fabi/Xenia-Canary
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Revert "Minor decoder optimizations, kernel fixes, cpu backend fixes"

Browse Source
This commit is contained in:
chrisps
2022-11-01 14:45:36 -07:00
committed by GitHub
parent 781871e2d5
commit 8186792113
41 changed files with 589 additions and 1002 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
src/xenia/cpu/ppc/ppc_decode_data.h
View File

@@ -46,7 +46,6 @@ struct PPCDecodeData {
uint32_t LEV() const { return bits_.LEV; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -75,7 +74,6 @@ struct PPCDecodeData {
uint32_t L() const { return bits_.RT & 0x1; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -97,7 +95,6 @@ struct PPCDecodeData {
int32_t ds() const { return static_cast<int32_t>(XEEXTS16(DS() << 2)); }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -177,7 +174,6 @@ struct PPCDecodeData {
uint32_t CRFS() const { return bits_.RA >> 2; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -204,7 +200,6 @@ struct PPCDecodeData {
uint32_t CRFS() const { return CRBA() >> 2; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -228,7 +223,6 @@ struct PPCDecodeData {
}
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -250,7 +244,6 @@ struct PPCDecodeData {
bool Rc() const { return bits_.Rc ? true : false; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -273,7 +266,6 @@ struct PPCDecodeData {
bool Rc() const { return bits_.Rc ? true : false; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -297,7 +289,6 @@ struct PPCDecodeData {
bool Rc() const { return bits_.Rc ? true : false; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -323,7 +314,6 @@ struct PPCDecodeData {
bool Rc() const { return bits_.Rc ? true : false; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -349,7 +339,6 @@ struct PPCDecodeData {
bool Rc() const { return bits_.Rc ? true : false; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -374,7 +363,6 @@ struct PPCDecodeData {
bool Rc() const { return bits_.Rc ? true : false; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -401,7 +389,6 @@ struct PPCDecodeData {
bool Rc() const { return bits_.Rc ? true : false; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -425,7 +412,6 @@ struct PPCDecodeData {
int32_t SIMM() const { return static_cast<int32_t>(XEEXTS16(VA())); }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -445,7 +431,6 @@ struct PPCDecodeData {
bool Rc() const { return bits_.Rc ? true : false; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -467,7 +452,6 @@ struct PPCDecodeData {
uint32_t SHB() const { return VC() & 0xF; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -489,7 +473,6 @@ struct PPCDecodeData {
uint32_t VB() const { return bits_.VB128l | (bits_.VB128h << 5); }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -515,7 +498,6 @@ struct PPCDecodeData {
uint32_t RB() const { return bits_.RB; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -539,7 +521,6 @@ struct PPCDecodeData {
uint32_t VC() const { return bits_.VC; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -565,7 +546,6 @@ struct PPCDecodeData {
int32_t SIMM() const { return static_cast<int32_t>(XEEXTS16(bits_.UIMM)); }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -587,7 +567,6 @@ struct PPCDecodeData {
uint32_t z() const { return bits_.z; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -613,7 +592,6 @@ struct PPCDecodeData {
uint32_t SH() const { return bits_.SH; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -640,7 +618,6 @@ struct PPCDecodeData {
bool Rc() const { return bits_.Rc ? true : false; }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;
@@ -665,7 +642,6 @@ struct PPCDecodeData {
uint32_t UIMM() const { return bits_.PERMl | (bits_.PERMh << 5); }
private:
XE_MAYBE_UNUSED
uint32_t address_;
union {
uint32_t value_;

6
src/xenia/cpu/ppc/ppc_emit_altivec.cc
View File

@@ -2014,7 +2014,8 @@ int InstrEmit_vupkhsh(PPCHIRBuilder& f, const InstrData& i) {
return InstrEmit_vupkhsh_(f, i.VX.VD, i.VX.VB);
}
int InstrEmit_vupkhsh128(PPCHIRBuilder& f, const InstrData& i) {
assert_zero(VX128_VA128);
uint32_t va = VX128_VA128;
assert_zero(va);
return InstrEmit_vupkhsh_(f, VX128_VD128, VX128_VB128);
}
@@ -2031,7 +2032,8 @@ int InstrEmit_vupklsh(PPCHIRBuilder& f, const InstrData& i) {
return InstrEmit_vupklsh_(f, i.VX.VD, i.VX.VB);
}
int InstrEmit_vupklsh128(PPCHIRBuilder& f, const InstrData& i) {
assert_zero(VX128_VA128);
uint32_t va = VX128_VA128;
assert_zero(va);
return InstrEmit_vupklsh_(f, VX128_VD128, VX128_VB128);
}

2
src/xenia/cpu/ppc/ppc_emit_memory.cc
View File

@@ -16,7 +16,7 @@
#include "xenia/cpu/ppc/ppc_hir_builder.h"
DEFINE_bool(
disable_prefetch_and_cachecontrol, true,
disable_prefetch_and_cachecontrol, false,
"Disables translating ppc prefetch/cache flush instructions to host "
"prefetch/cacheflush instructions. This may improve performance as these "
"instructions were written with the Xbox 360's cache in mind, and modern "

5
src/xenia/cpu/ppc/ppc_frontend.cc
View File

@@ -105,11 +105,6 @@ bool PPCFrontend::Initialize() {
}
bool PPCFrontend::DeclareFunction(GuestFunction* function) {
//chrispy: make sure we aren't declaring a function that is actually padding data, this will mess up PPCScanner and is hard to debug
//wow, this halo reach actually has branches into 0 opcodes, look into further
//xenia_assert(*reinterpret_cast<const uint32_t*>(
// this->memory()->TranslateVirtual(function->address())) != 0);
// Could scan or something here.
// Could also check to see if it's a well-known function type and classify
// for later.

60
src/xenia/cpu/ppc/ppc_hir_builder.cc
View File

@@ -34,11 +34,6 @@ DEFINE_bool(
"unimplemented PowerPC instruction is encountered.",
"CPU");
DEFINE_bool(
emit_useless_fpscr_updates, false,
"Emit useless fpscr update instructions (pre-10/30/2022 behavior). ",
"CPU");
namespace xe {
namespace cpu {
namespace ppc {
@@ -94,9 +89,6 @@ bool PPCHIRBuilder::Emit(GuestFunction* function, uint32_t flags) {
function_ = function;
start_address_ = function_->address();
//chrispy: i've seen this one happen, not sure why but i think from trying to precompile twice
//i've also seen ones with a start and end address that are the same...
assert_true(function_->address() <= function_->end_address());
instr_count_ = (function_->end_address() - function_->address()) / 4 + 1;
with_debug_info_ = (flags & EMIT_DEBUG_COMMENTS) == EMIT_DEBUG_COMMENTS;
@@ -250,7 +242,6 @@ void PPCHIRBuilder::MaybeBreakOnInstruction(uint32_t address) {
}
void PPCHIRBuilder::AnnotateLabel(uint32_t address, Label* label) {
//chrispy: label->name is unused, it would be nice to be able to remove the field and this code
char name_buffer[13];
auto format_result = fmt::format_to_n(name_buffer, 12, "loc_{:08X}", address);
name_buffer[format_result.size] = '\0';
@@ -456,38 +447,31 @@ void PPCHIRBuilder::StoreFPSCR(Value* value) {
void PPCHIRBuilder::UpdateFPSCR(Value* result, bool update_cr1) {
// TODO(benvanik): detect overflow and nan cases.
// fx and vx are the most important.
/*
chrispy: stubbed this out because right now all it does is waste
memory and CPU time
*/
if (cvars::emit_useless_fpscr_updates) {
Value* fx = LoadConstantInt8(0);
Value* fex = LoadConstantInt8(0);
Value* vx = LoadConstantInt8(0);
Value* ox = LoadConstantInt8(0);
Value* fx = LoadConstantInt8(0);
Value* fex = LoadConstantInt8(0);
Value* vx = LoadConstantInt8(0);
Value* ox = LoadConstantInt8(0);
if (update_cr1) {
// Store into the CR1 field.
// We do this instead of just calling CopyFPSCRToCR1 so that we don't
// have to read back the bits and do shifting work.
StoreContext(offsetof(PPCContext, cr1.cr1_fx), fx);
StoreContext(offsetof(PPCContext, cr1.cr1_fex), fex);
StoreContext(offsetof(PPCContext, cr1.cr1_vx), vx);
StoreContext(offsetof(PPCContext, cr1.cr1_ox), ox);
}
// Generate our new bits.
Value* new_bits = Shl(ZeroExtend(fx, INT32_TYPE), 31);
new_bits = Or(new_bits, Shl(ZeroExtend(fex, INT32_TYPE), 30));
new_bits = Or(new_bits, Shl(ZeroExtend(vx, INT32_TYPE), 29));
new_bits = Or(new_bits, Shl(ZeroExtend(ox, INT32_TYPE), 28));
// Mix into fpscr while preserving sticky bits (FX and OX).
Value* bits = LoadFPSCR();
bits = Or(And(bits, LoadConstantUint32(0x9FFFFFFF)), new_bits);
StoreFPSCR(bits);
if (update_cr1) {
// Store into the CR1 field.
// We do this instead of just calling CopyFPSCRToCR1 so that we don't
// have to read back the bits and do shifting work.
StoreContext(offsetof(PPCContext, cr1.cr1_fx), fx);
StoreContext(offsetof(PPCContext, cr1.cr1_fex), fex);
StoreContext(offsetof(PPCContext, cr1.cr1_vx), vx);
StoreContext(offsetof(PPCContext, cr1.cr1_ox), ox);
}
// Generate our new bits.
Value* new_bits = Shl(ZeroExtend(fx, INT32_TYPE), 31);
new_bits = Or(new_bits, Shl(ZeroExtend(fex, INT32_TYPE), 30));
new_bits = Or(new_bits, Shl(ZeroExtend(vx, INT32_TYPE), 29));
new_bits = Or(new_bits, Shl(ZeroExtend(ox, INT32_TYPE), 28));
// Mix into fpscr while preserving sticky bits (FX and OX).
Value* bits = LoadFPSCR();
bits = Or(And(bits, LoadConstantUint32(0x9FFFFFFF)), new_bits);
StoreFPSCR(bits);
}
void PPCHIRBuilder::CopyFPSCRToCR1() {

16
src/xenia/cpu/ppc/ppc_instr.h
View File

@@ -21,7 +21,13 @@ namespace xe {
namespace cpu {
namespace ppc {
struct PPCOpcodeBits {
// DEPRECATED
// TODO(benvanik): move code to PPCDecodeData.
struct InstrData {
PPCOpcode opcode;
const PPCOpcodeInfo* opcode_info;
uint32_t address;
union {
uint32_t code;
@@ -323,14 +329,6 @@ struct PPCOpcodeBits {
};
};
// DEPRECATED
// TODO(benvanik): move code to PPCDecodeData.
struct InstrData : public PPCOpcodeBits {
PPCOpcode opcode;
const PPCOpcodeInfo* opcode_info;
uint32_t address;
};
} // namespace ppc
} // namespace cpu
} // namespace xe