shader: Add NVN storage buffer fallbacks
When we can't track the SSBO origin of a global memory instruction, leave it as a global memory operation and assume these pointers are in the NVN storage buffer slots, then apply a linear search in the shader's runtime.
This commit is contained in:
ReinUsesLisp
ameerj
@@ -411,6 +411,7 @@ EmitContext::EmitContext(const Profile& profile_, IR::Program& program, u32& bin
|
||||
DefineTextures(program.info, binding);
|
||||
DefineImages(program.info, binding);
|
||||
DefineAttributeMemAccess(program.info);
|
||||
DefineGlobalMemoryFunctions(program.info);
|
||||
DefineLabels(program);
|
||||
}
|
||||
|
||||
@@ -762,6 +763,82 @@ void EmitContext::DefineAttributeMemAccess(const Info& info) {
|
||||
}
|
||||
}
|
||||
|
||||
void EmitContext::DefineGlobalMemoryFunctions(const Info& info) {
|
||||
if (!info.uses_global_memory) {
|
||||
return;
|
||||
}
|
||||
using DefPtr = Id StorageDefinitions::*;
|
||||
const Id zero{u32_zero_value};
|
||||
const auto define_body{[&](DefPtr ssbo_member, Id addr, Id element_pointer, u32 shift,
|
||||
auto&& callback) {
|
||||
AddLabel();
|
||||
const size_t num_buffers{info.storage_buffers_descriptors.size()};
|
||||
for (size_t index = 0; index < num_buffers; ++index) {
|
||||
const auto& ssbo{info.storage_buffers_descriptors[index]};
|
||||
const Id ssbo_addr_cbuf_offset{Const(ssbo.cbuf_offset / 8)};
|
||||
const Id ssbo_size_cbuf_offset{Const(ssbo.cbuf_offset / 4 + 2)};
|
||||
const Id ssbo_addr_pointer{OpAccessChain(
|
||||
uniform_types.U32x2, cbufs[ssbo.cbuf_index].U32x2, zero, ssbo_addr_cbuf_offset)};
|
||||
const Id ssbo_size_pointer{OpAccessChain(uniform_types.U32, cbufs[ssbo.cbuf_index].U32,
|
||||
zero, ssbo_size_cbuf_offset)};
|
||||
|
||||
const Id ssbo_addr{OpBitcast(U64, OpLoad(U32[2], ssbo_addr_pointer))};
|
||||
const Id ssbo_size{OpUConvert(U64, OpLoad(U32[1], ssbo_size_pointer))};
|
||||
const Id ssbo_end{OpIAdd(U64, ssbo_addr, ssbo_size)};
|
||||
const Id cond{OpLogicalAnd(U1, OpUGreaterThanEqual(U1, addr, ssbo_addr),
|
||||
OpULessThan(U1, addr, ssbo_end))};
|
||||
const Id then_label{OpLabel()};
|
||||
const Id else_label{OpLabel()};
|
||||
OpSelectionMerge(else_label, spv::SelectionControlMask::MaskNone);
|
||||
OpBranchConditional(cond, then_label, else_label);
|
||||
AddLabel(then_label);
|
||||
const Id ssbo_id{ssbos[index].*ssbo_member};
|
||||
const Id ssbo_offset{OpUConvert(U32[1], OpISub(U64, addr, ssbo_addr))};
|
||||
const Id ssbo_index{OpShiftRightLogical(U32[1], ssbo_offset, Const(shift))};
|
||||
const Id ssbo_pointer{OpAccessChain(element_pointer, ssbo_id, zero, ssbo_index)};
|
||||
callback(ssbo_pointer);
|
||||
AddLabel(else_label);
|
||||
}
|
||||
}};
|
||||
const auto define_load{[&](DefPtr ssbo_member, Id element_pointer, Id type, u32 shift) {
|
||||
const Id function_type{TypeFunction(type, U64)};
|
||||
const Id func_id{OpFunction(type, spv::FunctionControlMask::MaskNone, function_type)};
|
||||
const Id addr{OpFunctionParameter(U64)};
|
||||
define_body(ssbo_member, addr, element_pointer, shift,
|
||||
[&](Id ssbo_pointer) { OpReturnValue(OpLoad(type, ssbo_pointer)); });
|
||||
OpReturnValue(ConstantNull(type));
|
||||
OpFunctionEnd();
|
||||
return func_id;
|
||||
}};
|
||||
const auto define_write{[&](DefPtr ssbo_member, Id element_pointer, Id type, u32 shift) {
|
||||
const Id function_type{TypeFunction(void_id, U64, type)};
|
||||
const Id func_id{OpFunction(void_id, spv::FunctionControlMask::MaskNone, function_type)};
|
||||
const Id addr{OpFunctionParameter(U64)};
|
||||
const Id data{OpFunctionParameter(type)};
|
||||
define_body(ssbo_member, addr, element_pointer, shift, [&](Id ssbo_pointer) {
|
||||
OpStore(ssbo_pointer, data);
|
||||
OpReturn();
|
||||
});
|
||||
OpReturn();
|
||||
OpFunctionEnd();
|
||||
return func_id;
|
||||
}};
|
||||
const auto define{
|
||||
[&](DefPtr ssbo_member, const StorageTypeDefinition& type_def, Id type, size_t size) {
|
||||
const Id element_type{type_def.element};
|
||||
const u32 shift{static_cast<u32>(std::countr_zero(size))};
|
||||
const Id load_func{define_load(ssbo_member, element_type, type, shift)};
|
||||
const Id write_func{define_write(ssbo_member, element_type, type, shift)};
|
||||
return std::make_pair(load_func, write_func);
|
||||
}};
|
||||
std::tie(load_global_func_u32, write_global_func_u32) =
|
||||
define(&StorageDefinitions::U32, storage_types.U32, U32[1], sizeof(u32));
|
||||
std::tie(load_global_func_u32x2, write_global_func_u32x2) =
|
||||
define(&StorageDefinitions::U32x2, storage_types.U32x2, U32[2], sizeof(u32[2]));
|
||||
std::tie(load_global_func_u32x4, write_global_func_u32x4) =
|
||||
define(&StorageDefinitions::U32x4, storage_types.U32x4, U32[4], sizeof(u32[4]));
|
||||
}
|
||||
|
||||
void EmitContext::DefineConstantBuffers(const Info& info, u32& binding) {
|
||||
if (info.constant_buffer_descriptors.empty()) {
|
||||
return;
|
||||
|
||||
@@ -224,6 +224,13 @@ public:
|
||||
Id f32x2_min_cas{};
|
||||
Id f32x2_max_cas{};
|
||||
|
||||
Id load_global_func_u32{};
|
||||
Id load_global_func_u32x2{};
|
||||
Id load_global_func_u32x4{};
|
||||
Id write_global_func_u32{};
|
||||
Id write_global_func_u32x2{};
|
||||
Id write_global_func_u32x4{};
|
||||
|
||||
Id input_position{};
|
||||
std::array<Id, 32> input_generics{};
|
||||
|
||||
@@ -255,6 +262,7 @@ private:
|
||||
void DefineTextures(const Info& info, u32& binding);
|
||||
void DefineImages(const Info& info, u32& binding);
|
||||
void DefineAttributeMemAccess(const Info& info);
|
||||
void DefineGlobalMemoryFunctions(const Info& info);
|
||||
void DefineLabels(IR::Program& program);
|
||||
|
||||
void DefineInputs(const Info& info);
|
||||
|
||||
@@ -84,16 +84,16 @@ void EmitLoadGlobalU8(EmitContext& ctx);
|
||||
void EmitLoadGlobalS8(EmitContext& ctx);
|
||||
void EmitLoadGlobalU16(EmitContext& ctx);
|
||||
void EmitLoadGlobalS16(EmitContext& ctx);
|
||||
void EmitLoadGlobal32(EmitContext& ctx);
|
||||
void EmitLoadGlobal64(EmitContext& ctx);
|
||||
void EmitLoadGlobal128(EmitContext& ctx);
|
||||
Id EmitLoadGlobal32(EmitContext& ctx, Id address);
|
||||
Id EmitLoadGlobal64(EmitContext& ctx, Id address);
|
||||
Id EmitLoadGlobal128(EmitContext& ctx, Id address);
|
||||
void EmitWriteGlobalU8(EmitContext& ctx);
|
||||
void EmitWriteGlobalS8(EmitContext& ctx);
|
||||
void EmitWriteGlobalU16(EmitContext& ctx);
|
||||
void EmitWriteGlobalS16(EmitContext& ctx);
|
||||
void EmitWriteGlobal32(EmitContext& ctx);
|
||||
void EmitWriteGlobal64(EmitContext& ctx);
|
||||
void EmitWriteGlobal128(EmitContext& ctx);
|
||||
void EmitWriteGlobal32(EmitContext& ctx, Id address, Id value);
|
||||
void EmitWriteGlobal64(EmitContext& ctx, Id address, Id value);
|
||||
void EmitWriteGlobal128(EmitContext& ctx, Id address, Id value);
|
||||
Id EmitLoadStorageU8(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset);
|
||||
Id EmitLoadStorageS8(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset);
|
||||
Id EmitLoadStorageU16(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset);
|
||||
@@ -277,9 +277,9 @@ Id EmitFPIsNan16(EmitContext& ctx, Id value);
|
||||
Id EmitFPIsNan32(EmitContext& ctx, Id value);
|
||||
Id EmitFPIsNan64(EmitContext& ctx, Id value);
|
||||
Id EmitIAdd32(EmitContext& ctx, IR::Inst* inst, Id a, Id b);
|
||||
void EmitIAdd64(EmitContext& ctx);
|
||||
Id EmitIAdd64(EmitContext& ctx, Id a, Id b);
|
||||
Id EmitISub32(EmitContext& ctx, Id a, Id b);
|
||||
void EmitISub64(EmitContext& ctx);
|
||||
Id EmitISub64(EmitContext& ctx, Id a, Id b);
|
||||
Id EmitIMul32(EmitContext& ctx, Id a, Id b);
|
||||
Id EmitINeg32(EmitContext& ctx, Id value);
|
||||
Id EmitINeg64(EmitContext& ctx, Id value);
|
||||
|
||||
@@ -55,16 +55,16 @@ Id EmitIAdd32(EmitContext& ctx, IR::Inst* inst, Id a, Id b) {
|
||||
return result;
|
||||
}
|
||||
|
||||
void EmitIAdd64(EmitContext&) {
|
||||
throw NotImplementedException("SPIR-V Instruction");
|
||||
Id EmitIAdd64(EmitContext& ctx, Id a, Id b) {
|
||||
return ctx.OpIAdd(ctx.U64, a, b);
|
||||
}
|
||||
|
||||
Id EmitISub32(EmitContext& ctx, Id a, Id b) {
|
||||
return ctx.OpISub(ctx.U32[1], a, b);
|
||||
}
|
||||
|
||||
void EmitISub64(EmitContext&) {
|
||||
throw NotImplementedException("SPIR-V Instruction");
|
||||
Id EmitISub64(EmitContext& ctx, Id a, Id b) {
|
||||
return ctx.OpISub(ctx.U64, a, b);
|
||||
}
|
||||
|
||||
Id EmitIMul32(EmitContext& ctx, Id a, Id b) {
|
||||
|
||||
@@ -64,16 +64,16 @@ void EmitLoadGlobalS16(EmitContext&) {
|
||||
throw NotImplementedException("SPIR-V Instruction");
|
||||
}
|
||||
|
||||
void EmitLoadGlobal32(EmitContext&) {
|
||||
throw NotImplementedException("SPIR-V Instruction");
|
||||
Id EmitLoadGlobal32(EmitContext& ctx, Id address) {
|
||||
return ctx.OpFunctionCall(ctx.U32[1], ctx.load_global_func_u32, address);
|
||||
}
|
||||
|
||||
void EmitLoadGlobal64(EmitContext&) {
|
||||
throw NotImplementedException("SPIR-V Instruction");
|
||||
Id EmitLoadGlobal64(EmitContext& ctx, Id address) {
|
||||
return ctx.OpFunctionCall(ctx.U32[2], ctx.load_global_func_u32x2, address);
|
||||
}
|
||||
|
||||
void EmitLoadGlobal128(EmitContext&) {
|
||||
throw NotImplementedException("SPIR-V Instruction");
|
||||
Id EmitLoadGlobal128(EmitContext& ctx, Id address) {
|
||||
return ctx.OpFunctionCall(ctx.U32[4], ctx.load_global_func_u32x4, address);
|
||||
}
|
||||
|
||||
void EmitWriteGlobalU8(EmitContext&) {
|
||||
@@ -92,16 +92,16 @@ void EmitWriteGlobalS16(EmitContext&) {
|
||||
throw NotImplementedException("SPIR-V Instruction");
|
||||
}
|
||||
|
||||
void EmitWriteGlobal32(EmitContext&) {
|
||||
throw NotImplementedException("SPIR-V Instruction");
|
||||
void EmitWriteGlobal32(EmitContext& ctx, Id address, Id value) {
|
||||
ctx.OpFunctionCall(ctx.void_id, ctx.write_global_func_u32, address, value);
|
||||
}
|
||||
|
||||
void EmitWriteGlobal64(EmitContext&) {
|
||||
throw NotImplementedException("SPIR-V Instruction");
|
||||
void EmitWriteGlobal64(EmitContext& ctx, Id address, Id value) {
|
||||
ctx.OpFunctionCall(ctx.void_id, ctx.write_global_func_u32x2, address, value);
|
||||
}
|
||||
|
||||
void EmitWriteGlobal128(EmitContext&) {
|
||||
throw NotImplementedException("SPIR-V Instruction");
|
||||
void EmitWriteGlobal128(EmitContext& ctx, Id address, Id value) {
|
||||
ctx.OpFunctionCall(ctx.void_id, ctx.write_global_func_u32x4, address, value);
|
||||
}
|
||||
|
||||
Id EmitLoadStorageU8(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset) {
|
||||
|
||||
Reference in New Issue
Block a user