mirror of
https://git.eden-emu.dev/eden-emu/eden.git
synced 2026-08-16 05:21:22 +00:00
eb9280dedf
The approach on this PR moves around refactoring/ updating residual work; main issue was the whole logic behind msaa upload/download images with the constant image copy per pass required; which was gated via shaderStorageImageMultisample, which is not available on all the platforms supported including Android, making them to rely into a more heavier approach to convert msaa image copy into a non msaa image copy, not only losing precision in the conversion, but also creating bugs around the resolve of the image itself, if the supported path had to copy 2 times the same image to be actually presentable, when the fallback was enabled 3 copies were performed, not only decreasing the performance but also increasing race conditions due to the certain capabilities processed in this chain (storage_views). In order to resolve them more naturally, the constant copies were removed from the now "common" path along establishing a proper color resolve for Android and actually making them more direct. Yet this opened a new pack of outdated handling on our current backend; since we now pass msaa images more directly (whenever a game request a msaa type of image/ which seems to be not common + each fragment resolves images on how it's actually required, adding missing depth/stencil), how we often load and store them becomes a critical issue, since our current configuration renderpass/framebuffer forces to always load and load clear whenever a new attachment it's passing through the renderpass, making this really heavier on memory bandwidth limited devices as SteamDeck and others that doesn't rely on the inmediate rendering mode (Tilers vs iGPU); to prevent over allocating data, I had to re-structure part of the current renderpass and framebuffer to actually reuse part of the data/ attachments called, with flags as DON'T_CARE, simplifying partially on how we actually call attachments including their barriers and conditions (occlusion, etc), the DON'T_CARE it's specifically useful to not only prevent calling new loads (which is a flush and heavier on tiler devices) and not only reusing the actual data store, which works on the future time, this saves memory bandwidth and gives space to actually avoid the constant pressure on SYSMEM/GMEM shader resolver path. This maintenance also improved the current removal features logic on QCOM drivers, closing gaps between implementations on descriptors (VK_EXT_indexing_descriptor), adjusting the whole WMEL to avoid being so stricter if certain capabilities are missing to actually use this feature, leading to test and play more on how QCOM drivers are actually behaving; meanwhile the main objective is to open path for a robust and bigger implementation in the future, the changes made here also contains small bug fixes on SPIR-V logic, adding missing cases to resolve shaders (FP64 -> FP16), adding more robustness on how narrow features are actually being used and how to wired them whenever they're not available, meanwhile there are still some issues in regards on how Tegra vs Adreno can preserve unorms at FP32 and FP16, leaving better understanding on base for what VK_EXT_shader_float_controls is actually doing most of the times (flushes NaN's and Denorms equally). Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4212
581 lines
23 KiB
C++
581 lines
23 KiB
C++
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
|
|
// SPDX-License-Identifier: GPL-3.0-or-later
|
|
|
|
// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
|
|
// SPDX-License-Identifier: GPL-2.0-or-later
|
|
|
|
#include <bit>
|
|
|
|
#include "shader_recompiler/backend/spirv/emit_spirv.h"
|
|
#include "shader_recompiler/backend/spirv/emit_spirv_instructions.h"
|
|
#include "shader_recompiler/backend/spirv/spirv_emit_context.h"
|
|
|
|
namespace Shader::Backend::SPIRV {
|
|
namespace {
|
|
Id SharedPointer(EmitContext& ctx, Id offset, u32 index_offset = 0) {
|
|
const Id shift_id{ctx.Const(2U)};
|
|
Id index{ctx.OpShiftRightArithmetic(ctx.U32[1], offset, shift_id)};
|
|
if (index_offset > 0) {
|
|
index = ctx.OpIAdd(ctx.U32[1], index, ctx.Const(index_offset));
|
|
}
|
|
return ctx.uses_explicit_workgroup_layout
|
|
? ctx.OpAccessChain(ctx.shared_u32, ctx.shared_memory_u32, ctx.u32_zero_value, index)
|
|
: ctx.OpAccessChain(ctx.shared_u32, ctx.shared_memory_u32, index);
|
|
}
|
|
|
|
Id StorageIndex(EmitContext& ctx, const IR::Value& offset, size_t element_size) {
|
|
if (offset.IsImmediate()) {
|
|
const u32 imm_offset{static_cast<u32>(offset.U32() / element_size)};
|
|
return ctx.Const(imm_offset);
|
|
}
|
|
const u32 shift{static_cast<u32>(std::countr_zero(element_size))};
|
|
const Id index{ctx.Def(offset)};
|
|
if (shift == 0) {
|
|
return index;
|
|
}
|
|
const Id shift_id{ctx.Const(shift)};
|
|
return ctx.OpShiftRightLogical(ctx.U32[1], index, shift_id);
|
|
}
|
|
|
|
Id StoragePointer(EmitContext& ctx, const StorageTypeDefinition& type_def,
|
|
Id StorageDefinitions::*member_ptr, const IR::Value& binding,
|
|
const IR::Value& offset, size_t element_size) {
|
|
if (!binding.IsImmediate()) {
|
|
throw NotImplementedException("Dynamic storage buffer indexing");
|
|
}
|
|
const Id ssbo{ctx.ssbos[binding.U32()].*member_ptr};
|
|
const Id index{StorageIndex(ctx, offset, element_size)};
|
|
return ctx.OpAccessChain(type_def.element, ssbo, ctx.u32_zero_value, index);
|
|
}
|
|
|
|
std::pair<Id, Id> AtomicArgs(EmitContext& ctx) {
|
|
const Id scope{ctx.Const(static_cast<u32>(spv::Scope::Device))};
|
|
const Id semantics{ctx.u32_zero_value};
|
|
return {scope, semantics};
|
|
}
|
|
|
|
Id SharedAtomicU32(EmitContext& ctx, Id offset, Id value,
|
|
Id (Sirit::Module::*atomic_func)(Id, Id, Id, Id, Id)) {
|
|
const Id pointer{SharedPointer(ctx, offset)};
|
|
const auto [scope, semantics]{AtomicArgs(ctx)};
|
|
return (ctx.*atomic_func)(ctx.U32[1], pointer, scope, semantics, value);
|
|
}
|
|
|
|
Id StorageAtomicU32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset, Id value,
|
|
Id (Sirit::Module::*atomic_func)(Id, Id, Id, Id, Id)) {
|
|
const Id pointer{StoragePointer(ctx, ctx.storage_types.U32, &StorageDefinitions::U32, binding,
|
|
offset, sizeof(u32))};
|
|
const auto [scope, semantics]{AtomicArgs(ctx)};
|
|
return (ctx.*atomic_func)(ctx.U32[1], pointer, scope, semantics, value);
|
|
}
|
|
|
|
Id StorageAtomicU64(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset, Id value,
|
|
Id (Sirit::Module::*atomic_func)(Id, Id, Id, Id, Id),
|
|
Id (Sirit::Module::*non_atomic_func)(Id, Id, Id)) {
|
|
if (!ctx.profile.support_descriptor_aliasing) {
|
|
LOG_WARNING(Shader_SPIRV, "Descriptor aliasing not supported, this cannot be atomic.");
|
|
return ctx.ConstantNull(ctx.U64);
|
|
}
|
|
|
|
if (ctx.profile.support_int64_atomics) {
|
|
const Id pointer{StoragePointer(ctx, ctx.storage_types.U64, &StorageDefinitions::U64,
|
|
binding, offset, sizeof(u64))};
|
|
const auto [scope, semantics]{AtomicArgs(ctx)};
|
|
return (ctx.*atomic_func)(ctx.U64, pointer, scope, semantics, value);
|
|
}
|
|
LOG_WARNING(Shader_SPIRV, "Int64 atomics not supported, fallback to non-atomic");
|
|
const Id pointer{StoragePointer(ctx, ctx.storage_types.U32x2, &StorageDefinitions::U32x2,
|
|
binding, offset, sizeof(u32[2]))};
|
|
const Id original_value{ctx.OpBitcast(ctx.U64, ctx.OpLoad(ctx.U32[2], pointer))};
|
|
const Id result{(ctx.*non_atomic_func)(ctx.U64, value, original_value)};
|
|
ctx.OpStore(pointer, ctx.OpBitcast(ctx.U32[2], result));
|
|
return original_value;
|
|
}
|
|
|
|
Id StorageAtomicU32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset, Id value,
|
|
Id (Sirit::Module::*non_atomic_func)(Id, Id, Id)) {
|
|
if (!ctx.profile.support_descriptor_aliasing) {
|
|
LOG_WARNING(Shader_SPIRV, "Descriptor aliasing not supported, this cannot be atomic.");
|
|
return ctx.ConstantNull(ctx.U32[2]);
|
|
}
|
|
|
|
LOG_WARNING(Shader_SPIRV, "Int64 atomics not supported, fallback to non-atomic");
|
|
const Id pointer{StoragePointer(ctx, ctx.storage_types.U32x2, &StorageDefinitions::U32x2,
|
|
binding, offset, sizeof(u32[2]))};
|
|
const Id original_value{ctx.OpLoad(ctx.U32[2], pointer)};
|
|
const Id result{(ctx.*non_atomic_func)(ctx.U32[2], value, original_value)};
|
|
ctx.OpStore(pointer, result);
|
|
return original_value;
|
|
}
|
|
} // Anonymous namespace
|
|
|
|
Id EmitSharedAtomicIAdd32(EmitContext& ctx, Id offset, Id value) {
|
|
return SharedAtomicU32(ctx, offset, value, &Sirit::Module::OpAtomicIAdd);
|
|
}
|
|
|
|
Id EmitSharedAtomicSMin32(EmitContext& ctx, Id offset, Id value) {
|
|
return SharedAtomicU32(ctx, offset, value, &Sirit::Module::OpAtomicSMin);
|
|
}
|
|
|
|
Id EmitSharedAtomicUMin32(EmitContext& ctx, Id offset, Id value) {
|
|
return SharedAtomicU32(ctx, offset, value, &Sirit::Module::OpAtomicUMin);
|
|
}
|
|
|
|
Id EmitSharedAtomicSMax32(EmitContext& ctx, Id offset, Id value) {
|
|
return SharedAtomicU32(ctx, offset, value, &Sirit::Module::OpAtomicSMax);
|
|
}
|
|
|
|
Id EmitSharedAtomicUMax32(EmitContext& ctx, Id offset, Id value) {
|
|
return SharedAtomicU32(ctx, offset, value, &Sirit::Module::OpAtomicUMax);
|
|
}
|
|
|
|
Id EmitSharedAtomicInc32(EmitContext& ctx, Id offset, Id value) {
|
|
const Id shift_id{ctx.Const(2U)};
|
|
const Id index{ctx.OpShiftRightArithmetic(ctx.U32[1], offset, shift_id)};
|
|
return ctx.OpFunctionCall(ctx.U32[1], ctx.increment_cas_shared, index, value);
|
|
}
|
|
|
|
Id EmitSharedAtomicDec32(EmitContext& ctx, Id offset, Id value) {
|
|
const Id shift_id{ctx.Const(2U)};
|
|
const Id index{ctx.OpShiftRightArithmetic(ctx.U32[1], offset, shift_id)};
|
|
return ctx.OpFunctionCall(ctx.U32[1], ctx.decrement_cas_shared, index, value);
|
|
}
|
|
|
|
Id EmitSharedAtomicAnd32(EmitContext& ctx, Id offset, Id value) {
|
|
return SharedAtomicU32(ctx, offset, value, &Sirit::Module::OpAtomicAnd);
|
|
}
|
|
|
|
Id EmitSharedAtomicOr32(EmitContext& ctx, Id offset, Id value) {
|
|
return SharedAtomicU32(ctx, offset, value, &Sirit::Module::OpAtomicOr);
|
|
}
|
|
|
|
Id EmitSharedAtomicXor32(EmitContext& ctx, Id offset, Id value) {
|
|
return SharedAtomicU32(ctx, offset, value, &Sirit::Module::OpAtomicXor);
|
|
}
|
|
|
|
Id EmitSharedAtomicExchange32(EmitContext& ctx, Id offset, Id value) {
|
|
return SharedAtomicU32(ctx, offset, value, &Sirit::Module::OpAtomicExchange);
|
|
}
|
|
|
|
Id EmitSharedAtomicExchange64(EmitContext& ctx, Id offset, Id value) {
|
|
if (ctx.profile.support_shared_int64_atomics && ctx.uses_explicit_workgroup_layout) {
|
|
const Id shift_id{ctx.Const(3U)};
|
|
const Id index{ctx.OpShiftRightArithmetic(ctx.U32[1], offset, shift_id)};
|
|
const Id pointer{
|
|
ctx.OpAccessChain(ctx.shared_u64, ctx.shared_memory_u64, ctx.u32_zero_value, index)};
|
|
const auto [scope, semantics]{AtomicArgs(ctx)};
|
|
return ctx.OpAtomicExchange(ctx.U64, pointer, scope, semantics, value);
|
|
}
|
|
LOG_WARNING(Shader_SPIRV, "Int64 atomics not supported, fallback to non-atomic");
|
|
const Id pointer_1{SharedPointer(ctx, offset, 0)};
|
|
const Id pointer_2{SharedPointer(ctx, offset, 1)};
|
|
const Id value_1{ctx.OpLoad(ctx.U32[1], pointer_1)};
|
|
const Id value_2{ctx.OpLoad(ctx.U32[1], pointer_2)};
|
|
const Id new_vector{ctx.OpBitcast(ctx.U32[2], value)};
|
|
ctx.OpStore(pointer_1, ctx.OpCompositeExtract(ctx.U32[1], new_vector, 0U));
|
|
ctx.OpStore(pointer_2, ctx.OpCompositeExtract(ctx.U32[1], new_vector, 1U));
|
|
return ctx.OpBitcast(ctx.U64, ctx.OpCompositeConstruct(ctx.U32[2], value_1, value_2));
|
|
}
|
|
|
|
Id EmitSharedAtomicExchange32x2(EmitContext& ctx, Id offset, Id value) {
|
|
LOG_WARNING(Shader_SPIRV, "Int64 atomics not supported, fallback to non-atomic");
|
|
const Id pointer_1{SharedPointer(ctx, offset, 0)};
|
|
const Id pointer_2{SharedPointer(ctx, offset, 1)};
|
|
const Id value_1{ctx.OpLoad(ctx.U32[1], pointer_1)};
|
|
const Id value_2{ctx.OpLoad(ctx.U32[1], pointer_2)};
|
|
const Id new_vector{ctx.OpBitcast(ctx.U32[2], value)};
|
|
ctx.OpStore(pointer_1, ctx.OpCompositeExtract(ctx.U32[1], new_vector, 0U));
|
|
ctx.OpStore(pointer_2, ctx.OpCompositeExtract(ctx.U32[1], new_vector, 1U));
|
|
return ctx.OpCompositeConstruct(ctx.U32[2], value_1, value_2);
|
|
}
|
|
|
|
Id EmitStorageAtomicIAdd32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32(ctx, binding, offset, value, &Sirit::Module::OpAtomicIAdd);
|
|
}
|
|
|
|
Id EmitStorageAtomicSMin32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32(ctx, binding, offset, value, &Sirit::Module::OpAtomicSMin);
|
|
}
|
|
|
|
Id EmitStorageAtomicUMin32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32(ctx, binding, offset, value, &Sirit::Module::OpAtomicUMin);
|
|
}
|
|
|
|
Id EmitStorageAtomicSMax32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32(ctx, binding, offset, value, &Sirit::Module::OpAtomicSMax);
|
|
}
|
|
|
|
Id EmitStorageAtomicUMax32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32(ctx, binding, offset, value, &Sirit::Module::OpAtomicUMax);
|
|
}
|
|
|
|
Id EmitStorageAtomicInc32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
const Id ssbo{ctx.ssbos[binding.U32()].U32};
|
|
const Id base_index{StorageIndex(ctx, offset, sizeof(u32))};
|
|
return ctx.OpFunctionCall(ctx.U32[1], ctx.increment_cas_ssbo, base_index, value, ssbo);
|
|
}
|
|
|
|
Id EmitStorageAtomicDec32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
const Id ssbo{ctx.ssbos[binding.U32()].U32};
|
|
const Id base_index{StorageIndex(ctx, offset, sizeof(u32))};
|
|
return ctx.OpFunctionCall(ctx.U32[1], ctx.decrement_cas_ssbo, base_index, value, ssbo);
|
|
}
|
|
|
|
Id EmitStorageAtomicAnd32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32(ctx, binding, offset, value, &Sirit::Module::OpAtomicAnd);
|
|
}
|
|
|
|
Id EmitStorageAtomicOr32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32(ctx, binding, offset, value, &Sirit::Module::OpAtomicOr);
|
|
}
|
|
|
|
Id EmitStorageAtomicXor32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32(ctx, binding, offset, value, &Sirit::Module::OpAtomicXor);
|
|
}
|
|
|
|
Id EmitStorageAtomicExchange32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32(ctx, binding, offset, value, &Sirit::Module::OpAtomicExchange);
|
|
}
|
|
|
|
Id EmitStorageAtomicIAdd64(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU64(ctx, binding, offset, value, &Sirit::Module::OpAtomicIAdd,
|
|
&Sirit::Module::OpIAdd);
|
|
}
|
|
|
|
Id EmitStorageAtomicSMin64(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU64(ctx, binding, offset, value, &Sirit::Module::OpAtomicSMin,
|
|
&Sirit::Module::OpSMin);
|
|
}
|
|
|
|
Id EmitStorageAtomicUMin64(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU64(ctx, binding, offset, value, &Sirit::Module::OpAtomicUMin,
|
|
&Sirit::Module::OpUMin);
|
|
}
|
|
|
|
Id EmitStorageAtomicSMax64(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU64(ctx, binding, offset, value, &Sirit::Module::OpAtomicSMax,
|
|
&Sirit::Module::OpSMax);
|
|
}
|
|
|
|
Id EmitStorageAtomicUMax64(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU64(ctx, binding, offset, value, &Sirit::Module::OpAtomicUMax,
|
|
&Sirit::Module::OpUMax);
|
|
}
|
|
|
|
Id EmitStorageAtomicAnd64(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU64(ctx, binding, offset, value, &Sirit::Module::OpAtomicAnd,
|
|
&Sirit::Module::OpBitwiseAnd);
|
|
}
|
|
|
|
Id EmitStorageAtomicOr64(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU64(ctx, binding, offset, value, &Sirit::Module::OpAtomicOr,
|
|
&Sirit::Module::OpBitwiseOr);
|
|
}
|
|
|
|
Id EmitStorageAtomicXor64(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU64(ctx, binding, offset, value, &Sirit::Module::OpAtomicXor,
|
|
&Sirit::Module::OpBitwiseXor);
|
|
}
|
|
|
|
Id EmitStorageAtomicExchange64(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
if (ctx.profile.support_int64_atomics) {
|
|
const Id pointer{StoragePointer(ctx, ctx.storage_types.U64, &StorageDefinitions::U64,
|
|
binding, offset, sizeof(u64))};
|
|
const auto [scope, semantics]{AtomicArgs(ctx)};
|
|
return ctx.OpAtomicExchange(ctx.U64, pointer, scope, semantics, value);
|
|
}
|
|
LOG_WARNING(Shader_SPIRV, "Int64 atomics not supported, fallback to non-atomic");
|
|
const Id pointer{StoragePointer(ctx, ctx.storage_types.U32x2, &StorageDefinitions::U32x2,
|
|
binding, offset, sizeof(u32[2]))};
|
|
const Id original{ctx.OpBitcast(ctx.U64, ctx.OpLoad(ctx.U32[2], pointer))};
|
|
ctx.OpStore(pointer, value);
|
|
return original;
|
|
}
|
|
|
|
Id EmitStorageAtomicIAdd32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32x2(ctx, binding, offset, value, &Sirit::Module::OpIAdd);
|
|
}
|
|
|
|
Id EmitStorageAtomicSMin32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32x2(ctx, binding, offset, value, &Sirit::Module::OpSMin);
|
|
}
|
|
|
|
Id EmitStorageAtomicUMin32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32x2(ctx, binding, offset, value, &Sirit::Module::OpUMin);
|
|
}
|
|
|
|
Id EmitStorageAtomicSMax32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32x2(ctx, binding, offset, value, &Sirit::Module::OpSMax);
|
|
}
|
|
|
|
Id EmitStorageAtomicUMax32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32x2(ctx, binding, offset, value, &Sirit::Module::OpUMax);
|
|
}
|
|
|
|
Id EmitStorageAtomicAnd32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32x2(ctx, binding, offset, value, &Sirit::Module::OpBitwiseAnd);
|
|
}
|
|
|
|
Id EmitStorageAtomicOr32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32x2(ctx, binding, offset, value, &Sirit::Module::OpBitwiseOr);
|
|
}
|
|
|
|
Id EmitStorageAtomicXor32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
return StorageAtomicU32x2(ctx, binding, offset, value, &Sirit::Module::OpBitwiseXor);
|
|
}
|
|
|
|
Id EmitStorageAtomicExchange32x2(EmitContext& ctx, const IR::Value& binding,
|
|
const IR::Value& offset, Id value) {
|
|
LOG_WARNING(Shader_SPIRV, "Int64 atomics not supported, fallback to non-atomic");
|
|
const Id pointer{StoragePointer(ctx, ctx.storage_types.U32x2, &StorageDefinitions::U32x2,
|
|
binding, offset, sizeof(u32[2]))};
|
|
const Id original{ctx.OpLoad(ctx.U32[2], pointer)};
|
|
ctx.OpStore(pointer, value);
|
|
return original;
|
|
}
|
|
|
|
Id EmitStorageAtomicAddF32(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
const Id ssbo{ctx.ssbos[binding.U32()].U32};
|
|
const Id base_index{StorageIndex(ctx, offset, sizeof(u32))};
|
|
return ctx.OpFunctionCall(ctx.F32[1], ctx.f32_add_cas, base_index, value, ssbo);
|
|
}
|
|
|
|
Id EmitStorageAtomicAddF16x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
const Id ssbo{ctx.ssbos[binding.U32()].U32};
|
|
const Id base_index{StorageIndex(ctx, offset, sizeof(u32))};
|
|
const Id result{ctx.OpFunctionCall(ctx.F16[2], ctx.f16x2_add_cas, base_index, value, ssbo)};
|
|
return ctx.OpBitcast(ctx.U32[1], result);
|
|
}
|
|
|
|
Id EmitStorageAtomicAddF32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
const Id ssbo{ctx.ssbos[binding.U32()].U32};
|
|
const Id base_index{StorageIndex(ctx, offset, sizeof(u32))};
|
|
const Id result{ctx.OpFunctionCall(ctx.F32[2], ctx.f32x2_add_cas, base_index, value, ssbo)};
|
|
return ctx.OpPackHalf2x16(ctx.U32[1], result);
|
|
}
|
|
|
|
Id EmitStorageAtomicMinF16x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
const Id ssbo{ctx.ssbos[binding.U32()].U32};
|
|
const Id base_index{StorageIndex(ctx, offset, sizeof(u32))};
|
|
const Id result{ctx.OpFunctionCall(ctx.F16[2], ctx.f16x2_min_cas, base_index, value, ssbo)};
|
|
return ctx.OpBitcast(ctx.U32[1], result);
|
|
}
|
|
|
|
Id EmitStorageAtomicMinF32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
const Id ssbo{ctx.ssbos[binding.U32()].U32};
|
|
const Id base_index{StorageIndex(ctx, offset, sizeof(u32))};
|
|
const Id result{ctx.OpFunctionCall(ctx.F32[2], ctx.f32x2_min_cas, base_index, value, ssbo)};
|
|
return ctx.OpPackHalf2x16(ctx.U32[1], result);
|
|
}
|
|
|
|
Id EmitStorageAtomicMaxF16x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
const Id ssbo{ctx.ssbos[binding.U32()].U32};
|
|
const Id base_index{StorageIndex(ctx, offset, sizeof(u32))};
|
|
const Id result{ctx.OpFunctionCall(ctx.F16[2], ctx.f16x2_max_cas, base_index, value, ssbo)};
|
|
return ctx.OpBitcast(ctx.U32[1], result);
|
|
}
|
|
|
|
Id EmitStorageAtomicMaxF32x2(EmitContext& ctx, const IR::Value& binding, const IR::Value& offset,
|
|
Id value) {
|
|
const Id ssbo{ctx.ssbos[binding.U32()].U32};
|
|
const Id base_index{StorageIndex(ctx, offset, sizeof(u32))};
|
|
const Id result{ctx.OpFunctionCall(ctx.F32[2], ctx.f32x2_max_cas, base_index, value, ssbo)};
|
|
return ctx.OpPackHalf2x16(ctx.U32[1], result);
|
|
}
|
|
|
|
Id EmitGlobalAtomicIAdd32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicSMin32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicUMin32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicSMax32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicUMax32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicInc32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicDec32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicAnd32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicOr32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicXor32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicExchange32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicIAdd64(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicSMin64(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicUMin64(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicSMax64(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicUMax64(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicInc64(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicDec64(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicAnd64(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicOr64(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicXor64(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicExchange64(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicIAdd32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicSMin32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicUMin32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicSMax32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicUMax32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicInc32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicDec32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicAnd32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicOr32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicXor32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicExchange32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicAddF32(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicAddF16x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicAddF32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicMinF16x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicMinF32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicMaxF16x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
Id EmitGlobalAtomicMaxF32x2(EmitContext&) {
|
|
throw NotImplementedException("SPIR-V Instruction");
|
|
}
|
|
|
|
} // namespace Shader::Backend::SPIRV
|