Duckstation/src/core/cpu_code_cache.h

101 lines
2.6 KiB
C
Raw Normal View History

#pragma once
#include "common/bitfield.h"
JIT optimizations and refactoring (#675) * CPU/Recompiler: Use rel32 call where possible for no-args * JitCodeBuffer: Support using preallocated buffer * CPU/Recompiler/AArch64: Use bl instead of blr for short branches * CPU/CodeCache: Allocate recompiler buffer in program space This means we don't need 64-bit moves for every call out of the recompiler. * GTE: Don't store as u16 and load as u32 * CPU/Recompiler: Add methods to emit global load/stores * GTE: Convert class to namespace * CPU/Recompiler: Call GTE functions directly * Settings: Turn into a global variable * GPU: Replace local pointers with global * InterruptController: Turn into a global pointer * System: Replace local pointers with global * Timers: Turn into a global instance * DMA: Turn into a global instance * SPU: Turn into a global instance * CDROM: Turn into a global instance * MDEC: Turn into a global instance * Pad: Turn into a global instance * SIO: Turn into a global instance * CDROM: Move audio FIFO to the heap * CPU/Recompiler: Drop ASMFunctions No longer needed since we have code in the same 4GB window. * CPUCodeCache: Turn class into namespace * Bus: Local pointer -> global pointers * CPU: Turn class into namespace * Bus: Turn into namespace * GTE: Store registers in CPU state struct Allows relative addressing on ARM. * CPU/Recompiler: Align code storage to page size * CPU/Recompiler: Fix relative branches on A64 * HostInterface: Local references to global * System: Turn into a namespace, move events out * Add guard pages * Android: Fix build
2020-07-31 07:09:18 +00:00
#include "common/jit_code_buffer.h"
#include "cpu_types.h"
#include <array>
#include <memory>
#include <unordered_map>
#include <vector>
namespace CPU {
2019-11-27 15:55:33 +00:00
union CodeBlockKey
{
u32 bits;
BitField<u32, bool, 0, 1> user_mode;
BitField<u32, u32, 2, 30> aligned_pc;
ALWAYS_INLINE u32 GetPC() const { return aligned_pc << 2; }
ALWAYS_INLINE void SetPC(u32 pc) { aligned_pc = pc >> 2; }
ALWAYS_INLINE u32 GetPCPhysicalAddress() const { return (aligned_pc << 2) & PHYSICAL_MEMORY_ADDRESS_MASK; }
2019-11-27 15:55:33 +00:00
ALWAYS_INLINE CodeBlockKey& operator=(const CodeBlockKey& rhs)
{
bits = rhs.bits;
return *this;
}
ALWAYS_INLINE bool operator==(const CodeBlockKey& rhs) const { return bits == rhs.bits; }
ALWAYS_INLINE bool operator!=(const CodeBlockKey& rhs) const { return bits != rhs.bits; }
ALWAYS_INLINE bool operator<(const CodeBlockKey& rhs) const { return bits < rhs.bits; }
};
struct CodeBlockInstruction
{
Instruction instruction;
u32 pc;
bool is_branch_instruction : 1;
bool is_branch_delay_slot : 1;
bool is_load_instruction : 1;
bool is_store_instruction : 1;
bool is_load_delay_slot : 1;
bool is_last_instruction : 1;
bool has_load_delay : 1;
bool can_trap : 1;
};
struct CodeBlock
{
JIT optimizations and refactoring (#675) * CPU/Recompiler: Use rel32 call where possible for no-args * JitCodeBuffer: Support using preallocated buffer * CPU/Recompiler/AArch64: Use bl instead of blr for short branches * CPU/CodeCache: Allocate recompiler buffer in program space This means we don't need 64-bit moves for every call out of the recompiler. * GTE: Don't store as u16 and load as u32 * CPU/Recompiler: Add methods to emit global load/stores * GTE: Convert class to namespace * CPU/Recompiler: Call GTE functions directly * Settings: Turn into a global variable * GPU: Replace local pointers with global * InterruptController: Turn into a global pointer * System: Replace local pointers with global * Timers: Turn into a global instance * DMA: Turn into a global instance * SPU: Turn into a global instance * CDROM: Turn into a global instance * MDEC: Turn into a global instance * Pad: Turn into a global instance * SIO: Turn into a global instance * CDROM: Move audio FIFO to the heap * CPU/Recompiler: Drop ASMFunctions No longer needed since we have code in the same 4GB window. * CPUCodeCache: Turn class into namespace * Bus: Local pointer -> global pointers * CPU: Turn class into namespace * Bus: Turn into namespace * GTE: Store registers in CPU state struct Allows relative addressing on ARM. * CPU/Recompiler: Align code storage to page size * CPU/Recompiler: Fix relative branches on A64 * HostInterface: Local references to global * System: Turn into a namespace, move events out * Add guard pages * Android: Fix build
2020-07-31 07:09:18 +00:00
using HostCodePointer = void (*)();
2019-11-27 15:55:33 +00:00
CodeBlock(const CodeBlockKey key_) : key(key_) {}
CodeBlockKey key;
u32 host_code_size = 0;
HostCodePointer host_code = nullptr;
std::vector<CodeBlockInstruction> instructions;
std::vector<CodeBlock*> link_predecessors;
std::vector<CodeBlock*> link_successors;
bool invalidated = false;
const u32 GetPC() const { return key.GetPC(); }
const u32 GetSizeInBytes() const { return static_cast<u32>(instructions.size()) * sizeof(Instruction); }
const u32 GetStartPageIndex() const { return (key.GetPCPhysicalAddress() / CPU_CODE_CACHE_PAGE_SIZE); }
const u32 GetEndPageIndex() const
{
return ((key.GetPCPhysicalAddress() + GetSizeInBytes()) / CPU_CODE_CACHE_PAGE_SIZE);
}
2019-11-27 15:55:33 +00:00
bool IsInRAM() const
{
// TODO: Constant
return key.GetPCPhysicalAddress() < 0x200000;
2019-11-27 15:55:33 +00:00
}
};
JIT optimizations and refactoring (#675) * CPU/Recompiler: Use rel32 call where possible for no-args * JitCodeBuffer: Support using preallocated buffer * CPU/Recompiler/AArch64: Use bl instead of blr for short branches * CPU/CodeCache: Allocate recompiler buffer in program space This means we don't need 64-bit moves for every call out of the recompiler. * GTE: Don't store as u16 and load as u32 * CPU/Recompiler: Add methods to emit global load/stores * GTE: Convert class to namespace * CPU/Recompiler: Call GTE functions directly * Settings: Turn into a global variable * GPU: Replace local pointers with global * InterruptController: Turn into a global pointer * System: Replace local pointers with global * Timers: Turn into a global instance * DMA: Turn into a global instance * SPU: Turn into a global instance * CDROM: Turn into a global instance * MDEC: Turn into a global instance * Pad: Turn into a global instance * SIO: Turn into a global instance * CDROM: Move audio FIFO to the heap * CPU/Recompiler: Drop ASMFunctions No longer needed since we have code in the same 4GB window. * CPUCodeCache: Turn class into namespace * Bus: Local pointer -> global pointers * CPU: Turn class into namespace * Bus: Turn into namespace * GTE: Store registers in CPU state struct Allows relative addressing on ARM. * CPU/Recompiler: Align code storage to page size * CPU/Recompiler: Fix relative branches on A64 * HostInterface: Local references to global * System: Turn into a namespace, move events out * Add guard pages * Android: Fix build
2020-07-31 07:09:18 +00:00
namespace CodeCache {
JIT optimizations and refactoring (#675) * CPU/Recompiler: Use rel32 call where possible for no-args * JitCodeBuffer: Support using preallocated buffer * CPU/Recompiler/AArch64: Use bl instead of blr for short branches * CPU/CodeCache: Allocate recompiler buffer in program space This means we don't need 64-bit moves for every call out of the recompiler. * GTE: Don't store as u16 and load as u32 * CPU/Recompiler: Add methods to emit global load/stores * GTE: Convert class to namespace * CPU/Recompiler: Call GTE functions directly * Settings: Turn into a global variable * GPU: Replace local pointers with global * InterruptController: Turn into a global pointer * System: Replace local pointers with global * Timers: Turn into a global instance * DMA: Turn into a global instance * SPU: Turn into a global instance * CDROM: Turn into a global instance * MDEC: Turn into a global instance * Pad: Turn into a global instance * SIO: Turn into a global instance * CDROM: Move audio FIFO to the heap * CPU/Recompiler: Drop ASMFunctions No longer needed since we have code in the same 4GB window. * CPUCodeCache: Turn class into namespace * Bus: Local pointer -> global pointers * CPU: Turn class into namespace * Bus: Turn into namespace * GTE: Store registers in CPU state struct Allows relative addressing on ARM. * CPU/Recompiler: Align code storage to page size * CPU/Recompiler: Fix relative branches on A64 * HostInterface: Local references to global * System: Turn into a namespace, move events out * Add guard pages * Android: Fix build
2020-07-31 07:09:18 +00:00
void Initialize(bool use_recompiler);
void Shutdown();
void Execute();
JIT optimizations and refactoring (#675) * CPU/Recompiler: Use rel32 call where possible for no-args * JitCodeBuffer: Support using preallocated buffer * CPU/Recompiler/AArch64: Use bl instead of blr for short branches * CPU/CodeCache: Allocate recompiler buffer in program space This means we don't need 64-bit moves for every call out of the recompiler. * GTE: Don't store as u16 and load as u32 * CPU/Recompiler: Add methods to emit global load/stores * GTE: Convert class to namespace * CPU/Recompiler: Call GTE functions directly * Settings: Turn into a global variable * GPU: Replace local pointers with global * InterruptController: Turn into a global pointer * System: Replace local pointers with global * Timers: Turn into a global instance * DMA: Turn into a global instance * SPU: Turn into a global instance * CDROM: Turn into a global instance * MDEC: Turn into a global instance * Pad: Turn into a global instance * SIO: Turn into a global instance * CDROM: Move audio FIFO to the heap * CPU/Recompiler: Drop ASMFunctions No longer needed since we have code in the same 4GB window. * CPUCodeCache: Turn class into namespace * Bus: Local pointer -> global pointers * CPU: Turn class into namespace * Bus: Turn into namespace * GTE: Store registers in CPU state struct Allows relative addressing on ARM. * CPU/Recompiler: Align code storage to page size * CPU/Recompiler: Fix relative branches on A64 * HostInterface: Local references to global * System: Turn into a namespace, move events out * Add guard pages * Android: Fix build
2020-07-31 07:09:18 +00:00
/// Flushes the code cache, forcing all blocks to be recompiled.
void Flush();
JIT optimizations and refactoring (#675) * CPU/Recompiler: Use rel32 call where possible for no-args * JitCodeBuffer: Support using preallocated buffer * CPU/Recompiler/AArch64: Use bl instead of blr for short branches * CPU/CodeCache: Allocate recompiler buffer in program space This means we don't need 64-bit moves for every call out of the recompiler. * GTE: Don't store as u16 and load as u32 * CPU/Recompiler: Add methods to emit global load/stores * GTE: Convert class to namespace * CPU/Recompiler: Call GTE functions directly * Settings: Turn into a global variable * GPU: Replace local pointers with global * InterruptController: Turn into a global pointer * System: Replace local pointers with global * Timers: Turn into a global instance * DMA: Turn into a global instance * SPU: Turn into a global instance * CDROM: Turn into a global instance * MDEC: Turn into a global instance * Pad: Turn into a global instance * SIO: Turn into a global instance * CDROM: Move audio FIFO to the heap * CPU/Recompiler: Drop ASMFunctions No longer needed since we have code in the same 4GB window. * CPUCodeCache: Turn class into namespace * Bus: Local pointer -> global pointers * CPU: Turn class into namespace * Bus: Turn into namespace * GTE: Store registers in CPU state struct Allows relative addressing on ARM. * CPU/Recompiler: Align code storage to page size * CPU/Recompiler: Fix relative branches on A64 * HostInterface: Local references to global * System: Turn into a namespace, move events out * Add guard pages * Android: Fix build
2020-07-31 07:09:18 +00:00
/// Changes whether the recompiler is enabled.
void SetUseRecompiler(bool enable);
JIT optimizations and refactoring (#675) * CPU/Recompiler: Use rel32 call where possible for no-args * JitCodeBuffer: Support using preallocated buffer * CPU/Recompiler/AArch64: Use bl instead of blr for short branches * CPU/CodeCache: Allocate recompiler buffer in program space This means we don't need 64-bit moves for every call out of the recompiler. * GTE: Don't store as u16 and load as u32 * CPU/Recompiler: Add methods to emit global load/stores * GTE: Convert class to namespace * CPU/Recompiler: Call GTE functions directly * Settings: Turn into a global variable * GPU: Replace local pointers with global * InterruptController: Turn into a global pointer * System: Replace local pointers with global * Timers: Turn into a global instance * DMA: Turn into a global instance * SPU: Turn into a global instance * CDROM: Turn into a global instance * MDEC: Turn into a global instance * Pad: Turn into a global instance * SIO: Turn into a global instance * CDROM: Move audio FIFO to the heap * CPU/Recompiler: Drop ASMFunctions No longer needed since we have code in the same 4GB window. * CPUCodeCache: Turn class into namespace * Bus: Local pointer -> global pointers * CPU: Turn class into namespace * Bus: Turn into namespace * GTE: Store registers in CPU state struct Allows relative addressing on ARM. * CPU/Recompiler: Align code storage to page size * CPU/Recompiler: Fix relative branches on A64 * HostInterface: Local references to global * System: Turn into a namespace, move events out * Add guard pages * Android: Fix build
2020-07-31 07:09:18 +00:00
/// Invalidates all blocks which are in the range of the specified code page.
void InvalidateBlocksWithPageIndex(u32 page_index);
JIT optimizations and refactoring (#675) * CPU/Recompiler: Use rel32 call where possible for no-args * JitCodeBuffer: Support using preallocated buffer * CPU/Recompiler/AArch64: Use bl instead of blr for short branches * CPU/CodeCache: Allocate recompiler buffer in program space This means we don't need 64-bit moves for every call out of the recompiler. * GTE: Don't store as u16 and load as u32 * CPU/Recompiler: Add methods to emit global load/stores * GTE: Convert class to namespace * CPU/Recompiler: Call GTE functions directly * Settings: Turn into a global variable * GPU: Replace local pointers with global * InterruptController: Turn into a global pointer * System: Replace local pointers with global * Timers: Turn into a global instance * DMA: Turn into a global instance * SPU: Turn into a global instance * CDROM: Turn into a global instance * MDEC: Turn into a global instance * Pad: Turn into a global instance * SIO: Turn into a global instance * CDROM: Move audio FIFO to the heap * CPU/Recompiler: Drop ASMFunctions No longer needed since we have code in the same 4GB window. * CPUCodeCache: Turn class into namespace * Bus: Local pointer -> global pointers * CPU: Turn class into namespace * Bus: Turn into namespace * GTE: Store registers in CPU state struct Allows relative addressing on ARM. * CPU/Recompiler: Align code storage to page size * CPU/Recompiler: Fix relative branches on A64 * HostInterface: Local references to global * System: Turn into a namespace, move events out * Add guard pages * Android: Fix build
2020-07-31 07:09:18 +00:00
void InterpretCachedBlock(const CodeBlock& block);
void InterpretUncachedBlock();
JIT optimizations and refactoring (#675) * CPU/Recompiler: Use rel32 call where possible for no-args * JitCodeBuffer: Support using preallocated buffer * CPU/Recompiler/AArch64: Use bl instead of blr for short branches * CPU/CodeCache: Allocate recompiler buffer in program space This means we don't need 64-bit moves for every call out of the recompiler. * GTE: Don't store as u16 and load as u32 * CPU/Recompiler: Add methods to emit global load/stores * GTE: Convert class to namespace * CPU/Recompiler: Call GTE functions directly * Settings: Turn into a global variable * GPU: Replace local pointers with global * InterruptController: Turn into a global pointer * System: Replace local pointers with global * Timers: Turn into a global instance * DMA: Turn into a global instance * SPU: Turn into a global instance * CDROM: Turn into a global instance * MDEC: Turn into a global instance * Pad: Turn into a global instance * SIO: Turn into a global instance * CDROM: Move audio FIFO to the heap * CPU/Recompiler: Drop ASMFunctions No longer needed since we have code in the same 4GB window. * CPUCodeCache: Turn class into namespace * Bus: Local pointer -> global pointers * CPU: Turn class into namespace * Bus: Turn into namespace * GTE: Store registers in CPU state struct Allows relative addressing on ARM. * CPU/Recompiler: Align code storage to page size * CPU/Recompiler: Fix relative branches on A64 * HostInterface: Local references to global * System: Turn into a namespace, move events out * Add guard pages * Android: Fix build
2020-07-31 07:09:18 +00:00
}; // namespace CodeCache
JIT optimizations and refactoring (#675) * CPU/Recompiler: Use rel32 call where possible for no-args * JitCodeBuffer: Support using preallocated buffer * CPU/Recompiler/AArch64: Use bl instead of blr for short branches * CPU/CodeCache: Allocate recompiler buffer in program space This means we don't need 64-bit moves for every call out of the recompiler. * GTE: Don't store as u16 and load as u32 * CPU/Recompiler: Add methods to emit global load/stores * GTE: Convert class to namespace * CPU/Recompiler: Call GTE functions directly * Settings: Turn into a global variable * GPU: Replace local pointers with global * InterruptController: Turn into a global pointer * System: Replace local pointers with global * Timers: Turn into a global instance * DMA: Turn into a global instance * SPU: Turn into a global instance * CDROM: Turn into a global instance * MDEC: Turn into a global instance * Pad: Turn into a global instance * SIO: Turn into a global instance * CDROM: Move audio FIFO to the heap * CPU/Recompiler: Drop ASMFunctions No longer needed since we have code in the same 4GB window. * CPUCodeCache: Turn class into namespace * Bus: Local pointer -> global pointers * CPU: Turn class into namespace * Bus: Turn into namespace * GTE: Store registers in CPU state struct Allows relative addressing on ARM. * CPU/Recompiler: Align code storage to page size * CPU/Recompiler: Fix relative branches on A64 * HostInterface: Local references to global * System: Turn into a namespace, move events out * Add guard pages * Android: Fix build
2020-07-31 07:09:18 +00:00
} // namespace CPU