Merge pull request #812 from Nekotekina/master

cellSpursAttributeEnableSystemWorkload
This commit is contained in:
B1ackDaemon 2014-09-16 18:08:37 +03:00
commit ef1c5af608
8 changed files with 176 additions and 116 deletions

View file

@ -1,5 +1,4 @@
#pragma once
#include <array>
#include "Utilities/MTRingbuffer.h"
//#define BUFFERED_LOGGING 1

View file

@ -95,49 +95,37 @@ namespace cb_detail
return _bind_func_args<g, f, v>(CPU, args...) || (t == ARG_STACK);
}
template<typename RT>
template<typename T, _func_arg_type type>
struct _func_res
{
static_assert(sizeof(RT) <= 8, "Invalid callback result type");
static_assert(!std::is_pointer<RT>::value, "Invalid callback result type (pointer)");
static_assert(!std::is_reference<RT>::value, "Invalid callback result type (reference)");
static_assert(type == ARG_GENERAL, "Wrong use of _func_res template");
static_assert(sizeof(T) <= 8, "Invalid callback result type for ARG_GENERAL");
__forceinline static RT get_value(const PPUThread& CPU)
__forceinline static T get_value(const PPUThread& CPU)
{
if (std::is_floating_point<RT>::value)
{
return (RT)CPU.FPR[1];
}
else
{
return (RT&)CPU.GPR[3];
}
return (T&)CPU.GPR[3];
}
};
template<>
struct _func_res<u128>
template<typename T>
struct _func_res<T, ARG_FLOAT>
{
__forceinline static const u128 get_value(const PPUThread& CPU)
static_assert(sizeof(T) <= 8, "Invalid callback result type for ARG_FLOAT");
__forceinline static T get_value(const PPUThread& CPU)
{
return CPU.VPR[2];
return (T)CPU.FPR[1];
}
};
template<>
struct _func_res<const u128>
template<typename T>
struct _func_res<T, ARG_VECTOR>
{
__forceinline static const u128 get_value(const PPUThread& CPU)
{
return CPU.VPR[2];
}
};
static_assert(sizeof(T) == 16, "Invalid callback result type for ARG_VECTOR");
template<>
struct _func_res<void>
{
__forceinline static void get_value(const PPUThread& CPU)
__forceinline static T get_value(const PPUThread& CPU)
{
return (T&)CPU.VPR[2];
}
};
@ -150,7 +138,14 @@ namespace cb_detail
if (stack) CPU.GPR[1] -= FIXED_STACK_FRAME_SIZE;
CPU.FastCall2(pc, rtoc);
if (stack) CPU.GPR[1] += FIXED_STACK_FRAME_SIZE;
return _func_res<RT>::get_value(CPU);
static_assert(!std::is_pointer<RT>::value, "Invalid callback result type (pointer)");
static_assert(!std::is_reference<RT>::value, "Invalid callback result type (reference)");
const bool is_float = std::is_floating_point<RT>::value;
const bool is_vector = std::is_same<RT, u128>::value;
const _func_arg_type t = is_float ? ARG_FLOAT : (is_vector ? ARG_VECTOR : ARG_GENERAL);
return _func_res<RT, t>::get_value(CPU);
}
};

View file

@ -3,6 +3,7 @@
#include "Emu/SysCalls/Modules.h"
#include "Emu/SysCalls/Callback.h"
#include "Emu/SysCalls/lv2/sys_spu.h"
#include "Emu/Cell/SPURSManager.h"
#include "cellSpurs.h"
@ -20,18 +21,18 @@ s64 spursInit(
s32 nSpus,
s32 spuPriority,
s32 ppuPriority,
u32 flags,
u32 flags, // SpursAttrFlags
const char prefix[],
u32 prefixSize,
u32 container,
u32 arg11,
u32 arg12,
u32 arg13)
const u8 swlPriority[],
u32 swlMaxSpu,
u32 swlIsPreem)
{
// internal function
#ifdef PRX_DEBUG
return cb_caller<s32,vm::ptr<CellSpurs2>, u32, u32, s32, s32, s32, u32, u32, u32, u32, u32, u32, u32>::call(GetCurrentPPUThread(), libsre + 0x74E4, libsre_rtoc,
spurs, revision, sdkVersion, nSpus, spuPriority, ppuPriority, flags, Memory.RealToVirtualAddr(prefix), prefixSize, container, arg11, arg12, arg13);
return cb_call<s32, vm::ptr<CellSpurs2>, u32, u32, s32, s32, s32, u32, u32, u32, u32, u32, u32, u32>(GetCurrentPPUThread(), libsre + 0x74E4, libsre_rtoc,
spurs, revision, sdkVersion, nSpus, spuPriority, ppuPriority, flags, Memory.RealToVirtualAddr(prefix), prefixSize, container, Memory.RealToVirtualAddr(swlPriority), swlMaxSpu, swlIsPreem);
#else
//spurs->spurs = new SPURSManager(attr);
return CELL_OK;
@ -53,11 +54,11 @@ s64 cellSpursInitialize(vm::ptr<CellSpurs> spurs, s32 nSpus, s32 spuPriority, s3
nSpus,
spuPriority,
ppuPriority,
exitIfNoWork ? 1 : 0,
exitIfNoWork ? SAF_EXIT_IF_NO_WORK : SAF_NONE,
nullptr,
0,
0,
0,
nullptr,
0,
0);
#endif
@ -90,13 +91,13 @@ s64 cellSpursInitializeWithAttribute(vm::ptr<CellSpurs> spurs, vm::ptr<const Cel
attr->m.nSpus,
attr->m.spuPriority,
attr->m.ppuPriority,
(u32)attr->m.flags | (attr->m.exitIfNoWork ? 1 : 0),
attr->m.flags.ToLE() | (attr->m.exitIfNoWork ? SAF_EXIT_IF_NO_WORK : 0),
attr->m.prefix,
attr->m.prefixSize,
attr->m.container,
attr.addr() + 0x38,
attr->_u32[16],
attr->_u32[17]);
attr->m.swlPriority,
attr->m.swlMaxSpu,
attr->m.swlIsPreem);
#endif
}
@ -127,13 +128,13 @@ s64 cellSpursInitializeWithAttribute2(vm::ptr<CellSpurs2> spurs, vm::ptr<const C
attr->m.nSpus,
attr->m.spuPriority,
attr->m.ppuPriority,
(u32)attr->m.flags | (attr->m.exitIfNoWork ? 1 : 0) | 4,
attr->m.flags.ToLE() | (attr->m.exitIfNoWork ? SAF_EXIT_IF_NO_WORK : 0) | 4, // +add unknown flag
attr->m.prefix,
attr->m.prefixSize,
attr->m.container,
attr.addr() + 0x38,
attr->_u32[16],
attr->_u32[17]);
attr->m.swlPriority,
attr->m.swlMaxSpu,
attr->m.swlIsPreem);
#endif
}
@ -180,13 +181,14 @@ s64 cellSpursAttributeSetMemoryContainerForSpuThread(vm::ptr<CellSpursAttribute>
{
return CELL_SPURS_CORE_ERROR_ALIGN;
}
if ((u32)attr->m.flags & 0x20000000) // check unknown flag
if (attr->m.flags.ToLE() & SAF_SPU_TGT_EXCLUSIVE_NON_CONTEXT)
{
return CELL_SPURS_CORE_ERROR_STAT;
}
attr->_u32[11] = container;
attr->m.flags |= 0x40000000; // set unknown flag
attr->m.container = container;
attr->m.flags |= SAF_SPU_MEMORY_CONTAINER_SET;
return CELL_OK;
#endif
}
@ -220,11 +222,21 @@ s64 cellSpursAttributeSetNamePrefix(vm::ptr<CellSpursAttribute> attr, vm::ptr<co
s64 cellSpursAttributeEnableSpuPrintfIfAvailable(vm::ptr<CellSpursAttribute> attr)
{
#ifdef PRX_DEBUG
cellSpurs->Warning("cellSpursAttributeEnableSpuPrintfIfAvailable(attr_addr=0x%x)", attr.addr());
#ifdef PRX_DEBUG_XXX
return GetCurrentPPUThread().FastCall2(libsre + 0x7150, libsre_rtoc);
#else
UNIMPLEMENTED_FUNC(cellSpurs);
if (!attr)
{
return CELL_SPURS_CORE_ERROR_NULL_POINTER;
}
if (attr.addr() % 8)
{
return CELL_SPURS_CORE_ERROR_ALIGN;
}
attr->m.flags |= SAF_SPU_PRINTF_ENABLED;
return CELL_OK;
#endif
}
@ -245,45 +257,82 @@ s64 cellSpursAttributeSetSpuThreadGroupType(vm::ptr<CellSpursAttribute> attr, s3
return CELL_SPURS_CORE_ERROR_ALIGN;
}
if (type == 0x18)
if (type == SYS_SPU_THREAD_GROUP_TYPE_EXCLUSIVE_NON_CONTEXT)
{
if (attr->m.flags & 0x40000000) // check unknown flag
if (attr->m.flags.ToLE() & SAF_SPU_MEMORY_CONTAINER_SET)
{
return CELL_SPURS_CORE_ERROR_STAT;
}
attr->m.flags |= 0x20000000; // set unknown flag
return CELL_OK;
attr->m.flags |= SAF_SPU_TGT_EXCLUSIVE_NON_CONTEXT; // set
}
else if (type)
else if (type == SYS_SPU_THREAD_GROUP_TYPE_NORMAL)
{
attr->m.flags &= ~SAF_SPU_TGT_EXCLUSIVE_NON_CONTEXT; // clear
}
else
{
return CELL_SPURS_CORE_ERROR_INVAL;
}
else // if type == 0
{
attr->m.flags &= ~0x20000000; // clear unknown flag
return CELL_OK;
}
return CELL_OK;
#endif
}
s64 cellSpursAttributeEnableSystemWorkload(vm::ptr<CellSpursAttribute> attr, vm::ptr<const u8> priority, u32 maxSpu, vm::ptr<const bool> isPreemptible)
s64 cellSpursAttributeEnableSystemWorkload(vm::ptr<CellSpursAttribute> attr, vm::ptr<const u8[8]> priority, u32 maxSpu, vm::ptr<const bool[8]> isPreemptible)
{
cellSpurs->Warning("cellSpursAttributeEnableSystemWorkload(attr_addr=0x%x, priority_addr=0x%x, maxSpu=%d, isPreemptible_addr=0x%x)",
attr.addr(), priority.addr(), maxSpu, isPreemptible.addr());
#ifdef PRX_DEBUG
#ifdef PRX_DEBUG_XXX
return GetCurrentPPUThread().FastCall2(libsre + 0xF410, libsre_rtoc);
#else
for (s32 i = 0; i < CELL_SPURS_MAX_SPU; i++)
if (!attr)
{
if (priority[i] != 1 || maxSpu == 0)
return CELL_SPURS_CORE_ERROR_NULL_POINTER;
}
if (attr.addr() % 8)
{
return CELL_SPURS_CORE_ERROR_ALIGN;
}
const u32 nSpus = attr->_u32[2];
if (!nSpus)
{
return CELL_SPURS_CORE_ERROR_INVAL;
}
for (u32 i = 0; i < nSpus; i++)
{
if ((*priority)[i] == 1)
{
cellSpurs->Error("cellSpursAttributeEnableSystemWorkload : CELL_SPURS_CORE_ERROR_INVAL");
return CELL_SPURS_CORE_ERROR_INVAL;
if (!maxSpu)
{
return CELL_SPURS_CORE_ERROR_INVAL;
}
if (nSpus == 1 || attr->m.exitIfNoWork)
{
return CELL_SPURS_CORE_ERROR_PERM;
}
if (attr->m.flags.ToLE() & SAF_SYSTEM_WORKLOAD_ENABLED)
{
return CELL_SPURS_CORE_ERROR_BUSY;
}
attr->m.flags |= SAF_SYSTEM_WORKLOAD_ENABLED; // set flag
*(u64*)attr->m.swlPriority = *(u64*)*priority; // copy system workload priorities
u32 isPreem = 0; // generate mask from isPreemptible values
for (u32 j = 0; j < nSpus; j++)
{
if ((*isPreemptible)[j])
{
isPreem |= (1 << j);
}
}
attr->m.swlMaxSpu = maxSpu; // write max spu for system workload
attr->m.swlIsPreem = isPreem; // write isPreemptible mask
return CELL_OK;
}
}
return CELL_OK;
return CELL_SPURS_CORE_ERROR_INVAL;
#endif
}

View file

@ -91,6 +91,17 @@ struct CellSpurs2
SPURSManager *spurs;
};
enum SpursAttrFlags : u32
{
SAF_NONE = 0x0,
SAF_EXIT_IF_NO_WORK = 0x1,
SAF_SYSTEM_WORKLOAD_ENABLED = 0x02000000,
SAF_SPU_PRINTF_ENABLED = 0x10000000,
SAF_SPU_TGT_EXCLUSIVE_NON_CONTEXT = 0x20000000,
SAF_SPU_MEMORY_CONTAINER_SET = 0x40000000,
};
struct CellSpursAttribute
{
static const uint align = 8;
@ -113,9 +124,13 @@ struct CellSpursAttribute
bool exitIfNoWork; // 0x14
char prefix[15]; // 0x15 (not a NTS)
be_t<u32> prefixSize; // 0x24
be_t<u32> flags; // 0x28
be_t<u32> flags; // 0x28 (SpursAttrFlags)
be_t<u32> container; // 0x2C
// ...
be_t<u32> unk0; // 0x30
be_t<u32> unk1; // 0x34
u8 swlPriority[8]; // 0x38
be_t<u32> swlMaxSpu; // 0x40
be_t<u32> swlIsPreem; // 0x44
} m;
// alternative implementation

View file

@ -1535,7 +1535,7 @@ s32 _cellSyncLFQueuePushBody(vm::ptr<CellSyncLFQueue> queue, vm::ptr<const void>
else
{
#ifdef PRX_DEBUG
res = cb_caller<s32, vm::ptr<CellSyncLFQueue>, u32, u32, u64>::call(GetCurrentPPUThread(), libsre + 0x3050, libsre_rtoc,
res = cb_call<s32, vm::ptr<CellSyncLFQueue>, u32, u32, u64>(GetCurrentPPUThread(), libsre + 0x3050, libsre_rtoc,
queue, position_v.addr(), isBlocking, 0);
position = position_v->ToLE();
#else
@ -1580,7 +1580,7 @@ s32 _cellSyncLFQueuePushBody(vm::ptr<CellSyncLFQueue> queue, vm::ptr<const void>
else
{
#ifdef PRX_DEBUG
res = cb_caller<s32, vm::ptr<CellSyncLFQueue>, s32, u64>::call(GetCurrentPPUThread(), libsre + 0x355C, libsre_rtoc,
res = cb_call<s32, vm::ptr<CellSyncLFQueue>, s32, u64>(GetCurrentPPUThread(), libsre + 0x355C, libsre_rtoc,
queue, position, 0);
#else
res = syncLFQueueCompletePushPointer2(queue, position, nullptr);
@ -1930,7 +1930,7 @@ s32 _cellSyncLFQueuePopBody(vm::ptr<CellSyncLFQueue> queue, vm::ptr<void> buffer
else
{
#ifdef PRX_DEBUG
res = cb_caller<s32, vm::ptr<CellSyncLFQueue>, u32, u32, u64>::call(GetCurrentPPUThread(), libsre + 0x39AC, libsre_rtoc,
res = cb_call<s32, vm::ptr<CellSyncLFQueue>, u32, u32, u64>(GetCurrentPPUThread(), libsre + 0x39AC, libsre_rtoc,
queue, position_v.addr(), isBlocking, 0);
position = position_v->ToLE();
#else
@ -1972,7 +1972,7 @@ s32 _cellSyncLFQueuePopBody(vm::ptr<CellSyncLFQueue> queue, vm::ptr<void> buffer
else
{
#ifdef PRX_DEBUG
res = cb_caller<s32, vm::ptr<CellSyncLFQueue>, s32, u64, u64>::call(GetCurrentPPUThread(), libsre + 0x3EB8, libsre_rtoc,
res = cb_call<s32, vm::ptr<CellSyncLFQueue>, s32, u64, u64>(GetCurrentPPUThread(), libsre + 0x3EB8, libsre_rtoc,
queue, position, 0, 0);
#else
res = syncLFQueueCompletePopPointer2(queue, position, nullptr, 0);

View file

@ -46,7 +46,7 @@ namespace detail
template<typename T, int g_count, int f_count, int v_count>
struct bind_arg<T, ARG_VECTOR, g_count, f_count, v_count>
{
static_assert(std::is_same<T, u128>::value, "Invalid function argument type for ARG_VECTOR");
static_assert(sizeof(T) == 16, "Invalid function argument type for ARG_VECTOR");
static __forceinline T func(PPUThread& CPU)
{
@ -69,41 +69,37 @@ namespace detail
}
};
template<typename T>
template<typename T, bind_arg_type type>
struct bind_result
{
static_assert(sizeof(T) <= 8, "Invalid function result type");
static_assert(!std::is_pointer<T>::value, "Invalid function result type (pointer)");
static_assert(!std::is_reference<T>::value, "Invalid function result type (reference)");
static_assert(type == ARG_GENERAL, "Wrong use of bind_result template");
static_assert(sizeof(T) <= 8, "Invalid function result type for ARG_GENERAL");
static __forceinline void func(PPUThread& CPU, T result)
{
if (std::is_floating_point<T>::value)
{
CPU.FPR[1] = (double)result;
}
else
{
(T&)CPU.GPR[3] = result;
}
(T&)CPU.GPR[3] = result;
}
};
template<>
struct bind_result<u128>
template<typename T>
struct bind_result<T, ARG_FLOAT>
{
static __forceinline void func(PPUThread& CPU, const u128 result)
static_assert(sizeof(T) <= 8, "Invalid function result type for ARG_FLOAT");
static __forceinline void func(PPUThread& CPU, T result)
{
CPU.VPR[2] = result;
CPU.FPR[1] = (double)result;
}
};
template<>
struct bind_result<const u128>
template<typename T>
struct bind_result<T, ARG_VECTOR>
{
static __forceinline void func(PPUThread& CPU, const u128 result)
static_assert(sizeof(T) == 16, "Invalid function result type for ARG_VECTOR");
static __forceinline void func(PPUThread& CPU, const T result)
{
CPU.VPR[2] = result;
(T&)CPU.VPR[2] = result;
}
};
@ -157,13 +153,13 @@ namespace detail
return std::tuple_cat(std::tuple<T>(bind_arg<T, t, g, f, v>::func(CPU)), iterate<g, f, v, A...>(CPU));
}
template<typename RT, typename... TA>
template<typename RT, typename... T>
class func_binder;
template<typename... TA>
class func_binder<void, TA...> : public func_caller
template<typename... T>
class func_binder<void, T...> : public func_caller
{
typedef void(*func_t)(TA...);
typedef void(*func_t)(T...);
const func_t m_call;
public:
@ -176,14 +172,14 @@ namespace detail
virtual void operator()()
{
declCPU();
call<void>(m_call, iterate<0, 0, 0, TA...>(CPU));
call<void>(m_call, iterate<0, 0, 0, T...>(CPU));
}
};
template<typename TR, typename... TA>
template<typename RT, typename... T>
class func_binder : public func_caller
{
typedef TR(*func_t)(TA...);
typedef RT(*func_t)(T...);
const func_t m_call;
public:
@ -196,13 +192,19 @@ namespace detail
virtual void operator()()
{
declCPU();
bind_result<TR>::func(CPU, call<TR>(m_call, iterate<0, 0, 0, TA...>(CPU)));
static_assert(!std::is_pointer<RT>::value, "Invalid function result type (pointer)");
static_assert(!std::is_reference<RT>::value, "Invalid function result type (reference)");
const bool is_float = std::is_floating_point<RT>::value;
const bool is_vector = std::is_same<RT, u128>::value;
const bind_arg_type t = is_float ? ARG_FLOAT : (is_vector ? ARG_VECTOR : ARG_GENERAL);
bind_result<RT, t>::func(CPU, call<RT>(m_call, iterate<0, 0, 0, T...>(CPU)));
}
};
}
template<typename TR, typename... TA>
func_caller* bind_func(TR(*call)(TA...))
template<typename RT, typename... T>
func_caller* bind_func(RT(*call)(T...))
{
return new detail::func_binder<TR, TA...>(call);
return new detail::func_binder<RT, T...>(call);
}

View file

@ -11,8 +11,8 @@
#include "lv2/sys_event.h"
#include "lv2/sys_event_flag.h"
#include "lv2/sys_interrupt.h"
#include "lv2/sys_lwcond.h"
#include "lv2/sys_lwmutex.h"
//#include "lv2/sys_lwcond.h"
//#include "lv2/sys_lwmutex.h"
#include "lv2/sys_memory.h"
#include "lv2/sys_mmapper.h"
#include "lv2/sys_ppu_thread.h"
@ -21,7 +21,6 @@
#include "lv2/sys_rsx.h"
#include "lv2/sys_rwlock.h"
#include "lv2/sys_semaphore.h"
#include "lv2/sys_spinlock.h"
#include "lv2/sys_spu.h"
#include "lv2/sys_time.h"
#include "lv2/sys_timer.h"
@ -138,11 +137,11 @@ static func_caller* sc_table[kSyscallTableLength] =
bind_func(sys_semaphore_wait), //92 (0x05C)
bind_func(sys_semaphore_trywait), //93 (0x05D)
bind_func(sys_semaphore_post), //94 (0x05E)
null_func,//bind_func(sys_lwmutex_create), //95 (0x05F)
null_func,//bind_func(sys_lwmutex_destroy), //96 (0x060)
null_func,//bind_func(sys_lwmutex_lock), //97 (0x061)
null_func,//bind_func(sys_lwmutex_trylock), //98 (0x062)
null_func,//bind_func(sys_lwmutex_unlock), //99 (0x063)
null_func,//bind_func(sys_lwmutex_create), //95 (0x05F) // internal, used by sys_lwmutex_create
null_func,//bind_func(sys_lwmutex_destroy), //96 (0x060) // internal, used by sys_lwmutex_destroy
null_func,//bind_func(sys_lwmutex_lock), //97 (0x061) // internal, used by sys_lwmutex_lock
null_func,//bind_func(sys_lwmutex_trylock), //98 (0x062) // internal, used by sys_lwmutex_unlock
null_func,//bind_func(sys_lwmutex_unlock), //99 (0x063) // internal, used by sys_lwmutex_trylock
bind_func(sys_mutex_create), //100 (0x064)
bind_func(sys_mutex_destroy), //101 (0x065)
bind_func(sys_mutex_lock), //102 (0x066)
@ -160,7 +159,7 @@ static func_caller* sc_table[kSyscallTableLength] =
bind_func(sys_semaphore_get_value), //114 (0x072)
null_func,//bind_func(sys_semaphore_...) //115 (0x073)
null_func,//bind_func(sys_semaphore_...) //116 (0x074)
null_func,//bind_func(sys_semaphore_...) //117 (0x075)
null_func,//bind_func(sys_semaphore_...) //117 (0x075) // internal, used by sys_lwmutex_unlock
bind_func(sys_event_flag_clear), //118 (0x076)
null_func,//bind_func(sys_event_...) //119 (0x077) ROOT
bind_func(sys_rwlock_create), //120 (0x078)

View file

@ -26,6 +26,7 @@
#include <memory>
#include <vector>
#include <set>
#include <array>
#include <string>
#include <ostream>
#include <sstream>