampler,
SamplerInfo sampler_info) {
- const auto offset = static_cast<u32>(sampler.index.Value());
- const auto info = GetSamplerInfo(sampler_info, offset);
+ const u32 offset = static_cast<u32>(sampler.index.Value());
+ const auto info = GetSamplerInfo(sampler_info, registry.ObtainBoundSampler(offset));
// If this sampler has already been used, return the existing mapping.
const auto it = std::find_if(used_samplers.begin(), used_samplers.end(),
@@ -404,20 +402,19 @@ std::optional<Sampler> ShaderIR::GetBindlessSampler(Tegra::Shader::Register reg,
const Node sampler_register = GetRegister(reg);
const auto [base_node, tracked_sampler_info] =
TrackBindlessSampler(sampler_register, global_code, static_cast<s64>(global_code.size()));
- ASSERT(base_node != nullptr);
- if (base_node == nullptr) {
+ if (!base_node) {
+ UNREACHABLE();
return std::nullopt;
}
- if (const auto bindless_sampler_info =
- std::get_if<BindlessSamplerNode>(&*tracked_sampler_info)) {
- const u32 buffer = bindless_sampler_info->GetIndex();
- const u32 offset = bindless_sampler_info->GetOffset();
- info = GetSamplerInfo(info, offset, buffer);
+ if (const auto sampler_info = std::get_if<BindlessSamplerNode>(&*tracked_sampler_info)) {
+ const u32 buffer = sampler_info->index;
+ const u32 offset = sampler_info->offset;
+ info = GetSamplerInfo(info, registry.ObtainBindlessSampler(buffer, offset));
// If this sampler has already been used, return the existing mapping.
const auto it = std::find_if(used_samplers.begin(), used_samplers.end(),
- [buffer = buffer, offset = offset](const Sampler& entry) {
+ [buffer, offset](const Sampler& entry) {
return entry.buffer == buffer && entry.offset == offset;
});
if (it != used_samplers.end()) {
@@ -431,10 +428,32 @@ std::optional<Sampler> ShaderIR::GetBindlessSampler(Tegra::Shader::Register reg,
return used_samplers.emplace_back(next_index, offset, buffer, *info.type, *info.is_array,
*info.is_shadow, *info.is_buffer, false);
}
- if (const auto array_sampler_info = std::get_if<ArraySamplerNode>(&*tracked_sampler_info)) {
- const u32 base_offset = array_sampler_info->GetBaseOffset() / 4;
- index_var = GetCustomVariable(array_sampler_info->GetIndexVar());
- info = GetSamplerInfo(info, base_offset);
+ if (const auto sampler_info = std::get_if<SeparateSamplerNode>(&*tracked_sampler_info)) {
+ const std::pair indices = sampler_info->indices;
+ const std::pair offsets = sampler_info->offsets;
+ info = GetSamplerInfo(info, registry.ObtainSeparateSampler(indices, offsets));
+
+ // Try to use an already created sampler if it exists
+ const auto it = std::find_if(
+ used_samplers.begin(), used_samplers.end(), [indices, offsets](const Sampler& entry) {
+ return offsets == std::pair{entry.offset, entry.secondary_offset} &&
+ indices == std::pair{entry.buffer, entry.secondary_buffer};
+ });
+ if (it != used_samplers.end()) {
+ ASSERT(it->is_separated && it->type == info.type && it->is_array == info.is_array &&
+ it->is_shadow == info.is_shadow && it->is_buffer == info.is_buffer);
+ return *it;
+ }
+
+ // Otherwise create a new mapping for this sampler
+ const u32 next_index = static_cast<u32>(used_samplers.size());
+ return used_samplers.emplace_back(next_index, offsets, indices, *info.type, *info.is_array,
+ *info.is_shadow, *info.is_buffer);
+ }
+ if (const auto sampler_info = std::get_if<ArraySamplerNode>(&*tracked_sampler_info)) {
+ const u32 base_offset = sampler_info->base_offset / 4;
+ index_var = GetCustomVariable(sampler_info->bindless_var);
+ info = GetSamplerInfo(info, registry.ObtainBoundSampler(base_offset));
// If this sampler has already been used, return the existing mapping.
const auto it = std::find_if(
diff --git a/src/video_core/shader/decode/xmad.cpp b/src/video_core/shader/decode/xmad.cpp
index 6191ffba1..c83dc6615 100644
--- a/src/video_core/shader/decode/xmad.cpp
+++ b/src/video_core/shader/decode/xmad.cpp
@@ -97,19 +97,19 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
return SignedOperation(OperationCode::IAdd, is_signed_c, original_c, shifted_b);
}
case Tegra::Shader::XmadMode::CSfu: {
- const Node comp_a = GetPredicateComparisonInteger(PredCondition::Equal, is_signed_a,
- op_a, Immediate(0));
- const Node comp_b = GetPredicateComparisonInteger(PredCondition::Equal, is_signed_b,
- op_b, Immediate(0));
+ const Node comp_a =
+ GetPredicateComparisonInteger(PredCondition::EQ, is_signed_a, op_a, Immediate(0));
+ const Node comp_b =
+ GetPredicateComparisonInteger(PredCondition::EQ, is_signed_b, op_b, Immediate(0));
const Node comp = Operation(OperationCode::LogicalOr, comp_a, comp_b);
const Node comp_minus_a = GetPredicateComparisonInteger(
- PredCondition::NotEqual, is_signed_a,
+ PredCondition::NE, is_signed_a,
SignedOperation(OperationCode::IBitwiseAnd, is_signed_a, op_a,
Immediate(0x80000000)),
Immediate(0));
const Node comp_minus_b = GetPredicateComparisonInteger(
- PredCondition::NotEqual, is_signed_b,
+ PredCondition::NE, is_signed_b,
SignedOperation(OperationCode::IBitwiseAnd, is_signed_b, op_b,
Immediate(0x80000000)),
Immediate(0));
diff --git a/src/video_core/shader/memory_util.cpp b/src/video_core/shader/memory_util.cpp
index 074f21691..5071c83ca 100644
--- a/src/video_core/shader/memory_util.cpp
+++ b/src/video_core/shader/memory_util.cpp
@@ -66,12 +66,12 @@ ProgramCode GetShaderCode(Tegra::MemoryManager& memory_manager, GPUVAddr gpu_add
u64 GetUniqueIdentifier(Tegra::Engines::ShaderType shader_type, bool is_a, const ProgramCode& code,
const ProgramCode& code_b) {
- u64 unique_identifier = boost::hash_value(code);
+ size_t unique_identifier = boost::hash_value(code);
if (is_a) {
// VertexA programs include two programs
boost::hash_combine(unique_identifier, boost::hash_value(code_b));
}
- return unique_identifier;
+ return static_cast<u64>(unique_identifier);
}
} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h
index 601c822d2..8f230d57a 100644
--- a/src/video_core/shader/node.h
+++ b/src/video_core/shader/node.h
@@ -110,13 +110,20 @@ enum class OperationCode {
LogicalPick2, /// (bool2 pair, uint index) -> bool
LogicalAnd2, /// (bool2 a) -> bool
- LogicalFLessThan, /// (float a, float b) -> bool
- LogicalFEqual, /// (float a, float b) -> bool
- LogicalFLessEqual, /// (float a, float b) -> bool
- LogicalFGreaterThan, /// (float a, float b) -> bool
- LogicalFNotEqual, /// (float a, float b) -> bool
- LogicalFGreaterEqual, /// (float a, float b) -> bool
- LogicalFIsNan, /// (float a) -> bool
+ LogicalFOrdLessThan, /// (float a, float b) -> bool
+ LogicalFOrdEqual, /// (float a, float b) -> bool
+ LogicalFOrdLessEqual, /// (float a, float b) -> bool
+ LogicalFOrdGreaterThan, /// (float a, float b) -> bool
+ LogicalFOrdNotEqual, /// (float a, float b) -> bool
+ LogicalFOrdGreaterEqual, /// (float a, float b) -> bool
+ LogicalFOrdered, /// (float a, float b) -> bool
+ LogicalFUnordered, /// (float a, float b) -> bool
+ LogicalFUnordLessThan, /// (float a, float b) -> bool
+ LogicalFUnordEqual, /// (float a, float b) -> bool
+ LogicalFUnordLessEqual, /// (float a, float b) -> bool
+ LogicalFUnordGreaterThan, /// (float a, float b) -> bool
+ LogicalFUnordNotEqual, /// (float a, float b) -> bool
+ LogicalFUnordGreaterEqual, /// (float a, float b) -> bool
LogicalILessThan, /// (int a, int b) -> bool
LogicalIEqual, /// (int a, int b) -> bool
@@ -219,9 +226,16 @@ enum class OperationCode {
VoteEqual, /// (bool) -> bool
ThreadId, /// () -> uint
+ ThreadEqMask, /// () -> uint
+ ThreadGeMask, /// () -> uint
+ ThreadGtMask, /// () -> uint
+ ThreadLeMask, /// () -> uint
+ ThreadLtMask, /// () -> uint
ShuffleIndexed, /// (uint value, uint index) -> uint
- MemoryBarrierGL, /// () -> void
+ Barrier, /// () -> void
+ MemoryBarrierGroup, /// () -> void
+ MemoryBarrierGlobal, /// () -> void
Amount,
};
@@ -261,10 +275,11 @@ using Node = std::shared_ptr<NodeData>;
using Node4 = std::array<Node, 4>;
using NodeBlock = std::vector<Node>;
-class BindlessSamplerNode;
-class ArraySamplerNode;
+struct ArraySamplerNode;
+struct BindlessSamplerNode;
+struct SeparateSamplerNode;
-using TrackSamplerData = std::variant<BindlessSamplerNode, ArraySamplerNode>;
+using TrackSamplerData = std::variant<BindlessSamplerNode, SeparateSamplerNode, ArraySamplerNode>;
using TrackSampler = std::shared_ptr<TrackSamplerData>;
struct Sampler {
@@ -274,63 +289,51 @@ struct Sampler {
: index{index}, offset{offset}, type{type}, is_array{is_array}, is_shadow{is_shadow},
is_buffer{is_buffer}, is_indexed{is_indexed} {}
+ /// Separate sampler constructor
+ constexpr explicit Sampler(u32 index, std::pair<u32, u32> offsets, std::pair<u32, u32> buffers,
+ Tegra::Shader::TextureType type, bool is_array, bool is_shadow,
+ bool is_buffer)
+ : index{index}, offset{offsets.first}, secondary_offset{offsets.second},
+ buffer{buffers.first}, secondary_buffer{buffers.second}, type{type}, is_array{is_array},
+ is_shadow{is_shadow}, is_buffer{is_buffer}, is_separated{true} {}
+
/// Bindless samplers constructor
constexpr explicit Sampler(u32 index, u32 offset, u32 buffer, Tegra::Shader::TextureType type,
bool is_array, bool is_shadow, bool is_buffer, bool is_indexed)
: index{index}, offset{offset}, buffer{buffer}, type{type}, is_array{is_array},
is_shadow{is_shadow}, is_buffer{is_buffer}, is_bindless{true}, is_indexed{is_indexed} {}
- u32 index = 0; ///< Emulated index given for the this sampler.
- u32 offset = 0; ///< Offset in the const buffer from where the sampler is being read.
- u32 buffer = 0; ///< Buffer where the bindless sampler is being read (unused on bound samplers).
- u32 size = 1; ///< Size of the sampler.
+ u32 index = 0; ///< Emulated index given for the this sampler.
+ u32 offset = 0; ///< Offset in the const buffer from where the sampler is being read.
+ u32 secondary_offset = 0; ///< Secondary offset in the const buffer.
+ u32 buffer = 0; ///< Buffer where the bindless sampler is read.
+ u32 secondary_buffer = 0; ///< Secondary buffer where the bindless sampler is read.
+ u32 size = 1; ///< Size of the sampler.
Tegra::Shader::TextureType type{}; ///< The type used to sample this texture (Texture2D, etc)
- bool is_array = false; ///< Whether the texture is being sampled as an array texture or not.
- bool is_shadow = false; ///< Whether the texture is being sampled as a depth texture or not.
- bool is_buffer = false; ///< Whether the texture is a texture buffer without sampler.
- bool is_bindless = false; ///< Whether this sampler belongs to a bindless texture or not.
- bool is_indexed = false; ///< Whether this sampler is an indexed array of textures.
+ bool is_array = false; ///< Whether the texture is being sampled as an array texture or not.
+ bool is_shadow = false; ///< Whether the texture is being sampled as a depth texture or not.
+ bool is_buffer = false; ///< Whether the texture is a texture buffer without sampler.
+ bool is_bindless = false; ///< Whether this sampler belongs to a bindless texture or not.
+ bool is_indexed = false; ///< Whether this sampler is an indexed array of textures.
+ bool is_separated = false; ///< Whether the image and sampler is separated or not.
};
/// Represents a tracked bindless sampler into a direct const buffer
-class ArraySamplerNode final {
-public:
- explicit ArraySamplerNode(u32 index, u32 base_offset, u32 bindless_var)
- : index{index}, base_offset{base_offset}, bindless_var{bindless_var} {}
-
- constexpr u32 GetIndex() const {
- return index;
- }
-
- constexpr u32 GetBaseOffset() const {
- return base_offset;
- }
-
- constexpr u32 GetIndexVar() const {
- return bindless_var;
- }
-
-private:
+struct ArraySamplerNode {
u32 index;
u32 base_offset;
u32 bindless_var;
};
-/// Represents a tracked bindless sampler into a direct const buffer
-class BindlessSamplerNode final {
-public:
- explicit BindlessSamplerNode(u32 index, u32 offset) : index{index}, offset{offset} {}
-
- constexpr u32 GetIndex() const {
- return index;
- }
-
- constexpr u32 GetOffset() const {
- return offset;
- }
+/// Represents a tracked separate sampler image pair that was folded statically
+struct SeparateSamplerNode {
+ std::pair<u32, u32> indices;
+ std::pair<u32, u32> offsets;
+};
-private:
+/// Represents a tracked bindless sampler into a direct const buffer
+struct BindlessSamplerNode {
u32 index;
u32 offset;
};
diff --git a/src/video_core/shader/node_helper.h b/src/video_core/shader/node_helper.h
index 11231bbea..1e0886185 100644
--- a/src/video_core/shader/node_helper.h
+++ b/src/video_core/shader/node_helper.h
@@ -48,7 +48,7 @@ Node MakeNode(Args&&... args) {
template <typename T, typename... Args>
TrackSampler MakeTrackSampler(Args&&... args) {
static_assert(std::is_convertible_v<T, TrackSamplerData>);
- return std::make_shared<TrackSamplerData>(T(std::forward<Args>(args)...));
+ return std::make_shared<TrackSamplerData>(T{std::forward<Args>(args)...});
}
template <typename... Args>
diff --git a/src/video_core/shader/registry.cpp b/src/video_core/shader/registry.cpp
index af70b3f35..cdf274e54 100644
--- a/src/video_core/shader/registry.cpp
+++ b/src/video_core/shader/registry.cpp
@@ -93,6 +93,26 @@ std::optional<SamplerDescriptor> Registry::ObtainBoundSampler(u32 offset) {
return value;
}
+std::optional<Tegra::Engines::SamplerDescriptor> Registry::ObtainSeparateSampler(
+ std::pair<u32, u32> buffers, std::pair<u32, u32> offsets) {
+ SeparateSamplerKey key;
+ key.buffers = buffers;
+ key.offsets = offsets;
+ const auto iter = separate_samplers.find(key);
+ if (iter != separate_samplers.end()) {
+ return iter->second;
+ }
+ if (!engine) {
+ return std::nullopt;
+ }
+
+ const u32 handle_1 = engine->AccessConstBuffer32(stage, key.buffers.first, key.offsets.first);
+ const u32 handle_2 = engine->AccessConstBuffer32(stage, key.buffers.second, key.offsets.second);
+ const SamplerDescriptor value = engine->AccessSampler(handle_1 | handle_2);
+ separate_samplers.emplace(key, value);
+ return value;
+}
+
std::optional<Tegra::Engines::SamplerDescriptor> Registry::ObtainBindlessSampler(u32 buffer,
u32 offset) {
const std::pair key = {buffer, offset};
diff --git a/src/video_core/shader/registry.h b/src/video_core/shader/registry.h
index 0c80d35fd..231206765 100644
--- a/src/video_core/shader/registry.h
+++ b/src/video_core/shader/registry.h
@@ -19,8 +19,39 @@
namespace VideoCommon::Shader {
+struct SeparateSamplerKey {
+ std::pair<u32, u32> buffers;
+ std::pair<u32, u32> offsets;
+};
+
+} // namespace VideoCommon::Shader
+
+namespace std {
+
+template <>
+struct hash<VideoCommon::Shader::SeparateSamplerKey> {
+ std::size_t operator()(const VideoCommon::Shader::SeparateSamplerKey& key) const noexcept {
+ return std::hash<u32>{}(key.buffers.first ^ key.buffers.second ^ key.offsets.first ^
+ key.offsets.second);
+ }
+};
+
+template <>
+struct equal_to<VideoCommon::Shader::SeparateSamplerKey> {
+ bool operator()(const VideoCommon::Shader::SeparateSamplerKey& lhs,
+ const VideoCommon::Shader::SeparateSamplerKey& rhs) const noexcept {
+ return lhs.buffers == rhs.buffers && lhs.offsets == rhs.offsets;
+ }
+};
+
+} // namespace std
+
+namespace VideoCommon::Shader {
+
using KeyMap = std::unordered_map<std::pair<u32, u32>, u32, Common::PairHash>;
using BoundSamplerMap = std::unordered_map<u32, Tegra::Engines::SamplerDescriptor>;
+using SeparateSamplerMap =
+ std::unordered_map<SeparateSamplerKey, Tegra::Engines::SamplerDescriptor>;
using BindlessSamplerMap =
std::unordered_map<std::pair<u32, u32>, Tegra::Engines::SamplerDescriptor, Common::PairHash>;
@@ -73,6 +104,9 @@ public:
std::optional<Tegra::Engines::SamplerDescriptor> ObtainBoundSampler(u32 offset);
+ std::optional<Tegra::Engines::SamplerDescriptor> ObtainSeparateSampler(
+ std::pair<u32, u32> buffers, std::pair<u32, u32> offsets);
+
std::optional<Tegra::Engines::SamplerDescriptor> ObtainBindlessSampler(u32 buffer, u32 offset);
/// Inserts a key.
@@ -128,6 +162,7 @@ private:
Tegra::Engines::ConstBufferEngineInterface* engine = nullptr;
KeyMap keys;
BoundSamplerMap bound_samplers;
+ SeparateSamplerMap separate_samplers;
BindlessSamplerMap bindless_samplers;
u32 bound_buffer;
GraphicsInfo graphics_info;
diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp
index 822674926..e322c3402 100644
--- a/src/video_core/shader/shader_ir.cpp
+++ b/src/video_core/shader/shader_ir.cpp
@@ -10,6 +10,7 @@
#include "common/common_types.h"
#include "common/logging/log.h"
#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/node.h"
#include "video_core/shader/node_helper.h"
#include "video_core/shader/registry.h"
#include "video_core/shader/shader_ir.h"
@@ -243,56 +244,44 @@ Node ShaderIR::GetSaturatedHalfFloat(Node value, bool saturate) {
}
Node ShaderIR::GetPredicateComparisonFloat(PredCondition condition, Node op_a, Node op_b) {
+ if (condition == PredCondition::T) {
+ return GetPredicate(true);
+ } else if (condition == PredCondition::F) {
+ return GetPredicate(false);
+ }
+
static constexpr std::array comparison_table{
- std::pair{PredCondition::LessThan, OperationCode::LogicalFLessThan},
- std::pair{PredCondition::Equal, OperationCode::LogicalFEqual},
- std::pair{PredCondition::LessEqual, OperationCode::LogicalFLessEqual},
- std::pair{PredCondition::GreaterThan, OperationCode::LogicalFGreaterThan},
- std::pair{PredCondition::NotEqual, OperationCode::LogicalFNotEqual},
- std::pair{PredCondition::GreaterEqual, OperationCode::LogicalFGreaterEqual},
- std::pair{PredCondition::LessThanWithNan, OperationCode::LogicalFLessThan},
- std::pair{PredCondition::NotEqualWithNan, OperationCode::LogicalFNotEqual},
- std::pair{PredCondition::LessEqualWithNan, OperationCode::LogicalFLessEqual},
- std::pair{PredCondition::GreaterThanWithNan, OperationCode::LogicalFGreaterThan},
- std::pair{PredCondition::GreaterEqualWithNan, OperationCode::LogicalFGreaterEqual},
+ OperationCode(0),
+ OperationCode::LogicalFOrdLessThan, // LT
+ OperationCode::LogicalFOrdEqual, // EQ
+ OperationCode::LogicalFOrdLessEqual, // LE
+ OperationCode::LogicalFOrdGreaterThan, // GT
+ OperationCode::LogicalFOrdNotEqual, // NE
+ OperationCode::LogicalFOrdGreaterEqual, // GE
+ OperationCode::LogicalFOrdered, // NUM
+ OperationCode::LogicalFUnordered, // NAN
+ OperationCode::LogicalFUnordLessThan, // LTU
+ OperationCode::LogicalFUnordEqual, // EQU
+ OperationCode::LogicalFUnordLessEqual, // LEU
+ OperationCode::LogicalFUnordGreaterThan, // GTU
+ OperationCode::LogicalFUnordNotEqual, // NEU
+ OperationCode::LogicalFUnordGreaterEqual, // GEU
};
+ const std::size_t index = static_cast<std::size_t>(condition);
+ ASSERT_MSG(index < std::size(comparison_table), "Invalid condition={}", index);
- const auto comparison =
- std::find_if(comparison_table.cbegin(), comparison_table.cend(),
- [condition](const auto entry) { return condition == entry.first; });
- UNIMPLEMENTED_IF_MSG(comparison == comparison_table.cend(),
- "Unknown predicate comparison operation");
-
- Node predicate = Operation(comparison->second, NO_PRECISE, op_a, op_b);
-
- if (condition == PredCondition::LessThanWithNan ||
- condition == PredCondition::NotEqualWithNan ||
- condition == PredCondition::LessEqualWithNan ||
- condition == PredCondition::GreaterThanWithNan ||
- condition == PredCondition::GreaterEqualWithNan) {
- predicate = Operation(OperationCode::LogicalOr, predicate,
- Operation(OperationCode::LogicalFIsNan, op_a));
- predicate = Operation(OperationCode::LogicalOr, predicate,
- Operation(OperationCode::LogicalFIsNan, op_b));
- }
-
- return predicate;
+ return Operation(comparison_table[index], op_a, op_b);
}
Node ShaderIR::GetPredicateComparisonInteger(PredCondition condition, bool is_signed, Node op_a,
Node op_b) {
static constexpr std::array comparison_table{
- std::pair{PredCondition::LessThan, OperationCode::LogicalILessThan},
- std::pair{PredCondition::Equal, OperationCode::LogicalIEqual},
- std::pair{PredCondition::LessEqual, OperationCode::LogicalILessEqual},
- std::pair{PredCondition::GreaterThan, OperationCode::LogicalIGreaterThan},
- std::pair{PredCondition::NotEqual, OperationCode::LogicalINotEqual},
- std::pair{PredCondition::GreaterEqual, OperationCode::LogicalIGreaterEqual},
- std::pair{PredCondition::LessThanWithNan, OperationCode::LogicalILessThan},
- std::pair{PredCondition::NotEqualWithNan, OperationCode::LogicalINotEqual},
- std::pair{PredCondition::LessEqualWithNan, OperationCode::LogicalILessEqual},
- std::pair{PredCondition::GreaterThanWithNan, OperationCode::LogicalIGreaterThan},
- std::pair{PredCondition::GreaterEqualWithNan, OperationCode::LogicalIGreaterEqual},
+ std::pair{PredCondition::LT, OperationCode::LogicalILessThan},
+ std::pair{PredCondition::EQ, OperationCode::LogicalIEqual},
+ std::pair{PredCondition::LE, OperationCode::LogicalILessEqual},
+ std::pair{PredCondition::GT, OperationCode::LogicalIGreaterThan},
+ std::pair{PredCondition::NE, OperationCode::LogicalINotEqual},
+ std::pair{PredCondition::GE, OperationCode::LogicalIGreaterEqual},
};
const auto comparison =
@@ -301,32 +290,24 @@ Node ShaderIR::GetPredicateComparisonInteger(PredCondition condition, bool is_si
UNIMPLEMENTED_IF_MSG(comparison == comparison_table.cend(),
"Unknown predicate comparison operation");
- Node predicate = SignedOperation(comparison->second, is_signed, NO_PRECISE, std::move(op_a),
- std::move(op_b));
-
- UNIMPLEMENTED_IF_MSG(condition == PredCondition::LessThanWithNan ||
- condition == PredCondition::NotEqualWithNan ||
- condition == PredCondition::LessEqualWithNan ||
- condition == PredCondition::GreaterThanWithNan ||
- condition == PredCondition::GreaterEqualWithNan,
- "NaN comparisons for integers are not implemented");
- return predicate;
+ return SignedOperation(comparison->second, is_signed, NO_PRECISE, std::move(op_a),
+ std::move(op_b));
}
Node ShaderIR::GetPredicateComparisonHalf(Tegra::Shader::PredCondition condition, Node op_a,
Node op_b) {
static constexpr std::array comparison_table{
- std::pair{PredCondition::LessThan, OperationCode::Logical2HLessThan},
- std::pair{PredCondition::Equal, OperationCode::Logical2HEqual},
- std::pair{PredCondition::LessEqual, OperationCode::Logical2HLessEqual},
- std::pair{PredCondition::GreaterThan, OperationCode::Logical2HGreaterThan},
- std::pair{PredCondition::NotEqual, OperationCode::Logical2HNotEqual},
- std::pair{PredCondition::GreaterEqual, OperationCode::Logical2HGreaterEqual},
- std::pair{PredCondition::LessThanWithNan, OperationCode::Logical2HLessThanWithNan},
- std::pair{PredCondition::NotEqualWithNan, OperationCode::Logical2HNotEqualWithNan},
- std::pair{PredCondition::LessEqualWithNan, OperationCode::Logical2HLessEqualWithNan},
- std::pair{PredCondition::GreaterThanWithNan, OperationCode::Logical2HGreaterThanWithNan},
- std::pair{PredCondition::GreaterEqualWithNan, OperationCode::Logical2HGreaterEqualWithNan},
+ std::pair{PredCondition::LT, OperationCode::Logical2HLessThan},
+ std::pair{PredCondition::EQ, OperationCode::Logical2HEqual},
+ std::pair{PredCondition::LE, OperationCode::Logical2HLessEqual},
+ std::pair{PredCondition::GT, OperationCode::Logical2HGreaterThan},
+ std::pair{PredCondition::NE, OperationCode::Logical2HNotEqual},
+ std::pair{PredCondition::GE, OperationCode::Logical2HGreaterEqual},
+ std::pair{PredCondition::LTU, OperationCode::Logical2HLessThanWithNan},
+ std::pair{PredCondition::LEU, OperationCode::Logical2HLessEqualWithNan},
+ std::pair{PredCondition::GTU, OperationCode::Logical2HGreaterThanWithNan},
+ std::pair{PredCondition::NEU, OperationCode::Logical2HNotEqualWithNan},
+ std::pair{PredCondition::GEU, OperationCode::Logical2HGreaterEqualWithNan},
};
const auto comparison =
@@ -397,7 +378,7 @@ void ShaderIR::SetInternalFlagsFromFloat(NodeBlock& bb, Node value, bool sets_cc
if (!sets_cc) {
return;
}
- Node zerop = Operation(OperationCode::LogicalFEqual, std::move(value), Immediate(0.0f));
+ Node zerop = Operation(OperationCode::LogicalFOrdEqual, std::move(value), Immediate(0.0f));
SetInternalFlag(bb, InternalFlag::Zero, std::move(zerop));
LOG_WARNING(HW_GPU, "Condition codes implementation is incomplete");
}
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index 15ae152f2..3a98b2104 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -330,8 +330,8 @@ private:
OperationCode GetPredicateCombiner(Tegra::Shader::PredOperation operation);
/// Queries the missing sampler info from the execution context.
- SamplerInfo GetSamplerInfo(SamplerInfo info, u32 offset,
- std::optional<u32> buffer = std::nullopt);
+ SamplerInfo GetSamplerInfo(SamplerInfo info,
+ std::optional<Tegra::Engines::SamplerDescriptor> sampler);
/// Accesses a texture sampler.
std::optional<Sampler> GetSampler(Tegra::Shader::Sampler sampler, SamplerInfo info);
@@ -409,8 +409,14 @@ private:
std::tuple<Node, u32, u32> TrackCbuf(Node tracked, const NodeBlock& code, s64 cursor) const;
- std::tuple<Node, TrackSampler> TrackBindlessSampler(Node tracked, const NodeBlock& code,
- s64 cursor);
+ std::pair<Node, TrackSampler> TrackBindlessSampler(Node tracked, const NodeBlock& code,
+ s64 cursor);
+
+ std::pair<Node, TrackSampler> HandleBindlessIndirectRead(const CbufNode& cbuf,
+ const OperationNode& operation,
+ Node gpr, Node base_offset,
+ Node tracked, const NodeBlock& code,
+ s64 cursor);
std::optional<u32> TrackImmediate(Node tracked, const NodeBlock& code, s64 cursor) const;
diff --git a/src/video_core/shader/track.cpp b/src/video_core/shader/track.cpp
index eb97bfd41..d5ed81442 100644
--- a/src/video_core/shader/track.cpp
+++ b/src/video_core/shader/track.cpp
@@ -14,6 +14,7 @@
namespace VideoCommon::Shader {
namespace {
+
std::pair<Node, s64> FindOperation(const NodeBlock& code, s64 cursor,
OperationCode operation_code) {
for (; cursor >= 0; --cursor) {
@@ -63,7 +64,8 @@ bool AmendNodeCv(std::size_t amend_index, Node node) {
if (const auto operation = std::get_if<OperationNode>(&*node)) {
operation->SetAmendIndex(amend_index);
return true;
- } else if (const auto conditional = std::get_if<ConditionalNode>(&*node)) {
+ }
+ if (const auto conditional = std::get_if<ConditionalNode>(&*node)) {
conditional->SetAmendIndex(amend_index);
return true;
}
@@ -72,40 +74,27 @@ bool AmendNodeCv(std::size_t amend_index, Node node) {
} // Anonymous namespace
-std::tuple<Node, TrackSampler> ShaderIR::TrackBindlessSampler(Node tracked, const NodeBlock& code,
- s64 cursor) {
+std::pair<Node, TrackSampler> ShaderIR::TrackBindlessSampler(Node tracked, const NodeBlock& code,
+ s64 cursor) {
if (const auto cbuf = std::get_if<CbufNode>(&*tracked)) {
+ const u32 cbuf_index = cbuf->GetIndex();
+
// Constant buffer found, test if it's an immediate
const auto& offset = cbuf->GetOffset();
if (const auto immediate = std::get_if<ImmediateNode>(&*offset)) {
- auto track =
- MakeTrackSampler<BindlessSamplerNode>(cbuf->GetIndex(), immediate->GetValue());
+ auto track = MakeTrackSampler<BindlessSamplerNode>(cbuf_index, immediate->GetValue());
return {tracked, track};
}
if (const auto operation = std::get_if<OperationNode>(&*offset)) {
const u32 bound_buffer = registry.GetBoundBuffer();
- if (bound_buffer != cbuf->GetIndex()) {
+ if (bound_buffer != cbuf_index) {
return {};
}
- const auto pair = DecoupleIndirectRead(*operation);
- if (!pair) {
- return {};
+ if (const std::optional pair = DecoupleIndirectRead(*operation)) {
+ auto [gpr, base_offset] = *pair;
+ return HandleBindlessIndirectRead(*cbuf, *operation, gpr, base_offset, tracked,
+ code, cursor);
}
- auto [gpr, base_offset] = *pair;
- const auto offset_inm = std::get_if<ImmediateNode>(&*base_offset);
- const auto& gpu_driver = registry.AccessGuestDriverProfile();
- const u32 bindless_cv = NewCustomVariable();
- Node op =
- Operation(OperationCode::UDiv, gpr, Immediate(gpu_driver.GetTextureHandlerSize()));
-
- const Node cv_node = GetCustomVariable(bindless_cv);
- Node amend_op = Operation(OperationCode::Assign, cv_node, std::move(op));
- const std::size_t amend_index = DeclareAmend(std::move(amend_op));
- AmendNodeCv(amend_index, code[cursor]);
- // TODO Implement Bindless Index custom variable
- auto track = MakeTrackSampler<ArraySamplerNode>(cbuf->GetIndex(),
- offset_inm->GetValue(), bindless_cv);
- return {tracked, track};
}
return {};
}
@@ -122,10 +111,23 @@ std::tuple<Node, TrackSampler> ShaderIR::TrackBindlessSampler(Node tracked, cons
return TrackBindlessSampler(source, code, new_cursor);
}
if (const auto operation = std::get_if<OperationNode>(&*tracked)) {
- for (std::size_t i = operation->GetOperandsCount(); i > 0; --i) {
- if (auto found = TrackBindlessSampler((*operation)[i - 1], code, cursor);
- std::get<0>(found)) {
- // Cbuf found in operand.
+ const OperationNode& op = *operation;
+
+ const OperationCode opcode = operation->GetCode();
+ if (opcode == OperationCode::IBitwiseOr || opcode == OperationCode::UBitwiseOr) {
+ ASSERT(op.GetOperandsCount() == 2);
+ auto [node_a, index_a, offset_a] = TrackCbuf(op[0], code, cursor);
+ auto [node_b, index_b, offset_b] = TrackCbuf(op[1], code, cursor);
+ if (node_a && node_b) {
+ auto track = MakeTrackSampler<SeparateSamplerNode>(std::pair{index_a, index_b},
+ std::pair{offset_a, offset_b});
+ return {tracked, std::move(track)};
+ }
+ }
+ std::size_t i = op.GetOperandsCount();
+ while (i--) {
+ if (auto found = TrackBindlessSampler(op[i - 1], code, cursor); std::get<0>(found)) {
+ // Constant buffer found in operand.
return found;
}
}
@@ -139,6 +141,26 @@ std::tuple<Node, TrackSampler> ShaderIR::TrackBindlessSampler(Node tracked, cons
return {};
}
+std::pair<Node, TrackSampler> ShaderIR::HandleBindlessIndirectRead(
+ const CbufNode& cbuf, const OperationNode& operation, Node gpr, Node base_offset, Node tracked,
+ const NodeBlock& code, s64 cursor) {
+ const auto offset_imm = std::get<ImmediateNode>(*base_offset);
+ const auto& gpu_driver = registry.AccessGuestDriverProfile();
+ const u32 bindless_cv = NewCustomVariable();
+ const u32 texture_handler_size = gpu_driver.GetTextureHandlerSize();
+ Node op = Operation(OperationCode::UDiv, gpr, Immediate(texture_handler_size));
+
+ Node cv_node = GetCustomVariable(bindless_cv);
+ Node amend_op = Operation(OperationCode::Assign, std::move(cv_node), std::move(op));
+ const std::size_t amend_index = DeclareAmend(std::move(amend_op));
+ AmendNodeCv(amend_index, code[cursor]);
+
+ // TODO: Implement bindless index custom variable
+ auto track =
+ MakeTrackSampler<ArraySamplerNode>(cbuf.GetIndex(), offset_imm.GetValue(), bindless_cv);
+ return {tracked, track};
+}
+
std::tuple<Node, u32, u32> ShaderIR::TrackCbuf(Node tracked, const NodeBlock& code,
s64 cursor) const {
if (const auto cbuf = std::get_if<CbufNode>(&*tracked)) {
generated by cgit v1.2.3 (git 2.25.1) at 2026-01-01 08:24:55 +0000