shader_recompiler,video_core: Cleanup some GCC and Clang errors
Mostly fixing unused *, implicit conversion, braced scalar init, fpermissive, and some others. Some Clang errors likely remain in video_core, and std::ranges is still a pertinent issue in shader_recompiler shader_recompiler: cmake: Force bracket depth to 1024 on Clang Increases the maximum fold expression depth thread_worker: Include condition_variable Don't use list initializers in control flow Co-authored-by: ReinUsesLisp <reinuseslisp@airmail.cc>
This commit is contained in:
parent
5cd3d00167
commit
0bb85f6a75
66 changed files with 308 additions and 313 deletions
|
@ -6,6 +6,7 @@
|
|||
#include "shader_recompiler/frontend/ir/microinstruction.h"
|
||||
#include "shader_recompiler/frontend/ir/modifiers.h"
|
||||
#include "shader_recompiler/frontend/ir/program.h"
|
||||
#include "shader_recompiler/ir_opt/passes.h"
|
||||
#include "shader_recompiler/shader_info.h"
|
||||
|
||||
namespace Shader::Optimization {
|
||||
|
@ -22,8 +23,8 @@ void AddConstantBufferDescriptor(Info& info, u32 index, u32 count) {
|
|||
auto& cbufs{info.constant_buffer_descriptors};
|
||||
cbufs.insert(std::ranges::lower_bound(cbufs, index, {}, &ConstantBufferDescriptor::index),
|
||||
ConstantBufferDescriptor{
|
||||
.index{index},
|
||||
.count{1},
|
||||
.index = index,
|
||||
.count = 1,
|
||||
});
|
||||
}
|
||||
|
||||
|
@ -91,7 +92,7 @@ void SetAttribute(Info& info, IR::Attribute attribute) {
|
|||
}
|
||||
|
||||
void VisitUsages(Info& info, IR::Inst& inst) {
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::CompositeConstructF16x2:
|
||||
case IR::Opcode::CompositeConstructF16x3:
|
||||
case IR::Opcode::CompositeConstructF16x4:
|
||||
|
@ -209,7 +210,7 @@ void VisitUsages(Info& info, IR::Inst& inst) {
|
|||
default:
|
||||
break;
|
||||
}
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::GetCbufU8:
|
||||
case IR::Opcode::GetCbufS8:
|
||||
case IR::Opcode::UndefU8:
|
||||
|
@ -236,7 +237,7 @@ void VisitUsages(Info& info, IR::Inst& inst) {
|
|||
default:
|
||||
break;
|
||||
}
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::GetCbufU16:
|
||||
case IR::Opcode::GetCbufS16:
|
||||
case IR::Opcode::UndefU16:
|
||||
|
@ -271,7 +272,7 @@ void VisitUsages(Info& info, IR::Inst& inst) {
|
|||
default:
|
||||
break;
|
||||
}
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::UndefU64:
|
||||
case IR::Opcode::LoadGlobalU8:
|
||||
case IR::Opcode::LoadGlobalS8:
|
||||
|
@ -314,7 +315,7 @@ void VisitUsages(Info& info, IR::Inst& inst) {
|
|||
default:
|
||||
break;
|
||||
}
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::DemoteToHelperInvocation:
|
||||
info.uses_demote_to_helper_invocation = true;
|
||||
break;
|
||||
|
@ -361,7 +362,7 @@ void VisitUsages(Info& info, IR::Inst& inst) {
|
|||
} else {
|
||||
throw NotImplementedException("Constant buffer with non-immediate index");
|
||||
}
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::GetCbufU8:
|
||||
case IR::Opcode::GetCbufS8:
|
||||
info.used_constant_buffer_types |= IR::Type::U8;
|
||||
|
@ -443,7 +444,7 @@ void VisitUsages(Info& info, IR::Inst& inst) {
|
|||
}
|
||||
|
||||
void VisitFpModifiers(Info& info, IR::Inst& inst) {
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::FPAdd16:
|
||||
case IR::Opcode::FPFma16:
|
||||
case IR::Opcode::FPMul16:
|
||||
|
@ -540,7 +541,6 @@ void GatherInfoFromHeader(Environment& env, Info& info) {
|
|||
info.stores_position |= header.vtg.omap_systemb.position != 0;
|
||||
}
|
||||
}
|
||||
|
||||
} // Anonymous namespace
|
||||
|
||||
void CollectShaderInfoPass(Environment& env, IR::Program& program) {
|
||||
|
|
|
@ -58,7 +58,7 @@ bool FoldCommutative(IR::Inst& inst, ImmFn&& imm_fn) {
|
|||
}
|
||||
if (is_lhs_immediate && !is_rhs_immediate) {
|
||||
IR::Inst* const rhs_inst{rhs.InstRecursive()};
|
||||
if (rhs_inst->Opcode() == inst.Opcode() && rhs_inst->Arg(1).IsImmediate()) {
|
||||
if (rhs_inst->GetOpcode() == inst.GetOpcode() && rhs_inst->Arg(1).IsImmediate()) {
|
||||
const auto combined{imm_fn(Arg<T>(lhs), Arg<T>(rhs_inst->Arg(1)))};
|
||||
inst.SetArg(0, rhs_inst->Arg(0));
|
||||
inst.SetArg(1, IR::Value{combined});
|
||||
|
@ -70,7 +70,7 @@ bool FoldCommutative(IR::Inst& inst, ImmFn&& imm_fn) {
|
|||
}
|
||||
if (!is_lhs_immediate && is_rhs_immediate) {
|
||||
const IR::Inst* const lhs_inst{lhs.InstRecursive()};
|
||||
if (lhs_inst->Opcode() == inst.Opcode() && lhs_inst->Arg(1).IsImmediate()) {
|
||||
if (lhs_inst->GetOpcode() == inst.GetOpcode() && lhs_inst->Arg(1).IsImmediate()) {
|
||||
const auto combined{imm_fn(Arg<T>(rhs), Arg<T>(lhs_inst->Arg(1)))};
|
||||
inst.SetArg(0, lhs_inst->Arg(0));
|
||||
inst.SetArg(1, IR::Value{combined});
|
||||
|
@ -123,7 +123,8 @@ bool FoldXmadMultiply(IR::Block& block, IR::Inst& inst) {
|
|||
return false;
|
||||
}
|
||||
IR::Inst* const lhs_shl{lhs_arg.InstRecursive()};
|
||||
if (lhs_shl->Opcode() != IR::Opcode::ShiftLeftLogical32 || lhs_shl->Arg(1) != IR::Value{16U}) {
|
||||
if (lhs_shl->GetOpcode() != IR::Opcode::ShiftLeftLogical32 ||
|
||||
lhs_shl->Arg(1) != IR::Value{16U}) {
|
||||
return false;
|
||||
}
|
||||
if (lhs_shl->Arg(0).IsImmediate()) {
|
||||
|
@ -131,7 +132,7 @@ bool FoldXmadMultiply(IR::Block& block, IR::Inst& inst) {
|
|||
}
|
||||
IR::Inst* const lhs_mul{lhs_shl->Arg(0).InstRecursive()};
|
||||
IR::Inst* const rhs_mul{rhs_arg.InstRecursive()};
|
||||
if (lhs_mul->Opcode() != IR::Opcode::IMul32 || rhs_mul->Opcode() != IR::Opcode::IMul32) {
|
||||
if (lhs_mul->GetOpcode() != IR::Opcode::IMul32 || rhs_mul->GetOpcode() != IR::Opcode::IMul32) {
|
||||
return false;
|
||||
}
|
||||
if (lhs_mul->Arg(1).Resolve() != rhs_mul->Arg(1).Resolve()) {
|
||||
|
@ -143,10 +144,10 @@ bool FoldXmadMultiply(IR::Block& block, IR::Inst& inst) {
|
|||
}
|
||||
IR::Inst* const lhs_bfe{lhs_mul->Arg(0).InstRecursive()};
|
||||
IR::Inst* const rhs_bfe{rhs_mul->Arg(0).InstRecursive()};
|
||||
if (lhs_bfe->Opcode() != IR::Opcode::BitFieldUExtract) {
|
||||
if (lhs_bfe->GetOpcode() != IR::Opcode::BitFieldUExtract) {
|
||||
return false;
|
||||
}
|
||||
if (rhs_bfe->Opcode() != IR::Opcode::BitFieldUExtract) {
|
||||
if (rhs_bfe->GetOpcode() != IR::Opcode::BitFieldUExtract) {
|
||||
return false;
|
||||
}
|
||||
if (lhs_bfe->Arg(1) != IR::Value{16U} || lhs_bfe->Arg(2) != IR::Value{16U}) {
|
||||
|
@ -194,8 +195,9 @@ void FoldISub32(IR::Inst& inst) {
|
|||
// ISub32 is generally used to subtract two constant buffers, compare and replace this with
|
||||
// zero if they equal.
|
||||
const auto equal_cbuf{[](IR::Inst* a, IR::Inst* b) {
|
||||
return a->Opcode() == IR::Opcode::GetCbufU32 && b->Opcode() == IR::Opcode::GetCbufU32 &&
|
||||
a->Arg(0) == b->Arg(0) && a->Arg(1) == b->Arg(1);
|
||||
return a->GetOpcode() == IR::Opcode::GetCbufU32 &&
|
||||
b->GetOpcode() == IR::Opcode::GetCbufU32 && a->Arg(0) == b->Arg(0) &&
|
||||
a->Arg(1) == b->Arg(1);
|
||||
}};
|
||||
IR::Inst* op_a{inst.Arg(0).InstRecursive()};
|
||||
IR::Inst* op_b{inst.Arg(1).InstRecursive()};
|
||||
|
@ -204,15 +206,15 @@ void FoldISub32(IR::Inst& inst) {
|
|||
return;
|
||||
}
|
||||
// It's also possible a value is being added to a cbuf and then subtracted
|
||||
if (op_b->Opcode() == IR::Opcode::IAdd32) {
|
||||
if (op_b->GetOpcode() == IR::Opcode::IAdd32) {
|
||||
// Canonicalize local variables to simplify the following logic
|
||||
std::swap(op_a, op_b);
|
||||
}
|
||||
if (op_b->Opcode() != IR::Opcode::GetCbufU32) {
|
||||
if (op_b->GetOpcode() != IR::Opcode::GetCbufU32) {
|
||||
return;
|
||||
}
|
||||
IR::Inst* const inst_cbuf{op_b};
|
||||
if (op_a->Opcode() != IR::Opcode::IAdd32) {
|
||||
if (op_a->GetOpcode() != IR::Opcode::IAdd32) {
|
||||
return;
|
||||
}
|
||||
IR::Value add_op_a{op_a->Arg(0)};
|
||||
|
@ -250,7 +252,8 @@ void FoldFPMul32(IR::Inst& inst) {
|
|||
}
|
||||
IR::Inst* const lhs_op{lhs_value.InstRecursive()};
|
||||
IR::Inst* const rhs_op{rhs_value.InstRecursive()};
|
||||
if (lhs_op->Opcode() != IR::Opcode::FPMul32 || rhs_op->Opcode() != IR::Opcode::FPRecip32) {
|
||||
if (lhs_op->GetOpcode() != IR::Opcode::FPMul32 ||
|
||||
rhs_op->GetOpcode() != IR::Opcode::FPRecip32) {
|
||||
return;
|
||||
}
|
||||
const IR::Value recip_source{rhs_op->Arg(0)};
|
||||
|
@ -260,8 +263,8 @@ void FoldFPMul32(IR::Inst& inst) {
|
|||
}
|
||||
IR::Inst* const attr_a{recip_source.InstRecursive()};
|
||||
IR::Inst* const attr_b{lhs_mul_source.InstRecursive()};
|
||||
if (attr_a->Opcode() != IR::Opcode::GetAttribute ||
|
||||
attr_b->Opcode() != IR::Opcode::GetAttribute) {
|
||||
if (attr_a->GetOpcode() != IR::Opcode::GetAttribute ||
|
||||
attr_b->GetOpcode() != IR::Opcode::GetAttribute) {
|
||||
return;
|
||||
}
|
||||
if (attr_a->Arg(0).Attribute() == attr_b->Arg(0).Attribute()) {
|
||||
|
@ -304,7 +307,7 @@ void FoldLogicalNot(IR::Inst& inst) {
|
|||
return;
|
||||
}
|
||||
IR::Inst* const arg{value.InstRecursive()};
|
||||
if (arg->Opcode() == IR::Opcode::LogicalNot) {
|
||||
if (arg->GetOpcode() == IR::Opcode::LogicalNot) {
|
||||
inst.ReplaceUsesWith(arg->Arg(0));
|
||||
}
|
||||
}
|
||||
|
@ -317,12 +320,12 @@ void FoldBitCast(IR::Inst& inst, IR::Opcode reverse) {
|
|||
return;
|
||||
}
|
||||
IR::Inst* const arg_inst{value.InstRecursive()};
|
||||
if (arg_inst->Opcode() == reverse) {
|
||||
if (arg_inst->GetOpcode() == reverse) {
|
||||
inst.ReplaceUsesWith(arg_inst->Arg(0));
|
||||
return;
|
||||
}
|
||||
if constexpr (op == IR::Opcode::BitCastF32U32) {
|
||||
if (arg_inst->Opcode() == IR::Opcode::GetCbufU32) {
|
||||
if (arg_inst->GetOpcode() == IR::Opcode::GetCbufU32) {
|
||||
// Replace the bitcast with a typed constant buffer read
|
||||
inst.ReplaceOpcode(IR::Opcode::GetCbufF32);
|
||||
inst.SetArg(0, arg_inst->Arg(0));
|
||||
|
@ -338,7 +341,7 @@ void FoldInverseFunc(IR::Inst& inst, IR::Opcode reverse) {
|
|||
return;
|
||||
}
|
||||
IR::Inst* const arg_inst{value.InstRecursive()};
|
||||
if (arg_inst->Opcode() == reverse) {
|
||||
if (arg_inst->GetOpcode() == reverse) {
|
||||
inst.ReplaceUsesWith(arg_inst->Arg(0));
|
||||
return;
|
||||
}
|
||||
|
@ -347,7 +350,7 @@ void FoldInverseFunc(IR::Inst& inst, IR::Opcode reverse) {
|
|||
template <typename Func, size_t... I>
|
||||
IR::Value EvalImmediates(const IR::Inst& inst, Func&& func, std::index_sequence<I...>) {
|
||||
using Traits = LambdaTraits<decltype(func)>;
|
||||
return IR::Value{func(Arg<Traits::ArgType<I>>(inst.Arg(I))...)};
|
||||
return IR::Value{func(Arg<typename Traits::template ArgType<I>>(inst.Arg(I))...)};
|
||||
}
|
||||
|
||||
void FoldBranchConditional(IR::Inst& inst) {
|
||||
|
@ -357,7 +360,7 @@ void FoldBranchConditional(IR::Inst& inst) {
|
|||
return;
|
||||
}
|
||||
const IR::Inst* cond_inst{cond.InstRecursive()};
|
||||
if (cond_inst->Opcode() == IR::Opcode::LogicalNot) {
|
||||
if (cond_inst->GetOpcode() == IR::Opcode::LogicalNot) {
|
||||
const IR::Value true_label{inst.Arg(1)};
|
||||
const IR::Value false_label{inst.Arg(2)};
|
||||
// Remove negation on the conditional (take the parameter out of LogicalNot) and swap
|
||||
|
@ -371,10 +374,10 @@ void FoldBranchConditional(IR::Inst& inst) {
|
|||
std::optional<IR::Value> FoldCompositeExtractImpl(IR::Value inst_value, IR::Opcode insert,
|
||||
IR::Opcode construct, u32 first_index) {
|
||||
IR::Inst* const inst{inst_value.InstRecursive()};
|
||||
if (inst->Opcode() == construct) {
|
||||
if (inst->GetOpcode() == construct) {
|
||||
return inst->Arg(first_index);
|
||||
}
|
||||
if (inst->Opcode() != insert) {
|
||||
if (inst->GetOpcode() != insert) {
|
||||
return std::nullopt;
|
||||
}
|
||||
IR::Value value_index{inst->Arg(2)};
|
||||
|
@ -410,7 +413,7 @@ void FoldCompositeExtract(IR::Inst& inst, IR::Opcode construct, IR::Opcode inser
|
|||
}
|
||||
|
||||
void ConstantPropagation(IR::Block& block, IR::Inst& inst) {
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::GetRegister:
|
||||
return FoldGetRegister(inst);
|
||||
case IR::Opcode::GetPred:
|
||||
|
|
|
@ -57,7 +57,7 @@ struct StorageInfo {
|
|||
|
||||
/// Returns true when the instruction is a global memory instruction
|
||||
bool IsGlobalMemory(const IR::Inst& inst) {
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::LoadGlobalS8:
|
||||
case IR::Opcode::LoadGlobalU8:
|
||||
case IR::Opcode::LoadGlobalS16:
|
||||
|
@ -80,7 +80,7 @@ bool IsGlobalMemory(const IR::Inst& inst) {
|
|||
|
||||
/// Returns true when the instruction is a global memory instruction
|
||||
bool IsGlobalMemoryWrite(const IR::Inst& inst) {
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::WriteGlobalS8:
|
||||
case IR::Opcode::WriteGlobalU8:
|
||||
case IR::Opcode::WriteGlobalS16:
|
||||
|
@ -140,7 +140,7 @@ bool MeetsBias(const StorageBufferAddr& storage_buffer, const Bias& bias) noexce
|
|||
void DiscardGlobalMemory(IR::Block& block, IR::Inst& inst) {
|
||||
IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
|
||||
const IR::Value zero{u32{0}};
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::LoadGlobalS8:
|
||||
case IR::Opcode::LoadGlobalU8:
|
||||
case IR::Opcode::LoadGlobalS16:
|
||||
|
@ -164,7 +164,7 @@ void DiscardGlobalMemory(IR::Block& block, IR::Inst& inst) {
|
|||
inst.Invalidate();
|
||||
break;
|
||||
default:
|
||||
throw LogicError("Invalid opcode to discard its global memory operation {}", inst.Opcode());
|
||||
throw LogicError("Invalid opcode to discard its global memory operation {}", inst.GetOpcode());
|
||||
}
|
||||
}
|
||||
|
||||
|
@ -184,7 +184,7 @@ std::optional<LowAddrInfo> TrackLowAddress(IR::Inst* inst) {
|
|||
// This address is expected to either be a PackUint2x32 or a IAdd64
|
||||
IR::Inst* addr_inst{addr.InstRecursive()};
|
||||
s32 imm_offset{0};
|
||||
if (addr_inst->Opcode() == IR::Opcode::IAdd64) {
|
||||
if (addr_inst->GetOpcode() == IR::Opcode::IAdd64) {
|
||||
// If it's an IAdd64, get the immediate offset it is applying and grab the address
|
||||
// instruction. This expects for the instruction to be canonicalized having the address on
|
||||
// the first argument and the immediate offset on the second one.
|
||||
|
@ -200,7 +200,7 @@ std::optional<LowAddrInfo> TrackLowAddress(IR::Inst* inst) {
|
|||
addr_inst = iadd_addr.Inst();
|
||||
}
|
||||
// With IAdd64 handled, now PackUint2x32 is expected without exceptions
|
||||
if (addr_inst->Opcode() != IR::Opcode::PackUint2x32) {
|
||||
if (addr_inst->GetOpcode() != IR::Opcode::PackUint2x32) {
|
||||
return std::nullopt;
|
||||
}
|
||||
// PackUint2x32 is expected to be generated from a vector
|
||||
|
@ -210,20 +210,20 @@ std::optional<LowAddrInfo> TrackLowAddress(IR::Inst* inst) {
|
|||
}
|
||||
// This vector is expected to be a CompositeConstructU32x2
|
||||
IR::Inst* const vector_inst{vector.InstRecursive()};
|
||||
if (vector_inst->Opcode() != IR::Opcode::CompositeConstructU32x2) {
|
||||
if (vector_inst->GetOpcode() != IR::Opcode::CompositeConstructU32x2) {
|
||||
return std::nullopt;
|
||||
}
|
||||
// Grab the first argument from the CompositeConstructU32x2, this is the low address.
|
||||
return LowAddrInfo{
|
||||
.value{IR::U32{vector_inst->Arg(0)}},
|
||||
.imm_offset{imm_offset},
|
||||
.imm_offset = imm_offset,
|
||||
};
|
||||
}
|
||||
|
||||
/// Tries to track the storage buffer address used by a global memory instruction
|
||||
std::optional<StorageBufferAddr> Track(const IR::Value& value, const Bias* bias) {
|
||||
const auto pred{[bias](const IR::Inst* inst) -> std::optional<StorageBufferAddr> {
|
||||
if (inst->Opcode() != IR::Opcode::GetCbufU32) {
|
||||
if (inst->GetOpcode() != IR::Opcode::GetCbufU32) {
|
||||
return std::nullopt;
|
||||
}
|
||||
const IR::Value index{inst->Arg(0)};
|
||||
|
@ -256,9 +256,9 @@ void CollectStorageBuffers(IR::Block& block, IR::Inst& inst, StorageInfo& info)
|
|||
// NVN puts storage buffers in a specific range, we have to bias towards these addresses to
|
||||
// avoid getting false positives
|
||||
static constexpr Bias nvn_bias{
|
||||
.index{0},
|
||||
.offset_begin{0x110},
|
||||
.offset_end{0x610},
|
||||
.index = 0,
|
||||
.offset_begin = 0x110,
|
||||
.offset_end = 0x610,
|
||||
};
|
||||
// Track the low address of the instruction
|
||||
const std::optional<LowAddrInfo> low_addr_info{TrackLowAddress(&inst)};
|
||||
|
@ -286,8 +286,8 @@ void CollectStorageBuffers(IR::Block& block, IR::Inst& inst, StorageInfo& info)
|
|||
info.set.insert(*storage_buffer);
|
||||
info.to_replace.push_back(StorageInst{
|
||||
.storage_buffer{*storage_buffer},
|
||||
.inst{&inst},
|
||||
.block{&block},
|
||||
.inst = &inst,
|
||||
.block = &block,
|
||||
});
|
||||
}
|
||||
|
||||
|
@ -312,7 +312,7 @@ IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer
|
|||
/// Replace a global memory load instruction with its storage buffer equivalent
|
||||
void ReplaceLoad(IR::Block& block, IR::Inst& inst, const IR::U32& storage_index,
|
||||
const IR::U32& offset) {
|
||||
const IR::Opcode new_opcode{GlobalToStorage(inst.Opcode())};
|
||||
const IR::Opcode new_opcode{GlobalToStorage(inst.GetOpcode())};
|
||||
const auto it{IR::Block::InstructionList::s_iterator_to(inst)};
|
||||
const IR::Value value{&*block.PrependNewInst(it, new_opcode, {storage_index, offset})};
|
||||
inst.ReplaceUsesWith(value);
|
||||
|
@ -321,7 +321,7 @@ void ReplaceLoad(IR::Block& block, IR::Inst& inst, const IR::U32& storage_index,
|
|||
/// Replace a global memory write instruction with its storage buffer equivalent
|
||||
void ReplaceWrite(IR::Block& block, IR::Inst& inst, const IR::U32& storage_index,
|
||||
const IR::U32& offset) {
|
||||
const IR::Opcode new_opcode{GlobalToStorage(inst.Opcode())};
|
||||
const IR::Opcode new_opcode{GlobalToStorage(inst.GetOpcode())};
|
||||
const auto it{IR::Block::InstructionList::s_iterator_to(inst)};
|
||||
block.PrependNewInst(it, new_opcode, {storage_index, offset, inst.Arg(1)});
|
||||
inst.Invalidate();
|
||||
|
@ -330,7 +330,7 @@ void ReplaceWrite(IR::Block& block, IR::Inst& inst, const IR::U32& storage_index
|
|||
/// Replace a global memory instruction with its storage buffer equivalent
|
||||
void Replace(IR::Block& block, IR::Inst& inst, const IR::U32& storage_index,
|
||||
const IR::U32& offset) {
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::LoadGlobalS8:
|
||||
case IR::Opcode::LoadGlobalU8:
|
||||
case IR::Opcode::LoadGlobalS16:
|
||||
|
@ -348,7 +348,7 @@ void Replace(IR::Block& block, IR::Inst& inst, const IR::U32& storage_index,
|
|||
case IR::Opcode::WriteGlobal128:
|
||||
return ReplaceWrite(block, inst, storage_index, offset);
|
||||
default:
|
||||
throw InvalidArgument("Invalid global memory opcode {}", inst.Opcode());
|
||||
throw InvalidArgument("Invalid global memory opcode {}", inst.GetOpcode());
|
||||
}
|
||||
}
|
||||
} // Anonymous namespace
|
||||
|
@ -366,9 +366,9 @@ void GlobalMemoryToStorageBufferPass(IR::Program& program) {
|
|||
u32 storage_index{};
|
||||
for (const StorageBufferAddr& storage_buffer : info.set) {
|
||||
program.info.storage_buffers_descriptors.push_back({
|
||||
.cbuf_index{storage_buffer.index},
|
||||
.cbuf_offset{storage_buffer.offset},
|
||||
.count{1},
|
||||
.cbuf_index = storage_buffer.index,
|
||||
.cbuf_offset = storage_buffer.offset,
|
||||
.count = 1,
|
||||
.is_written{info.writes.contains(storage_buffer)},
|
||||
});
|
||||
++storage_index;
|
||||
|
|
|
@ -22,7 +22,8 @@ void IdentityRemovalPass(IR::Program& program) {
|
|||
inst->SetArg(i, arg.Inst()->Arg(0));
|
||||
}
|
||||
}
|
||||
if (inst->Opcode() == IR::Opcode::Identity || inst->Opcode() == IR::Opcode::Void) {
|
||||
if (inst->GetOpcode() == IR::Opcode::Identity ||
|
||||
inst->GetOpcode() == IR::Opcode::Void) {
|
||||
to_invalidate.push_back(&*inst);
|
||||
inst = block->Instructions().erase(inst);
|
||||
} else {
|
||||
|
|
|
@ -123,7 +123,7 @@ IR::Opcode Replace(IR::Opcode op) {
|
|||
void LowerFp16ToFp32(IR::Program& program) {
|
||||
for (IR::Block* const block : program.blocks) {
|
||||
for (IR::Inst& inst : block->Instructions()) {
|
||||
inst.ReplaceOpcode(Replace(inst.Opcode()));
|
||||
inst.ReplaceOpcode(Replace(inst.GetOpcode()));
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
|
@ -116,7 +116,7 @@ IR::Opcode UndefOpcode(IndirectBranchVariable) noexcept {
|
|||
}
|
||||
|
||||
[[nodiscard]] bool IsPhi(const IR::Inst& inst) noexcept {
|
||||
return inst.Opcode() == IR::Opcode::Phi;
|
||||
return inst.GetOpcode() == IR::Opcode::Phi;
|
||||
}
|
||||
|
||||
enum class Status {
|
||||
|
@ -278,7 +278,7 @@ private:
|
|||
};
|
||||
|
||||
void VisitInst(Pass& pass, IR::Block* block, IR::Inst& inst) {
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::SetRegister:
|
||||
if (const IR::Reg reg{inst.Arg(0).Reg()}; reg != IR::Reg::RZ) {
|
||||
pass.WriteVariable(reg, block, inst.Arg(1));
|
||||
|
|
|
@ -30,7 +30,7 @@ struct TextureInst {
|
|||
using TextureInstVector = boost::container::small_vector<TextureInst, 24>;
|
||||
|
||||
IR::Opcode IndexedInstruction(const IR::Inst& inst) {
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::BindlessImageSampleImplicitLod:
|
||||
case IR::Opcode::BoundImageSampleImplicitLod:
|
||||
return IR::Opcode::ImageSampleImplicitLod;
|
||||
|
@ -67,7 +67,7 @@ IR::Opcode IndexedInstruction(const IR::Inst& inst) {
|
|||
}
|
||||
|
||||
bool IsBindless(const IR::Inst& inst) {
|
||||
switch (inst.Opcode()) {
|
||||
switch (inst.GetOpcode()) {
|
||||
case IR::Opcode::BindlessImageSampleImplicitLod:
|
||||
case IR::Opcode::BindlessImageSampleExplicitLod:
|
||||
case IR::Opcode::BindlessImageSampleDrefImplicitLod:
|
||||
|
@ -91,7 +91,7 @@ bool IsBindless(const IR::Inst& inst) {
|
|||
case IR::Opcode::BoundImageGradient:
|
||||
return false;
|
||||
default:
|
||||
throw InvalidArgument("Invalid opcode {}", inst.Opcode());
|
||||
throw InvalidArgument("Invalid opcode {}", inst.GetOpcode());
|
||||
}
|
||||
}
|
||||
|
||||
|
@ -100,7 +100,7 @@ bool IsTextureInstruction(const IR::Inst& inst) {
|
|||
}
|
||||
|
||||
std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst) {
|
||||
if (inst->Opcode() != IR::Opcode::GetCbufU32) {
|
||||
if (inst->GetOpcode() != IR::Opcode::GetCbufU32) {
|
||||
return std::nullopt;
|
||||
}
|
||||
const IR::Value index{inst->Arg(0)};
|
||||
|
@ -134,14 +134,14 @@ TextureInst MakeInst(Environment& env, IR::Block* block, IR::Inst& inst) {
|
|||
addr = *track_addr;
|
||||
} else {
|
||||
addr = ConstBufferAddr{
|
||||
.index{env.TextureBoundBuffer()},
|
||||
.offset{inst.Arg(0).U32()},
|
||||
.index = env.TextureBoundBuffer(),
|
||||
.offset = inst.Arg(0).U32(),
|
||||
};
|
||||
}
|
||||
return TextureInst{
|
||||
.cbuf{addr},
|
||||
.inst{&inst},
|
||||
.block{block},
|
||||
.inst = &inst,
|
||||
.block = block,
|
||||
};
|
||||
}
|
||||
|
||||
|
@ -211,7 +211,7 @@ void TexturePass(Environment& env, IR::Program& program) {
|
|||
|
||||
const auto& cbuf{texture_inst.cbuf};
|
||||
auto flags{inst->Flags<IR::TextureInstInfo>()};
|
||||
switch (inst->Opcode()) {
|
||||
switch (inst->GetOpcode()) {
|
||||
case IR::Opcode::ImageQueryDimensions:
|
||||
flags.type.Assign(env.ReadTextureType(cbuf.index, cbuf.offset));
|
||||
inst->SetFlags(flags);
|
||||
|
@ -235,16 +235,16 @@ void TexturePass(Environment& env, IR::Program& program) {
|
|||
u32 index;
|
||||
if (flags.type == TextureType::Buffer) {
|
||||
index = descriptors.Add(TextureBufferDescriptor{
|
||||
.cbuf_index{cbuf.index},
|
||||
.cbuf_offset{cbuf.offset},
|
||||
.count{1},
|
||||
.cbuf_index = cbuf.index,
|
||||
.cbuf_offset = cbuf.offset,
|
||||
.count = 1,
|
||||
});
|
||||
} else {
|
||||
index = descriptors.Add(TextureDescriptor{
|
||||
.type{flags.type},
|
||||
.cbuf_index{cbuf.index},
|
||||
.cbuf_offset{cbuf.offset},
|
||||
.count{1},
|
||||
.type = flags.type,
|
||||
.cbuf_index = cbuf.index,
|
||||
.cbuf_offset = cbuf.offset,
|
||||
.count = 1,
|
||||
});
|
||||
}
|
||||
inst->SetArg(0, IR::Value{index});
|
||||
|
|
|
@ -14,14 +14,14 @@ namespace Shader::Optimization {
|
|||
static void ValidateTypes(const IR::Program& program) {
|
||||
for (const auto& block : program.blocks) {
|
||||
for (const IR::Inst& inst : *block) {
|
||||
if (inst.Opcode() == IR::Opcode::Phi) {
|
||||
if (inst.GetOpcode() == IR::Opcode::Phi) {
|
||||
// Skip validation on phi nodes
|
||||
continue;
|
||||
}
|
||||
const size_t num_args{inst.NumArgs()};
|
||||
for (size_t i = 0; i < num_args; ++i) {
|
||||
const IR::Type t1{inst.Arg(i).Type()};
|
||||
const IR::Type t2{IR::ArgTypeOf(inst.Opcode(), i)};
|
||||
const IR::Type t2{IR::ArgTypeOf(inst.GetOpcode(), i)};
|
||||
if (!IR::AreTypesCompatible(t1, t2)) {
|
||||
throw LogicError("Invalid types in block:\n{}", IR::DumpBlock(*block));
|
||||
}
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue