Merge branch 'master' into d3d12
This commit is contained in:
commit
ff363d85be
|
@ -176,9 +176,9 @@ int xenia_main(const std::vector<std::wstring>& args) {
|
|||
#endif
|
||||
content_root = xe::join_paths(content_root, L"content");
|
||||
}
|
||||
|
||||
content_root = xe::to_absolute_path(content_root);
|
||||
}
|
||||
content_root = xe::to_absolute_path(content_root);
|
||||
XELOGI("Content root: %S", content_root.c_str());
|
||||
|
||||
// Create the emulator but don't initialize so we can setup the window.
|
||||
auto emulator = std::make_unique<Emulator>(L"", content_root);
|
||||
|
|
|
@ -5691,6 +5691,15 @@ struct VECTOR_SHL_V128
|
|||
return _mm_load_si128(reinterpret_cast<__m128i*>(value));
|
||||
}
|
||||
static void EmitInt32(X64Emitter& e, const EmitArgType& i) {
|
||||
Xmm src1;
|
||||
if (i.src1.is_constant) {
|
||||
src1 = e.xmm2;
|
||||
e.LoadConstantXmm(src1, i.src1.constant());
|
||||
}
|
||||
else {
|
||||
src1 = i.src1;
|
||||
}
|
||||
|
||||
if (i.src2.is_constant) {
|
||||
const auto& shamt = i.src2.constant();
|
||||
bool all_same = true;
|
||||
|
@ -5702,7 +5711,7 @@ struct VECTOR_SHL_V128
|
|||
}
|
||||
if (all_same) {
|
||||
// Every count is the same, so we can use vpslld.
|
||||
e.vpslld(i.dest, i.src1, shamt.u8[0] & 0x1F);
|
||||
e.vpslld(i.dest, src1, shamt.u8[0] & 0x1F);
|
||||
return;
|
||||
}
|
||||
}
|
||||
|
@ -5716,13 +5725,13 @@ struct VECTOR_SHL_V128
|
|||
masked.u32[n] &= 0x1F;
|
||||
}
|
||||
e.LoadConstantXmm(e.xmm0, masked);
|
||||
e.vpsllvd(i.dest, i.src1, e.xmm0);
|
||||
e.vpsllvd(i.dest, src1, e.xmm0);
|
||||
} else {
|
||||
// Fully variable shift.
|
||||
// src shift mask may have values >31, and x86 sets to zero when
|
||||
// that happens so we mask.
|
||||
e.vandps(e.xmm0, i.src2, e.GetXmmConstPtr(XMMShiftMaskPS));
|
||||
e.vpsllvd(i.dest, i.src1, e.xmm0);
|
||||
e.vpsllvd(i.dest, src1, e.xmm0);
|
||||
}
|
||||
} else {
|
||||
// Shift 4 words in src1 by amount specified in src2.
|
||||
|
@ -5740,7 +5749,8 @@ struct VECTOR_SHL_V128
|
|||
e.mov(e.rax, 0x1F);
|
||||
e.vmovq(e.xmm1, e.rax);
|
||||
e.vpand(e.xmm0, e.xmm0, e.xmm1);
|
||||
e.vpslld(i.dest, i.src1, e.xmm0);
|
||||
|
||||
e.vpslld(i.dest, src1, e.xmm0);
|
||||
e.jmp(end);
|
||||
}
|
||||
|
||||
|
@ -5752,7 +5762,7 @@ struct VECTOR_SHL_V128
|
|||
} else {
|
||||
e.lea(e.r9, e.StashXmm(1, i.src2));
|
||||
}
|
||||
e.lea(e.r8, e.StashXmm(0, i.src1));
|
||||
e.lea(e.r8, e.StashXmm(0, src1));
|
||||
e.CallNativeSafe(reinterpret_cast<void*>(EmulateVectorShlI32));
|
||||
e.vmovaps(i.dest, e.xmm0);
|
||||
|
||||
|
@ -5877,6 +5887,15 @@ struct VECTOR_SHR_V128
|
|||
return _mm_load_si128(reinterpret_cast<__m128i*>(value));
|
||||
}
|
||||
static void EmitInt32(X64Emitter& e, const EmitArgType& i) {
|
||||
Xmm src1;
|
||||
if (i.src1.is_constant) {
|
||||
src1 = e.xmm2;
|
||||
e.LoadConstantXmm(src1, i.src1.constant());
|
||||
}
|
||||
else {
|
||||
src1 = i.src1;
|
||||
}
|
||||
|
||||
if (i.src2.is_constant) {
|
||||
const auto& shamt = i.src2.constant();
|
||||
bool all_same = true;
|
||||
|
@ -5888,7 +5907,7 @@ struct VECTOR_SHR_V128
|
|||
}
|
||||
if (all_same) {
|
||||
// Every count is the same, so we can use vpsrld.
|
||||
e.vpsrld(i.dest, i.src1, shamt.u8[0] & 0x1F);
|
||||
e.vpsrld(i.dest, src1, shamt.u8[0] & 0x1F);
|
||||
return;
|
||||
} else {
|
||||
if (e.IsFeatureEnabled(kX64EmitAVX2)) {
|
||||
|
@ -5898,7 +5917,7 @@ struct VECTOR_SHR_V128
|
|||
masked.u32[n] &= 0x1F;
|
||||
}
|
||||
e.LoadConstantXmm(e.xmm0, masked);
|
||||
e.vpsrlvd(i.dest, i.src1, e.xmm0);
|
||||
e.vpsrlvd(i.dest, src1, e.xmm0);
|
||||
return;
|
||||
}
|
||||
}
|
||||
|
@ -5909,7 +5928,7 @@ struct VECTOR_SHR_V128
|
|||
// src shift mask may have values >31, and x86 sets to zero when
|
||||
// that happens so we mask.
|
||||
e.vandps(e.xmm0, i.src2, e.GetXmmConstPtr(XMMShiftMaskPS));
|
||||
e.vpsrlvd(i.dest, i.src1, e.xmm0);
|
||||
e.vpsrlvd(i.dest, src1, e.xmm0);
|
||||
} else {
|
||||
// Shift 4 words in src1 by amount specified in src2.
|
||||
Xbyak::Label emu, end;
|
||||
|
@ -5926,7 +5945,7 @@ struct VECTOR_SHR_V128
|
|||
e.mov(e.rax, 0x1F);
|
||||
e.vmovq(e.xmm1, e.rax);
|
||||
e.vpand(e.xmm0, e.xmm0, e.xmm1);
|
||||
e.vpsrld(i.dest, i.src1, e.xmm0);
|
||||
e.vpsrld(i.dest, src1, e.xmm0);
|
||||
e.jmp(end);
|
||||
}
|
||||
|
||||
|
@ -5938,7 +5957,7 @@ struct VECTOR_SHR_V128
|
|||
} else {
|
||||
e.lea(e.r9, e.StashXmm(1, i.src2));
|
||||
}
|
||||
e.lea(e.r8, e.StashXmm(0, i.src1));
|
||||
e.lea(e.r8, e.StashXmm(0, src1));
|
||||
e.CallNativeSafe(reinterpret_cast<void*>(EmulateVectorShrI32));
|
||||
e.vmovaps(i.dest, e.xmm0);
|
||||
|
||||
|
|
|
@ -278,11 +278,13 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder) {
|
|||
if (i->src1.value->IsConstant()) {
|
||||
if (i->src1.value->type != VEC128_TYPE) {
|
||||
if (i->src1.value->IsConstantTrue()) {
|
||||
v->set_from(i->src2.value);
|
||||
i->Remove();
|
||||
auto src2 = i->src2.value;
|
||||
i->Replace(&OPCODE_ASSIGN_info, 0);
|
||||
i->set_src1(src2);
|
||||
} else if (i->src1.value->IsConstantFalse()) {
|
||||
v->set_from(i->src3.value);
|
||||
i->Remove();
|
||||
auto src3 = i->src3.value;
|
||||
i->Replace(&OPCODE_ASSIGN_info, 0);
|
||||
i->set_src1(src3);
|
||||
} else if (i->src2.value->IsConstant() &&
|
||||
i->src3.value->IsConstant()) {
|
||||
// TODO: Select
|
||||
|
@ -616,6 +618,10 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder) {
|
|||
v->set_from(i->src1.value);
|
||||
v->Shl(i->src2.value);
|
||||
i->Remove();
|
||||
} else if (i->src2.value->IsConstantZero()) {
|
||||
auto src1 = i->src1.value;
|
||||
i->Replace(&OPCODE_ASSIGN_info, 0);
|
||||
i->set_src1(src1);
|
||||
}
|
||||
break;
|
||||
case OPCODE_SHR:
|
||||
|
@ -623,6 +629,10 @@ bool ConstantPropagationPass::Run(HIRBuilder* builder) {
|
|||
v->set_from(i->src1.value);
|
||||
v->Shr(i->src2.value);
|
||||
i->Remove();
|
||||
} else if (i->src2.value->IsConstantZero()) {
|
||||
auto src1 = i->src1.value;
|
||||
i->Replace(&OPCODE_ASSIGN_info, 0);
|
||||
i->set_src1(src1);
|
||||
}
|
||||
break;
|
||||
case OPCODE_SHA:
|
||||
|
|
|
@ -764,7 +764,7 @@ void HIRBuilder::CommentFormat(const char* format, ...) {
|
|||
va_start(args, format);
|
||||
size_t chars_written = vsnprintf(p, kMaxCommentSize - 1, format, args);
|
||||
va_end(args);
|
||||
size_t rewind = kMaxCommentSize - chars_written;
|
||||
size_t rewind = kMaxCommentSize - chars_written - 1;
|
||||
arena_->Rewind(rewind);
|
||||
Instr* i = AppendInstr(OPCODE_COMMENT_info, 0);
|
||||
i->src1.offset = (uint64_t)p;
|
||||
|
|
|
@ -53,10 +53,15 @@ PPCTranslator::PPCTranslator(PPCFrontend* frontend) : frontend_(frontend) {
|
|||
if (validate) compiler_->AddPass(std::make_unique<passes::ValidationPass>());
|
||||
compiler_->AddPass(std::make_unique<passes::ContextPromotionPass>());
|
||||
if (validate) compiler_->AddPass(std::make_unique<passes::ValidationPass>());
|
||||
// TODO(gibbed): loop until these passes stop making changes?
|
||||
for (int i = 0; i < 5; ++i) {
|
||||
compiler_->AddPass(std::make_unique<passes::SimplificationPass>());
|
||||
if (validate) compiler_->AddPass(std::make_unique<passes::ValidationPass>());
|
||||
if (validate)
|
||||
compiler_->AddPass(std::make_unique<passes::ValidationPass>());
|
||||
compiler_->AddPass(std::make_unique<passes::ConstantPropagationPass>());
|
||||
if (validate) compiler_->AddPass(std::make_unique<passes::ValidationPass>());
|
||||
if (validate)
|
||||
compiler_->AddPass(std::make_unique<passes::ValidationPass>());
|
||||
}
|
||||
if (backend->machine_info()->supports_extended_load_store) {
|
||||
// Backend supports the advanced LOAD/STORE instructions.
|
||||
// These will save us a lot of HIR opcodes.
|
||||
|
|
|
@ -159,3 +159,44 @@ test_slw_9_constant:
|
|||
#_ REGISTER_OUT r3 0
|
||||
#_ REGISTER_OUT r4 0xFFFFFFFFFFFFFFFF
|
||||
#_ REGISTER_OUT r5 32
|
||||
|
||||
test_slw_10:
|
||||
#_ REGISTER_IN r4 99
|
||||
#_ REGISTER_IN r5 1
|
||||
cntlzw r5, r5
|
||||
subi r5, r5, 28
|
||||
slw r3, r4, r5
|
||||
blr
|
||||
#_ REGISTER_OUT r3 792
|
||||
#_ REGISTER_OUT r4 99
|
||||
#_ REGISTER_OUT r5 3
|
||||
|
||||
test_slw_10_constant:
|
||||
#_ REGISTER_IN r4 99
|
||||
li r5, 1
|
||||
cntlzw r5, r5
|
||||
subi r5, r5, 28
|
||||
slw r3, r4, r5
|
||||
blr
|
||||
#_ REGISTER_OUT r3 792
|
||||
#_ REGISTER_OUT r4 99
|
||||
#_ REGISTER_OUT r5 3
|
||||
|
||||
test_slw_11:
|
||||
#_ REGISTER_IN r4 99
|
||||
#_ REGISTER_IN r5 3
|
||||
li r5, 3
|
||||
slw r3, r4, r5
|
||||
blr
|
||||
#_ REGISTER_OUT r3 792
|
||||
#_ REGISTER_OUT r4 99
|
||||
#_ REGISTER_OUT r5 3
|
||||
|
||||
test_slw_11_constant:
|
||||
#_ REGISTER_IN r4 99
|
||||
li r5, 3
|
||||
slw r3, r4, r5
|
||||
blr
|
||||
#_ REGISTER_OUT r3 792
|
||||
#_ REGISTER_OUT r4 99
|
||||
#_ REGISTER_OUT r5 3
|
||||
|
|
|
@ -198,12 +198,16 @@ dword_result_t XamEnumerate(dword_t handle, dword_t flags, lpvoid_t buffer,
|
|||
}
|
||||
}
|
||||
|
||||
size_t actual_buffer_length = e->item_size() * e->items_per_enumerate();
|
||||
if (actual_buffer_length != buffer_length) {
|
||||
size_t actual_buffer_length = (uint32_t)buffer_length;
|
||||
if (buffer_length == e->items_per_enumerate()) {
|
||||
actual_buffer_length = e->item_size() * e->items_per_enumerate();
|
||||
// Known culprits:
|
||||
// Final Fight: Double Impact
|
||||
XELOGW("Broken usage of XamEnumerate! %.X vs %.X", buffer_length,
|
||||
actual_buffer_length);
|
||||
// Final Fight: Double Impact (saves)
|
||||
XELOGW(
|
||||
"Broken usage of XamEnumerate! buffer length=%.X vs actual length=%.X "
|
||||
"(item size=%.X, items per enumerate=%u)",
|
||||
(uint32_t)buffer_length, actual_buffer_length, e->item_size(),
|
||||
e->items_per_enumerate());
|
||||
}
|
||||
|
||||
buffer.Zero(actual_buffer_length);
|
||||
|
|
|
@ -49,7 +49,8 @@ dword_result_t XamUserGetSigninState(dword_t user_index) {
|
|||
return 0;
|
||||
}
|
||||
}
|
||||
DECLARE_XAM_EXPORT1(XamUserGetSigninState, kUserProfiles, kImplemented);
|
||||
DECLARE_XAM_EXPORT2(XamUserGetSigninState, kUserProfiles, kImplemented,
|
||||
kHighFrequency);
|
||||
|
||||
typedef struct {
|
||||
xe::be<uint64_t> xuid;
|
||||
|
|
Loading…
Reference in New Issue