Work

src/core/address_space.h

@@ -28,7 +28,7 @@ constexpr VAddr USER_MAX = 0xFBFFFFFFFFULL;
 // User area size is normally larger than this. However games are unlikely to map to high
 // regions of that area, so by default we allocate a smaller virtual address space (about 1/4th).
 // to save space on page tables.
-static constexpr size_t UserSize = 1ULL << 38;
+static constexpr size_t UserSize = 1ULL << 39;
 static constexpr size_t SystemSize = USER_MIN - SYSTEM_MANAGED_MIN;
 
 /**

src/core/file_sys/fs.cpp

@@ -13,6 +13,7 @@ void MntPoints::Mount(const std::filesystem::path& host_folder, const std::strin
 
     MntPair pair;
     pair.host_path = host_folder.string();
+    std::replace(pair.host_path.begin(), pair.host_path.end(), '\\', '/');
     pair.guest_path = guest_folder;
 
     m_mnt_pairs.push_back(pair);
@@ -40,17 +41,36 @@ std::string MntPoints::GetHostDirectory(const std::string& guest_directory) {
     return "";
 }
 
+std::string ToLower(std::string str) {
+    std::transform(str.begin(), str.end(), str.begin(),
+                   [](unsigned char c) { return static_cast<char>(std::tolower(c)); });
+    return str;
+}
+
 std::string MntPoints::GetHostFile(const std::string& guest_file) {
     std::scoped_lock lock{m_mutex};
 
     for (auto& pair : m_mnt_pairs) {
         // horrible code but it works :D
         int find = guest_file.find(pair.guest_path);
-        if (find == 0) {
+        if (find != 0) {
-            std::string npath = guest_file.substr(pair.guest_path.size(), guest_file.size() - 1);
+            continue;
-            std::replace(pair.host_path.begin(), pair.host_path.end(), '\\', '/');
-            return pair.host_path + npath;
         }
+        std::string npath = guest_file.substr(pair.guest_path.size(), guest_file.size() - 1);
+        const auto host_path = pair.host_path + npath;
+#ifndef _WIN64
+        const std::filesystem::path path{host_path};
+        if (!std::filesystem::exists(path)) {
+            const auto filename = ToLower(path.filename());
+            for (const auto& file : std::filesystem::directory_iterator(path.parent_path())) {
+                const auto exist_filename = ToLower(file.path().filename());
+                if (filename == exist_filename) {
+                    return file.path();
+                }
+            }
+        }
+#endif
+        return host_path;
     }
     return "";
 }

src/core/libraries/gnmdriver/gnmdriver.cpp

@@ -344,8 +344,8 @@ s32 PS4_SYSV_ABI sceGnmAddEqEvent(SceKernelEqueue eq, u64 id, void* udata) {
 }
 
 int PS4_SYSV_ABI sceGnmAreSubmitsAllowed() {
-    LOG_TRACE(Lib_GnmDriver, "called");
+    LOG_TRACE(Lib_GnmDriver, "called submission_lock = {}", submission_lock);
-    return submission_lock == 0;
+    return liverpool->IsGpuIdle();
 }
 
 int PS4_SYSV_ABI sceGnmBeginWorkload() {
@@ -803,9 +803,9 @@ int PS4_SYSV_ABI sceGnmDrawOpaqueAuto() {
     return ORBIS_OK;
 }
 
-int PS4_SYSV_ABI sceGnmDriverCaptureInProgress() {
+bool PS4_SYSV_ABI sceGnmDriverCaptureInProgress() {
-    LOG_ERROR(Lib_GnmDriver, "(STUBBED) called");
+    LOG_TRACE(Lib_GnmDriver, "called");
-    return ORBIS_OK;
+    return false;
 }
 
 int PS4_SYSV_ABI sceGnmDriverInternalRetrieveGnmInterface() {
@@ -1930,8 +1930,10 @@ s32 PS4_SYSV_ABI sceGnmSubmitCommandBuffers(u32 count, const u32* dcb_gpu_addrs[
         }
     }
 
+    LOG_INFO(Lib_GnmDriver, "called submission_lock = {}", submission_lock);
     if (submission_lock != 0) {
         liverpool->WaitGpuIdle();
+        LOG_INFO(Lib_GnmDriver, "Done waiting for GPU");
 
         // Suspend logic goes here
 

src/core/libraries/gnmdriver/gnmdriver.h

@@ -63,7 +63,7 @@ u32 PS4_SYSV_ABI sceGnmDrawInitDefaultHardwareState350(u32* cmdbuf, u32 size);
 u32 PS4_SYSV_ABI sceGnmDrawInitToDefaultContextState(u32* cmdbuf, u32 size);
 u32 PS4_SYSV_ABI sceGnmDrawInitToDefaultContextState400(u32* cmdbuf, u32 size);
 int PS4_SYSV_ABI sceGnmDrawOpaqueAuto();
-int PS4_SYSV_ABI sceGnmDriverCaptureInProgress();
+bool PS4_SYSV_ABI sceGnmDriverCaptureInProgress();
 int PS4_SYSV_ABI sceGnmDriverInternalRetrieveGnmInterface();
 int PS4_SYSV_ABI sceGnmDriverInternalRetrieveGnmInterfaceForGpuDebugger();
 int PS4_SYSV_ABI sceGnmDriverInternalRetrieveGnmInterfaceForGpuException();

src/core/libraries/kernel/file_system.cpp

@@ -75,7 +75,7 @@ int PS4_SYSV_ABI sceKernelOpen(const char* path, int flags, u16 mode) {
         file->m_host_name = mnt->GetHostFile(file->m_guest_name);
         if (read) {
             file->f.Open(file->m_host_name, Common::FS::FileAccessMode::Read);
-        } else if (write && create && truncate) {
+        } else if (write && create) {
             file->f.Open(file->m_host_name, Common::FS::FileAccessMode::Write);
         } else if (write && create && append) { // CUSA04729 (appends app0/shaderlist.txt)
             file->f.Open(file->m_host_name, Common::FS::FileAccessMode::Append);
@@ -90,7 +90,7 @@ int PS4_SYSV_ABI sceKernelOpen(const char* path, int flags, u16 mode) {
         }
         if (!file->f.IsOpen()) {
             h->DeleteHandle(handle);
-            return SCE_KERNEL_ERROR_EACCES;
+            return SCE_KERNEL_ERROR_ENOENT;
         }
     }
     file->is_opened = true;

src/core/libraries/kernel/memory_management.cpp

@@ -110,10 +110,13 @@ int PS4_SYSV_ABI sceKernelMapNamedDirectMemory(void** addr, u64 len, int prot, i
         }
     }
 
-    const VAddr in_addr = reinterpret_cast<VAddr>(*addr);
+    VAddr in_addr = reinterpret_cast<VAddr>(*addr);
     const auto mem_prot = static_cast<Core::MemoryProt>(prot);
     const auto map_flags = static_cast<Core::MemoryMapFlags>(flags);
     auto* memory = Core::Memory::Instance();
+    if (False(map_flags & Core::MemoryMapFlags::Fixed) && in_addr == 0) {
+        //in_addr = 0x880000000;
+    }
     return memory->MapMemory(addr, in_addr, len, mem_prot, map_flags, Core::VMAType::Direct, "",
                              false, directMemoryStart, alignment);
 }
@@ -143,10 +146,13 @@ s32 PS4_SYSV_ABI sceKernelMapNamedFlexibleMemory(void** addr_in_out, std::size_t
         return ORBIS_KERNEL_ERROR_EFAULT;
     }
 
-    const VAddr in_addr = reinterpret_cast<VAddr>(*addr_in_out);
+    VAddr in_addr = reinterpret_cast<VAddr>(*addr_in_out);
     const auto mem_prot = static_cast<Core::MemoryProt>(prot);
     const auto map_flags = static_cast<Core::MemoryMapFlags>(flags);
     auto* memory = Core::Memory::Instance();
+    if (False(map_flags & Core::MemoryMapFlags::Fixed) && in_addr == 0) {
+        //in_addr = 0x880000000;
+    }
     const int ret = memory->MapMemory(addr_in_out, in_addr, len, mem_prot, map_flags,
                                       Core::VMAType::Flexible, name);
 
@@ -161,7 +167,6 @@ s32 PS4_SYSV_ABI sceKernelMapFlexibleMemory(void** addr_in_out, std::size_t len,
 }
 
 int PS4_SYSV_ABI sceKernelQueryMemoryProtection(void* addr, void** start, void** end, u32* prot) {
-    LOG_WARNING(Kernel_Vmm, "called");
     auto* memory = Core::Memory::Instance();
     return memory->QueryProtection(std::bit_cast<VAddr>(addr), start, end, prot);
 }

src/core/libraries/kernel/thread_management.cpp

@@ -5,6 +5,7 @@
 #include <thread>
 #include <semaphore.h>
 #include "common/assert.h"
+#include "common/alignment.h"
 #include "common/error.h"
 #include "common/logging/log.h"
 #include "common/singleton.h"
@@ -16,6 +17,8 @@
 #include "core/linker.h"
 #ifdef _WIN64
 #include <windows.h>
+#else
+#include <sys/mman.h>
 #endif
 
 namespace Libraries::Kernel {
@@ -46,7 +49,8 @@ void init_pthreads() {
 }
 
 void pthreadInitSelfMainThread() {
-    g_pthread_self = new PthreadInternal{};
+    auto* pthread_pool = g_pthread_cxt->GetPthreadPool();
+    g_pthread_self = pthread_pool->Create();
     scePthreadAttrInit(&g_pthread_self->attr);
     g_pthread_self->pth = pthread_self();
     g_pthread_self->name = "Main_Thread";
@@ -978,7 +982,14 @@ ScePthread PThreadPool::Create() {
         }
     }
 
+#ifndef _WIN64
     auto* ret = new PthreadInternal{};
+#else
+    static u8* hint_address = reinterpret_cast<u8*>(0x7FFFFC000ULL);
+    auto* ret = reinterpret_cast<PthreadInternal*>(mmap(hint_address, sizeof(PthreadInternal),
+                                                        PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED, -1, 0));
+    hint_address += Common::AlignUp(sizeof(PthreadInternal), 4_KB);
+#endif
 
     ret->is_free = false;
     ret->is_detached = false;

src/core/libraries/libs.h

@@ -42,7 +42,7 @@ struct wrapper_impl<name, PS4_SYSV_ABI R (*)(Args...), f> {
 template <StringLiteral name, class F, F f>
 constexpr auto wrapper = wrapper_impl<name, F, f>::wrap;
 
-// #define W(foo) wrapper<#foo, decltype(&foo), foo>
+//#define W(foo) wrapper<#foo, decltype(&foo), foo>
 #define W(foo) foo
 
 #define LIB_FUNCTION(nid, lib, libversion, mod, moduleVersionMajor, moduleVersionMinor, function)  \

src/core/libraries/network/net.cpp

@@ -10,6 +10,7 @@
 #include <arpa/inet.h>
 #endif
 
+#include <thread>
 #include <common/assert.h>
 #include "common/logging/log.h"
 #include "core/libraries/error_codes.h"
@@ -59,6 +60,7 @@ int PS4_SYSV_ABI sce_net_in6addr_nodelocal_allnodes() {
 }
 
 OrbisNetId PS4_SYSV_ABI sceNetAccept(OrbisNetId s, OrbisNetSockaddr* addr, u32* paddrlen) {
+    std::this_thread::sleep_for(std::chrono::seconds(60));
     LOG_ERROR(Lib_Net, "(STUBBED) called");
     return ORBIS_OK;
 }
@@ -559,7 +561,7 @@ int PS4_SYSV_ABI sceNetEpollDestroy() {
 }
 
 int PS4_SYSV_ABI sceNetEpollWait() {
-    LOG_ERROR(Lib_Net, "(STUBBED) called");
+    //LOG_ERROR(Lib_Net, "(STUBBED) called");
     return ORBIS_OK;
 }
 

src/core/libraries/save_data/savedata.cpp

@@ -341,6 +341,7 @@ s32 saveDataMount(u32 user_id, std::string dir_name, u32 mount_mode,
     switch (mount_mode) {
     case ORBIS_SAVE_DATA_MOUNT_MODE_RDONLY:
     case ORBIS_SAVE_DATA_MOUNT_MODE_RDWR:
+    case ORBIS_SAVE_DATA_MOUNT_MODE_RDWR | ORBIS_SAVE_DATA_MOUNT_MODE_DESTRUCT_OFF:
     case ORBIS_SAVE_DATA_MOUNT_MODE_RDONLY | ORBIS_SAVE_DATA_MOUNT_MODE_DESTRUCT_OFF: {
         if (!std::filesystem::exists(mount_dir)) {
             return ORBIS_SAVE_DATA_ERROR_NOT_FOUND;
@@ -349,11 +350,13 @@ s32 saveDataMount(u32 user_id, std::string dir_name, u32 mount_mode,
         mnt->Mount(mount_dir, g_mount_point);
 
         mount_result->mount_status = 0;
-        strncpy(mount_result->mount_point.data, g_mount_point.c_str(), 16);
+        std::strncpy(mount_result->mount_point.data, g_mount_point.c_str(), 16);
-    } break;
+        break;
+    }
     case ORBIS_SAVE_DATA_MOUNT_MODE_CREATE:
     case ORBIS_SAVE_DATA_MOUNT_MODE_CREATE | ORBIS_SAVE_DATA_MOUNT_MODE_RDONLY:
     case ORBIS_SAVE_DATA_MOUNT_MODE_CREATE | ORBIS_SAVE_DATA_MOUNT_MODE_RDWR:
+    case ORBIS_SAVE_DATA_MOUNT_MODE_CREATE | ORBIS_SAVE_DATA_MOUNT_MODE_RDWR | ORBIS_SAVE_DATA_MOUNT_MODE_DESTRUCT_OFF:
     case ORBIS_SAVE_DATA_MOUNT_MODE_CREATE | ORBIS_SAVE_DATA_MOUNT_MODE_RDWR |
         ORBIS_SAVE_DATA_MOUNT_MODE_COPY_ICON:
     case ORBIS_SAVE_DATA_MOUNT_MODE_CREATE | ORBIS_SAVE_DATA_MOUNT_MODE_DESTRUCT_OFF |

src/emulator.cpp

@@ -137,8 +137,9 @@ void Emulator::Run(const std::filesystem::path& file) {
 }
 
 void Emulator::LoadSystemModules(const std::filesystem::path& file) {
-    constexpr std::array<SysModules, 6> ModulesToLoad{
+    constexpr std::array<SysModules, 7> ModulesToLoad{
         {{"libSceNgs2.sprx", nullptr},
+         {"libSceFiber.sprx", nullptr},
          {"libSceLibcInternal.sprx", &Libraries::LibcInternal::RegisterlibSceLibcInternal},
          {"libSceDiscMap.sprx", &Libraries::DiscMap::RegisterlibSceDiscMap},
          {"libSceRtc.sprx", &Libraries::Rtc::RegisterlibSceRtc},

src/shader_recompiler/backend/spirv/emit_spirv.cpp

@@ -176,6 +176,10 @@ Id DefineMain(EmitContext& ctx, IR::Program& program) {
 void DefineEntryPoint(const IR::Program& program, EmitContext& ctx, Id main) {
     const std::span interfaces(ctx.interfaces.data(), ctx.interfaces.size());
     spv::ExecutionModel execution_model{};
+    ctx.AddCapability(spv::Capability::Image1D);
+    ctx.AddCapability(spv::Capability::Sampled1D);
+    ctx.AddCapability(spv::Capability::Float16);
+    ctx.AddCapability(spv::Capability::Int16);
     ctx.AddCapability(spv::Capability::StorageImageWriteWithoutFormat);
     ctx.AddCapability(spv::Capability::StorageImageExtendedFormats);
     switch (program.info.stage) {

src/shader_recompiler/backend/spirv/emit_spirv_bitwise_conversion.cpp

@@ -18,8 +18,8 @@ void EmitBitCastU64F64(EmitContext&) {
     UNREACHABLE_MSG("SPIR-V Instruction");
 }
 
-void EmitBitCastF16U16(EmitContext&) {
+Id EmitBitCastF16U16(EmitContext& ctx, Id value) {
-    UNREACHABLE_MSG("SPIR-V Instruction");
+    return ctx.OpBitcast(ctx.F16[1], value);
 }
 
 Id EmitBitCastF32U32(EmitContext& ctx, Id value) {

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -60,7 +60,11 @@ Id OutputAttrPointer(EmitContext& ctx, IR::Attribute attr, u32 element) {
     case IR::Attribute::RenderTarget0:
     case IR::Attribute::RenderTarget1:
     case IR::Attribute::RenderTarget2:
-    case IR::Attribute::RenderTarget3: {
+    case IR::Attribute::RenderTarget3:
+    case IR::Attribute::RenderTarget4:
+    case IR::Attribute::RenderTarget5:
+    case IR::Attribute::RenderTarget6:
+    case IR::Attribute::RenderTarget7: {
         const u32 index = u32(attr) - u32(IR::Attribute::RenderTarget0);
         if (ctx.frag_num_comp[index] > 1) {
             return ctx.OpAccessChain(ctx.output_f32, ctx.frag_color[index], ctx.ConstU32(element));
@@ -196,7 +200,15 @@ Id EmitLoadBufferU32(EmitContext& ctx, IR::Inst* inst, u32 handle, Id address) {
 }
 
 Id EmitLoadBufferF32x2(EmitContext& ctx, IR::Inst* inst, u32 handle, Id address) {
-    UNREACHABLE();
+    const auto info = inst->Flags<IR::BufferInstInfo>();
+    const auto& buffer = ctx.buffers[handle];
+    boost::container::static_vector<Id, 2> ids;
+    for (u32 i = 0; i < 2; i++) {
+        const Id index{ctx.OpIAdd(ctx.U32[1], address, ctx.ConstU32(i))};
+        const Id ptr{ctx.OpAccessChain(buffer.pointer_type, buffer.id, ctx.u32_zero_value, index)};
+        ids.push_back(ctx.OpLoad(buffer.data_types->Get(1), ptr));
+    }
+    return ctx.OpCompositeConstruct(buffer.data_types->Get(2), ids);
 }
 
 Id EmitLoadBufferF32x3(EmitContext& ctx, IR::Inst* inst, u32 handle, Id address) {

src/shader_recompiler/backend/spirv/emit_spirv_convert.cpp

@@ -255,4 +255,8 @@ Id EmitConvertF64U64(EmitContext& ctx, Id value) {
     return ctx.OpConvertUToF(ctx.F64[1], value);
 }
 
+Id EmitConvertU16U32(EmitContext& ctx, Id value) {
+    return ctx.OpUConvert(ctx.U16, value);
+}
+
 } // namespace Shader::Backend::SPIRV

src/shader_recompiler/backend/spirv/emit_spirv_floating_point.cpp

@@ -6,6 +6,11 @@
 
 namespace Shader::Backend::SPIRV {
 
+Id Decorate(EmitContext& ctx, IR::Inst* inst, Id op) {
+    ctx.Decorate(op, spv::Decoration::NoContraction);
+    return op;
+}
+
 Id EmitFPAbs16(EmitContext& ctx, Id value) {
     return ctx.OpFAbs(ctx.F16[1], value);
 }
@@ -19,31 +24,31 @@ Id EmitFPAbs64(EmitContext& ctx, Id value) {
 }
 
 Id EmitFPAdd16(EmitContext& ctx, IR::Inst* inst, Id a, Id b) {
-    return ctx.OpFAdd(ctx.F16[1], a, b);
+    return Decorate(ctx, inst, ctx.OpFAdd(ctx.F16[1], a, b));
 }
 
 Id EmitFPAdd32(EmitContext& ctx, IR::Inst* inst, Id a, Id b) {
-    return ctx.OpFAdd(ctx.F32[1], a, b);
+    return Decorate(ctx, inst, ctx.OpFAdd(ctx.F32[1], a, b));
 }
 
 Id EmitFPAdd64(EmitContext& ctx, IR::Inst* inst, Id a, Id b) {
-    return ctx.OpFAdd(ctx.F64[1], a, b);
+    return Decorate(ctx, inst, ctx.OpFAdd(ctx.F64[1], a, b));
 }
 
 Id EmitFPSub32(EmitContext& ctx, IR::Inst* inst, Id a, Id b) {
-    return ctx.OpFSub(ctx.F32[1], a, b);
+    return Decorate(ctx, inst, ctx.OpFSub(ctx.F32[1], a, b));
 }
 
 Id EmitFPFma16(EmitContext& ctx, IR::Inst* inst, Id a, Id b, Id c) {
-    return ctx.OpFma(ctx.F16[1], a, b, c);
+    return Decorate(ctx, inst, ctx.OpFma(ctx.F16[1], a, b, c));
 }
 
 Id EmitFPFma32(EmitContext& ctx, IR::Inst* inst, Id a, Id b, Id c) {
-    return ctx.OpFma(ctx.F32[1], a, b, c);
+    return Decorate(ctx, inst, ctx.OpFma(ctx.F32[1], a, b, c));
 }
 
 Id EmitFPFma64(EmitContext& ctx, IR::Inst* inst, Id a, Id b, Id c) {
-    return ctx.OpFma(ctx.F64[1], a, b, c);
+    return Decorate(ctx, inst, ctx.OpFma(ctx.F64[1], a, b, c));
 }
 
 Id EmitFPMax32(EmitContext& ctx, Id a, Id b) {
@@ -63,15 +68,15 @@ Id EmitFPMin64(EmitContext& ctx, Id a, Id b) {
 }
 
 Id EmitFPMul16(EmitContext& ctx, IR::Inst* inst, Id a, Id b) {
-    return ctx.OpFMul(ctx.F16[1], a, b);
+    return Decorate(ctx, inst, ctx.OpFMul(ctx.F16[1], a, b));
 }
 
 Id EmitFPMul32(EmitContext& ctx, IR::Inst* inst, Id a, Id b) {
-    return ctx.OpFMul(ctx.F32[1], a, b);
+    return Decorate(ctx, inst, ctx.OpFMul(ctx.F32[1], a, b));
 }
 
 Id EmitFPMul64(EmitContext& ctx, IR::Inst* inst, Id a, Id b) {
-    return ctx.OpFMul(ctx.F64[1], a, b);
+    return Decorate(ctx, inst, ctx.OpFMul(ctx.F64[1], a, b));
 }
 
 Id EmitFPNeg16(EmitContext& ctx, Id value) {

src/shader_recompiler/backend/spirv/emit_spirv_image.cpp

@@ -141,7 +141,7 @@ void EmitImageWrite(EmitContext& ctx, IR::Inst* inst, u32 handle, Id coords, Id
     const auto& texture = ctx.images[handle & 0xFFFF];
     const Id image = ctx.OpLoad(texture.image_type, texture.id);
     const Id color_type = texture.data_types->Get(4);
-    ctx.OpImageWrite(image, ctx.OpBitcast(ctx.S32[2], coords), ctx.OpBitcast(color_type, color));
+    ctx.OpImageWrite(image, coords, ctx.OpBitcast(color_type, color));
 }
 
 } // namespace Shader::Backend::SPIRV

src/shader_recompiler/backend/spirv/emit_spirv_instructions.h

@@ -146,7 +146,7 @@ Id EmitSelectF64(EmitContext& ctx, Id cond, Id true_value, Id false_value);
 void EmitBitCastU16F16(EmitContext& ctx);
 Id EmitBitCastU32F32(EmitContext& ctx, Id value);
 void EmitBitCastU64F64(EmitContext& ctx);
-void EmitBitCastF16U16(EmitContext&);
+Id EmitBitCastF16U16(EmitContext& ctx, Id value);
 Id EmitBitCastF32U32(EmitContext& ctx, Id value);
 void EmitBitCastF64U64(EmitContext& ctx);
 Id EmitPackUint2x32(EmitContext& ctx, Id value);
@@ -343,6 +343,7 @@ Id EmitConvertF64U8(EmitContext& ctx, Id value);
 Id EmitConvertF64U16(EmitContext& ctx, Id value);
 Id EmitConvertF64U32(EmitContext& ctx, Id value);
 Id EmitConvertF64U64(EmitContext& ctx, Id value);
+Id EmitConvertU16U32(EmitContext& ctx, Id value);
 
 Id EmitImageSampleImplicitLod(EmitContext& ctx, IR::Inst* inst, u32 handle, Id coords, Id bias_lc,
                               Id offset);

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -74,19 +74,19 @@ Id EmitContext::Def(const IR::Value& value) {
 void EmitContext::DefineArithmeticTypes() {
     void_id = Name(TypeVoid(), "void_id");
     U1[1] = Name(TypeBool(), "bool_id");
-    // F16[1] = Name(TypeFloat(16), "f16_id");
+    F16[1] = Name(TypeFloat(16), "f16_id");
     F32[1] = Name(TypeFloat(32), "f32_id");
     // F64[1] = Name(TypeFloat(64), "f64_id");
     S32[1] = Name(TypeSInt(32), "i32_id");
     U32[1] = Name(TypeUInt(32), "u32_id");
     // U8 = Name(TypeSInt(8), "u8");
     // S8 = Name(TypeUInt(8), "s8");
-    // U16 = Name(TypeUInt(16), "u16_id");
+    U16 = Name(TypeUInt(16), "u16_id");
     // S16 = Name(TypeSInt(16), "s16_id");
     // U64 = Name(TypeUInt(64), "u64_id");
 
     for (u32 i = 2; i <= 4; i++) {
-        // F16[i] = Name(TypeVector(F16[1], i), fmt::format("f16vec{}_id", i));
+        F16[i] = Name(TypeVector(F16[1], i), fmt::format("f16vec{}_id", i));
         F32[i] = Name(TypeVector(F32[1], i), fmt::format("f32vec{}_id", i));
         // F64[i] = Name(TypeVector(F64[1], i), fmt::format("f64vec{}_id", i));
         S32[i] = Name(TypeVector(S32[1], i), fmt::format("i32vec{}_id", i));

src/shader_recompiler/frontend/fetch_shader.cpp

@@ -32,7 +32,7 @@ namespace Shader::Gcn {
  * We take the reverse way, extract the original input semantics from these instructions.
  **/
 
-std::vector<VertexAttribute> ParseFetchShader(const u32* code) {
+std::vector<VertexAttribute> ParseFetchShader(const u32* code, u32* out_size) {
     std::vector<VertexAttribute> attributes;
     GcnCodeSlice code_slice(code, code + std::numeric_limits<u32>::max());
     GcnDecodeContext decoder;
@@ -47,6 +47,8 @@ std::vector<VertexAttribute> ParseFetchShader(const u32* code) {
     u32 semantic_index = 0;
     while (!code_slice.atEnd()) {
         const auto inst = decoder.decodeInstruction(code_slice);
+        *out_size += inst.length;
+
         if (inst.opcode == Opcode::S_SETPC_B64) {
             break;
         }

src/shader_recompiler/frontend/fetch_shader.h

@@ -17,6 +17,6 @@ struct VertexAttribute {
     u8 instance_data; ///< Indicates that the buffer will be accessed in instance rate
 };
 
-std::vector<VertexAttribute> ParseFetchShader(const u32* code);
+std::vector<VertexAttribute> ParseFetchShader(const u32* code, u32* out_size);
 
 } // namespace Shader::Gcn

src/shader_recompiler/frontend/translate/scalar_memory.cpp

@@ -5,20 +5,29 @@
 
 namespace Shader::Gcn {
 
+static constexpr u32 SQ_SRC_LITERAL = 0xFF;
+
 void Translator::S_LOAD_DWORD(int num_dwords, const GcnInst& inst) {
     const auto& smrd = inst.control.smrd;
-    ASSERT_MSG(smrd.imm, "Bindless texture loads unsupported");
+    const u32 dword_offset = [&] -> u32 {
+        if (smrd.imm) {
+            return smrd.offset;
+        }
+        if (smrd.offset == SQ_SRC_LITERAL) {
+            return inst.src[1].code;
+        }
+        UNREACHABLE();
+    }();
     const IR::ScalarReg sbase{inst.src[0].code * 2};
     const IR::Value base =
         ir.CompositeConstruct(ir.GetScalarReg(sbase), ir.GetScalarReg(sbase + 1));
     IR::ScalarReg dst_reg{inst.dst[0].code};
     for (u32 i = 0; i < num_dwords; i++) {
-        ir.SetScalarReg(dst_reg++, ir.ReadConst(base, ir.Imm32(smrd.offset + i)));
+        ir.SetScalarReg(dst_reg++, ir.ReadConst(base, ir.Imm32(dword_offset + i)));
     }
 }
 
 void Translator::S_BUFFER_LOAD_DWORD(int num_dwords, const GcnInst& inst) {
-    static constexpr u32 SQ_SRC_LITERAL = 0xFF;
     const auto& smrd = inst.control.smrd;
     const IR::ScalarReg sbase{inst.src[0].code * 2};
     const IR::U32 dword_offset = [&] -> IR::U32 {

src/shader_recompiler/frontend/translate/translate.cpp

@@ -1,6 +1,9 @@
 // SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include "common/config.h"
+#include "common/io_file.h"
+#include "common/path_util.h"
 #include "shader_recompiler/exception.h"
 #include "shader_recompiler/frontend/fetch_shader.h"
 #include "shader_recompiler/frontend/translate/translate.h"
@@ -190,7 +193,20 @@ void Translator::EmitFetch(const GcnInst& inst) {
     std::memcpy(&code, &info.user_data[sgpr_base], sizeof(code));
 
     // Parse the assembly to generate a list of attributes.
-    const auto attribs = ParseFetchShader(code);
+    u32 fetch_size{};
+    const auto attribs = ParseFetchShader(code, &fetch_size);
+
+    if (Config::dumpShaders()) {
+        using namespace Common::FS;
+        const auto dump_dir = GetUserPath(PathType::ShaderDir) / "dumps";
+        if (!std::filesystem::exists(dump_dir)) {
+            std::filesystem::create_directories(dump_dir);
+        }
+        const auto filename = fmt::format("vs_fetch_{:#018x}.bin", info.pgm_hash);
+        const auto file = IOFile{dump_dir / filename, FileAccessMode::Write};
+        file.WriteRaw<u8>(code, fetch_size);
+    }
+
     for (const auto& attrib : attribs) {
         const IR::Attribute attr{IR::Attribute::Param0 + attrib.semantic};
         IR::VectorReg dst_reg{attrib.dest_vgpr};
@@ -242,6 +258,7 @@ void Translate(IR::Block* block, u32 block_base, std::span<const GcnInst> inst_l
     }
     Translator translator{block, info};
     for (const auto& inst : inst_list) {
+        block_base += inst.length;
         switch (inst.opcode) {
         case Opcode::S_MOVK_I32:
             translator.S_MOVK(inst);
@@ -378,6 +395,7 @@ void Translate(IR::Block* block, u32 block_base, std::span<const GcnInst> inst_l
         case Opcode::IMAGE_SAMPLE:
         case Opcode::IMAGE_SAMPLE_L:
         case Opcode::IMAGE_SAMPLE_C_O:
+        case Opcode::IMAGE_SAMPLE_B:
             translator.IMAGE_SAMPLE(inst);
             break;
         case Opcode::IMAGE_ATOMIC_ADD:
@@ -527,6 +545,12 @@ void Translate(IR::Block* block, u32 block_base, std::span<const GcnInst> inst_l
         case Opcode::V_CNDMASK_B32:
             translator.V_CNDMASK_B32(inst);
             break;
+        case Opcode::TBUFFER_LOAD_FORMAT_X:
+            translator.BUFFER_LOAD_FORMAT(1, true, inst);
+            break;
+        case Opcode::TBUFFER_LOAD_FORMAT_XY:
+            translator.BUFFER_LOAD_FORMAT(2, true, inst);
+            break;
         case Opcode::TBUFFER_LOAD_FORMAT_XYZ:
             translator.BUFFER_LOAD_FORMAT(3, true, inst);
             break;
@@ -901,7 +925,6 @@ void Translate(IR::Block* block, u32 block_base, std::span<const GcnInst> inst_l
                       magic_enum::enum_name(inst.opcode), opcode);
             info.translation_failed = true;
         }
-        block_base += inst.length;
     }
 }
 

src/shader_recompiler/frontend/translate/vector_alu.cpp

@@ -28,7 +28,8 @@ void Translator::V_CVT_PKRTZ_F16_F32(const GcnInst& inst) {
 
 void Translator::V_CVT_F32_F16(const GcnInst& inst) {
     const IR::U32 src0 = GetSrc(inst.src[0]);
-    SetDst(inst.dst[0], ir.ConvertUToF(32, 16, src0));
+    const IR::U16 src0l = ir.UConvert(16, src0);
+    SetDst(inst.dst[0], ir.FPConvert(32, ir.BitCast<IR::F16>(src0l)));
 }
 
 void Translator::V_MUL_F32(const GcnInst& inst) {
@@ -519,7 +520,7 @@ void Translator::V_CVT_FLR_I32_F32(const GcnInst& inst) {
 void Translator::V_CMP_CLASS_F32(const GcnInst& inst) {
     const IR::F32 src0{GetSrc(inst.src[0], true)};
     const IR::U32 src1{GetSrc(inst.src[1])};
-    ir.SetVcc(ir.Imm1(false));
+    ir.SetVcc(ir.FPIsInf(src0));
     // TODO
 }
 

src/shader_recompiler/ir/ir_emitter.cpp

@@ -1194,6 +1194,13 @@ F32F64 IREmitter::ConvertIToF(size_t dest_bitsize, size_t src_bitsize, bool is_s
 }
 
 U16U32U64 IREmitter::UConvert(size_t result_bitsize, const U16U32U64& value) {
+    switch (result_bitsize) {
+    case 16:
+        switch (value.Type()) {
+        case Type::U32:
+            return Inst<U16>(Opcode::ConvertU16U32, value);
+        }
+    }
     throw NotImplementedException("Conversion from {} to {} bits", value.Type(), result_bitsize);
 }
 

src/shader_recompiler/ir/opcodes.inc

@@ -283,6 +283,7 @@ OPCODE(ConvertF32U32,                                       F32,            U32,
 OPCODE(ConvertF64S32,                                       F64,            U32,                                                                            )
 OPCODE(ConvertF64U32,                                       F64,            U32,                                                                            )
 OPCODE(ConvertF32U16,                                       F32,            U16,                                                                            )
+OPCODE(ConvertU16U32,                                       U16,            U32,                                                                            )
 
 // Image operations
 OPCODE(ImageSampleImplicitLod,                              F32x4,          Opaque,         Opaque,         Opaque,         Opaque,                         )

src/shader_recompiler/ir/passes/resource_tracking_pass.cpp

@@ -234,7 +234,7 @@ s32 TryHandleInlineCbuf(IR::Inst& inst, Info& info, Descriptors& descriptors, Am
      **/
     IR::Inst* handle = inst.Arg(0).InstRecursive();
     IR::Inst* p0 = handle->Arg(0).InstRecursive();
-    if (p0->GetOpcode() != IR::Opcode::IAdd32 || !p0->Arg(0).IsImmediate()) {
+    if (p0->GetOpcode() != IR::Opcode::IAdd32 || !p0->Arg(0).IsImmediate() || !p0->Arg(1).IsImmediate()) {
         return -1;
     }
     IR::Inst* p1 = handle->Arg(1).InstRecursive();
@@ -286,7 +286,9 @@ void PatchBufferInstruction(IR::Block& block, IR::Inst& inst, Info& info,
     if (inst_info.is_typed) {
         ASSERT(inst_info.nfmt == AmdGpu::NumberFormat::Float &&
                (inst_info.dmft == AmdGpu::DataFormat::Format32_32_32_32 ||
-                inst_info.dmft == AmdGpu::DataFormat::Format32_32_32));
+                inst_info.dmft == AmdGpu::DataFormat::Format32_32_32 ||
+                inst_info.dmft == AmdGpu::DataFormat::Format32_32 ||
+                inst_info.dmft == AmdGpu::DataFormat::Format32));
     }
     if (inst.GetOpcode() == IR::Opcode::ReadConstBuffer ||
         inst.GetOpcode() == IR::Opcode::ReadConstBufferU32) {

src/shader_recompiler/runtime_info.h

@@ -165,6 +165,7 @@ struct Info {
     Stage stage;
 
     uintptr_t pgm_base{};
+    u64 pgm_hash{};
     u32 shared_memory_size{};
     bool uses_group_quad{};
     bool uses_shared_u8{};

src/video_core/amdgpu/liverpool.h

@@ -479,7 +479,7 @@ struct Liverpool {
 
         template <typename T = VAddr>
         T Address() const {
-            return reinterpret_cast<T>(base_addr_lo | u64(base_addr_hi) << 32);
+            return reinterpret_cast<T>((base_addr_lo & ~1U) | u64(base_addr_hi) << 32);
         }
     };
 

src/video_core/amdgpu/pm4_cmds.h

@@ -329,6 +329,10 @@ struct PM4CmdEventWriteEop {
             *Address<u64>() = Common::FencedRDTSC();
             break;
         }
+        case DataSelect::GpuClock64: {
+            *Address<u64>() = 0;
+            break;
+        }
         default: {
             UNREACHABLE();
         }
@@ -549,8 +553,8 @@ struct PM4DumpConstRam {
     u32 addr_hi;
 
     template <typename T>
-    T* Address() const {
+    T Address() const {
-        return reinterpret_cast<T*>((u64(addr_hi) << 32u) | addr_lo);
+        return reinterpret_cast<T>((u64(addr_hi) << 32u) | addr_lo);
     }
 
     [[nodiscard]] u32 Offset() const {

src/video_core/amdgpu/resource.h

@@ -63,7 +63,7 @@ struct Buffer {
         if (stride == 0) {
             return 1U;
         }
-        ASSERT(stride % element_size == 0);
+        //ASSERT(stride % element_size == 0);
         return stride / element_size;
     }
 

src/video_core/renderer_vulkan/liverpool_to_vk.cpp

@@ -404,6 +404,26 @@ vk::Format SurfaceFormat(AmdGpu::DataFormat data_format, AmdGpu::NumberFormat nu
         num_format == AmdGpu::NumberFormat::Unorm) {
         return vk::Format::eR4G4B4A4UnormPack16;
     }
+    if (data_format == AmdGpu::DataFormat::Format16_16_16_16 &&
+        num_format == AmdGpu::NumberFormat::Uint) {
+        return vk::Format::eR16G16B16A16Uint;
+    }
+    if (data_format == AmdGpu::DataFormat::Format32_32_32_32 &&
+        num_format == AmdGpu::NumberFormat::Uint) {
+        return vk::Format::eR32G32B32A32Uint;
+    }
+    if (data_format == AmdGpu::DataFormat::Format8 &&
+        num_format == AmdGpu::NumberFormat::Sint) {
+        return vk::Format::eR8Sint;
+    }
+    if (data_format == AmdGpu::DataFormat::FormatBc1 &&
+        num_format == AmdGpu::NumberFormat::Srgb) {
+        return vk::Format::eBc1RgbaSrgbBlock;
+    }
+    if (data_format == AmdGpu::DataFormat::Format16_16 &&
+        num_format == AmdGpu::NumberFormat::Sint) {
+        return vk::Format::eR16G16Sint;
+    }
     UNREACHABLE_MSG("Unknown data_format={} and num_format={}", u32(data_format), u32(num_format));
 }
 

src/video_core/renderer_vulkan/vk_instance.cpp

@@ -209,12 +209,14 @@ bool Instance::CreateDevice() {
                 .shaderImageGatherExtended = true,
                 .shaderStorageImageMultisample = true,
                 .shaderClipDistance = features.shaderClipDistance,
+                .shaderInt16 = true,
             },
         },
         vk::PhysicalDeviceVulkan11Features{
             .shaderDrawParameters = true,
         },
         vk::PhysicalDeviceVulkan12Features{
+            .shaderFloat16 = true,
             .scalarBlockLayout = true,
             .uniformBufferStandardLayout = true,
             .hostQueryReset = true,

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -251,11 +251,16 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline() {
         block_pool.ReleaseContents();
         inst_pool.ReleaseContents();
 
+        if (hash == 0x43ade46898f820e2 || hash == 0xbcf2be6c546ad35a) {
+            return nullptr;
+        }
+
         // Recompile shader to IR.
         try {
             LOG_INFO(Render_Vulkan, "Compiling {} shader {:#x}", stage, hash);
             Shader::Info info = MakeShaderInfo(stage, pgm->user_data, regs);
             info.pgm_base = pgm->Address<uintptr_t>();
+            info.pgm_hash = hash;
             programs[i] = Shader::TranslateProgram(inst_pool, block_pool, code, std::move(info));
 
             // Compile IR to SPIR-V

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -23,7 +23,7 @@ Rasterizer::Rasterizer(const Instance& instance_, Scheduler& scheduler_,
     : instance{instance_}, scheduler{scheduler_}, texture_cache{texture_cache_},
       liverpool{liverpool_}, memory{Core::Memory::Instance()},
       pipeline_cache{instance, scheduler, liverpool},
-      vertex_index_buffer{instance, scheduler, VertexIndexFlags, 1_GB, BufferType::Upload} {
+      vertex_index_buffer{instance, scheduler, VertexIndexFlags, 3_GB, BufferType::Upload} {
     if (!Config::nullGpu()) {
         liverpool->BindRasterizer(this);
     }
@@ -174,11 +174,9 @@ u32 Rasterizer::SetupIndexBuffer(bool& is_indexed, u32 index_offset) {
 
     // Upload index data to stream buffer.
     const auto index_address = regs.index_base_address.Address<const void*>();
-    const u32 index_buffer_size = regs.num_indices * index_size;
+    const u32 index_buffer_size = (index_offset + regs.num_indices) * index_size;
     const auto [data, offset, _] = vertex_index_buffer.Map(index_buffer_size);
-    static constexpr std::array<u16, 4> test{};
     std::memcpy(data, index_address, index_buffer_size);
-    ASSERT(std::memcmp(data, test.data(), sizeof(test)) != 0);
     vertex_index_buffer.Commit(index_buffer_size);
 
     // Bind index buffer.

src/video_core/renderer_vulkan/vk_stream_buffer.cpp

@@ -226,7 +226,7 @@ void StreamBuffer::WaitPendingOperations(u64 requested_upper_bound) {
     while (requested_upper_bound > wait_bound && wait_cursor < *invalidation_mark) {
         auto& watch = previous_watches[wait_cursor];
         wait_bound = watch.upper_bound;
-        scheduler.Wait(watch.tick);
+        //scheduler.Wait(watch.tick);
         ++wait_cursor;
     }
 }

src/video_core/texture_cache/image_view.cpp

@@ -77,8 +77,8 @@ ImageView::ImageView(const Vulkan::Instance& instance, const ImageViewInfo& info
     if (usage_override) {
         usage_ci.usage = usage_override.value();
     }
-    if (info.format == vk::Format::eR32Sfloat) {
+    if (image.info.type == vk::ImageType::e1D) {
-        printf("stop\n");
+        printf("bad\n");
     }
     // When sampling D32 texture from shader, the T# specifies R32 Float format so adjust it.
     vk::Format format = info.format;

src/video_core/texture_cache/tile_manager.cpp

@@ -185,6 +185,7 @@ vk::Format DemoteImageFormatForDetiling(vk::Format format) {
     case vk::Format::eB8G8R8A8Srgb:
     case vk::Format::eB8G8R8A8Unorm:
     case vk::Format::eR8G8B8A8Unorm:
+    case vk::Format::eR32Sfloat:
     case vk::Format::eR32Uint:
         return vk::Format::eR32Uint;
     case vk::Format::eBc1RgbaUnormBlock: