oops forgot optimize off

This commit is contained in:
IndecisiveTurtle 2024-08-08 14:13:30 +03:00
parent b9938cb1d8
commit 2c98063e76
8 changed files with 19 additions and 21 deletions

View File

@ -1056,16 +1056,7 @@ ScePthread PThreadPool::Create() {
} }
} }
#ifdef _WIN64
auto* ret = new PthreadInternal{}; auto* ret = new PthreadInternal{};
#else
// TODO: Linux specific hack
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_free = false;
ret->is_detached = false; ret->is_detached = false;
ret->is_almost_done = false; ret->is_almost_done = false;

View File

@ -129,7 +129,7 @@ Id EmitReadConst(EmitContext& ctx) {
Id EmitReadConstBuffer(EmitContext& ctx, u32 handle, Id index) { Id EmitReadConstBuffer(EmitContext& ctx, u32 handle, Id index) {
auto& buffer = ctx.buffers[handle]; auto& buffer = ctx.buffers[handle];
if (!Sirit::ValidId(buffer.offset)) { if (!Sirit::ValidId(buffer.offset)) {
buffer.offset = ctx.GetBufferOffset(buffer.global_binding); buffer.offset = ctx.GetBufferOffset(handle);
} }
const Id offset_dwords{ctx.OpShiftRightLogical(ctx.U32[1], buffer.offset, ctx.ConstU32(2U))}; const Id offset_dwords{ctx.OpShiftRightLogical(ctx.U32[1], buffer.offset, ctx.ConstU32(2U))};
index = ctx.OpIAdd(ctx.U32[1], index, offset_dwords); index = ctx.OpIAdd(ctx.U32[1], index, offset_dwords);
@ -230,7 +230,7 @@ template <u32 N>
static Id EmitLoadBufferF32xN(EmitContext& ctx, u32 handle, Id address) { static Id EmitLoadBufferF32xN(EmitContext& ctx, u32 handle, Id address) {
auto& buffer = ctx.buffers[handle]; auto& buffer = ctx.buffers[handle];
if (!Sirit::ValidId(buffer.offset)) { if (!Sirit::ValidId(buffer.offset)) {
buffer.offset = ctx.GetBufferOffset(buffer.global_binding); buffer.offset = ctx.GetBufferOffset(handle);
} }
address = ctx.OpIAdd(ctx.U32[1], address, buffer.offset); address = ctx.OpIAdd(ctx.U32[1], address, buffer.offset);
const Id index = ctx.OpShiftRightLogical(ctx.U32[1], address, ctx.ConstU32(2u)); const Id index = ctx.OpShiftRightLogical(ctx.U32[1], address, ctx.ConstU32(2u));
@ -412,7 +412,7 @@ template <u32 N>
static Id EmitLoadBufferFormatF32xN(EmitContext& ctx, IR::Inst* inst, u32 handle, Id address) { static Id EmitLoadBufferFormatF32xN(EmitContext& ctx, IR::Inst* inst, u32 handle, Id address) {
auto& buffer = ctx.buffers[handle]; auto& buffer = ctx.buffers[handle];
if (!Sirit::ValidId(buffer.offset)) { if (!Sirit::ValidId(buffer.offset)) {
buffer.offset = ctx.GetBufferOffset(buffer.global_binding); buffer.offset = ctx.GetBufferOffset(handle);
} }
address = ctx.OpIAdd(ctx.U32[1], address, buffer.offset); address = ctx.OpIAdd(ctx.U32[1], address, buffer.offset);
if constexpr (N == 1) { if constexpr (N == 1) {
@ -446,7 +446,7 @@ template <u32 N>
static void EmitStoreBufferF32xN(EmitContext& ctx, u32 handle, Id address, Id value) { static void EmitStoreBufferF32xN(EmitContext& ctx, u32 handle, Id address, Id value) {
auto& buffer = ctx.buffers[handle]; auto& buffer = ctx.buffers[handle];
if (!Sirit::ValidId(buffer.offset)) { if (!Sirit::ValidId(buffer.offset)) {
buffer.offset = ctx.GetBufferOffset(buffer.global_binding); buffer.offset = ctx.GetBufferOffset(handle);
} }
address = ctx.OpIAdd(ctx.U32[1], address, buffer.offset); address = ctx.OpIAdd(ctx.U32[1], address, buffer.offset);
const Id index = ctx.OpShiftRightLogical(ctx.U32[1], address, ctx.ConstU32(2u)); const Id index = ctx.OpShiftRightLogical(ctx.U32[1], address, ctx.ConstU32(2u));

View File

@ -352,9 +352,9 @@ void EmitContext::DefineBuffers() {
Decorate(id, spv::Decoration::DescriptorSet, 0U); Decorate(id, spv::Decoration::DescriptorSet, 0U);
Name(id, fmt::format("{}_{}", buffer.is_storage ? "ssbo" : "cbuf", buffer.sgpr_base)); Name(id, fmt::format("{}_{}", buffer.is_storage ? "ssbo" : "cbuf", buffer.sgpr_base));
binding++;
buffers.push_back({ buffers.push_back({
.id = id, .id = id,
.global_binding = binding++,
.data_types = data_types, .data_types = data_types,
.pointer_type = pointer_type, .pointer_type = pointer_type,
.buffer = buffer.GetVsharp(info), .buffer = buffer.GetVsharp(info),

View File

@ -203,7 +203,6 @@ public:
struct BufferDefinition { struct BufferDefinition {
Id id; Id id;
Id offset; Id offset;
u32 global_binding;
const VectorIds* data_types; const VectorIds* data_types;
Id pointer_type; Id pointer_type;
AmdGpu::Buffer buffer; AmdGpu::Buffer buffer;

View File

@ -1,6 +1,6 @@
// SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project // SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later // SPDX-License-Identifier: GPL-2.0-or-later
#pragma clang optimize off
#include <algorithm> #include <algorithm>
#include "common/alignment.h" #include "common/alignment.h"
#include "common/scope_exit.h" #include "common/scope_exit.h"

View File

@ -10,11 +10,13 @@
#ifndef _WIN64 #ifndef _WIN64
#include <fcntl.h> #include <fcntl.h>
#include <linux/userfaultfd.h>
#include <poll.h> #include <poll.h>
#include <signal.h> #include <signal.h>
#include <sys/ioctl.h> #include <sys/ioctl.h>
#include <sys/mman.h> #include <sys/mman.h>
#ifdef ENABLE_USERFAULTFD
#include <linux/userfaultfd.h>
#endif
#else #else
#include <windows.h> #include <windows.h>
#endif #endif
@ -191,7 +193,11 @@ struct PageManager::Impl {
static void GuestFaultSignalHandler(int sig, siginfo_t* info, void* raw_context) { static void GuestFaultSignalHandler(int sig, siginfo_t* info, void* raw_context) {
ucontext_t* ctx = reinterpret_cast<ucontext_t*>(raw_context); ucontext_t* ctx = reinterpret_cast<ucontext_t*>(raw_context);
const VAddr address = reinterpret_cast<VAddr>(info->si_addr); const VAddr address = reinterpret_cast<VAddr>(info->si_addr);
#ifdef __APPLE__
const u32 err = ctx->uc_mcontext->__es.__err;
#else
const greg_t err = ctx->uc_mcontext.gregs[REG_ERR]; const greg_t err = ctx->uc_mcontext.gregs[REG_ERR];
#endif
if (err & 0x2) { if (err & 0x2) {
rasterizer->InvalidateMemory(address, sizeof(u64)); rasterizer->InvalidateMemory(address, sizeof(u64));
} else { } else {

View File

@ -96,7 +96,7 @@ bool ComputePipeline::BindResources(VideoCore::BufferCache& buffer_cache,
Shader::PushData push_data{}; Shader::PushData push_data{};
u32 binding{}; u32 binding{};
for (const auto& buffer : info.buffers) { for (u32 i = 0; const auto& buffer : info.buffers) {
const auto vsharp = buffer.GetVsharp(info); const auto vsharp = buffer.GetVsharp(info);
const VAddr address = vsharp.base_address; const VAddr address = vsharp.base_address;
// Most of the time when a metadata is updated with a shader it gets cleared. It means we // Most of the time when a metadata is updated with a shader it gets cleared. It means we
@ -125,7 +125,7 @@ bool ComputePipeline::BindResources(VideoCore::BufferCache& buffer_cache,
const u32 adjust = offset - offset_aligned; const u32 adjust = offset - offset_aligned;
if (adjust != 0) { if (adjust != 0) {
ASSERT(adjust % 4 == 0); ASSERT(adjust % 4 == 0);
push_data.AddOffset(binding, adjust); push_data.AddOffset(i, adjust);
} }
buffer_infos.emplace_back(vk_buffer->Handle(), offset_aligned, size + adjust); buffer_infos.emplace_back(vk_buffer->Handle(), offset_aligned, size + adjust);
set_writes.push_back({ set_writes.push_back({
@ -137,6 +137,7 @@ bool ComputePipeline::BindResources(VideoCore::BufferCache& buffer_cache,
: vk::DescriptorType::eUniformBuffer, : vk::DescriptorType::eUniformBuffer,
.pBufferInfo = &buffer_infos.back(), .pBufferInfo = &buffer_infos.back(),
}); });
i++;
} }
for (const auto& image_desc : info.images) { for (const auto& image_desc : info.images) {

View File

@ -343,7 +343,7 @@ void GraphicsPipeline::BindResources(const Liverpool::Regs& regs,
push_data.step0 = regs.vgt_instance_step_rate_0; push_data.step0 = regs.vgt_instance_step_rate_0;
push_data.step1 = regs.vgt_instance_step_rate_1; push_data.step1 = regs.vgt_instance_step_rate_1;
} }
for (const auto& buffer : stage.buffers) { for (u32 i = 0; const auto& buffer : stage.buffers) {
const auto vsharp = buffer.GetVsharp(stage); const auto vsharp = buffer.GetVsharp(stage);
if (vsharp) { if (vsharp) {
const VAddr address = vsharp.base_address; const VAddr address = vsharp.base_address;
@ -359,7 +359,7 @@ void GraphicsPipeline::BindResources(const Liverpool::Regs& regs,
const u32 adjust = offset - offset_aligned; const u32 adjust = offset - offset_aligned;
if (adjust != 0) { if (adjust != 0) {
ASSERT(adjust % 4 == 0); ASSERT(adjust % 4 == 0);
push_data.AddOffset(binding, adjust); push_data.AddOffset(i, adjust);
} }
buffer_infos.emplace_back(vk_buffer->Handle(), offset_aligned, size + adjust); buffer_infos.emplace_back(vk_buffer->Handle(), offset_aligned, size + adjust);
} else { } else {
@ -374,6 +374,7 @@ void GraphicsPipeline::BindResources(const Liverpool::Regs& regs,
: vk::DescriptorType::eUniformBuffer, : vk::DescriptorType::eUniformBuffer,
.pBufferInfo = &buffer_infos.back(), .pBufferInfo = &buffer_infos.back(),
}); });
i++;
} }
boost::container::static_vector<AmdGpu::Image, 16> tsharps; boost::container::static_vector<AmdGpu::Image, 16> tsharps;