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

use Sleep(0) instead of SwitchToThread, should waste less power and help the os with scheduling.

Browse Source 
PM4 buffer handling made a virtual member of commandprocessor, place the implementation/declaration into reusable macro files. this is probably the biggest boost here.
Optimized SET_CONSTANT/ LOAD_CONSTANT pm4 ops based on the register range they start writing at, this was also a nice boost

Expose X64 extension flags to code outside of x64 backend, so we can detect and use things like avx512, xop, avx2, etc in normal code
Add freelists for HIR structures to try to reduce the number of last level cache misses during optimization (currently disabled... fixme later)

Analyzed PGO feedback and reordered branches, uninlined functions, moved code out into different functions based on info from it in the PM4 functions, this gave like a 2% boost at best.

Added support for the db16cyc opcode, which is used often in xb360 spinlocks. before it was just being translated to nop, now on x64 we translate it to _mm_pause but may change that in the future to reduce cpu time wasted

texture util - all our divisors were powers of 2, instead we look up a shift. this made texture scaling slightly faster, more so on intel processors which seem to be worse at int divs. GetGuestTextureLayout is now a little faster, although it is still one of the heaviest functions in the emulator when scaling is on.

xe_unlikely_mutex was not a good choice for the guest clock lock, (running theory) on intel processors another thread may take a significant time to update the clock? maybe because of the uint64 division? really not sure, but switched it to xe_mutex. This fixed audio stutter that i had introduced to 1 or 2 games, fixed performance on that n64 rare game with the monkeys.
Took another crack at DMA implementation, another failure.
Instead of passing as a parameter, keep the ringbuffer reader as the first member of commandprocessor so it can be accessed through this
Added macro for noalias
Applied noalias to Memory::LookupHeap. This reduced the size of the executable by 7 kb.
Reworked kernel shim template, this shaved like 100kb off the exe and eliminated the indirect calls from the shim to the actual implementation. We still unconditionally generate string representations of kernel calls though :(, unless it is kHighFrequency

Add nvapi extensions support, currently unused. Will use CPUVISIBLE memory at some point
Inserted prefetches in a few places based on feedback from vtune.
Add native implementation of SHA int8 if all elements are the same

Vectorized comparisons for SetViewport, SetScissorRect
Vectorized ranged comparisons for WriteRegister
Add XE_MSVC_ASSUME
Move FormatInfo::name out of the structure, instead look up the name in a different table. Debug related data and critical runtime data are best kept apart
Templated UpdateSystemConstantValues based on ROV/RTV and primitive_polygonal
Add ArchFloatMask functions, these are for storing the results of floating point comparisons without doing costly float->int pipeline transfers (vucomiss/setb)
Use floatmasks in UpdateSystemConstantValues for checking if dirty, only transfer to int at end of function.
Instead of dirty |= (x == y) in UpdateSystemConstantValues, now we do dirty_u32 |= (x^y). if any of them are not equal, dirty_u32 will be nz, else if theyre all equal it will be zero. This is more friendly to register renaming and the lack of dependencies on EFLAGS lets the compiler reorder better
Add PrefetchSamplerParameters to D3D12TextureCache
use PrefetchSamplerParameters in UpdateBindings to eliminate cache misses that vtune detected

Add PrefetchTextureBinding to D3D12TextureCache
Prefetch texture bindings to get rid of more misses vtune detected (more accesses out of order with random strides)
Rewrote DMAC, still terrible though and have disabled it for now.
Replace tiny memcmp of 6 U64 in render_target_cache with inline loop, msvc fails to make it a loop and instead does a thunk to their memcmp function, which is optimized for larger sizes

PrefetchTextureBinding in AreActiveTextureSRVKeysUpToDate
Replace memcmp calls for pipelinedescription with handwritten cmp
Directly write some registers that dont have special handling in PM4 functions
Changed EstimateMaxY to try to eliminate mispredictions that vtune was reporting, msvc ended up turning the changed code into a series of blends

in ExecutePacketType3_EVENT_WRITE_EXT, instead of writing extents to an array on the stack and then doing xe_copy_and_swap_16 of the data to its dest, pre-swap each constant and then store those. msvc manages to unroll that into wider stores
stop logging XE_SWAP every time we receive XE_SWAP, stop logging the start and end of each viz query

Prefetch watch nodes in FireWatches based on feedback from vtune
Removed dead code from texture_info.cc
NOINLINE on GpuSwap, PGO builds did it so we should too.
This commit is contained in:
chss95cs@gmail.com
2022-09-11 14:14:48 -07:00
parent 9a6dd4cd6f
commit 20638c2e61
67 changed files with 3369 additions and 2184 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
src/xenia/cpu/backend/x64/x64_backend.cc
View File

@@ -26,24 +26,6 @@
#include "xenia/cpu/processor.h"
#include "xenia/cpu/stack_walker.h"
#include "xenia/cpu/xex_module.h"
DEFINE_int32(x64_extension_mask, -1,
"Allow the detection and utilization of specific instruction set "
"features.\n"
" 0 = x86_64 + AVX1\n"
" 1 = AVX2\n"
" 2 = FMA\n"
" 4 = LZCNT\n"
" 8 = BMI1\n"
" 16 = BMI2\n"
" 32 = F16C\n"
" 64 = Movbe\n"
" 128 = GFNI\n"
" 256 = AVX512F\n"
" 512 = AVX512VL\n"
" 1024 = AVX512BW\n"
" 2048 = AVX512DQ\n"
" -1 = Detect and utilize all possible processor features\n",
"x64");
DEFINE_bool(record_mmio_access_exceptions, true,
"For guest addresses records whether we caught any mmio accesses "

10
src/xenia/cpu/backend/x64/x64_emitter.cc
View File

@@ -103,7 +103,9 @@ X64Emitter::X64Emitter(X64Backend* backend, XbyakAllocator* allocator)
"FAQ for system requirements at https://xenia.jp");
return;
}
#if 1
feature_flags_ = amd64::GetFeatureFlags();
#else
#define TEST_EMIT_FEATURE(emit, ext) \
if ((cvars::x64_extension_mask & emit) == emit) { \
feature_flags_ |= (cpu_.has(ext) ? emit : 0); \
@@ -168,6 +170,7 @@ X64Emitter::X64Emitter(X64Backend* backend, XbyakAllocator* allocator)
feature_flags_ |= kX64FlagsIndependentVars;
}
}
#endif
may_use_membase32_as_zero_reg_ =
static_cast<uint32_t>(reinterpret_cast<uintptr_t>(
processor()->memory()->virtual_membase())) == 0;
@@ -913,6 +916,8 @@ static inline vec128_t v128_setr_bytes(unsigned char v0, unsigned char v1,
static const vec128_t xmm_consts[] = {
/* XMMZero */ vec128f(0.0f),
/* XMMByteSwapMask */
vec128i(0x00010203u, 0x04050607u, 0x08090A0Bu, 0x0C0D0E0Fu),
/* XMMOne */ vec128f(1.0f),
/* XMMOnePD */ vec128d(1.0),
/* XMMNegativeOne */ vec128f(-1.0f, -1.0f, -1.0f, -1.0f),
@@ -937,8 +942,7 @@ static const vec128_t xmm_consts[] = {
vec128i(0x7FFFFFFFu, 0x7FFFFFFFu, 0x7FFFFFFFu, 0x7FFFFFFFu),
/* XMMAbsMaskPD */
vec128i(0xFFFFFFFFu, 0x7FFFFFFFu, 0xFFFFFFFFu, 0x7FFFFFFFu),
/* XMMByteSwapMask */
vec128i(0x00010203u, 0x04050607u, 0x08090A0Bu, 0x0C0D0E0Fu),
/* XMMByteOrderMask */
vec128i(0x01000302u, 0x05040706u, 0x09080B0Au, 0x0D0C0F0Eu),
/* XMMPermuteControl15 */ vec128b(15),

41
src/xenia/cpu/backend/x64/x64_emitter.h
View File

@@ -34,7 +34,7 @@ namespace xe {
namespace cpu {
namespace backend {
namespace x64 {
using namespace amd64;
class X64Backend;
class X64CodeCache;
@@ -81,6 +81,7 @@ static SimdDomain PickDomain2(SimdDomain dom1, SimdDomain dom2) {
}
enum XmmConst {
XMMZero = 0,
XMMByteSwapMask,
XMMOne,
XMMOnePD,
XMMNegativeOne,
@@ -97,7 +98,7 @@ enum XmmConst {
XMMSignMaskPD,
XMMAbsMaskPS,
XMMAbsMaskPD,
XMMByteSwapMask,
XMMByteOrderMask,
XMMPermuteControl15,
XMMPermuteByteMask,
@@ -189,42 +190,6 @@ class XbyakAllocator : public Xbyak::Allocator {
virtual bool useProtect() const { return false; }
};
enum X64EmitterFeatureFlags {
kX64EmitAVX2 = 1 << 0,
kX64EmitFMA = 1 << 1,
kX64EmitLZCNT = 1 << 2, // this is actually ABM and includes popcount
kX64EmitBMI1 = 1 << 3,
kX64EmitBMI2 = 1 << 4,
kX64EmitPrefetchW = 1 << 5,
kX64EmitMovbe = 1 << 6,
kX64EmitGFNI = 1 << 7,
kX64EmitAVX512F = 1 << 8,
kX64EmitAVX512VL = 1 << 9,
kX64EmitAVX512BW = 1 << 10,
kX64EmitAVX512DQ = 1 << 11,
kX64EmitAVX512Ortho = kX64EmitAVX512F | kX64EmitAVX512VL,
kX64EmitAVX512Ortho64 = kX64EmitAVX512Ortho | kX64EmitAVX512DQ,
kX64FastJrcx = 1 << 12, // jrcxz is as fast as any other jump ( >= Zen1)
kX64FastLoop =
1 << 13, // loop/loope/loopne is as fast as any other jump ( >= Zen2)
kX64EmitAVX512VBMI = 1 << 14,
kX64FlagsIndependentVars =
1 << 15, // if true, instructions that only modify some flags (like
// inc/dec) do not introduce false dependencies on EFLAGS
// because the individual flags are treated as different vars by
// the processor. (this applies to zen)
kX64EmitXOP = 1 << 16, // chrispy: xop maps really well to many vmx
// instructions, and FX users need the boost
kX64EmitFMA4 = 1 << 17, // todo: also use on zen1?
kX64EmitTBM = 1 << 18,
// kX64XMMRegisterMergeOptimization = 1 << 19, //section 2.11.5, amd family
// 17h/19h optimization manuals. allows us to save 1 byte on certain xmm
// instructions by using the legacy sse version if we recently cleared the
// high 128 bits of the
};
class ResolvableGuestCall {
public:
bool is_jump_;

31
src/xenia/cpu/backend/x64/x64_seq_vector.cc
View File

@@ -1354,15 +1354,17 @@ struct VECTOR_SHA_V128
static void EmitInt8(X64Emitter& e, const EmitArgType& i) {
// TODO(benvanik): native version (with shift magic).
if (i.src2.is_constant) {
if (e.IsFeatureEnabled(kX64EmitGFNI)) {
const auto& shamt = i.src2.constant();
bool all_same = true;
for (size_t n = 0; n < 16 - n; ++n) {
if (shamt.u8[n] != shamt.u8[n + 1]) {
all_same = false;
break;
}
const auto& shamt = i.src2.constant();
bool all_same = true;
for (size_t n = 0; n < 16 - n; ++n) {
if (shamt.u8[n] != shamt.u8[n + 1]) {
all_same = false;
break;
}
}
if (e.IsFeatureEnabled(kX64EmitGFNI)) {
if (all_same) {
// Every count is the same, so we can use gf2p8affineqb.
const uint8_t shift_amount = shamt.u8[0] & 0b111;
@@ -1375,6 +1377,19 @@ struct VECTOR_SHA_V128
return;
}
}
else if (all_same) {
Xmm to_be_shifted = GetInputRegOrConstant(e, i.src1, e.xmm1);
e.vpmovsxbw(e.xmm0, to_be_shifted); //_mm_srai_epi16 / psraw
e.vpunpckhqdq(e.xmm2, to_be_shifted, to_be_shifted);
e.vpmovsxbw(e.xmm1, e.xmm2);
e.vpsraw(e.xmm0, shamt.u8[0]);
e.vpsraw(e.xmm1, shamt.u8[0]);
e.vpacksswb(i.dest, e.xmm0, e.xmm1);
return;
}
e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));
} else {
e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));

12
src/xenia/cpu/backend/x64/x64_sequences.cc
View File

@@ -3234,7 +3234,17 @@ struct SET_ROUNDING_MODE_I32
}
};
EMITTER_OPCODE_TABLE(OPCODE_SET_ROUNDING_MODE, SET_ROUNDING_MODE_I32);
// ============================================================================
// OPCODE_DELAY_EXECUTION
// ============================================================================
struct DELAY_EXECUTION
: Sequence<DELAY_EXECUTION, I<OPCODE_DELAY_EXECUTION, VoidOp>> {
static void Emit(X64Emitter& e, const EmitArgType& i) {
// todo: what if they dont have smt?
e.pause();
}
};
EMITTER_OPCODE_TABLE(OPCODE_DELAY_EXECUTION, DELAY_EXECUTION);
// Include anchors to other sequence sources so they get included in the build.
extern volatile int anchor_control;
static int anchor_control_dest = anchor_control;

152
src/xenia/cpu/compiler/passes/constant_propagation_pass.cc
View File

@@ -98,7 +98,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstantTrue()) {
i->Replace(&OPCODE_DEBUG_BREAK_info, i->flags);
} else {
i->Remove();
i->UnlinkAndNOP();
}
result = true;
}
@@ -109,7 +109,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstantTrue()) {
i->Replace(&OPCODE_TRAP_info, i->flags);
} else {
i->Remove();
i->UnlinkAndNOP();
}
result = true;
}
@@ -122,7 +122,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
i->Replace(&OPCODE_CALL_info, i->flags);
i->src1.symbol = symbol;
} else {
i->Remove();
i->UnlinkAndNOP();
}
result = true;
}
@@ -146,7 +146,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
i->Replace(&OPCODE_CALL_INDIRECT_info, i->flags);
i->set_src1(value);
} else {
i->Remove();
i->UnlinkAndNOP();
}
result = true;
} else if (i->src2.value->IsConstant()) { // chrispy: fix h3 bug from
@@ -172,7 +172,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
i->Replace(&OPCODE_BRANCH_info, i->flags);
i->src1.label = label;
} else {
i->Remove();
i->UnlinkAndNOP();
}
result = true;
}
@@ -184,7 +184,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
i->Replace(&OPCODE_BRANCH_info, i->flags);
i->src1.label = label;
} else {
i->Remove();
i->UnlinkAndNOP();
}
result = true;
}
@@ -195,7 +195,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
TypeName target_type = v->type;
v->set_from(i->src1.value);
v->Cast(target_type);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -204,7 +204,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
TypeName target_type = v->type;
v->set_from(i->src1.value);
v->Convert(target_type, RoundMode(i->flags));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -212,7 +212,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_from(i->src1.value);
v->Round(RoundMode(i->flags));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -221,7 +221,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
TypeName target_type = v->type;
v->set_from(i->src1.value);
v->ZeroExtend(target_type);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -230,7 +230,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
TypeName target_type = v->type;
v->set_from(i->src1.value);
v->SignExtend(target_type);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -239,7 +239,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
TypeName target_type = v->type;
v->set_from(i->src1.value);
v->Truncate(target_type);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -247,7 +247,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
if (!(i->src1.value->AsUint32() & 0xF)) {
v->set_zero(VEC128_TYPE);
i->Remove();
i->UnlinkAndNOP();
result = true;
break;
}
@@ -281,22 +281,22 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
switch (v->type) {
case INT8_TYPE:
v->set_constant(xe::load<uint8_t>(host_addr));
i->Remove();
i->UnlinkAndNOP();
result = true;
break;
case INT16_TYPE:
v->set_constant(xe::load<uint16_t>(host_addr));
i->Remove();
i->UnlinkAndNOP();
result = true;
break;
case INT32_TYPE:
v->set_constant(xe::load<uint32_t>(host_addr));
i->Remove();
i->UnlinkAndNOP();
result = true;
break;
case INT64_TYPE:
v->set_constant(xe::load<uint64_t>(host_addr));
i->Remove();
i->UnlinkAndNOP();
result = true;
break;
case VEC128_TYPE:
@@ -304,7 +304,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
val.low = xe::load<uint64_t>(host_addr);
val.high = xe::load<uint64_t>(host_addr + 8);
v->set_constant(val);
i->Remove();
i->UnlinkAndNOP();
result = true;
break;
default:
@@ -357,14 +357,14 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
i->src3.value->IsConstant()) {
v->set_from(i->src2.value);
v->Select(i->src3.value, i->src1.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
} else {
if (i->src2.value->IsConstant() && i->src3.value->IsConstant()) {
v->set_from(i->src2.value);
v->Select(i->src3.value, i->src1.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
}
@@ -381,7 +381,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
} else {
v->set_constant(uint8_t(0));
}
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -391,7 +391,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
bool value = i->src1.value->IsConstantEQ(i->src2.value);
i->dest->set_constant(uint8_t(value));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -399,7 +399,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
bool value = i->src1.value->IsConstantNE(i->src2.value);
i->dest->set_constant(uint8_t(value));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -407,7 +407,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
bool value = i->src1.value->IsConstantSLT(i->src2.value);
i->dest->set_constant(uint8_t(value));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -415,7 +415,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
bool value = i->src1.value->IsConstantSLE(i->src2.value);
i->dest->set_constant(uint8_t(value));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -423,7 +423,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
bool value = i->src1.value->IsConstantSGT(i->src2.value);
i->dest->set_constant(uint8_t(value));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -431,7 +431,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
bool value = i->src1.value->IsConstantSGE(i->src2.value);
i->dest->set_constant(uint8_t(value));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -439,7 +439,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
bool value = i->src1.value->IsConstantULT(i->src2.value);
i->dest->set_constant(uint8_t(value));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -447,7 +447,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
bool value = i->src1.value->IsConstantULE(i->src2.value);
i->dest->set_constant(uint8_t(value));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -455,7 +455,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
bool value = i->src1.value->IsConstantUGT(i->src2.value);
i->dest->set_constant(uint8_t(value));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -463,7 +463,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
bool value = i->src1.value->IsConstantUGE(i->src2.value);
i->dest->set_constant(uint8_t(value));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -477,7 +477,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
!should_skip_because_of_float) {
v->set_from(i->src1.value);
v->Add(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -489,7 +489,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
TypeName target_type = v->type;
v->set_from(ca);
v->ZeroExtend(target_type);
i->Remove();
i->UnlinkAndNOP();
} else {
if (i->dest->type == ca->type) {
i->Replace(&OPCODE_ASSIGN_info, 0);
@@ -507,7 +507,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
!should_skip_because_of_float) {
v->set_from(i->src1.value);
v->Sub(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -516,7 +516,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->Mul(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
} else if (i->src1.value->IsConstant() ||
i->src2.value->IsConstant()) {
@@ -548,7 +548,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->MulHi(i->src2.value, (i->flags & ARITHMETIC_UNSIGNED) != 0);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -557,13 +557,13 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->Div(i->src2.value, (i->flags & ARITHMETIC_UNSIGNED) != 0);
i->Remove();
i->UnlinkAndNOP();
result = true;
} else if (!i->src2.value->MaybeFloaty() &&
i->src2.value->IsConstantZero()) {
// division by 0 == 0 every time,
v->set_zero(i->src2.value->type);
i->Remove();
i->UnlinkAndNOP();
result = true;
} else if (i->src2.value->IsConstant()) {
// Division by one = no-op.
@@ -592,7 +592,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
}
v->set_from(i->src1.value);
v->Max(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -600,7 +600,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_from(i->src1.value);
v->Neg();
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -608,7 +608,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_from(i->src1.value);
v->Abs();
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -616,7 +616,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_from(i->src1.value);
v->Sqrt();
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -624,7 +624,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_from(i->src1.value);
v->RSqrt();
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -632,7 +632,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_from(i->src1.value);
v->Recip();
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -640,7 +640,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->And(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -648,7 +648,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->AndNot(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -656,7 +656,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->Or(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -664,13 +664,13 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->Xor(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
} else if (!i->src1.value->IsConstant() &&
!i->src2.value->IsConstant() &&
i->src1.value == i->src2.value) {
v->set_zero(v->type);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -678,7 +678,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_from(i->src1.value);
v->Not();
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -687,7 +687,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->Shl(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
} else if (i->src2.value->IsConstantZero()) {
auto src1 = i->src1.value;
@@ -702,7 +702,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->Shr(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
} else if (i->src2.value->IsConstantZero()) {
auto src1 = i->src1.value;
@@ -716,7 +716,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->Sha(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -724,7 +724,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->RotateLeft(i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -732,7 +732,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_from(i->src1.value);
v->ByteSwap();
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -740,7 +740,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_zero(v->type);
v->CountLeadingZeros(i->src1.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -751,7 +751,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
(i->flags == INT8_TYPE || i->flags == INT16_TYPE)) {
v->set_from(i->src1.value);
v->Permute(i->src2.value, i->src3.value, (TypeName)i->flags);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
@@ -765,7 +765,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
*/
v->set_zero(VEC128_TYPE);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
@@ -777,7 +777,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
i->src3.value->IsConstant()) {
v->set_from(i->src1.value);
v->Insert(i->src2.value, i->src3.value, (TypeName)i->flags);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -785,7 +785,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_from(i->src1.value);
v->Swizzle((uint32_t)i->src2.offset, (TypeName)i->flags);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -793,7 +793,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_zero(v->type);
v->Extract(i->src1.value, i->src2.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -801,7 +801,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_zero(v->type);
v->Splat(i->src1.value);
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -809,7 +809,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->VectorCompareEQ(i->src2.value, hir::TypeName(i->flags));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -817,7 +817,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->VectorCompareSGT(i->src2.value, hir::TypeName(i->flags));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -825,7 +825,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->VectorCompareSGE(i->src2.value, hir::TypeName(i->flags));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -833,7 +833,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->VectorCompareUGT(i->src2.value, hir::TypeName(i->flags));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -841,7 +841,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->VectorCompareUGE(i->src2.value, hir::TypeName(i->flags));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -850,7 +850,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
v->set_zero(VEC128_TYPE);
v->VectorConvertF2I(i->src1.value,
!!(i->flags & ARITHMETIC_UNSIGNED));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -859,7 +859,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
v->set_zero(VEC128_TYPE);
v->VectorConvertI2F(i->src1.value,
!!(i->flags & ARITHMETIC_UNSIGNED));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -867,7 +867,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->VectorShl(i->src2.value, hir::TypeName(i->flags));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -875,7 +875,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->VectorShr(i->src2.value, hir::TypeName(i->flags));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -883,7 +883,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {
v->set_from(i->src1.value);
v->VectorRol(i->src2.value, hir::TypeName(i->flags));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -894,7 +894,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
v->VectorAdd(i->src2.value, hir::TypeName(i->flags & 0xFF),
!!(arith_flags & ARITHMETIC_UNSIGNED),
!!(arith_flags & ARITHMETIC_SATURATE));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -905,7 +905,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
v->VectorSub(i->src2.value, hir::TypeName(i->flags & 0xFF),
!!(arith_flags & ARITHMETIC_UNSIGNED),
!!(arith_flags & ARITHMETIC_SATURATE));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -917,7 +917,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
v->VectorAverage(i->src2.value, hir::TypeName(i->flags & 0xFF),
!!(arith_flags & ARITHMETIC_UNSIGNED),
!!(arith_flags & ARITHMETIC_SATURATE));
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;
@@ -926,7 +926,7 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {
if (i->src1.value->IsConstant()) {
v->set_from(i->src1.value);
v->DenormalFlush();
i->Remove();
i->UnlinkAndNOP();
result = true;
}
break;

2
src/xenia/cpu/compiler/passes/context_promotion_pass.cc
View File

@@ -146,7 +146,7 @@ void ContextPromotionPass::RemoveDeadStoresBlock(Block* block) {
validity.set(static_cast<uint32_t>(offset));
} else {
// Already written to. Remove this store.
i->Remove();
i->UnlinkAndNOP();
}
}
i = prev;

18
src/xenia/cpu/compiler/passes/dead_code_elimination_pass.cc
View File

@@ -120,7 +120,8 @@ bool DeadCodeEliminationPass::Run(HIRBuilder* builder) {
Instr* next = i->next;
if (i->opcode == &OPCODE_NOP_info) {
// Nop - remove!
i->Remove();
i->UnlinkAndNOP();
i->Deallocate();
}
i = next;
}
@@ -148,7 +149,9 @@ bool DeadCodeEliminationPass::Run(HIRBuilder* builder) {
void DeadCodeEliminationPass::MakeNopRecursive(Instr* i) {
i->opcode = &hir::OPCODE_NOP_info;
i->dest->def = NULL;
if (i->dest) {
i->dest->def = NULL;
}
i->dest = NULL;
#define MAKE_NOP_SRC(n) \
@@ -163,7 +166,9 @@ void DeadCodeEliminationPass::MakeNopRecursive(Instr* i) {
if (value->def && value->def != i) { \
MakeNopRecursive(value->def); \
} \
HIRBuilder::GetCurrent()->DeallocateValue(value); \
} \
HIRBuilder::GetCurrent()->DeallocateUse(use); \
}
MAKE_NOP_SRC(1);
MAKE_NOP_SRC(2);
@@ -189,7 +194,8 @@ void DeadCodeEliminationPass::ReplaceAssignment(Instr* i) {
use = use->next;
}
i->Remove();
i->UnlinkAndNOP();
i->Deallocate();
}
bool DeadCodeEliminationPass::CheckLocalUse(Instr* i) {
@@ -204,11 +210,11 @@ bool DeadCodeEliminationPass::CheckLocalUse(Instr* i) {
}
// Load/store are paired. They can both be removed.
use_instr->Remove();
use_instr->UnlinkAndNOP();
}
i->Remove();
i->UnlinkAndNOP();
i->Deallocate();
return false;
}

2
src/xenia/cpu/compiler/passes/finalization_pass.cc
View File

@@ -61,7 +61,7 @@ bool FinalizationPass::Run(HIRBuilder* builder) {
auto target = tail->src1.label;
if (target->block == block->next) {
// Jumping to subsequent block. Remove.
tail->Remove();
tail->UnlinkAndNOP();
}
}

57
src/xenia/cpu/hir/hir_builder.cc
View File

@@ -46,15 +46,27 @@ namespace hir {
(value->type) == FLOAT64_TYPE || (value->type) == VEC128_TYPE)
#define ASSERT_TYPES_EQUAL(value1, value2) \
assert_true((value1->type) == (value2->type))
thread_local HIRBuilder* thrd_current_hirfunction = nullptr;
HIRBuilder::HIRBuilder() {
arena_ = new Arena();
Reset();
}
HIRBuilder* HIRBuilder::GetCurrent() { return thrd_current_hirfunction; }
void HIRBuilder::MakeCurrent() { thrd_current_hirfunction = this; }
void HIRBuilder::RemoveCurrent() {
if (thrd_current_hirfunction == this) {
thrd_current_hirfunction = nullptr;
}
}
HIRBuilder::~HIRBuilder() {
Reset();
delete arena_;
if (thrd_current_hirfunction == this) {
thrd_current_hirfunction = nullptr;
}
}
void HIRBuilder::Reset() {
@@ -105,7 +117,37 @@ bool HIRBuilder::Finalize() {
}
return true;
}
Instr* HIRBuilder::AllocateInstruction() {
Instr* result = free_instrs_.NewEntry();
if (result) {
return result;
}
return arena()->Alloc<Instr>();
}
Value* HIRBuilder::AllocateValue() {
Value* result = free_values_.NewEntry();
if (result) {
return result;
}
return arena()->Alloc<Value>();
}
Value::Use* HIRBuilder::AllocateUse() {
Value::Use* result = free_uses_.NewEntry();
if (result) {
return result;
}
return arena()->Alloc<Value::Use>();
}
void HIRBuilder::DeallocateInstruction(Instr* instr) {
// free_instrs_.DeleteEntry(instr);
}
void HIRBuilder::DeallocateValue(Value* value) {
// free_values_.DeleteEntry(value);
}
void HIRBuilder::DeallocateUse(Value::Use* use) {
// free_uses_.DeleteEntry(use);
}
void HIRBuilder::DumpValue(StringBuffer* str, Value* value) {
if (value->IsConstant()) {
switch (value->type) {
@@ -545,12 +587,12 @@ void HIRBuilder::MergeAdjacentBlocks(Block* left, Block* right) {
auto sig = left->instr_tail->opcode->signature;
if (GET_OPCODE_SIG_TYPE_SRC1(sig) == OPCODE_SIG_TYPE_L) {
if (left->instr_tail->src1.label->block == right) {
left->instr_tail->Remove();
left->instr_tail->UnlinkAndNOP();
}
}
if (GET_OPCODE_SIG_TYPE_SRC2(sig) == OPCODE_SIG_TYPE_L) {
if (left->instr_tail->src2.label->block == right) {
left->instr_tail->Remove();
left->instr_tail->UnlinkAndNOP();
}
}
}
@@ -678,7 +720,7 @@ Instr* HIRBuilder::AppendInstr(const OpcodeInfo& opcode_info, uint16_t flags,
}
Block* block = current_block_;
Instr* instr = arena_->Alloc<Instr>();
Instr* instr = AllocateInstruction();
instr->next = NULL;
instr->prev = block->instr_tail;
if (block->instr_tail) {
@@ -705,7 +747,7 @@ Instr* HIRBuilder::AppendInstr(const OpcodeInfo& opcode_info, uint16_t flags,
}
Value* HIRBuilder::AllocValue(TypeName type) {
Value* value = arena_->Alloc<Value>();
Value* value = AllocateValue();
value->ordinal = next_value_ordinal_++;
value->type = type;
value->flags = 0;
@@ -719,7 +761,7 @@ Value* HIRBuilder::AllocValue(TypeName type) {
}
Value* HIRBuilder::CloneValue(Value* source) {
Value* value = arena_->Alloc<Value>();
Value* value = AllocateValue();
value->ordinal = next_value_ordinal_++;
value->type = source->type;
value->flags = source->flags;
@@ -1295,6 +1337,9 @@ void HIRBuilder::CacheControl(Value* address, size_t cache_line_size,
void HIRBuilder::MemoryBarrier() { AppendInstr(OPCODE_MEMORY_BARRIER_info, 0); }
void HIRBuilder::DelayExecution() {
AppendInstr(OPCODE_DELAY_EXECUTION_info, 0);
}
void HIRBuilder::SetRoundingMode(Value* value) {
ASSERT_INTEGER_TYPE(value);
Instr* i = AppendInstr(OPCODE_SET_ROUNDING_MODE_info, 0);

25
src/xenia/cpu/hir/hir_builder.h
View File

@@ -15,6 +15,8 @@
#include "third_party/fmt/include/fmt/format.h"
#include "xenia/base/arena.h"
#include "xenia/base/string_buffer.h"
#include "xenia/base/simple_freelist.h"
#include "xenia/cpu/hir/block.h"
#include "xenia/cpu/hir/instr.h"
#include "xenia/cpu/hir/label.h"
@@ -31,11 +33,20 @@ enum FunctionAttributes {
};
class HIRBuilder {
SimpleFreelist<Instr> free_instrs_;
SimpleFreelist<Value> free_values_;
SimpleFreelist<Value::Use> free_uses_;
public:
HIRBuilder();
virtual ~HIRBuilder();
static HIRBuilder* GetCurrent();
void MakeCurrent();
void RemoveCurrent();
virtual void Reset();
virtual bool Finalize();
void Dump(StringBuffer* str);
@@ -66,6 +77,18 @@ class HIRBuilder {
void RemoveBlock(Block* block);
void MergeAdjacentBlocks(Block* left, Block* right);
Instr* AllocateInstruction();
Value* AllocateValue();
Value::Use* AllocateUse();
void DeallocateInstruction(Instr* instr);
void DeallocateValue(Value* value);
void DeallocateUse(Value::Use* use);
void ResetPools() {
free_instrs_.Reset();
free_uses_.Reset();
free_values_.Reset();
}
// static allocations:
// Value* AllocStatic(size_t length);
@@ -176,7 +199,7 @@ class HIRBuilder {
void CacheControl(Value* address, size_t cache_line_size,
CacheControlType type);
void MemoryBarrier();
void DelayExecution();
void SetRoundingMode(Value* value);
Value* Max(Value* value1, Value* value2);
Value* VectorMax(Value* value1, Value* value2, TypeName part_type,

28
src/xenia/cpu/hir/instr.cc
View File

@@ -10,7 +10,7 @@
#include "xenia/cpu/hir/instr.h"
#include "xenia/cpu/hir/block.h"
#include "xenia/cpu/hir/hir_builder.h"
namespace xe {
namespace cpu {
namespace hir {
@@ -62,21 +62,35 @@ void Instr::Replace(const OpcodeInfo* new_opcode, uint16_t new_flags) {
if (src1_use) {
src1.value->RemoveUse(src1_use);
src1.value = NULL;
src1_use = NULL;
// src1_use = NULL;
}
if (src2_use) {
src2.value->RemoveUse(src2_use);
src2.value = NULL;
src2_use = NULL;
// src2_use = NULL;
}
if (src3_use) {
src3.value->RemoveUse(src3_use);
src3.value = NULL;
src3_use = NULL;
// src3_use = NULL;
}
if (src1_use) {
HIRBuilder::GetCurrent()->DeallocateUse(src1_use);
src1_use = nullptr;
}
if (src2_use) {
HIRBuilder::GetCurrent()->DeallocateUse(src2_use);
src2_use = nullptr;
}
if (src3_use) {
HIRBuilder::GetCurrent()->DeallocateUse(src3_use);
src3_use = nullptr;
}
}
void Instr::Remove() {
void Instr::UnlinkAndNOP() {
// Remove all srcs/dest.
Replace(&OPCODE_NOP_info, 0);
@@ -91,6 +105,10 @@ void Instr::Remove() {
block->instr_tail = prev;
}
}
void Instr::Deallocate() {
HIRBuilder::GetCurrent()->DeallocateInstruction(this);
}
Instr* Instr::GetDestDefSkipAssigns() {
Instr* current_def = this;

7
src/xenia/cpu/hir/instr.h
View File

@@ -78,7 +78,12 @@ class Instr {
void MoveBefore(Instr* other);
void Replace(const OpcodeInfo* new_opcode, uint16_t new_flags);
void Remove();
void UnlinkAndNOP();
//chrispy: wanted to change this one to Remove, but i changed Remove's name to UnlinkAndNOP,
//so if changes happened in master that we wanted to port over, and those changes used Remove, then we would have a lot of issues that the cause of would
//be difficult to track
//^todo: rework this comment, im frazzled
void Deallocate();
const OpcodeInfo* GetOpcodeInfo() const { return opcode; }
// if opcode is null, we have bigger problems
Opcode GetOpcodeNum() const { return GetOpcodeInfo()->num; }

2
src/xenia/cpu/hir/opcodes.h
View File

@@ -292,7 +292,7 @@ enum Opcode {
// as we already have OPCODE_ROUND. round double to float (
// ppc "single" fpu instruction result rounding behavior )
OPCODE_SET_NJM,
OPCODE_DELAY_EXECUTION, //for db16cyc
__OPCODE_MAX_VALUE, // Keep at end.
};

2
src/xenia/cpu/hir/opcodes.inl
View File

@@ -218,7 +218,7 @@ DEFINE_OPCODE(
"context_barrier",
OPCODE_SIG_X,
0)
DEFINE_OPCODE(OPCODE_DELAY_EXECUTION, "delay_execution", OPCODE_SIG_X, 0)
DEFINE_OPCODE(
OPCODE_LOAD_MMIO,
"load_mmio",

6
src/xenia/cpu/hir/value.cc
View File

@@ -16,13 +16,13 @@
#include "xenia/base/assert.h"
#include "xenia/base/byte_order.h"
#include "xenia/base/math.h"
#include "xenia/cpu/hir/hir_builder.h"
namespace xe {
namespace cpu {
namespace hir {
Value::Use* Value::AddUse(Arena* arena, Instr* instr) {
Use* use = arena->Alloc<Use>();
Use* use = HIRBuilder::GetCurrent()->AllocateUse();
use->instr = instr;
use->prev = NULL;
use->next = use_head;
@@ -42,6 +42,8 @@ void Value::RemoveUse(Use* use) {
if (use->next) {
use->next->prev = use->prev;
}
//HIRBuilder::GetCurrent()->DeallocateUse(use);
}
uint32_t Value::AsUint32() {

11
src/xenia/cpu/ppc/ppc_emit_alu.cc
View File

@@ -789,8 +789,15 @@ int InstrEmit_norx(PPCHIRBuilder& f, const InstrData& i) {
int InstrEmit_orx(PPCHIRBuilder& f, const InstrData& i) {
// RA <- (RS) | (RB)
if (i.X.RT == i.X.RB && i.X.RT == i.X.RA && !i.X.Rc) {
// Sometimes used as no-op.
f.Nop();
// chrispy: this special version of orx is db16cyc and is heavily used in
// spinlocks. since we do not emit any code for this we end up wasting a ton
// of power
if (i.code == 0x7FFFFB78) {
f.DelayExecution();
} else {
// Sometimes used as no-op.
f.Nop();
}
return 0;
}
Value* ra;

1
src/xenia/cpu/ppc/ppc_frontend.cc
View File

@@ -117,6 +117,7 @@ bool PPCFrontend::DefineFunction(GuestFunction* function,
uint32_t debug_info_flags) {
auto translator = translator_pool_.Allocate(this);
bool result = translator->Translate(function, debug_info_flags);
translator->Reset();
translator_pool_.Release(translator);
return result;
}

19
src/xenia/cpu/ppc/ppc_translator.cc
View File

@@ -96,10 +96,25 @@ PPCTranslator::PPCTranslator(PPCFrontend* frontend) : frontend_(frontend) {
PPCTranslator::~PPCTranslator() = default;
class HirBuilderScope {
PPCHIRBuilder* builder_;
public:
HirBuilderScope(PPCHIRBuilder* builder) : builder_(builder) {
builder_->MakeCurrent();
}
~HirBuilderScope() {
if (builder_) {
builder_->RemoveCurrent();
}
}
};
bool PPCTranslator::Translate(GuestFunction* function,
uint32_t debug_info_flags) {
SCOPE_profile_cpu_f("cpu");
HirBuilderScope hir_build_scope{builder_.get()};
// Reset() all caching when we leave.
xe::make_reset_scope(builder_);
xe::make_reset_scope(compiler_);
@@ -196,7 +211,7 @@ bool PPCTranslator::Translate(GuestFunction* function,
return true;
}
void PPCTranslator::Reset() { builder_->ResetPools(); }
void PPCTranslator::DumpSource(GuestFunction* function,
StringBuffer* string_buffer) {
Memory* memory = frontend_->memory();

2
src/xenia/cpu/ppc/ppc_translator.h
View File

@@ -31,7 +31,7 @@ class PPCTranslator {
~PPCTranslator();
bool Translate(GuestFunction* function, uint32_t debug_info_flags);
void Reset();
private:
void DumpSource(GuestFunction* function, StringBuffer* string_buffer);