Merge pull request #2784 from ReinUsesLisp/smem

shader_ir: Implement shared memory
master
bunnei 2019-09-18 16:26:05 +07:00 committed by GitHub
commit b31880dc5e
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 81 additions and 21 deletions

@ -325,6 +325,7 @@ public:
DeclareRegisters();
DeclarePredicates();
DeclareLocalMemory();
DeclareSharedMemory();
DeclareInternalFlags();
DeclareInputAttributes();
DeclareOutputAttributes();
@ -499,6 +500,13 @@ private:
code.AddNewLine();
}
void DeclareSharedMemory() {
if (stage != ProgramType::Compute) {
return;
}
code.AddLine("shared uint {}[];", GetSharedMemory());
}
void DeclareInternalFlags() {
for (u32 flag = 0; flag < static_cast<u32>(InternalFlag::Amount); flag++) {
const auto flag_code = static_cast<InternalFlag>(flag);
@ -881,6 +889,12 @@ private:
Type::Uint};
}
if (const auto smem = std::get_if<SmemNode>(&*node)) {
return {
fmt::format("{}[{} >> 2]", GetSharedMemory(), Visit(smem->GetAddress()).AsUint()),
Type::Uint};
}
if (const auto internal_flag = std::get_if<InternalFlagNode>(&*node)) {
return {GetInternalFlag(internal_flag->GetFlag()), Type::Bool};
}
@ -1286,6 +1300,11 @@ private:
target = {
fmt::format("{}[{} >> 2]", GetLocalMemory(), Visit(lmem->GetAddress()).AsUint()),
Type::Uint};
} else if (const auto smem = std::get_if<SmemNode>(&*dest)) {
ASSERT(stage == ProgramType::Compute);
target = {
fmt::format("{}[{} >> 2]", GetSharedMemory(), Visit(smem->GetAddress()).AsUint()),
Type::Uint};
} else if (const auto gmem = std::get_if<GmemNode>(&*dest)) {
const std::string real = Visit(gmem->GetRealAddress()).AsUint();
const std::string base = Visit(gmem->GetBaseAddress()).AsUint();
@ -2175,6 +2194,10 @@ private:
return "lmem_" + suffix;
}
std::string GetSharedMemory() const {
return fmt::format("smem_{}", suffix);
}
std::string GetInternalFlag(InternalFlag flag) const {
constexpr std::array InternalFlagNames = {"zero_flag", "sign_flag", "carry_flag",
"overflow_flag"};

@ -35,7 +35,7 @@ u32 GetUniformTypeElementsCount(Tegra::Shader::UniformType uniform_type) {
return 1;
}
}
} // namespace
} // Anonymous namespace
u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
@ -106,16 +106,17 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
}
break;
}
case OpCode::Id::LD_L: {
LOG_DEBUG(HW_GPU, "LD_L cache management mode: {}",
static_cast<u64>(instr.ld_l.unknown.Value()));
const auto GetLmem = [&](s32 offset) {
case OpCode::Id::LD_L:
LOG_DEBUG(HW_GPU, "LD_L cache management mode: {}", static_cast<u64>(instr.ld_l.unknown));
[[fallthrough]];
case OpCode::Id::LD_S: {
const auto GetMemory = [&](s32 offset) {
ASSERT(offset % 4 == 0);
const Node immediate_offset = Immediate(static_cast<s32>(instr.smem_imm) + offset);
const Node address = Operation(OperationCode::IAdd, NO_PRECISE, GetRegister(instr.gpr8),
immediate_offset);
return GetLocalMemory(address);
return opcode->get().GetId() == OpCode::Id::LD_S ? GetSharedMemory(address)
: GetLocalMemory(address);
};
switch (instr.ldst_sl.type.Value()) {
@ -135,14 +136,16 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
return 0;
}
}();
for (u32 i = 0; i < count; ++i)
SetTemporary(bb, i, GetLmem(i * 4));
for (u32 i = 0; i < count; ++i)
for (u32 i = 0; i < count; ++i) {
SetTemporary(bb, i, GetMemory(i * 4));
}
for (u32 i = 0; i < count; ++i) {
SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i));
}
break;
}
default:
UNIMPLEMENTED_MSG("LD_L Unhandled type: {}",
UNIMPLEMENTED_MSG("{} Unhandled type: {}", opcode->get().GetName(),
static_cast<u32>(instr.ldst_sl.type.Value()));
}
break;
@ -209,27 +212,34 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
break;
}
case OpCode::Id::ST_L: {
case OpCode::Id::ST_L:
LOG_DEBUG(HW_GPU, "ST_L cache management mode: {}",
static_cast<u64>(instr.st_l.cache_management.Value()));
const auto GetLmemAddr = [&](s32 offset) {
[[fallthrough]];
case OpCode::Id::ST_S: {
const auto GetAddress = [&](s32 offset) {
ASSERT(offset % 4 == 0);
const Node immediate = Immediate(static_cast<s32>(instr.smem_imm) + offset);
return Operation(OperationCode::IAdd, NO_PRECISE, GetRegister(instr.gpr8), immediate);
};
const auto set_memory = opcode->get().GetId() == OpCode::Id::ST_L
? &ShaderIR::SetLocalMemory
: &ShaderIR::SetSharedMemory;
switch (instr.ldst_sl.type.Value()) {
case Tegra::Shader::StoreType::Bits128:
SetLocalMemory(bb, GetLmemAddr(12), GetRegister(instr.gpr0.Value() + 3));
SetLocalMemory(bb, GetLmemAddr(8), GetRegister(instr.gpr0.Value() + 2));
(this->*set_memory)(bb, GetAddress(12), GetRegister(instr.gpr0.Value() + 3));
(this->*set_memory)(bb, GetAddress(8), GetRegister(instr.gpr0.Value() + 2));
[[fallthrough]];
case Tegra::Shader::StoreType::Bits64:
SetLocalMemory(bb, GetLmemAddr(4), GetRegister(instr.gpr0.Value() + 1));
(this->*set_memory)(bb, GetAddress(4), GetRegister(instr.gpr0.Value() + 1));
[[fallthrough]];
case Tegra::Shader::StoreType::Bits32:
SetLocalMemory(bb, GetLmemAddr(0), GetRegister(instr.gpr0));
(this->*set_memory)(bb, GetAddress(0), GetRegister(instr.gpr0));
break;
default:
UNIMPLEMENTED_MSG("ST_L Unhandled type: {}",
UNIMPLEMENTED_MSG("{} unhandled type: {}", opcode->get().GetName(),
static_cast<u32>(instr.ldst_sl.type.Value()));
}
break;

@ -206,12 +206,13 @@ class PredicateNode;
class AbufNode;
class CbufNode;
class LmemNode;
class SmemNode;
class GmemNode;
class CommentNode;
using NodeData =
std::variant<OperationNode, ConditionalNode, GprNode, ImmediateNode, InternalFlagNode,
PredicateNode, AbufNode, CbufNode, LmemNode, GmemNode, CommentNode>;
PredicateNode, AbufNode, CbufNode, LmemNode, SmemNode, GmemNode, CommentNode>;
using Node = std::shared_ptr<NodeData>;
using Node4 = std::array<Node, 4>;
using NodeBlock = std::vector<Node>;
@ -583,6 +584,19 @@ private:
Node address;
};
/// Shared memory node
class SmemNode final {
public:
explicit SmemNode(Node address) : address{std::move(address)} {}
const Node& GetAddress() const {
return address;
}
private:
Node address;
};
/// Global memory node
class GmemNode final {
public:

@ -137,6 +137,10 @@ Node ShaderIR::GetLocalMemory(Node address) {
return MakeNode<LmemNode>(std::move(address));
}
Node ShaderIR::GetSharedMemory(Node address) {
return MakeNode<SmemNode>(std::move(address));
}
Node ShaderIR::GetTemporary(u32 id) {
return GetRegister(Register::ZeroIndex + 1 + id);
}
@ -378,6 +382,11 @@ void ShaderIR::SetLocalMemory(NodeBlock& bb, Node address, Node value) {
Operation(OperationCode::Assign, GetLocalMemory(std::move(address)), std::move(value)));
}
void ShaderIR::SetSharedMemory(NodeBlock& bb, Node address, Node value) {
bb.push_back(
Operation(OperationCode::Assign, GetSharedMemory(std::move(address)), std::move(value)));
}
void ShaderIR::SetTemporary(NodeBlock& bb, u32 id, Node value) {
SetRegister(bb, Register::ZeroIndex + 1 + id, std::move(value));
}

@ -208,6 +208,8 @@ private:
Node GetInternalFlag(InternalFlag flag, bool negated = false);
/// Generates a node representing a local memory address
Node GetLocalMemory(Node address);
/// Generates a node representing a shared memory address
Node GetSharedMemory(Node address);
/// Generates a temporary, internally it uses a post-RZ register
Node GetTemporary(u32 id);
@ -217,8 +219,10 @@ private:
void SetPredicate(NodeBlock& bb, u64 dest, Node src);
/// Sets an internal flag. src value must be a bool-evaluated node
void SetInternalFlag(NodeBlock& bb, InternalFlag flag, Node value);
/// Sets a local memory address. address and value must be a number-evaluated node
/// Sets a local memory address with a value.
void SetLocalMemory(NodeBlock& bb, Node address, Node value);
/// Sets a shared memory address with a value.
void SetSharedMemory(NodeBlock& bb, Node address, Node value);
/// Sets a temporary. Internally it uses a post-RZ register
void SetTemporary(NodeBlock& bb, u32 id, Node value);