diff options
Diffstat (limited to 'src/core/arm/skyeye_common')
| -rw-r--r-- | src/core/arm/skyeye_common/arm_regformat.h | 4 | ||||
| -rw-r--r-- | src/core/arm/skyeye_common/armdefs.h | 344 | ||||
| -rw-r--r-- | src/core/arm/skyeye_common/armemu.h | 556 | ||||
| -rw-r--r-- | src/core/arm/skyeye_common/vfp/asm_vfp.h | 141 | ||||
| -rw-r--r-- | src/core/arm/skyeye_common/vfp/vfp.cpp | 241 | ||||
| -rw-r--r-- | src/core/arm/skyeye_common/vfp/vfp.h | 112 | ||||
| -rw-r--r-- | src/core/arm/skyeye_common/vfp/vfp_helper.h | 705 | ||||
| -rw-r--r-- | src/core/arm/skyeye_common/vfp/vfpdouble.cpp | 202 | ||||
| -rw-r--r-- | src/core/arm/skyeye_common/vfp/vfpinstr.cpp | 1733 | ||||
| -rw-r--r-- | src/core/arm/skyeye_common/vfp/vfpsingle.cpp | 57 |
10 files changed, 630 insertions, 3465 deletions
diff --git a/src/core/arm/skyeye_common/arm_regformat.h b/src/core/arm/skyeye_common/arm_regformat.h index 997874764..5be3a561f 100644 --- a/src/core/arm/skyeye_common/arm_regformat.h +++ b/src/core/arm/skyeye_common/arm_regformat.h @@ -86,7 +86,9 @@ enum { CP15_IFAR, CP15_PID, CP15_CONTEXT_ID, - CP15_THREAD_URO, + CP15_THREAD_UPRW, // Thread ID register - User/Privileged Read/Write + CP15_THREAD_URO, // Thread ID register - User Read Only (Privileged R/W) + CP15_THREAD_PRW, // Thread ID register - Privileged R/W only. CP15_TLB_FAULT_ADDR, /* defined by SkyEye */ CP15_TLB_FAULT_STATUS, /* defined by SkyEye */ /* VFP registers */ diff --git a/src/core/arm/skyeye_common/armdefs.h b/src/core/arm/skyeye_common/armdefs.h index a9c41ce5a..ff9296e0f 100644 --- a/src/core/arm/skyeye_common/armdefs.h +++ b/src/core/arm/skyeye_common/armdefs.h @@ -32,10 +32,8 @@ #include "core/arm/skyeye_common/armmmu.h" #include "core/arm/skyeye_common/skyeye_defs.h" -#ifndef FALSE -#define FALSE 0 -#define TRUE 1 -#endif +#define BITS(s, a, b) ((s << ((sizeof(s) * 8 - 1) - b)) >> (sizeof(s) * 8 - b + a - 1)) +#define BIT(s, n) ((s >> (n)) & 1) #define LOW 0 #define HIGH 1 @@ -71,88 +69,68 @@ typedef unsigned ARMul_CPWrites(ARMul_State* state, unsigned reg, ARMword value) #define VFP_REG_NUM 64 struct ARMul_State { - ARMword Emulate; /* to start and stop emulation */ - unsigned EndCondition; /* reason for stopping */ - unsigned ErrorCode; /* type of illegal instruction */ + ARMword Emulate; // To start and stop emulation + unsigned EndCondition; // Reason for stopping + unsigned ErrorCode; // Type of illegal instruction - /* Order of the following register should not be modified */ - ARMword Reg[16]; /* the current register file */ - ARMword Cpsr; /* the current psr */ + // Order of the following register should not be modified + ARMword Reg[16]; // The current register file + ARMword Cpsr; // The current PSR ARMword Spsr_copy; ARMword phys_pc; ARMword Reg_usr[2]; - ARMword Reg_svc[2]; /* R13_SVC R14_SVC */ - ARMword Reg_abort[2]; /* R13_ABORT R14_ABORT */ - ARMword Reg_undef[2]; /* R13 UNDEF R14 UNDEF */ - ARMword Reg_irq[2]; /* R13_IRQ R14_IRQ */ - ARMword Reg_firq[7]; /* R8---R14 FIRQ */ - ARMword Spsr[7]; /* the exception psr's */ - ARMword Mode; /* the current mode */ - ARMword Bank; /* the current register bank */ - ARMword exclusive_tag; /* the address for which the local monitor is in exclusive access mode */ + ARMword Reg_svc[2]; // R13_SVC R14_SVC + ARMword Reg_abort[2]; // R13_ABORT R14_ABORT + ARMword Reg_undef[2]; // R13 UNDEF R14 UNDEF + ARMword Reg_irq[2]; // R13_IRQ R14_IRQ + ARMword Reg_firq[7]; // R8---R14 FIRQ + ARMword Spsr[7]; // The exception psr's + ARMword Mode; // The current mode + ARMword Bank; // The current register bank + ARMword exclusive_tag; // The address for which the local monitor is in exclusive access mode ARMword exclusive_state; ARMword exclusive_result; ARMword CP15[VFP_BASE - CP15_BASE]; - ARMword VFP[3]; /* FPSID, FPSCR, and FPEXC */ - /* VFPv2 and VFPv3-D16 has 16 doubleword registers (D0-D16 or S0-S31). - VFPv3-D32/ASIMD may have up to 32 doubleword registers (D0-D31), - and only 32 singleword registers are accessible (S0-S31). */ + ARMword VFP[3]; // FPSID, FPSCR, and FPEXC + // VFPv2 and VFPv3-D16 has 16 doubleword registers (D0-D16 or S0-S31). + // VFPv3-D32/ASIMD may have up to 32 doubleword registers (D0-D31), + // and only 32 singleword registers are accessible (S0-S31). ARMword ExtReg[VFP_REG_NUM]; /* ---- End of the ordered registers ---- */ - ARMword RegBank[7][16]; /* all the registers */ - //chy:2003-08-19, used in arm xscale - /* 40 bit accumulator. We always keep this 64 bits wide, - and move only 40 bits out of it in an MRA insn. */ - ARMdword Accumulator; - - ARMword NFlag, ZFlag, CFlag, VFlag, IFFlags; /* dummy flags for speed */ - unsigned long long int icounter, debug_icounter, kernel_icounter; + ARMword RegBank[7][16]; // all the registers + + ARMword NFlag, ZFlag, CFlag, VFlag, IFFlags; // Dummy flags for speed unsigned int shifter_carry_out; - /* add armv6 flags dyf:2010-08-09 */ + // Add armv6 flags dyf:2010-08-09 ARMword GEFlag, EFlag, AFlag, QFlag; - //chy:2003-08-19, used in arm v5e|xscale - ARMword SFlag; + #ifdef MODET - ARMword TFlag; /* Thumb state */ + ARMword TFlag; // Thumb state #endif - ARMword instr, pc, temp; /* saved register state */ - ARMword loaded, decoded; /* saved pipeline state */ - //chy 2006-04-12 for ICE breakpoint - ARMword loaded_addr, decoded_addr; /* saved pipeline state addr*/ - unsigned int NumScycles, NumNcycles, NumIcycles, NumCcycles, NumFcycles; /* emulated cycles used */ - unsigned long long NumInstrs; /* the number of instructions executed */ - unsigned NumInstrsToExecute; - ARMword currentexaddr; - ARMword currentexval; - ARMword currentexvald; - ARMword servaddr; + unsigned long long NumInstrs; // The number of instructions executed + unsigned NumInstrsToExecute; unsigned NextInstr; - unsigned VectorCatch; /* caught exception mask */ - unsigned CallDebug; /* set to call the debugger */ - unsigned CanWatch; /* set by memory interface if its willing to suffer the - overhead of checking for watchpoints on each memory - access */ - - ARMul_CPInits *CPInit[16]; /* coprocessor initialisers */ - ARMul_CPExits *CPExit[16]; /* coprocessor finalisers */ - ARMul_LDCs *LDC[16]; /* LDC instruction */ - ARMul_STCs *STC[16]; /* STC instruction */ - ARMul_MRCs *MRC[16]; /* MRC instruction */ - ARMul_MCRs *MCR[16]; /* MCR instruction */ - ARMul_MRRCs *MRRC[16]; /* MRRC instruction */ - ARMul_MCRRs *MCRR[16]; /* MCRR instruction */ - ARMul_CDPs *CDP[16]; /* CDP instruction */ - ARMul_CPReads *CPRead[16]; /* Read CP register */ - ARMul_CPWrites *CPWrite[16]; /* Write CP register */ - unsigned char *CPData[16]; /* Coprocessor data */ - unsigned char const *CPRegWords[16]; /* map of coprocessor register sizes */ - - unsigned Debug; /* show instructions as they are executed */ - unsigned NresetSig; /* reset the processor */ + unsigned VectorCatch; // Caught exception mask + + ARMul_CPInits* CPInit[16]; // Coprocessor initialisers + ARMul_CPExits* CPExit[16]; // Coprocessor finalisers + ARMul_LDCs* LDC[16]; // LDC instruction + ARMul_STCs* STC[16]; // STC instruction + ARMul_MRCs* MRC[16]; // MRC instruction + ARMul_MCRs* MCR[16]; // MCR instruction + ARMul_MRRCs* MRRC[16]; // MRRC instruction + ARMul_MCRRs* MCRR[16]; // MCRR instruction + ARMul_CDPs* CDP[16]; // CDP instruction + ARMul_CPReads* CPRead[16]; // Read CP register + ARMul_CPWrites* CPWrite[16]; // Write CP register + unsigned char* CPData[16]; // Coprocessor data + unsigned char const* CPRegWords[16]; // Map of coprocessor register sizes + + unsigned NresetSig; // Reset the processor unsigned NfiqSig; unsigned NirqSig; @@ -196,54 +174,34 @@ So, if lateabtSig=1, then it means Late Abort Model(Base Updated Abort Model) */ unsigned lateabtSig; - ARMword Vector; /* synthesize aborts in cycle modes */ - ARMword Aborted; /* sticky flag for aborts */ - ARMword Reseted; /* sticky flag for Reset */ - ARMword Inted, LastInted; /* sticky flags for interrupts */ - ARMword Base; /* extra hand for base writeback */ - ARMword AbortAddr; /* to keep track of Prefetch aborts */ - - int verbose; /* non-zero means print various messages like the banner */ - - int mmu_inited; - - //chy: 2003-08-11, for different arm core type - unsigned is_v4; /* Are we emulating a v4 architecture (or higher) ? */ - unsigned is_v5; /* Are we emulating a v5 architecture ? */ - unsigned is_v5e; /* Are we emulating a v5e architecture ? */ - unsigned is_v6; /* Are we emulating a v6 architecture ? */ - unsigned is_v7; /* Are we emulating a v7 architecture ? */ - unsigned is_XScale; /* Are we emulating an XScale architecture ? */ - unsigned is_iWMMXt; /* Are we emulating an iWMMXt co-processor ? */ - unsigned is_ep9312; /* Are we emulating a Cirrus Maverick co-processor ? */ - unsigned is_pxa27x; /* Are we emulating a Intel PXA27x co-processor ? */ - - //chy: seems only used in xscale's CP14 - ARMword CP14R0_CCD; /* used to count 64 clock cycles with CP14 R0 bit 3 set */ - - //teawater add for arm2x86 2005.07.05------------------------------------------- - //arm_arm A2-18 - int abort_model; //0 Base Restored Abort Model, 1 the Early Abort Model, 2 Base Updated Abort Model - - /*added by ksh in 2005-10-1*/ - cpu_config_t *cpu; - - /* added LPC remap function */ - int vector_remap_flag; - u32 vector_remap_addr; - u32 vector_remap_size; - - u32 step; - u32 cycle; - - /* monitored memory for exclusice access */ - ARMword exclusive_tag_array[128]; - /* 1 means exclusive access and 0 means open access */ - ARMword exclusive_access_state; + bool Aborted; // Sticky flag for aborts + bool Reseted; // Sticky flag for Reset + ARMword Inted, LastInted; // Sticky flags for interrupts + ARMword Base; // Extra hand for base writeback + ARMword AbortAddr; // To keep track of Prefetch aborts + ARMword Vector; // Synthesize aborts in cycle modes + + // For differentiating ARM core emulaiton. + bool is_v4; // Are we emulating a v4 architecture (or higher)? + bool is_v5; // Are we emulating a v5 architecture? + bool is_v5e; // Are we emulating a v5e architecture? + bool is_v6; // Are we emulating a v6 architecture? + bool is_v7; // Are we emulating a v7 architecture? + bool is_XScale; // Are we emulating an XScale architecture? + bool is_iWMMXt; // Are we emulating an iWMMXt co-processor? + bool is_ep9312; // Are we emulating a Cirrus Maverick co-processor? + bool is_pxa27x; // Are we emulating a Intel PXA27x co-processor? + + // ARM_ARM A2-18 + // 0 Base Restored Abort Model, 1 the Early Abort Model, 2 Base Updated Abort Model + int abort_model; + + // Added by ksh in 2005-10-1 + cpu_config_t* cpu; u32 CurrInstr; - u32 last_pc; /* the last pc executed */ - u32 last_instr; /* the last inst executed */ + u32 last_pc; // The last PC executed + u32 last_instr; // The last instruction executed u32 WriteAddr[17]; u32 WriteData[17]; u32 WritePc[17]; @@ -287,15 +245,6 @@ enum { ARM620 = ARM6 }; - -/***************************************************************************\ -* Macros to extract instruction fields * -\***************************************************************************/ - -#define BIT(n) ( (ARMword)(instr>>(n))&1) /* bit n of instruction */ -#define BITS(m,n) ( (ARMword)(instr<<(31-(n))) >> ((31-(n))+(m)) ) /* bits m to n of instr */ -#define TOPBITS(n) (instr >> (n)) /* bits 31 to n of instr */ - /***************************************************************************\ * The hardware vector addresses * \***************************************************************************/ @@ -339,13 +288,6 @@ enum { SYSTEM32MODE = 31 }; -#define ARM32BITMODE (state->Mode > 3) -#define ARM26BITMODE (state->Mode <= 3) -#define ARMMODE (state->Mode) -#define ARMul_MODEBITS 0x1fL -#define ARMul_MODE32BIT ARM32BITMODE -#define ARMul_MODE26BIT ARM26BITMODE - enum { USERBANK = 0, FIQBANK = 1, @@ -357,10 +299,6 @@ enum { SYSTEMBANK = USERBANK }; -#define BANK_CAN_ACCESS_SPSR(bank) \ - ((bank) != USERBANK && (bank) != SYSTEMBANK && (bank) != DUMMYBANK) - - /***************************************************************************\ * Definitons of things in the emulator * \***************************************************************************/ @@ -372,85 +310,7 @@ extern void ARMul_Reset(ARMul_State* state); #ifdef __cplusplus } #endif -extern ARMul_State *ARMul_NewState(ARMul_State* state); -extern ARMword ARMul_DoProg(ARMul_State* state); -extern ARMword ARMul_DoInstr(ARMul_State* state); - -/***************************************************************************\ -* Useful support routines * -\***************************************************************************/ - -extern ARMword ARMul_GetReg (ARMul_State* state, unsigned mode, unsigned reg); -extern void ARMul_SetReg (ARMul_State* state, unsigned mode, unsigned reg, ARMword value); -extern ARMword ARMul_GetPC(ARMul_State* state); -extern ARMword ARMul_GetNextPC(ARMul_State* state); -extern void ARMul_SetPC(ARMul_State* state, ARMword value); -extern ARMword ARMul_GetR15(ARMul_State* state); -extern void ARMul_SetR15(ARMul_State* state, ARMword value); - -extern ARMword ARMul_GetCPSR(ARMul_State* state); -extern void ARMul_SetCPSR(ARMul_State* state, ARMword value); -extern ARMword ARMul_GetSPSR(ARMul_State* state, ARMword mode); -extern void ARMul_SetSPSR(ARMul_State* state, ARMword mode, ARMword value); - -/***************************************************************************\ -* Definitons of things to handle aborts * -\***************************************************************************/ - -extern void ARMul_Abort(ARMul_State* state, ARMword address); -#ifdef MODET -#define ARMul_ABORTWORD (state->TFlag ? 0xefffdfff : 0xefffffff) /* SWI -1 */ -#define ARMul_PREFETCHABORT(address) if (state->AbortAddr == 1) \ - state->AbortAddr = (address & (state->TFlag ? ~1L : ~3L)) -#else -#define ARMul_ABORTWORD 0xefffffff /* SWI -1 */ -#define ARMul_PREFETCHABORT(address) if (state->AbortAddr == 1) \ - state->AbortAddr = (address & ~3L) -#endif -#define ARMul_DATAABORT(address) state->abortSig = HIGH ; \ - state->Aborted = ARMul_DataAbortV ; -#define ARMul_CLEARABORT state->abortSig = LOW - -/***************************************************************************\ -* Definitons of things in the memory interface * -\***************************************************************************/ - -extern unsigned ARMul_MemoryInit(ARMul_State* state, unsigned int initmemsize); -extern void ARMul_MemoryExit(ARMul_State* state); - -extern ARMword ARMul_LoadInstrS(ARMul_State* state, ARMword address, ARMword isize); -extern ARMword ARMul_LoadInstrN(ARMul_State* state, ARMword address, ARMword isize); -#ifdef __cplusplus -extern "C" { -#endif -extern ARMword ARMul_ReLoadInstr(ARMul_State* state, ARMword address, ARMword isize); -#ifdef __cplusplus - } -#endif -extern ARMword ARMul_LoadWordS(ARMul_State* state, ARMword address); -extern ARMword ARMul_LoadWordN(ARMul_State* state, ARMword address); -extern ARMword ARMul_LoadHalfWord(ARMul_State* state, ARMword address); -extern ARMword ARMul_LoadByte(ARMul_State* state, ARMword address); - -extern void ARMul_StoreWordS(ARMul_State* state, ARMword address, ARMword data); -extern void ARMul_StoreWordN(ARMul_State* state, ARMword address, ARMword data); -extern void ARMul_StoreHalfWord(ARMul_State* state, ARMword address, ARMword data); -extern void ARMul_StoreByte(ARMul_State* state, ARMword address, ARMword data); - -extern ARMword ARMul_SwapWord(ARMul_State* state, ARMword address, ARMword data); -extern ARMword ARMul_SwapByte(ARMul_State* state, ARMword address, ARMword data); - -extern void ARMul_Icycles(ARMul_State* state, unsigned number, ARMword address); -extern void ARMul_Ccycles(ARMul_State* state, unsigned number, ARMword address); - -extern ARMword ARMul_ReadWord(ARMul_State* state, ARMword address); -extern ARMword ARMul_ReadByte(ARMul_State* state, ARMword address); -extern void ARMul_WriteWord(ARMul_State* state, ARMword address, ARMword data); -extern void ARMul_WriteByte(ARMul_State* state, ARMword address, ARMword data); - -extern ARMword ARMul_MemAccess(ARMul_State* state, ARMword, ARMword, - ARMword, ARMword, ARMword, ARMword, ARMword, - ARMword, ARMword, ARMword); +extern ARMul_State* ARMul_NewState(ARMul_State* state); /***************************************************************************\ * Definitons of things in the co-processor interface * @@ -495,37 +355,10 @@ enum { ARMul_CP15_DBCON_E0 = 0x0003 }; -extern unsigned ARMul_CoProInit(ARMul_State* state); -extern void ARMul_CoProExit(ARMul_State* state); -extern void ARMul_CoProAttach (ARMul_State* state, unsigned number, - ARMul_CPInits* init, ARMul_CPExits* exit, - ARMul_LDCs* ldc, ARMul_STCs* stc, - ARMul_MRCs* mrc, ARMul_MCRs* mcr, - ARMul_MRRCs* mrrc, ARMul_MCRRs* mcrr, - ARMul_CDPs* cdp, - ARMul_CPReads* read, ARMul_CPWrites* write); -extern void ARMul_CoProDetach(ARMul_State* state, unsigned number); - /***************************************************************************\ * Definitons of things in the host environment * \***************************************************************************/ -extern unsigned ARMul_OSInit(ARMul_State* state); -extern void ARMul_OSExit(ARMul_State* state); - -#ifdef __cplusplus - extern "C" { -#endif - -extern unsigned ARMul_OSHandleSWI(ARMul_State* state, ARMword number); -#ifdef __cplusplus -} -#endif - -extern ARMword ARMul_OSLastErrorP(ARMul_State* state); -extern ARMword ARMul_Debug(ARMul_State* state, ARMword pc, ARMword instr); -extern unsigned ARMul_OSException(ARMul_State* state, ARMword vector, ARMword pc); - enum ConditionCode { EQ = 0, NE = 1, @@ -545,40 +378,9 @@ enum ConditionCode { NV = 15, }; -#ifndef NFLAG -#define NFLAG state->NFlag -#endif //NFLAG - -#ifndef ZFLAG -#define ZFLAG state->ZFlag -#endif //ZFLAG - -#ifndef CFLAG -#define CFLAG state->CFlag -#endif //CFLAG - -#ifndef VFLAG -#define VFLAG state->VFlag -#endif //VFLAG - -#ifndef IFLAG -#define IFLAG (state->IFFlags >> 1) -#endif //IFLAG - -#ifndef FFLAG -#define FFLAG (state->IFFlags & 1) -#endif //FFLAG - -#ifndef IFFLAGS -#define IFFLAGS state->IFFlags -#endif //VFLAG - extern bool AddOverflow(ARMword, ARMword, ARMword); extern bool SubOverflow(ARMword, ARMword, ARMword); -extern void ARMul_UndefInstr(ARMul_State*, ARMword); -extern void ARMul_FixCPSR(ARMul_State*, ARMword, ARMword); -extern void ARMul_FixSPSR(ARMul_State*, ARMword, ARMword); extern void ARMul_SelectProcessor(ARMul_State*, unsigned); extern u32 AddWithCarry(u32, u32, u32, bool*, bool*); diff --git a/src/core/arm/skyeye_common/armemu.h b/src/core/arm/skyeye_common/armemu.h index 7e10dad86..beee54c9a 100644 --- a/src/core/arm/skyeye_common/armemu.h +++ b/src/core/arm/skyeye_common/armemu.h @@ -19,12 +19,6 @@ #include "core/arm/skyeye_common/armdefs.h" -/* Shift Opcodes. */ -#define LSL 0 -#define LSR 1 -#define ASR 2 -#define ROR 3 - /* Macros to twiddle the status flags and mode. */ #define NBIT ((unsigned)1L << 31) #define ZBIT (1L << 30) @@ -38,73 +32,6 @@ #define R15FBIT (1L << 26) #define R15IFBITS (3L << 26) -#ifdef MODET /* Thumb support. */ -/* ??? This bit is actually in the low order bit of the PC in the hardware. - It isn't clear if the simulator needs to model that or not. */ -#define TBIT (1L << 5) -#define TFLAG state->TFlag -#define SETT state->TFlag = 1 -#define CLEART state->TFlag = 0 -#define ASSIGNT(res) state->TFlag = res -#define INSN_SIZE (TFLAG ? 2 : 4) -#else -#define INSN_SIZE 4 -#endif - -/*add armv6 CPSR feature*/ -#define EFLAG state->EFlag -#define SETE state->EFlag = 1 -#define CLEARE state->EFlag = 0 -#define ASSIGNE(res) state->NFlag = res - -#define AFLAG state->AFlag -#define SETA state->AFlag = 1 -#define CLEARA state->AFlag = 0 -#define ASSIGNA(res) state->NFlag = res - -#define QFLAG state->QFlag -#define SETQ state->QFlag = 1 -#define CLEARQ state->AFlag = 0 -#define ASSIGNQ(res) state->QFlag = res - -/* add end */ - -#define NFLAG state->NFlag -#define SETN state->NFlag = 1 -#define CLEARN state->NFlag = 0 -#define ASSIGNN(res) state->NFlag = res - -#define ZFLAG state->ZFlag -#define SETZ state->ZFlag = 1 -#define CLEARZ state->ZFlag = 0 -#define ASSIGNZ(res) state->ZFlag = res - -#define CFLAG state->CFlag -#define SETC state->CFlag = 1 -#define CLEARC state->CFlag = 0 -#define ASSIGNC(res) state->CFlag = res - -#define VFLAG state->VFlag -#define SETV state->VFlag = 1 -#define CLEARV state->VFlag = 0 -#define ASSIGNV(res) state->VFlag = res - -#define SFLAG state->SFlag -#define SETS state->SFlag = 1 -#define CLEARS state->SFlag = 0 -#define ASSIGNS(res) state->SFlag = res - -#define IFLAG (state->IFFlags >> 1) -#define FFLAG (state->IFFlags & 1) -#define IFFLAGS state->IFFlags -#define ASSIGNINT(res) state->IFFlags = (((res) >> 6) & 3) -#define ASSIGNR15INT(res) state->IFFlags = (((res) >> 26) & 3) ; - -#define PSR_FBITS (0xff000000L) -#define PSR_SBITS (0x00ff0000L) -#define PSR_XBITS (0x0000ff00L) -#define PSR_CBITS (0x000000ffL) - #if defined MODE32 || defined MODET #define CCBITS (0xf8000000L) #else @@ -128,7 +55,6 @@ #define R15PCBITS (0x03fffffcL) #endif -#define R15PCMODEBITS (0x03ffffffL) #define R15MODEBITS (0x3L) #ifdef MODE32 @@ -149,106 +75,7 @@ #define R15PCMODE (state->Reg[15] & (R15PCBITS | R15MODEBITS)) #define R15MODE (state->Reg[15] & R15MODEBITS) -#define ECC ((NFLAG << 31) | (ZFLAG << 30) | (CFLAG << 29) | (VFLAG << 28) | (QFLAG << 27)) -#define EINT (IFFLAGS << 6) -#define ER15INT (IFFLAGS << 26) -#define EMODE (state->Mode) -#define EGEBITS (state->GEFlag & 0x000F0000) - -#ifdef MODET -#define CPSR (ECC | EGEBITS | (EFLAG << 9) | (AFLAG << 8) | EINT | (TFLAG << 5) | EMODE) -#else -#define CPSR (ECC | EINT | EMODE) -#endif - -#ifdef MODE32 -#define PATCHR15 -#else -#define PATCHR15 state->Reg[15] = ECC | ER15INT | EMODE | R15PC -#endif - -#define GETSPSR(bank) (ARMul_GetSPSR (state, EMODE)) -#define SETPSR_F(d,s) d = ((d) & ~PSR_FBITS) | ((s) & PSR_FBITS) -#define SETPSR_S(d,s) d = ((d) & ~PSR_SBITS) | ((s) & PSR_SBITS) -#define SETPSR_X(d,s) d = ((d) & ~PSR_XBITS) | ((s) & PSR_XBITS) -#define SETPSR_C(d,s) d = ((d) & ~PSR_CBITS) | ((s) & PSR_CBITS) - -#define SETR15PSR(s) \ - do \ - { \ - if (state->Mode == USER26MODE) \ - { \ - state->Reg[15] = ((s) & CCBITS) | R15PC | ER15INT | EMODE; \ - ASSIGNN ((state->Reg[15] & NBIT) != 0); \ - ASSIGNZ ((state->Reg[15] & ZBIT) != 0); \ - ASSIGNC ((state->Reg[15] & CBIT) != 0); \ - ASSIGNV ((state->Reg[15] & VBIT) != 0); \ - } \ - else \ - { \ - state->Reg[15] = R15PC | ((s) & (CCBITS | R15INTBITS | R15MODEBITS)); \ - ARMul_R15Altered (state); \ - } \ - } \ - while (0) - -#define SETABORT(i, m, d) \ - do \ - { \ - int SETABORT_mode = (m); \ - \ - ARMul_SetSPSR (state, SETABORT_mode, ARMul_GetCPSR (state)); \ - ARMul_SetCPSR (state, ((ARMul_GetCPSR (state) & ~(EMODE | TBIT)) \ - | (i) | SETABORT_mode)); \ - state->Reg[14] = temp - (d); \ - } \ - while (0) - -#ifndef MODE32 -#define VECTORS 0x20 -#define LEGALADDR 0x03ffffff -#define VECTORACCESS(address) (address < VECTORS && ARMul_MODE26BIT && state->prog32Sig) -#define ADDREXCEPT(address) (address > LEGALADDR && !state->data32Sig) -#endif - -#define INTERNALABORT(address) \ - do \ - { \ - if (address < VECTORS) \ - state->Aborted = ARMul_DataAbortV; \ - else \ - state->Aborted = ARMul_AddrExceptnV; \ - } \ - while (0) - -#ifdef MODE32 -#define TAKEABORT ARMul_Abort (state, ARMul_DataAbortV) -#else -#define TAKEABORT \ - do \ - { \ - if (state->Aborted == ARMul_AddrExceptnV) \ - ARMul_Abort (state, ARMul_AddrExceptnV); \ - else \ - ARMul_Abort (state, ARMul_DataAbortV); \ - } \ - while (0) -#endif - -#define CPTAKEABORT \ - do \ - { \ - if (!state->Aborted) \ - ARMul_Abort (state, ARMul_UndefinedInstrV); \ - else if (state->Aborted == ARMul_AddrExceptnV) \ - ARMul_Abort (state, ARMul_AddrExceptnV); \ - else \ - ARMul_Abort (state, ARMul_DataAbortV); \ - } \ - while (0); - - -/* Different ways to start the next instruction. */ +// Different ways to start the next instruction. #define SEQ 0 #define NONSEQ 1 #define PCINCEDSEQ 2 @@ -256,368 +83,27 @@ #define PRIMEPIPE 4 #define RESUME 8 -/************************************/ -/* shenoubang 2012-3-11 */ -/* for armv7 DBG DMB DSB instr*/ -/************************************/ -#define MBReqTypes_Writes 0 -#define MBReqTypes_All 1 - -#define NORMALCYCLE state->NextInstr = 0 -#define BUSUSEDN state->NextInstr |= 1 /* The next fetch will be an N cycle. */ -#define BUSUSEDINCPCS \ - do \ - { \ - if (! state->is_v4) \ - { \ - /* A standard PC inc and an S cycle. */ \ - state->Reg[15] += INSN_SIZE; \ - state->NextInstr = (state->NextInstr & 0xff) | 2; \ - } \ - } \ - while (0) - -#define BUSUSEDINCPCN \ - do \ - { \ - if (state->is_v4) \ - BUSUSEDN; \ - else \ - { \ - /* A standard PC inc and an N cycle. */ \ - state->Reg[15] += INSN_SIZE; \ - state->NextInstr |= 3; \ - } \ - } \ - while (0) - -#define INCPC \ - do \ - { \ - /* A standard PC inc. */ \ - state->Reg[15] += INSN_SIZE; \ - state->NextInstr |= 2; \ - } \ - while (0) - #define FLUSHPIPE state->NextInstr |= PRIMEPIPE -/* Cycle based emulation. */ - -#define OUTPUTCP(i,a,b) -#define NCYCLE -#define SCYCLE -#define ICYCLE -#define CCYCLE -#define NEXTCYCLE(c) - -/* Macros to extract parts of instructions. */ -#define DESTReg (BITS (12, 15)) -#define LHSReg (BITS (16, 19)) -#define RHSReg (BITS ( 0, 3)) - -#define DEST (state->Reg[DESTReg]) - -#ifdef MODE32 -#ifdef MODET -#define LHS ((LHSReg == 15) ? (state->Reg[15] & 0xFFFFFFFC) : (state->Reg[LHSReg])) -#define RHS ((RHSReg == 15) ? (state->Reg[15] & 0xFFFFFFFC) : (state->Reg[RHSReg])) -#else -#define LHS (state->Reg[LHSReg]) -#define RHS (state->Reg[RHSReg]) -#endif -#else -#define LHS ((LHSReg == 15) ? R15PC : (state->Reg[LHSReg])) -#define RHS ((RHSReg == 15) ? R15PC : (state->Reg[RHSReg])) -#endif - -#define MULDESTReg (BITS (16, 19)) -#define MULLHSReg (BITS ( 0, 3)) -#define MULRHSReg (BITS ( 8, 11)) -#define MULACCReg (BITS (12, 15)) - -#define DPImmRHS (ARMul_ImmedTable[BITS(0, 11)]) -#define DPSImmRHS temp = BITS(0,11) ; \ - rhs = ARMul_ImmedTable[temp] ; \ - if (temp > 255) /* There was a shift. */ \ - ASSIGNC (rhs >> 31) ; - -#ifdef MODE32 -#define DPRegRHS ((BITS (4,11) == 0) ? state->Reg[RHSReg] \ - : GetDPRegRHS (state, instr)) -#define DPSRegRHS ((BITS (4,11) == 0) ? state->Reg[RHSReg] \ - : GetDPSRegRHS (state, instr)) -#else -#define DPRegRHS ((BITS (0, 11) < 15) ? state->Reg[RHSReg] \ - : GetDPRegRHS (state, instr)) -#define DPSRegRHS ((BITS (0, 11) < 15) ? state->Reg[RHSReg] \ - : GetDPSRegRHS (state, instr)) -#endif - -#define LSBase state->Reg[LHSReg] -#define LSImmRHS (BITS(0,11)) - -#ifdef MODE32 -#define LSRegRHS ((BITS (4, 11) == 0) ? state->Reg[RHSReg] \ - : GetLSRegRHS (state, instr)) -#else -#define LSRegRHS ((BITS (0, 11) < 15) ? state->Reg[RHSReg] \ - : GetLSRegRHS (state, instr)) -#endif - -#define LSMNumRegs ((ARMword) ARMul_BitList[BITS (0, 7)] + \ - (ARMword) ARMul_BitList[BITS (8, 15)] ) -#define LSMBaseFirst ((LHSReg == 0 && BIT (0)) || \ - (BIT (LHSReg) && BITS (0, LHSReg - 1) == 0)) - -#define SWAPSRC (state->Reg[RHSReg]) - -#define LSCOff (BITS (0, 7) << 2) -#define CPNum BITS (8, 11) - -/* Determine if access to coprocessor CP is permitted. - The XScale has a register in CP15 which controls access to CP0 - CP13. */ -//chy 2003-09-03, new CP_ACCESS_ALLOWED -/* -#define CP_ACCESS_ALLOWED(STATE, CP) \ - ( ((CP) >= 14) \ - || (! (STATE)->is_XScale) \ - || (read_cp15_reg (15, 0, 1) & (1 << (CP)))) -*/ -#define CP_ACCESS_ALLOWED(STATE, CP) \ - ( ((CP) >= 14) ) \ - -/* Macro to rotate n right by b bits. */ +// Macro to rotate n right by b bits. #define ROTATER(n, b) (((n) >> (b)) | ((n) << (32 - (b)))) -/* Macros to store results of instructions. */ -#define WRITEDEST(d) \ - do \ - { \ - if (DESTReg == 15) \ - WriteR15 (state, d); \ - else \ - DEST = d; \ - } \ - while (0) - -#define WRITESDEST(d) \ - do \ - { \ - if (DESTReg == 15) \ - WriteSR15 (state, d); \ - else \ - { \ - DEST = d; \ - ARMul_NegZero (state, d); \ - } \ - } \ - while (0) - -#define WRITEDESTB(d) \ - do \ - { \ - if (DESTReg == 15){ \ - WriteR15Branch (state, d); \ - } \ - else{ \ - DEST = d; \ - } \ - } \ - while (0) - -#define BYTETOBUS(data) ((data & 0xff) | \ - ((data & 0xff) << 8) | \ - ((data & 0xff) << 16) | \ - ((data & 0xff) << 24)) - -#define BUSTOBYTE(address, data) \ - do \ - { \ - if (state->bigendSig) \ - temp = (data >> (((address ^ 3) & 3) << 3)) & 0xff; \ - else \ - temp = (data >> ((address & 3) << 3)) & 0xff; \ - } \ - while (0) - -#define LOADMULT(instr, address, wb) LoadMult (state, instr, address, wb) -#define LOADSMULT(instr, address, wb) LoadSMult (state, instr, address, wb) -#define STOREMULT(instr, address, wb) StoreMult (state, instr, address, wb) -#define STORESMULT(instr, address, wb) StoreSMult (state, instr, address, wb) - -#define POSBRANCH ((instr & 0x7fffff) << 2) -#define NEGBRANCH ((0xff000000 |(instr & 0xffffff)) << 2) - - -/* Values for Emulate. */ -#define STOP 0 /* stop */ -#define CHANGEMODE 1 /* change mode */ -#define ONCE 2 /* execute just one interation */ -#define RUN 3 /* continuous execution */ - -/* Stuff that is shared across modes. */ -extern unsigned ARMul_MultTable[]; /* Number of I cycles for a mult. */ -extern ARMword ARMul_ImmedTable[]; /* Immediate DP LHS values. */ -extern char ARMul_BitList[]; /* Number of bits in a byte table. */ - -#define EVENTLISTSIZE 1024L - -/* Thumb support. */ -typedef enum -{ - t_undefined, /* Undefined Thumb instruction. */ - t_decoded, /* Instruction decoded to ARM equivalent. */ - t_branch /* Thumb branch (already processed). */ -} -tdstate; - -/********************************************************************************* - * Check all the possible undef or unpredict behavior, Some of them probably is - * out-of-updated with the newer ISA. - * -- Michael.Kang - ********************************************************************************/ -#define UNDEF_WARNING LOG_WARNING(Core_ARM11, "undefined or unpredicted behavior for arm instruction."); - -/* Macros to scrutinize instructions. */ -#define UNDEF_Test UNDEF_WARNING -//#define UNDEF_Test - -//#define UNDEF_Shift UNDEF_WARNING -#define UNDEF_Shift - -//#define UNDEF_MSRPC UNDEF_WARNING -#define UNDEF_MSRPC - -//#define UNDEF_MRSPC UNDEF_WARNING -#define UNDEF_MRSPC - -#define UNDEF_MULPCDest UNDEF_WARNING -//#define UNDEF_MULPCDest - -#define UNDEF_MULDestEQOp1 UNDEF_WARNING -//#define UNDEF_MULDestEQOp1 - -//#define UNDEF_LSRBPC UNDEF_WARNING -#define UNDEF_LSRBPC - -//#define UNDEF_LSRBaseEQOffWb UNDEF_WARNING -#define UNDEF_LSRBaseEQOffWb - -//#define UNDEF_LSRBaseEQDestWb UNDEF_WARNING -#define UNDEF_LSRBaseEQDestWb - -//#define UNDEF_LSRPCBaseWb UNDEF_WARNING -#define UNDEF_LSRPCBaseWb - -//#define UNDEF_LSRPCOffWb UNDEF_WARNING -#define UNDEF_LSRPCOffWb - -//#define UNDEF_LSMNoRegs UNDEF_WARNING -#define UNDEF_LSMNoRegs - -//#define UNDEF_LSMPCBase UNDEF_WARNING -#define UNDEF_LSMPCBase - -//#define UNDEF_LSMUserBankWb UNDEF_WARNING -#define UNDEF_LSMUserBankWb - -//#define UNDEF_LSMBaseInListWb UNDEF_WARNING -#define UNDEF_LSMBaseInListWb - -#define UNDEF_SWPPC UNDEF_WARNING -//#define UNDEF_SWPPC - -#define UNDEF_CoProHS UNDEF_WARNING -//#define UNDEF_CoProHS - -#define UNDEF_MCRPC UNDEF_WARNING -//#define UNDEF_MCRPC - -//#define UNDEF_LSCPCBaseWb UNDEF_WARNING -#define UNDEF_LSCPCBaseWb - -#define UNDEF_UndefNotBounced UNDEF_WARNING -//#define UNDEF_UndefNotBounced - -#define UNDEF_ShortInt UNDEF_WARNING -//#define UNDEF_ShortInt - -#define UNDEF_IllegalMode UNDEF_WARNING -//#define UNDEF_IllegalMode - -#define UNDEF_Prog32SigChange UNDEF_WARNING -//#define UNDEF_Prog32SigChange - -#define UNDEF_Data32SigChange UNDEF_WARNING -//#define UNDEF_Data32SigChange - -/* Prototypes for exported functions. */ -extern unsigned ARMul_NthReg (ARMword, unsigned); - -/* Prototypes for exported functions. */ -#ifdef __cplusplus - extern "C" { -#endif -extern ARMword ARMul_Emulate26 (ARMul_State *); -extern ARMword ARMul_Emulate32 (ARMul_State *); -#ifdef __cplusplus - } -#endif -extern unsigned IntPending (ARMul_State *); -extern void ARMul_CPSRAltered (ARMul_State *); -extern void ARMul_R15Altered (ARMul_State *); -extern ARMword ARMul_GetPC (ARMul_State *); -extern ARMword ARMul_GetNextPC (ARMul_State *); -extern ARMword ARMul_GetR15 (ARMul_State *); -extern ARMword ARMul_GetCPSR (ARMul_State *); -extern void ARMul_NegZero (ARMul_State *, ARMword); -extern void ARMul_SetPC (ARMul_State *, ARMword); -extern void ARMul_SetR15 (ARMul_State *, ARMword); -extern void ARMul_SetCPSR (ARMul_State *, ARMword); -extern ARMword ARMul_GetSPSR (ARMul_State *, ARMword); -extern void ARMul_Abort26 (ARMul_State *, ARMword); -extern void ARMul_Abort32 (ARMul_State *, ARMword); -extern ARMword ARMul_MRC (ARMul_State *, ARMword); -extern void ARMul_MRRC (ARMul_State *, ARMword, ARMword *, ARMword *); -extern void ARMul_CDP (ARMul_State *, ARMword); -extern void ARMul_LDC (ARMul_State *, ARMword, ARMword); -extern void ARMul_STC (ARMul_State *, ARMword, ARMword); -extern void ARMul_MCR (ARMul_State *, ARMword, ARMword); -extern void ARMul_MCRR (ARMul_State *, ARMword, ARMword, ARMword); -extern void ARMul_SetSPSR (ARMul_State *, ARMword, ARMword); -extern ARMword ARMul_SwitchMode (ARMul_State *, ARMword, ARMword); -extern ARMword ARMul_Align (ARMul_State *, ARMword, ARMword); -extern ARMword ARMul_SwitchMode (ARMul_State *, ARMword, ARMword); -extern void ARMul_MSRCpsr (ARMul_State *, ARMword, ARMword); -extern void ARMul_SubOverflow (ARMul_State *, ARMword, ARMword, ARMword); -extern void ARMul_AddOverflow (ARMul_State *, ARMword, ARMword, ARMword); -extern void ARMul_SubCarry (ARMul_State *, ARMword, ARMword, ARMword); -extern void ARMul_AddCarry (ARMul_State *, ARMword, ARMword, ARMword); -extern tdstate ARMul_ThumbDecode (ARMul_State *, ARMword, ARMword, ARMword *); -extern ARMword ARMul_GetReg (ARMul_State *, unsigned, unsigned); -extern void ARMul_SetReg (ARMul_State *, unsigned, unsigned, ARMword); - -/* Coprocessor support functions. */ -extern unsigned ARMul_CoProInit (ARMul_State *); -extern void ARMul_CoProExit (ARMul_State *); -extern void ARMul_CoProAttach (ARMul_State *, unsigned, ARMul_CPInits *, - ARMul_CPExits *, ARMul_LDCs *, ARMul_STCs *, - ARMul_MRCs *, ARMul_MCRs *, ARMul_MRRCs *, ARMul_MCRRs *, - ARMul_CDPs *, ARMul_CPReads *, ARMul_CPWrites *); -extern void ARMul_CoProDetach (ARMul_State *, unsigned); -extern ARMword read_cp15_reg (unsigned, unsigned, unsigned); - -extern unsigned DSPLDC4 (ARMul_State *, unsigned, ARMword, ARMword); -extern unsigned DSPMCR4 (ARMul_State *, unsigned, ARMword, ARMword); -extern unsigned DSPMRC4 (ARMul_State *, unsigned, ARMword, ARMword *); -extern unsigned DSPSTC4 (ARMul_State *, unsigned, ARMword, ARMword *); -extern unsigned DSPCDP4 (ARMul_State *, unsigned, ARMword); -extern unsigned DSPMCR5 (ARMul_State *, unsigned, ARMword, ARMword); -extern unsigned DSPMRC5 (ARMul_State *, unsigned, ARMword, ARMword *); -extern unsigned DSPLDC5 (ARMul_State *, unsigned, ARMword, ARMword); -extern unsigned DSPSTC5 (ARMul_State *, unsigned, ARMword, ARMword *); -extern unsigned DSPCDP5 (ARMul_State *, unsigned, ARMword); -extern unsigned DSPMCR6 (ARMul_State *, unsigned, ARMword, ARMword); -extern unsigned DSPMRC6 (ARMul_State *, unsigned, ARMword, ARMword *); -extern unsigned DSPCDP6 (ARMul_State *, unsigned, ARMword); +// Values for Emulate. +#define STOP 0 // stop +#define CHANGEMODE 1 // change mode +#define ONCE 2 // execute just one interation +#define RUN 3 // continuous execution + +// Stuff that is shared across modes. +extern unsigned ARMul_MultTable[]; // Number of I cycles for a mult. +extern ARMword ARMul_ImmedTable[]; // Immediate DP LHS values. +extern char ARMul_BitList[]; // Number of bits in a byte table. + +// Coprocessor support functions. +extern void ARMul_CoProInit(ARMul_State*); +extern void ARMul_CoProExit(ARMul_State*); +extern void ARMul_CoProAttach(ARMul_State*, unsigned, ARMul_CPInits*, + ARMul_CPExits*, ARMul_LDCs*, ARMul_STCs*, + ARMul_MRCs*, ARMul_MCRs*, ARMul_MRRCs*, ARMul_MCRRs*, + ARMul_CDPs*, ARMul_CPReads*, ARMul_CPWrites*); +extern void ARMul_CoProDetach(ARMul_State*, unsigned); diff --git a/src/core/arm/skyeye_common/vfp/asm_vfp.h b/src/core/arm/skyeye_common/vfp/asm_vfp.h index e113eaf29..ccb7cf4d7 100644 --- a/src/core/arm/skyeye_common/vfp/asm_vfp.h +++ b/src/core/arm/skyeye_common/vfp/asm_vfp.h @@ -7,80 +7,77 @@ #pragma once -#define FPSID cr0 -#define FPSCR cr1 -#define MVFR1 cr6 -#define MVFR0 cr7 -#define FPEXC cr8 -#define FPINST cr9 -#define FPINST2 cr10 +// FPSID Information +// Note that these are used as values and not as flags. +enum : u32 { + VFP_FPSID_IMPLMEN = 0, // Implementation code. Should be the same as cp15 0 c0 0 + VFP_FPSID_SW = 0, // Software emulation bit value + VFP_FPSID_SUBARCH = 0x2, // Subarchitecture version number + VFP_FPSID_PARTNUM = 0x1, // Part number + VFP_FPSID_VARIANT = 0x1, // Variant number + VFP_FPSID_REVISION = 0x1 // Revision number +}; -/* FPSID bits */ -#define FPSID_IMPLEMENTER_BIT (24) -#define FPSID_IMPLEMENTER_MASK (0xff << FPSID_IMPLEMENTER_BIT) -#define FPSID_SOFTWARE (1<<23) -#define FPSID_FORMAT_BIT (21) -#define FPSID_FORMAT_MASK (0x3 << FPSID_FORMAT_BIT) -#define FPSID_NODOUBLE (1<<20) -#define FPSID_ARCH_BIT (16) -#define FPSID_ARCH_MASK (0xF << FPSID_ARCH_BIT) -#define FPSID_PART_BIT (8) -#define FPSID_PART_MASK (0xFF << FPSID_PART_BIT) -#define FPSID_VARIANT_BIT (4) -#define FPSID_VARIANT_MASK (0xF << FPSID_VARIANT_BIT) -#define FPSID_REV_BIT (0) -#define FPSID_REV_MASK (0xF << FPSID_REV_BIT) +// FPEXC bits +enum : u32 { + FPEXC_EX = (1U << 31U), + FPEXC_EN = (1 << 30), + FPEXC_DEX = (1 << 29), + FPEXC_FP2V = (1 << 28), + FPEXC_VV = (1 << 27), + FPEXC_TFV = (1 << 26), + FPEXC_LENGTH_BIT = (8), + FPEXC_LENGTH_MASK = (7 << FPEXC_LENGTH_BIT), + FPEXC_IDF = (1 << 7), + FPEXC_IXF = (1 << 4), + FPEXC_UFF = (1 << 3), + FPEXC_OFF = (1 << 2), + FPEXC_DZF = (1 << 1), + FPEXC_IOF = (1 << 0), + FPEXC_TRAP_MASK = (FPEXC_IDF|FPEXC_IXF|FPEXC_UFF|FPEXC_OFF|FPEXC_DZF|FPEXC_IOF) +}; -/* FPEXC bits */ -#define FPEXC_EX (1 << 31) -#define FPEXC_EN (1 << 30) -#define FPEXC_DEX (1 << 29) -#define FPEXC_FP2V (1 << 28) -#define FPEXC_VV (1 << 27) -#define FPEXC_TFV (1 << 26) -#define FPEXC_LENGTH_BIT (8) -#define FPEXC_LENGTH_MASK (7 << FPEXC_LENGTH_BIT) -#define FPEXC_IDF (1 << 7) -#define FPEXC_IXF (1 << 4) -#define FPEXC_UFF (1 << 3) -#define FPEXC_OFF (1 << 2) -#define FPEXC_DZF (1 << 1) -#define FPEXC_IOF (1 << 0) -#define FPEXC_TRAP_MASK (FPEXC_IDF|FPEXC_IXF|FPEXC_UFF|FPEXC_OFF|FPEXC_DZF|FPEXC_IOF) +// FPSCR Flags +enum : u32 { + FPSCR_NFLAG = (1U << 31U), // Negative condition flag + FPSCR_ZFLAG = (1 << 30), // Zero condition flag + FPSCR_CFLAG = (1 << 29), // Carry condition flag + FPSCR_VFLAG = (1 << 28), // Overflow condition flag -/* FPSCR bits */ -#define FPSCR_DEFAULT_NAN (1<<25) -#define FPSCR_FLUSHTOZERO (1<<24) -#define FPSCR_ROUND_NEAREST (0<<22) -#define FPSCR_ROUND_PLUSINF (1<<22) -#define FPSCR_ROUND_MINUSINF (2<<22) -#define FPSCR_ROUND_TOZERO (3<<22) -#define FPSCR_RMODE_BIT (22) -#define FPSCR_RMODE_MASK (3 << FPSCR_RMODE_BIT) -#define FPSCR_STRIDE_BIT (20) -#define FPSCR_STRIDE_MASK (3 << FPSCR_STRIDE_BIT) -#define FPSCR_LENGTH_BIT (16) -#define FPSCR_LENGTH_MASK (7 << FPSCR_LENGTH_BIT) -#define FPSCR_IOE (1<<8) -#define FPSCR_DZE (1<<9) -#define FPSCR_OFE (1<<10) -#define FPSCR_UFE (1<<11) -#define FPSCR_IXE (1<<12) -#define FPSCR_IDE (1<<15) -#define FPSCR_IOC (1<<0) -#define FPSCR_DZC (1<<1) -#define FPSCR_OFC (1<<2) -#define FPSCR_UFC (1<<3) -#define FPSCR_IXC (1<<4) -#define FPSCR_IDC (1<<7) + FPSCR_QC = (1 << 27), // Cumulative saturation bit + FPSCR_AHP = (1 << 26), // Alternative half-precision control bit + FPSCR_DEFAULT_NAN = (1 << 25), // Default NaN mode control bit + FPSCR_FLUSH_TO_ZERO = (1 << 24), // Flush-to-zero mode control bit + FPSCR_RMODE_MASK = (3 << 22), // Rounding Mode bit mask + FPSCR_STRIDE_MASK = (3 << 20), // Vector stride bit mask + FPSCR_LENGTH_MASK = (7 << 16), // Vector length bit mask -/* MVFR0 bits */ -#define MVFR0_A_SIMD_BIT (0) -#define MVFR0_A_SIMD_MASK (0xf << MVFR0_A_SIMD_BIT) + FPSCR_IDE = (1 << 15), // Input Denormal exception trap enable. + FPSCR_IXE = (1 << 12), // Inexact exception trap enable + FPSCR_UFE = (1 << 11), // Undeflow exception trap enable + FPSCR_OFE = (1 << 10), // Overflow exception trap enable + FPSCR_DZE = (1 << 9), // Division by Zero exception trap enable + FPSCR_IOE = (1 << 8), // Invalid Operation exception trap enable -/* Bit patterns for decoding the packaged operation descriptors */ -#define VFPOPDESC_LENGTH_BIT (9) -#define VFPOPDESC_LENGTH_MASK (0x07 << VFPOPDESC_LENGTH_BIT) -#define VFPOPDESC_UNUSED_BIT (24) -#define VFPOPDESC_UNUSED_MASK (0xFF << VFPOPDESC_UNUSED_BIT) -#define VFPOPDESC_OPDESC_MASK (~(VFPOPDESC_LENGTH_MASK | VFPOPDESC_UNUSED_MASK)) + FPSCR_IDC = (1 << 7), // Input Denormal cumulative exception bit + FPSCR_IXC = (1 << 4), // Inexact cumulative exception bit + FPSCR_UFC = (1 << 3), // Undeflow cumulative exception bit + FPSCR_OFC = (1 << 2), // Overflow cumulative exception bit + FPSCR_DZC = (1 << 1), // Division by Zero cumulative exception bit + FPSCR_IOC = (1 << 0), // Invalid Operation cumulative exception bit +}; + +// FPSCR bit offsets +enum : u32 { + FPSCR_RMODE_BIT = 22, + FPSCR_STRIDE_BIT = 20, + FPSCR_LENGTH_BIT = 16, +}; + +// FPSCR rounding modes +enum : u32 { + FPSCR_ROUND_NEAREST = (0 << 22), + FPSCR_ROUND_PLUSINF = (1 << 22), + FPSCR_ROUND_MINUSINF = (2 << 22), + FPSCR_ROUND_TOZERO = (3 << 22) +}; diff --git a/src/core/arm/skyeye_common/vfp/vfp.cpp b/src/core/arm/skyeye_common/vfp/vfp.cpp index bff296448..1cf146c53 100644 --- a/src/core/arm/skyeye_common/vfp/vfp.cpp +++ b/src/core/arm/skyeye_common/vfp/vfp.cpp @@ -23,6 +23,7 @@ #include "common/common.h" #include "core/arm/skyeye_common/armdefs.h" +#include "core/arm/skyeye_common/vfp/asm_vfp.h" #include "core/arm/skyeye_common/vfp/vfp.h" //ARMul_State* persistent_state; /* function calls from SoftFloat lib don't have an access to ARMul_state. */ @@ -43,12 +44,12 @@ unsigned VFPInit(ARMul_State* state) unsigned VFPMRC(ARMul_State* state, unsigned type, u32 instr, u32* value) { /* MRC<c> <coproc>,<opc1>,<Rt>,<CRn>,<CRm>{,<opc2>} */ - int CoProc = BITS (8, 11); /* 10 or 11 */ - int OPC_1 = BITS (21, 23); - int Rt = BITS (12, 15); - int CRn = BITS (16, 19); - int CRm = BITS (0, 3); - int OPC_2 = BITS (5, 7); + int CoProc = BITS(instr, 8, 11); /* 10 or 11 */ + int OPC_1 = BITS(instr, 21, 23); + int Rt = BITS(instr, 12, 15); + int CRn = BITS(instr, 16, 19); + int CRm = BITS(instr, 0, 3); + int OPC_2 = BITS(instr, 5, 7); /* TODO check access permission */ @@ -60,7 +61,7 @@ unsigned VFPMRC(ARMul_State* state, unsigned type, u32 instr, u32* value) { /* VMOV r to s */ /* Transfering Rt is not mandatory, as the value of interest is pointed by value */ - VMOVBRS(state, BIT(20), Rt, BIT(7)|CRn<<1, value); + VMOVBRS(state, BIT(instr, 20), Rt, BIT(instr, 7)|CRn<<1, value); return ARMul_DONE; } @@ -79,12 +80,12 @@ unsigned VFPMRC(ARMul_State* state, unsigned type, u32 instr, u32* value) unsigned VFPMCR(ARMul_State* state, unsigned type, u32 instr, u32 value) { /* MCR<c> <coproc>,<opc1>,<Rt>,<CRn>,<CRm>{,<opc2>} */ - int CoProc = BITS (8, 11); /* 10 or 11 */ - int OPC_1 = BITS (21, 23); - int Rt = BITS (12, 15); - int CRn = BITS (16, 19); - int CRm = BITS (0, 3); - int OPC_2 = BITS (5, 7); + int CoProc = BITS(instr, 8, 11); /* 10 or 11 */ + int OPC_1 = BITS(instr, 21, 23); + int Rt = BITS(instr, 12, 15); + int CRn = BITS(instr, 16, 19); + int CRm = BITS(instr, 0, 3); + int OPC_2 = BITS(instr, 5, 7); /* TODO check access permission */ @@ -95,7 +96,7 @@ unsigned VFPMCR(ARMul_State* state, unsigned type, u32 instr, u32 value) { /* VMOV s to r */ /* Transfering Rt is not mandatory, as the value of interest is pointed by value */ - VMOVBRS(state, BIT(20), Rt, BIT(7)|CRn<<1, &value); + VMOVBRS(state, BIT(instr, 20), Rt, BIT(instr, 7)|CRn<<1, &value); return ARMul_DONE; } @@ -126,24 +127,24 @@ unsigned VFPMCR(ARMul_State* state, unsigned type, u32 instr, u32 value) unsigned VFPMRRC(ARMul_State* state, unsigned type, u32 instr, u32* value1, u32* value2) { /* MCRR<c> <coproc>,<opc1>,<Rt>,<Rt2>,<CRm> */ - int CoProc = BITS (8, 11); /* 10 or 11 */ - int OPC_1 = BITS (4, 7); - int Rt = BITS (12, 15); - int Rt2 = BITS (16, 19); - int CRm = BITS (0, 3); + int CoProc = BITS(instr, 8, 11); /* 10 or 11 */ + int OPC_1 = BITS(instr, 4, 7); + int Rt = BITS(instr, 12, 15); + int Rt2 = BITS(instr, 16, 19); + int CRm = BITS(instr, 0, 3); if (CoProc == 10 || CoProc == 11) { if (CoProc == 10 && (OPC_1 & 0xD) == 1) { - VMOVBRRSS(state, BIT(20), Rt, Rt2, BIT(5)<<4|CRm, value1, value2); + VMOVBRRSS(state, BIT(instr, 20), Rt, Rt2, BIT(instr, 5)<<4|CRm, value1, value2); return ARMul_DONE; } if (CoProc == 11 && (OPC_1 & 0xD) == 1) { /* Transfering Rt and Rt2 is not mandatory, as the value of interest is pointed by value1 and value2 */ - VMOVBRRD(state, BIT(20), Rt, Rt2, BIT(5)<<4|CRm, value1, value2); + VMOVBRRD(state, BIT(instr, 20), Rt, Rt2, BIT(instr, 5)<<4|CRm, value1, value2); return ARMul_DONE; } } @@ -156,11 +157,11 @@ unsigned VFPMRRC(ARMul_State* state, unsigned type, u32 instr, u32* value1, u32* unsigned VFPMCRR(ARMul_State* state, unsigned type, u32 instr, u32 value1, u32 value2) { /* MCRR<c> <coproc>,<opc1>,<Rt>,<Rt2>,<CRm> */ - int CoProc = BITS (8, 11); /* 10 or 11 */ - int OPC_1 = BITS (4, 7); - int Rt = BITS (12, 15); - int Rt2 = BITS (16, 19); - int CRm = BITS (0, 3); + int CoProc = BITS(instr, 8, 11); /* 10 or 11 */ + int OPC_1 = BITS(instr, 4, 7); + int Rt = BITS(instr, 12, 15); + int Rt2 = BITS(instr, 16, 19); + int CRm = BITS(instr, 0, 3); /* TODO check access permission */ @@ -170,14 +171,14 @@ unsigned VFPMCRR(ARMul_State* state, unsigned type, u32 instr, u32 value1, u32 v { if (CoProc == 10 && (OPC_1 & 0xD) == 1) { - VMOVBRRSS(state, BIT(20), Rt, Rt2, BIT(5)<<4|CRm, &value1, &value2); + VMOVBRRSS(state, BIT(instr, 20), Rt, Rt2, BIT(instr, 5)<<4|CRm, &value1, &value2); return ARMul_DONE; } if (CoProc == 11 && (OPC_1 & 0xD) == 1) { /* Transfering Rt and Rt2 is not mandatory, as the value of interest is pointed by value1 and value2 */ - VMOVBRRD(state, BIT(20), Rt, Rt2, BIT(5)<<4|CRm, &value1, &value2); + VMOVBRRD(state, BIT(instr, 20), Rt, Rt2, BIT(instr, 5)<<4|CRm, &value1, &value2); return ARMul_DONE; } } @@ -190,14 +191,14 @@ unsigned VFPMCRR(ARMul_State* state, unsigned type, u32 instr, u32 value1, u32 v unsigned VFPSTC(ARMul_State* state, unsigned type, u32 instr, u32 * value) { /* STC{L}<c> <coproc>,<CRd>,[<Rn>],<option> */ - int CoProc = BITS (8, 11); /* 10 or 11 */ - int CRd = BITS (12, 15); - int Rn = BITS (16, 19); - int imm8 = BITS (0, 7); - int P = BIT(24); - int U = BIT(23); - int D = BIT(22); - int W = BIT(21); + int CoProc = BITS(instr, 8, 11); /* 10 or 11 */ + int CRd = BITS(instr, 12, 15); + int Rn = BITS(instr, 16, 19); + int imm8 = BITS(instr, 0, 7); + int P = BIT(instr, 24); + int U = BIT(instr, 23); + int D = BIT(instr, 22); + int W = BIT(instr, 21); /* TODO check access permission */ @@ -239,14 +240,14 @@ unsigned VFPSTC(ARMul_State* state, unsigned type, u32 instr, u32 * value) unsigned VFPLDC(ARMul_State* state, unsigned type, u32 instr, u32 value) { /* LDC{L}<c> <coproc>,<CRd>,[<Rn>] */ - int CoProc = BITS (8, 11); /* 10 or 11 */ - int CRd = BITS (12, 15); - int Rn = BITS (16, 19); - int imm8 = BITS (0, 7); - int P = BIT(24); - int U = BIT(23); - int D = BIT(22); - int W = BIT(21); + int CoProc = BITS(instr, 8, 11); /* 10 or 11 */ + int CRd = BITS(instr, 12, 15); + int Rn = BITS(instr, 16, 19); + int imm8 = BITS(instr, 0, 7); + int P = BIT(instr, 24); + int U = BIT(instr, 23); + int D = BIT(instr, 22); + int W = BIT(instr, 21); /* TODO check access permission */ @@ -277,57 +278,12 @@ unsigned VFPLDC(ARMul_State* state, unsigned type, u32 instr, u32 value) unsigned VFPCDP(ARMul_State* state, unsigned type, u32 instr) { /* CDP<c> <coproc>,<opc1>,<CRd>,<CRn>,<CRm>,<opc2> */ - int CoProc = BITS (8, 11); /* 10 or 11 */ - int OPC_1 = BITS (20, 23); - int CRd = BITS (12, 15); - int CRn = BITS (16, 19); - int CRm = BITS (0, 3); - int OPC_2 = BITS (5, 7); - - //ichfly - /*if ((instr & 0x0FBF0FD0) == 0x0EB70AC0) //vcvt.f64.f32 d8, s16 (s is bit 0-3 and LSB bit 22) (d is bit 12 - 15 MSB is Bit 6) - { - struct vfp_double vdd; - struct vfp_single vsd; - int dn = BITS(12, 15) + (BIT(22) << 4); - int sd = (BITS(0, 3) << 1) + BIT(5); - s32 n = vfp_get_float(state, sd); - vfp_single_unpack(&vsd, n); - if (vsd.exponent & 0x80) - { - vdd.exponent = (vsd.exponent&~0x80) | 0x400; - } - else - { - vdd.exponent = vsd.exponent | 0x380; - } - vdd.sign = vsd.sign; - vdd.significand = (u64)(vsd.significand & ~0xC0000000) << 32; // I have no idea why but the 2 uppern bits are not from the significand - vfp_put_double(state, vfp_double_pack(&vdd), dn); - return ARMul_DONE; - } - if ((instr & 0x0FBF0FD0) == 0x0EB70BC0) //vcvt.f32.f64 s15, d6 - { - struct vfp_double vdd; - struct vfp_single vsd; - int sd = BITS(0, 3) + (BIT(5) << 4); - int dn = (BITS(12, 15) << 1) + BIT(22); - vfp_double_unpack(&vdd, vfp_get_double(state, sd)); - if (vdd.exponent & 0x400) //todo if the exponent is to low or to high for this convert - { - vsd.exponent = (vdd.exponent) | 0x80; - } - else - { - vsd.exponent = vdd.exponent & ~0x80; - } - vsd.exponent &= 0xFF; - // vsd.exponent = vdd.exponent >> 3; - vsd.sign = vdd.sign; - vsd.significand = ((u64)(vdd.significand ) >> 32)& ~0xC0000000; - vfp_put_float(state, vfp_single_pack(&vsd), dn); - return ARMul_DONE; - }*/ + int CoProc = BITS(instr, 8, 11); /* 10 or 11 */ + int OPC_1 = BITS(instr, 20, 23); + int CRd = BITS(instr, 12, 15); + int CRn = BITS(instr, 16, 19); + int CRm = BITS(instr, 0, 3); + int OPC_2 = BITS(instr, 5, 7); /* TODO check access permission */ @@ -335,17 +291,17 @@ unsigned VFPCDP(ARMul_State* state, unsigned type, u32 instr) if (CoProc == 10 || CoProc == 11) { - if ((OPC_1 & 0xB) == 0xB && BITS(4, 7) == 0) + if ((OPC_1 & 0xB) == 0xB && BITS(instr, 4, 7) == 0) { - unsigned int single = BIT(8) == 0; - unsigned int d = (single ? BITS(12,15)<<1 | BIT(22) : BITS(12,15) | BIT(22)<<4); + unsigned int single = BIT(instr, 8) == 0; + unsigned int d = (single ? BITS(instr, 12,15)<<1 | BIT(instr, 22) : BITS(instr, 12,15) | BIT(instr, 22)<<4); unsigned int imm; - instr = BITS(16, 19) << 4 | BITS(0, 3); /* FIXME dirty workaround to get a correct imm */ + instr = BITS(instr, 16, 19) << 4 | BITS(instr, 0, 3); // FIXME dirty workaround to get a correct imm if (single) - imm = BIT(7)<<31 | (BIT(6)==0)<<30 | (BIT(6) ? 0x1f : 0)<<25 | BITS(0, 5)<<19; + imm = BIT(instr, 7)<<31 | (BIT(instr, 6)==0)<<30 | (BIT(instr, 6) ? 0x1f : 0)<<25 | BITS(instr, 0, 5)<<19; else - imm = BIT(7)<<31 | (BIT(6)==0)<<30 | (BIT(6) ? 0xff : 0)<<22 | BITS(0, 5)<<16; + imm = BIT(instr, 7)<<31 | (BIT(instr, 6)==0)<<30 | (BIT(instr, 6) ? 0xff : 0)<<22 | BITS(instr, 0, 5)<<16; VMOVI(state, single, d, imm); return ARMul_DONE; @@ -353,9 +309,9 @@ unsigned VFPCDP(ARMul_State* state, unsigned type, u32 instr) if ((OPC_1 & 0xB) == 0xB && CRn == 0 && (OPC_2 & 0x6) == 0x2) { - unsigned int single = BIT(8) == 0; - unsigned int d = (single ? BITS(12,15)<<1 | BIT(22) : BITS(12,15) | BIT(22)<<4); - unsigned int m = (single ? BITS( 0, 3)<<1 | BIT( 5) : BITS( 0, 3) | BIT( 5)<<4);; + unsigned int single = BIT(instr, 8) == 0; + unsigned int d = (single ? BITS(instr, 12,15)<<1 | BIT(instr, 22) : BITS(instr, 12,15) | BIT(instr, 22)<<4); + unsigned int m = (single ? BITS(instr, 0, 3)<<1 | BIT(instr, 5) : BITS(instr, 0, 3) | BIT(instr, 5)<<4); VMOVR(state, single, d, m); return ARMul_DONE; } @@ -477,11 +433,11 @@ int VSTR(ARMul_State* state, int type, ARMword instr, ARMword* value) static int single_reg, add, d, n, imm32, regs; if (type == ARMul_FIRST) { - single_reg = BIT(8) == 0; /* Double precision */ - add = BIT(23); /* */ - imm32 = BITS(0,7)<<2; /* may not be used */ - d = single_reg ? BITS(12, 15)<<1|BIT(22) : BIT(22)<<4|BITS(12, 15); /* Base register */ - n = BITS(16, 19); /* destination register */ + single_reg = BIT(instr, 8) == 0; // Double precision + add = BIT(instr, 23); + imm32 = BITS(instr, 0,7)<<2; // may not be used + d = single_reg ? BITS(instr, 12, 15)<<1|BIT(instr, 22) : BIT(instr, 22)<<4|BITS(instr, 12, 15); /* Base register */ + n = BITS(instr, 16, 19); // destination register i = 0; regs = 1; @@ -516,13 +472,13 @@ int VSTR(ARMul_State* state, int type, ARMword instr, ARMword* value) int VPUSH(ARMul_State* state, int type, ARMword instr, ARMword* value) { static int i = 0; - static int single_regs, add, wback, d, n, imm32, regs; + static int single_regs, d, imm32, regs; if (type == ARMul_FIRST) { - single_regs = BIT(8) == 0; /* Single precision */ - d = single_regs ? BITS(12, 15)<<1|BIT(22) : BIT(22)<<4|BITS(12, 15); /* Base register */ - imm32 = BITS(0,7)<<2; /* may not be used */ - regs = single_regs ? BITS(0, 7) : BITS(1, 7); /* FSTMX if regs is odd */ + single_regs = BIT(instr, 8) == 0; // Single precision + d = single_regs ? BITS(instr, 12, 15)<<1|BIT(instr, 22) : BIT(instr, 22)<<4|BITS(instr, 12, 15); // Base register + imm32 = BITS(instr, 0,7)<<2; // may not be used + regs = single_regs ? BITS(instr, 0, 7) : BITS(instr, 1, 7); // FSTMX if regs is odd state->Reg[R13] = state->Reg[R13] - imm32; @@ -561,13 +517,13 @@ int VSTM(ARMul_State* state, int type, ARMword instr, ARMword* value) static int single_regs, add, wback, d, n, imm32, regs; if (type == ARMul_FIRST) { - single_regs = BIT(8) == 0; /* Single precision */ - add = BIT(23); /* */ - wback = BIT(21); /* write-back */ - d = single_regs ? BITS(12, 15)<<1|BIT(22) : BIT(22)<<4|BITS(12, 15); /* Base register */ - n = BITS(16, 19); /* destination register */ - imm32 = BITS(0,7) * 4; /* may not be used */ - regs = single_regs ? BITS(0, 7) : BITS(0, 7)>>1; /* FSTMX if regs is odd */ + single_regs = BIT(instr, 8) == 0; // Single precision + add = BIT(instr, 23); + wback = BIT(instr, 21); // write-back + d = single_regs ? BITS(instr, 12, 15)<<1|BIT(instr, 22) : BIT(instr, 22)<<4|BITS(instr, 12, 15); // Base register + n = BITS(instr, 16, 19); // destination register + imm32 = BITS(instr, 0,7) * 4; // may not be used + regs = single_regs ? BITS(instr, 0, 7) : BITS(instr, 0, 7)>>1; // FSTMX if regs is odd if (wback) { state->Reg[n] = (add ? state->Reg[n] + imm32 : state->Reg[n] - imm32); @@ -607,13 +563,13 @@ int VSTM(ARMul_State* state, int type, ARMword instr, ARMword* value) int VPOP(ARMul_State* state, int type, ARMword instr, ARMword value) { static int i = 0; - static int single_regs, add, wback, d, n, imm32, regs; + static int single_regs, d, imm32, regs; if (type == ARMul_FIRST) { - single_regs = BIT(8) == 0; /* Single precision */ - d = single_regs ? BITS(12, 15)<<1|BIT(22) : BIT(22)<<4|BITS(12, 15); /* Base register */ - imm32 = BITS(0,7)<<2; /* may not be used */ - regs = single_regs ? BITS(0, 7) : BITS(1, 7); /* FLDMX if regs is odd */ + single_regs = BIT(instr, 8) == 0; // Single precision + d = single_regs ? BITS(instr, 12, 15)<<1|BIT(instr, 22) : BIT(instr, 22)<<4|BITS(instr, 12, 15); // Base register + imm32 = BITS(instr, 0, 7)<<2; // may not be used + regs = single_regs ? BITS(instr, 0, 7) : BITS(instr, 1, 7); // FLDMX if regs is odd state->Reg[R13] = state->Reg[R13] + imm32; @@ -656,11 +612,11 @@ int VLDR(ARMul_State* state, int type, ARMword instr, ARMword value) static int single_reg, add, d, n, imm32, regs; if (type == ARMul_FIRST) { - single_reg = BIT(8) == 0; /* Double precision */ - add = BIT(23); /* */ - imm32 = BITS(0,7)<<2; /* may not be used */ - d = single_reg ? BITS(12, 15)<<1|BIT(22) : BIT(22)<<4|BITS(12, 15); /* Base register */ - n = BITS(16, 19); /* destination register */ + single_reg = BIT(instr, 8) == 0; // Double precision + add = BIT(instr, 23); + imm32 = BITS(instr, 0, 7)<<2; // may not be used + d = single_reg ? BITS(instr, 12, 15)<<1|BIT(instr, 22) : BIT(instr, 22)<<4|BITS(instr, 12, 15); // Base register + n = BITS(instr, 16, 19); // destination register i = 0; regs = 1; @@ -702,13 +658,13 @@ int VLDM(ARMul_State* state, int type, ARMword instr, ARMword value) static int single_regs, add, wback, d, n, imm32, regs; if (type == ARMul_FIRST) { - single_regs = BIT(8) == 0; /* Single precision */ - add = BIT(23); /* */ - wback = BIT(21); /* write-back */ - d = single_regs ? BITS(12, 15)<<1|BIT(22) : BIT(22)<<4|BITS(12, 15); /* Base register */ - n = BITS(16, 19); /* destination register */ - imm32 = BITS(0,7) * 4; /* may not be used */ - regs = single_regs ? BITS(0, 7) : BITS(0, 7)>>1; /* FLDMX if regs is odd */ + single_regs = BIT(instr, 8) == 0; // Single precision + add = BIT(instr, 23); + wback = BIT(instr, 21); // write-back + d = single_regs ? BITS(instr, 12, 15)<<1|BIT(instr, 22) : BIT(instr, 22)<<4|BITS(instr, 12, 15); // Base register + n = BITS(instr, 16, 19); // destination register + imm32 = BITS(instr, 0, 7) * 4; // may not be used + regs = single_regs ? BITS(instr, 0, 7) : BITS(instr, 0, 7)>>1; // FLDMX if regs is odd if (wback) { state->Reg[n] = (add ? state->Reg[n] + imm32 : state->Reg[n] - imm32); @@ -787,8 +743,7 @@ void vfp_put_float(arm_core_t* state, int32_t val, unsigned int reg) uint64_t vfp_get_double(arm_core_t* state, unsigned int reg) { - uint64_t result; - result = ((uint64_t) state->ExtReg[reg*2+1])<<32 | state->ExtReg[reg*2]; + uint64_t result = ((uint64_t) state->ExtReg[reg*2+1])<<32 | state->ExtReg[reg*2]; LOG_TRACE(Core_ARM11, "VFP get double: s[%d-%d]=[%016llx]\n", reg * 2 + 1, reg * 2, result); return result; } @@ -805,8 +760,6 @@ void vfp_put_double(arm_core_t* state, uint64_t val, unsigned int reg) */ void vfp_raise_exceptions(ARMul_State* state, u32 exceptions, u32 inst, u32 fpscr) { - int si_code = 0; - LOG_TRACE(Core_ARM11, "VFP: raising exceptions %08x\n", exceptions); if (exceptions == VFP_EXCEPTION_ERROR) { @@ -820,8 +773,8 @@ void vfp_raise_exceptions(ARMul_State* state, u32 exceptions, u32 inst, u32 fpsc * Comparison instructions always return at least one of * these flags set. */ - if (exceptions & (FPSCR_N|FPSCR_Z|FPSCR_C|FPSCR_V)) - fpscr &= ~(FPSCR_N|FPSCR_Z|FPSCR_C|FPSCR_V); + if (exceptions & (FPSCR_NFLAG|FPSCR_ZFLAG|FPSCR_CFLAG|FPSCR_VFLAG)) + fpscr &= ~(FPSCR_NFLAG|FPSCR_ZFLAG|FPSCR_CFLAG|FPSCR_VFLAG); fpscr |= exceptions; diff --git a/src/core/arm/skyeye_common/vfp/vfp.h b/src/core/arm/skyeye_common/vfp/vfp.h index 09c7520db..445a224bc 100644 --- a/src/core/arm/skyeye_common/vfp/vfp.h +++ b/src/core/arm/skyeye_common/vfp/vfp.h @@ -25,86 +25,42 @@ #define VFP_DEBUG_UNIMPLEMENTED(x) LOG_ERROR(Core_ARM11, "in func %s, " #x " unimplemented\n", __FUNCTION__); exit(-1); #define VFP_DEBUG_UNTESTED(x) LOG_TRACE(Core_ARM11, "in func %s, " #x " untested\n", __FUNCTION__); #define CHECK_VFP_ENABLED -#define CHECK_VFP_CDP_RET vfp_raise_exceptions(cpu, ret, inst_cream->instr, cpu->VFP[VFP_OFFSET(VFP_FPSCR)]); //if (ret == -1) {printf("VFP CDP FAILURE %x\n", inst_cream->instr); exit(-1);} - -unsigned VFPInit (ARMul_State *state); -unsigned VFPMRC (ARMul_State * state, unsigned type, ARMword instr, ARMword * value); -unsigned VFPMCR (ARMul_State * state, unsigned type, ARMword instr, ARMword value); -unsigned VFPMRRC (ARMul_State * state, unsigned type, ARMword instr, ARMword * value1, ARMword * value2); -unsigned VFPMCRR (ARMul_State * state, unsigned type, ARMword instr, ARMword value1, ARMword value2); -unsigned VFPSTC (ARMul_State * state, unsigned type, ARMword instr, ARMword * value); -unsigned VFPLDC (ARMul_State * state, unsigned type, ARMword instr, ARMword value); -unsigned VFPCDP (ARMul_State * state, unsigned type, ARMword instr); - -/* FPSID Information */ -#define VFP_FPSID_IMPLMEN 0 /* should be the same as cp15 0 c0 0*/ -#define VFP_FPSID_SW 0 -#define VFP_FPSID_SUBARCH 0x2 /* VFP version. Current is v3 (not strict) */ -#define VFP_FPSID_PARTNUM 0x1 -#define VFP_FPSID_VARIANT 0x1 -#define VFP_FPSID_REVISION 0x1 - -/* FPEXC Flags */ -#define VFP_FPEXC_EX 1<<31 -#define VFP_FPEXC_EN 1<<30 - -/* FPSCR Flags */ -#define VFP_FPSCR_NFLAG 1<<31 -#define VFP_FPSCR_ZFLAG 1<<30 -#define VFP_FPSCR_CFLAG 1<<29 -#define VFP_FPSCR_VFLAG 1<<28 - -#define VFP_FPSCR_AHP 1<<26 /* Alternative Half Precision */ -#define VFP_FPSCR_DN 1<<25 /* Default NaN */ -#define VFP_FPSCR_FZ 1<<24 /* Flush-to-zero */ -#define VFP_FPSCR_RMODE 3<<22 /* Rounding Mode */ -#define VFP_FPSCR_STRIDE 3<<20 /* Stride (vector) */ -#define VFP_FPSCR_LEN 7<<16 /* Stride (vector) */ - -#define VFP_FPSCR_IDE 1<<15 /* Input Denormal exc */ -#define VFP_FPSCR_IXE 1<<12 /* Inexact exc */ -#define VFP_FPSCR_UFE 1<<11 /* Undeflow exc */ -#define VFP_FPSCR_OFE 1<<10 /* Overflow exc */ -#define VFP_FPSCR_DZE 1<<9 /* Division by Zero exc */ -#define VFP_FPSCR_IOE 1<<8 /* Invalid Operation exc */ - -#define VFP_FPSCR_IDC 1<<7 /* Input Denormal cum exc */ -#define VFP_FPSCR_IXC 1<<4 /* Inexact cum exc */ -#define VFP_FPSCR_UFC 1<<3 /* Undeflow cum exc */ -#define VFP_FPSCR_OFC 1<<2 /* Overflow cum exc */ -#define VFP_FPSCR_DZC 1<<1 /* Division by Zero cum exc */ -#define VFP_FPSCR_IOC 1<<0 /* Invalid Operation cum exc */ - -/* Inline instructions. Note: Used in a cpp file as well */ -#ifdef __cplusplus - extern "C" { -#endif -int32_t vfp_get_float(ARMul_State * state, unsigned int reg); -void vfp_put_float(ARMul_State * state, int32_t val, unsigned int reg); -uint64_t vfp_get_double(ARMul_State * state, unsigned int reg); -void vfp_put_double(ARMul_State * state, uint64_t val, unsigned int reg); -void vfp_raise_exceptions(ARMul_State * state, uint32_t exceptions, uint32_t inst, uint32_t fpscr); +#define CHECK_VFP_CDP_RET vfp_raise_exceptions(cpu, ret, inst_cream->instr, cpu->VFP[VFP_OFFSET(VFP_FPSCR)]); //if (ret == -1) {printf("VFP CDP FAILURE %x\n", inst_cream->instr); exit(-1);} + +unsigned VFPInit(ARMul_State* state); +unsigned VFPMRC(ARMul_State* state, unsigned type, ARMword instr, ARMword* value); +unsigned VFPMCR(ARMul_State* state, unsigned type, ARMword instr, ARMword value); +unsigned VFPMRRC(ARMul_State* state, unsigned type, ARMword instr, ARMword* value1, ARMword* value2); +unsigned VFPMCRR(ARMul_State* state, unsigned type, ARMword instr, ARMword value1, ARMword value2); +unsigned VFPSTC(ARMul_State* state, unsigned type, ARMword instr, ARMword* value); +unsigned VFPLDC(ARMul_State* state, unsigned type, ARMword instr, ARMword value); +unsigned VFPCDP(ARMul_State* state, unsigned type, ARMword instr); + +s32 vfp_get_float(ARMul_State* state, u32 reg); +void vfp_put_float(ARMul_State* state, s32 val, u32 reg); +u64 vfp_get_double(ARMul_State* state, u32 reg); +void vfp_put_double(ARMul_State* state, u64 val, u32 reg); +void vfp_raise_exceptions(ARMul_State* state, u32 exceptions, u32 inst, u32 fpscr); u32 vfp_single_cpdo(ARMul_State* state, u32 inst, u32 fpscr); u32 vfp_double_cpdo(ARMul_State* state, u32 inst, u32 fpscr); -/* MRC */ -void VMRS(ARMul_State * state, ARMword reg, ARMword Rt, ARMword *value); -void VMOVBRS(ARMul_State * state, ARMword to_arm, ARMword t, ARMword n, ARMword *value); -void VMOVBRRD(ARMul_State * state, ARMword to_arm, ARMword t, ARMword t2, ARMword n, ARMword *value1, ARMword *value2); +// MRC +void VMRS(ARMul_State* state, ARMword reg, ARMword Rt, ARMword* value); +void VMOVBRS(ARMul_State* state, ARMword to_arm, ARMword t, ARMword n, ARMword* value); +void VMOVBRRD(ARMul_State* state, ARMword to_arm, ARMword t, ARMword t2, ARMword n, ARMword* value1, ARMword* value2); void VMOVBRRSS(ARMul_State* state, ARMword to_arm, ARMword t, ARMword t2, ARMword n, ARMword* value1, ARMword* value2); -void VMOVI(ARMul_State * state, ARMword single, ARMword d, ARMword imm); -void VMOVR(ARMul_State * state, ARMword single, ARMword d, ARMword imm); -/* MCR */ -void VMSR(ARMul_State * state, ARMword reg, ARMword Rt); -/* STC */ -int VSTM(ARMul_State * state, int type, ARMword instr, ARMword* value); -int VPUSH(ARMul_State * state, int type, ARMword instr, ARMword* value); -int VSTR(ARMul_State * state, int type, ARMword instr, ARMword* value); -/* LDC */ -int VLDM(ARMul_State * state, int type, ARMword instr, ARMword value); -int VPOP(ARMul_State * state, int type, ARMword instr, ARMword value); -int VLDR(ARMul_State * state, int type, ARMword instr, ARMword value); +void VMOVI(ARMul_State* state, ARMword single, ARMword d, ARMword imm); +void VMOVR(ARMul_State* state, ARMword single, ARMword d, ARMword imm); + +// MCR +void VMSR(ARMul_State* state, ARMword reg, ARMword Rt); + +// STC +int VSTM(ARMul_State* state, int type, ARMword instr, ARMword* value); +int VPUSH(ARMul_State* state, int type, ARMword instr, ARMword* value); +int VSTR(ARMul_State* state, int type, ARMword instr, ARMword* value); -#ifdef __cplusplus - } -#endif +// LDC +int VLDM(ARMul_State* state, int type, ARMword instr, ARMword value); +int VPOP(ARMul_State* state, int type, ARMword instr, ARMword value); +int VLDR(ARMul_State* state, int type, ARMword instr, ARMword value); diff --git a/src/core/arm/skyeye_common/vfp/vfp_helper.h b/src/core/arm/skyeye_common/vfp/vfp_helper.h index 55e63f553..b68090b80 100644 --- a/src/core/arm/skyeye_common/vfp/vfp_helper.h +++ b/src/core/arm/skyeye_common/vfp/vfp_helper.h @@ -43,495 +43,402 @@ #define pr_info //printf #define pr_debug //printf -static u32 fls(ARMword x); #define do_div(n, base) {n/=base;} -/* From vfpinstr.h */ - -#define INST_CPRTDO(inst) (((inst) & 0x0f000000) == 0x0e000000) -#define INST_CPRT(inst) ((inst) & (1 << 4)) -#define INST_CPRT_L(inst) ((inst) & (1 << 20)) -#define INST_CPRT_Rd(inst) (((inst) & (15 << 12)) >> 12) -#define INST_CPRT_OP(inst) (((inst) >> 21) & 7) -#define INST_CPNUM(inst) ((inst) & 0xf00) -#define CPNUM(cp) ((cp) << 8) - -#define FOP_MASK (0x00b00040) -#define FOP_FMAC (0x00000000) -#define FOP_FNMAC (0x00000040) -#define FOP_FMSC (0x00100000) -#define FOP_FNMSC (0x00100040) -#define FOP_FMUL (0x00200000) -#define FOP_FNMUL (0x00200040) -#define FOP_FADD (0x00300000) -#define FOP_FSUB (0x00300040) -#define FOP_FDIV (0x00800000) -#define FOP_EXT (0x00b00040) - -#define FOP_TO_IDX(inst) ((inst & 0x00b00000) >> 20 | (inst & (1 << 6)) >> 4) - -#define FEXT_MASK (0x000f0080) -#define FEXT_FCPY (0x00000000) -#define FEXT_FABS (0x00000080) -#define FEXT_FNEG (0x00010000) -#define FEXT_FSQRT (0x00010080) -#define FEXT_FCMP (0x00040000) -#define FEXT_FCMPE (0x00040080) -#define FEXT_FCMPZ (0x00050000) -#define FEXT_FCMPEZ (0x00050080) -#define FEXT_FCVT (0x00070080) -#define FEXT_FUITO (0x00080000) -#define FEXT_FSITO (0x00080080) -#define FEXT_FTOUI (0x000c0000) -#define FEXT_FTOUIZ (0x000c0080) -#define FEXT_FTOSI (0x000d0000) -#define FEXT_FTOSIZ (0x000d0080) - -#define FEXT_TO_IDX(inst) ((inst & 0x000f0000) >> 15 | (inst & (1 << 7)) >> 7) - -#define vfp_get_sd(inst) ((inst & 0x0000f000) >> 11 | (inst & (1 << 22)) >> 22) -#define vfp_get_dd(inst) ((inst & 0x0000f000) >> 12 | (inst & (1 << 22)) >> 18) -#define vfp_get_sm(inst) ((inst & 0x0000000f) << 1 | (inst & (1 << 5)) >> 5) -#define vfp_get_dm(inst) ((inst & 0x0000000f) | (inst & (1 << 5)) >> 1) -#define vfp_get_sn(inst) ((inst & 0x000f0000) >> 15 | (inst & (1 << 7)) >> 7) -#define vfp_get_dn(inst) ((inst & 0x000f0000) >> 16 | (inst & (1 << 7)) >> 3) - -#define vfp_single(inst) (((inst) & 0x0000f00) == 0xa00) - -#define FPSCR_N (1 << 31) -#define FPSCR_Z (1 << 30) -#define FPSCR_C (1 << 29) -#define FPSCR_V (1 << 28) - -/* -------------- */ - -/* From asm/include/vfp.h */ - -/* FPSCR bits */ -#define FPSCR_DEFAULT_NAN (1<<25) -#define FPSCR_FLUSHTOZERO (1<<24) -#define FPSCR_ROUND_NEAREST (0<<22) -#define FPSCR_ROUND_PLUSINF (1<<22) -#define FPSCR_ROUND_MINUSINF (2<<22) -#define FPSCR_ROUND_TOZERO (3<<22) -#define FPSCR_RMODE_BIT (22) -#define FPSCR_RMODE_MASK (3 << FPSCR_RMODE_BIT) -#define FPSCR_STRIDE_BIT (20) -#define FPSCR_STRIDE_MASK (3 << FPSCR_STRIDE_BIT) -#define FPSCR_LENGTH_BIT (16) -#define FPSCR_LENGTH_MASK (7 << FPSCR_LENGTH_BIT) -#define FPSCR_IOE (1<<8) -#define FPSCR_DZE (1<<9) -#define FPSCR_OFE (1<<10) -#define FPSCR_UFE (1<<11) -#define FPSCR_IXE (1<<12) -#define FPSCR_IDE (1<<15) -#define FPSCR_IOC (1<<0) -#define FPSCR_DZC (1<<1) -#define FPSCR_OFC (1<<2) -#define FPSCR_UFC (1<<3) -#define FPSCR_IXC (1<<4) -#define FPSCR_IDC (1<<7) - -/* ---------------- */ +enum : u32 { + FOP_MASK = 0x00b00040, + FOP_FMAC = 0x00000000, + FOP_FNMAC = 0x00000040, + FOP_FMSC = 0x00100000, + FOP_FNMSC = 0x00100040, + FOP_FMUL = 0x00200000, + FOP_FNMUL = 0x00200040, + FOP_FADD = 0x00300000, + FOP_FSUB = 0x00300040, + FOP_FDIV = 0x00800000, + FOP_EXT = 0x00b00040 +}; + +#define FOP_TO_IDX(inst) ((inst & 0x00b00000) >> 20 | (inst & (1 << 6)) >> 4) + +enum : u32 { + FEXT_MASK = 0x000f0080, + FEXT_FCPY = 0x00000000, + FEXT_FABS = 0x00000080, + FEXT_FNEG = 0x00010000, + FEXT_FSQRT = 0x00010080, + FEXT_FCMP = 0x00040000, + FEXT_FCMPE = 0x00040080, + FEXT_FCMPZ = 0x00050000, + FEXT_FCMPEZ = 0x00050080, + FEXT_FCVT = 0x00070080, + FEXT_FUITO = 0x00080000, + FEXT_FSITO = 0x00080080, + FEXT_FTOUI = 0x000c0000, + FEXT_FTOUIZ = 0x000c0080, + FEXT_FTOSI = 0x000d0000, + FEXT_FTOSIZ = 0x000d0080 +}; + +#define FEXT_TO_IDX(inst) ((inst & 0x000f0000) >> 15 | (inst & (1 << 7)) >> 7) + +#define vfp_get_sd(inst) ((inst & 0x0000f000) >> 11 | (inst & (1 << 22)) >> 22) +#define vfp_get_dd(inst) ((inst & 0x0000f000) >> 12 | (inst & (1 << 22)) >> 18) +#define vfp_get_sm(inst) ((inst & 0x0000000f) << 1 | (inst & (1 << 5)) >> 5) +#define vfp_get_dm(inst) ((inst & 0x0000000f) | (inst & (1 << 5)) >> 1) +#define vfp_get_sn(inst) ((inst & 0x000f0000) >> 15 | (inst & (1 << 7)) >> 7) +#define vfp_get_dn(inst) ((inst & 0x000f0000) >> 16 | (inst & (1 << 7)) >> 3) + +#define vfp_single(inst) (((inst) & 0x0000f00) == 0xa00) static inline u32 vfp_shiftright32jamming(u32 val, unsigned int shift) { - if (shift) { - if (shift < 32) - val = val >> shift | ((val << (32 - shift)) != 0); - else - val = val != 0; - } - return val; + if (shift) { + if (shift < 32) + val = val >> shift | ((val << (32 - shift)) != 0); + else + val = val != 0; + } + return val; } static inline u64 vfp_shiftright64jamming(u64 val, unsigned int shift) { - if (shift) { - if (shift < 64) - val = val >> shift | ((val << (64 - shift)) != 0); - else - val = val != 0; - } - return val; + if (shift) { + if (shift < 64) + val = val >> shift | ((val << (64 - shift)) != 0); + else + val = val != 0; + } + return val; } static inline u32 vfp_hi64to32jamming(u64 val) { - u32 v; - u32 highval = val >> 32; - u32 lowval = val & 0xffffffff; + u32 v; + u32 highval = val >> 32; + u32 lowval = val & 0xffffffff; - if (lowval >= 1) - v = highval | 1; - else - v = highval; + if (lowval >= 1) + v = highval | 1; + else + v = highval; - return v; + return v; } -static inline void add128(u64 *resh, u64 *resl, u64 nh, u64 nl, u64 mh, u64 ml) +static inline void add128(u64* resh, u64* resl, u64 nh, u64 nl, u64 mh, u64 ml) { - *resl = nl + ml; - *resh = nh + mh; - if (*resl < nl) - *resh += 1; + *resl = nl + ml; + *resh = nh + mh; + if (*resl < nl) + *resh += 1; } -static inline void sub128(u64 *resh, u64 *resl, u64 nh, u64 nl, u64 mh, u64 ml) +static inline void sub128(u64* resh, u64* resl, u64 nh, u64 nl, u64 mh, u64 ml) { - *resl = nl - ml; - *resh = nh - mh; - if (*resl > nl) - *resh -= 1; + *resl = nl - ml; + *resh = nh - mh; + if (*resl > nl) + *resh -= 1; } -static inline void mul64to128(u64 *resh, u64 *resl, u64 n, u64 m) +static inline void mul64to128(u64* resh, u64* resl, u64 n, u64 m) { - u32 nh, nl, mh, ml; - u64 rh, rma, rmb, rl; + u32 nh, nl, mh, ml; + u64 rh, rma, rmb, rl; - nl = n; - ml = m; - rl = (u64)nl * ml; + nl = n; + ml = m; + rl = (u64)nl * ml; - nh = n >> 32; - rma = (u64)nh * ml; + nh = n >> 32; + rma = (u64)nh * ml; - mh = m >> 32; - rmb = (u64)nl * mh; - rma += rmb; + mh = m >> 32; + rmb = (u64)nl * mh; + rma += rmb; - rh = (u64)nh * mh; - rh += ((u64)(rma < rmb) << 32) + (rma >> 32); + rh = (u64)nh * mh; + rh += ((u64)(rma < rmb) << 32) + (rma >> 32); - rma <<= 32; - rl += rma; - rh += (rl < rma); + rma <<= 32; + rl += rma; + rh += (rl < rma); - *resl = rl; - *resh = rh; + *resl = rl; + *resh = rh; } -static inline void shift64left(u64 *resh, u64 *resl, u64 n) +static inline void shift64left(u64* resh, u64* resl, u64 n) { - *resh = n >> 63; - *resl = n << 1; + *resh = n >> 63; + *resl = n << 1; } static inline u64 vfp_hi64multiply64(u64 n, u64 m) { - u64 rh, rl; - mul64to128(&rh, &rl, n, m); - return rh | (rl != 0); + u64 rh, rl; + mul64to128(&rh, &rl, n, m); + return rh | (rl != 0); } static inline u64 vfp_estimate_div128to64(u64 nh, u64 nl, u64 m) { - u64 mh, ml, remh, reml, termh, terml, z; - - if (nh >= m) - return ~0ULL; - mh = m >> 32; - if (mh << 32 <= nh) { - z = 0xffffffff00000000ULL; - } else { - z = nh; - do_div(z, mh); - z <<= 32; - } - mul64to128(&termh, &terml, m, z); - sub128(&remh, &reml, nh, nl, termh, terml); - ml = m << 32; - while ((s64)remh < 0) { - z -= 0x100000000ULL; - add128(&remh, &reml, remh, reml, mh, ml); - } - remh = (remh << 32) | (reml >> 32); - if (mh << 32 <= remh) { - z |= 0xffffffff; - } else { - do_div(remh, mh); - z |= remh; - } - return z; + u64 mh, ml, remh, reml, termh, terml, z; + + if (nh >= m) + return ~0ULL; + mh = m >> 32; + if (mh << 32 <= nh) { + z = 0xffffffff00000000ULL; + } else { + z = nh; + do_div(z, mh); + z <<= 32; + } + mul64to128(&termh, &terml, m, z); + sub128(&remh, &reml, nh, nl, termh, terml); + ml = m << 32; + while ((s64)remh < 0) { + z -= 0x100000000ULL; + add128(&remh, &reml, remh, reml, mh, ml); + } + remh = (remh << 32) | (reml >> 32); + if (mh << 32 <= remh) { + z |= 0xffffffff; + } else { + do_div(remh, mh); + z |= remh; + } + return z; } -/* - * Operations on unpacked elements - */ -#define vfp_sign_negate(sign) (sign ^ 0x8000) +// Operations on unpacked elements +#define vfp_sign_negate(sign) (sign ^ 0x8000) -/* - * Single-precision - */ +// Single-precision struct vfp_single { - s16 exponent; - u16 sign; - u32 significand; + s16 exponent; + u16 sign; + u32 significand; }; -#ifdef __cplusplus - extern "C" { -#endif -extern s32 vfp_get_float(ARMul_State * state, unsigned int reg); -extern void vfp_put_float(ARMul_State * state, s32 val, unsigned int reg); -#ifdef __cplusplus - } -#endif - -/* - * VFP_SINGLE_MANTISSA_BITS - number of bits in the mantissa - * VFP_SINGLE_EXPONENT_BITS - number of bits in the exponent - * VFP_SINGLE_LOW_BITS - number of low bits in the unpacked significand - * which are not propagated to the float upon packing. - */ -#define VFP_SINGLE_MANTISSA_BITS (23) -#define VFP_SINGLE_EXPONENT_BITS (8) -#define VFP_SINGLE_LOW_BITS (32 - VFP_SINGLE_MANTISSA_BITS - 2) -#define VFP_SINGLE_LOW_BITS_MASK ((1 << VFP_SINGLE_LOW_BITS) - 1) +// VFP_SINGLE_MANTISSA_BITS - number of bits in the mantissa +// VFP_SINGLE_EXPONENT_BITS - number of bits in the exponent +// VFP_SINGLE_LOW_BITS - number of low bits in the unpacked significand +// which are not propagated to the float upon packing. +#define VFP_SINGLE_MANTISSA_BITS (23) +#define VFP_SINGLE_EXPONENT_BITS (8) +#define VFP_SINGLE_LOW_BITS (32 - VFP_SINGLE_MANTISSA_BITS - 2) +#define VFP_SINGLE_LOW_BITS_MASK ((1 << VFP_SINGLE_LOW_BITS) - 1) -/* - * The bit in an unpacked float which indicates that it is a quiet NaN - */ +// The bit in an unpacked float which indicates that it is a quiet NaN #define VFP_SINGLE_SIGNIFICAND_QNAN (1 << (VFP_SINGLE_MANTISSA_BITS - 1 + VFP_SINGLE_LOW_BITS)) -/* - * Operations on packed single-precision numbers - */ -#define vfp_single_packed_sign(v) ((v) & 0x80000000) -#define vfp_single_packed_negate(v) ((v) ^ 0x80000000) -#define vfp_single_packed_abs(v) ((v) & ~0x80000000) -#define vfp_single_packed_exponent(v) (((v) >> VFP_SINGLE_MANTISSA_BITS) & ((1 << VFP_SINGLE_EXPONENT_BITS) - 1)) -#define vfp_single_packed_mantissa(v) ((v) & ((1 << VFP_SINGLE_MANTISSA_BITS) - 1)) - -/* - * Unpack a single-precision float. Note that this returns the magnitude - * of the single-precision float mantissa with the 1. if necessary, - * aligned to bit 30. - */ -static inline void vfp_single_unpack(struct vfp_single *s, s32 val) +// Operations on packed single-precision numbers +#define vfp_single_packed_sign(v) ((v) & 0x80000000) +#define vfp_single_packed_negate(v) ((v) ^ 0x80000000) +#define vfp_single_packed_abs(v) ((v) & ~0x80000000) +#define vfp_single_packed_exponent(v) (((v) >> VFP_SINGLE_MANTISSA_BITS) & ((1 << VFP_SINGLE_EXPONENT_BITS) - 1)) +#define vfp_single_packed_mantissa(v) ((v) & ((1 << VFP_SINGLE_MANTISSA_BITS) - 1)) + +// Unpack a single-precision float. Note that this returns the magnitude +// of the single-precision float mantissa with the 1. if necessary, +// aligned to bit 30. +static inline void vfp_single_unpack(vfp_single* s, s32 val) { - u32 significand; + u32 significand; - s->sign = vfp_single_packed_sign(val) >> 16, - s->exponent = vfp_single_packed_exponent(val); + s->sign = vfp_single_packed_sign(val) >> 16, + s->exponent = vfp_single_packed_exponent(val); - significand = (u32) val; - significand = (significand << (32 - VFP_SINGLE_MANTISSA_BITS)) >> 2; - if (s->exponent && s->exponent != 255) - significand |= 0x40000000; - s->significand = significand; + significand = (u32) val; + significand = (significand << (32 - VFP_SINGLE_MANTISSA_BITS)) >> 2; + if (s->exponent && s->exponent != 255) + significand |= 0x40000000; + s->significand = significand; } -/* - * Re-pack a single-precision float. This assumes that the float is - * already normalised such that the MSB is bit 30, _not_ bit 31. - */ -static inline s32 vfp_single_pack(struct vfp_single *s) +// Re-pack a single-precision float. This assumes that the float is +// already normalised such that the MSB is bit 30, _not_ bit 31. +static inline s32 vfp_single_pack(vfp_single* s) { - u32 val; - val = (s->sign << 16) + - (s->exponent << VFP_SINGLE_MANTISSA_BITS) + - (s->significand >> VFP_SINGLE_LOW_BITS); - return (s32)val; + u32 val = (s->sign << 16) + + (s->exponent << VFP_SINGLE_MANTISSA_BITS) + + (s->significand >> VFP_SINGLE_LOW_BITS); + return (s32)val; } -#define VFP_NUMBER (1<<0) -#define VFP_ZERO (1<<1) -#define VFP_DENORMAL (1<<2) -#define VFP_INFINITY (1<<3) -#define VFP_NAN (1<<4) -#define VFP_NAN_SIGNAL (1<<5) +enum : u32 { + VFP_NUMBER = (1 << 0), + VFP_ZERO = (1 << 1), + VFP_DENORMAL = (1 << 2), + VFP_INFINITY = (1 << 3), + VFP_NAN = (1 << 4), + VFP_NAN_SIGNAL = (1 << 5), -#define VFP_QNAN (VFP_NAN) -#define VFP_SNAN (VFP_NAN|VFP_NAN_SIGNAL) + VFP_QNAN = (VFP_NAN), + VFP_SNAN = (VFP_NAN|VFP_NAN_SIGNAL) +}; -static inline int vfp_single_type(struct vfp_single *s) +static inline int vfp_single_type(vfp_single* s) { - int type = VFP_NUMBER; - if (s->exponent == 255) { - if (s->significand == 0) - type = VFP_INFINITY; - else if (s->significand & VFP_SINGLE_SIGNIFICAND_QNAN) - type = VFP_QNAN; - else - type = VFP_SNAN; - } else if (s->exponent == 0) { - if (s->significand == 0) - type |= VFP_ZERO; - else - type |= VFP_DENORMAL; - } - return type; + int type = VFP_NUMBER; + if (s->exponent == 255) { + if (s->significand == 0) + type = VFP_INFINITY; + else if (s->significand & VFP_SINGLE_SIGNIFICAND_QNAN) + type = VFP_QNAN; + else + type = VFP_SNAN; + } else if (s->exponent == 0) { + if (s->significand == 0) + type |= VFP_ZERO; + else + type |= VFP_DENORMAL; + } + return type; } -u32 vfp_single_normaliseround(ARMul_State* state, int sd, struct vfp_single *vs, u32 fpscr, u32 exceptions, const char *func); +u32 vfp_single_normaliseround(ARMul_State* state, int sd, vfp_single* vs, u32 fpscr, u32 exceptions, const char* func); -/* - * Double-precision - */ +// Double-precision struct vfp_double { - s16 exponent; - u16 sign; - u64 significand; + s16 exponent; + u16 sign; + u64 significand; }; -/* - * VFP_REG_ZERO is a special register number for vfp_get_double - * which returns (double)0.0. This is useful for the compare with - * zero instructions. - */ +// VFP_REG_ZERO is a special register number for vfp_get_double +// which returns (double)0.0. This is useful for the compare with +// zero instructions. #ifdef CONFIG_VFPv3 -#define VFP_REG_ZERO 32 +#define VFP_REG_ZERO 32 #else -#define VFP_REG_ZERO 16 -#endif -#ifdef __cplusplus - extern "C" { +#define VFP_REG_ZERO 16 #endif -extern u64 vfp_get_double(ARMul_State * state, unsigned int reg); -extern void vfp_put_double(ARMul_State * state, u64 val, unsigned int reg); -#ifdef __cplusplus - } -#endif -#define VFP_DOUBLE_MANTISSA_BITS (52) -#define VFP_DOUBLE_EXPONENT_BITS (11) -#define VFP_DOUBLE_LOW_BITS (64 - VFP_DOUBLE_MANTISSA_BITS - 2) -#define VFP_DOUBLE_LOW_BITS_MASK ((1 << VFP_DOUBLE_LOW_BITS) - 1) -/* - * The bit in an unpacked double which indicates that it is a quiet NaN - */ -#define VFP_DOUBLE_SIGNIFICAND_QNAN (1ULL << (VFP_DOUBLE_MANTISSA_BITS - 1 + VFP_DOUBLE_LOW_BITS)) - -/* - * Operations on packed single-precision numbers - */ -#define vfp_double_packed_sign(v) ((v) & (1ULL << 63)) -#define vfp_double_packed_negate(v) ((v) ^ (1ULL << 63)) -#define vfp_double_packed_abs(v) ((v) & ~(1ULL << 63)) -#define vfp_double_packed_exponent(v) (((v) >> VFP_DOUBLE_MANTISSA_BITS) & ((1 << VFP_DOUBLE_EXPONENT_BITS) - 1)) -#define vfp_double_packed_mantissa(v) ((v) & ((1ULL << VFP_DOUBLE_MANTISSA_BITS) - 1)) - -/* - * Unpack a double-precision float. Note that this returns the magnitude - * of the double-precision float mantissa with the 1. if necessary, - * aligned to bit 62. - */ -static inline void vfp_double_unpack(struct vfp_double *s, s64 val) +#define VFP_DOUBLE_MANTISSA_BITS (52) +#define VFP_DOUBLE_EXPONENT_BITS (11) +#define VFP_DOUBLE_LOW_BITS (64 - VFP_DOUBLE_MANTISSA_BITS - 2) +#define VFP_DOUBLE_LOW_BITS_MASK ((1 << VFP_DOUBLE_LOW_BITS) - 1) + +// The bit in an unpacked double which indicates that it is a quiet NaN +#define VFP_DOUBLE_SIGNIFICAND_QNAN (1ULL << (VFP_DOUBLE_MANTISSA_BITS - 1 + VFP_DOUBLE_LOW_BITS)) + +// Operations on packed single-precision numbers +#define vfp_double_packed_sign(v) ((v) & (1ULL << 63)) +#define vfp_double_packed_negate(v) ((v) ^ (1ULL << 63)) +#define vfp_double_packed_abs(v) ((v) & ~(1ULL << 63)) +#define vfp_double_packed_exponent(v) (((v) >> VFP_DOUBLE_MANTISSA_BITS) & ((1 << VFP_DOUBLE_EXPONENT_BITS) - 1)) +#define vfp_double_packed_mantissa(v) ((v) & ((1ULL << VFP_DOUBLE_MANTISSA_BITS) - 1)) + +// Unpack a double-precision float. Note that this returns the magnitude +// of the double-precision float mantissa with the 1. if necessary, +// aligned to bit 62. +static inline void vfp_double_unpack(vfp_double* s, s64 val) { - u64 significand; + u64 significand; - s->sign = vfp_double_packed_sign(val) >> 48; - s->exponent = vfp_double_packed_exponent(val); + s->sign = vfp_double_packed_sign(val) >> 48; + s->exponent = vfp_double_packed_exponent(val); - significand = (u64) val; - significand = (significand << (64 - VFP_DOUBLE_MANTISSA_BITS)) >> 2; - if (s->exponent && s->exponent != 2047) - significand |= (1ULL << 62); - s->significand = significand; + significand = (u64) val; + significand = (significand << (64 - VFP_DOUBLE_MANTISSA_BITS)) >> 2; + if (s->exponent && s->exponent != 2047) + significand |= (1ULL << 62); + s->significand = significand; } -/* - * Re-pack a double-precision float. This assumes that the float is - * already normalised such that the MSB is bit 30, _not_ bit 31. - */ -static inline s64 vfp_double_pack(struct vfp_double *s) +// Re-pack a double-precision float. This assumes that the float is +// already normalised such that the MSB is bit 30, _not_ bit 31. +static inline s64 vfp_double_pack(vfp_double* s) { - u64 val; - val = ((u64)s->sign << 48) + - ((u64)s->exponent << VFP_DOUBLE_MANTISSA_BITS) + - (s->significand >> VFP_DOUBLE_LOW_BITS); - return (s64)val; + u64 val = ((u64)s->sign << 48) + + ((u64)s->exponent << VFP_DOUBLE_MANTISSA_BITS) + + (s->significand >> VFP_DOUBLE_LOW_BITS); + return (s64)val; } -static inline int vfp_double_type(struct vfp_double *s) +static inline int vfp_double_type(vfp_double* s) { - int type = VFP_NUMBER; - if (s->exponent == 2047) { - if (s->significand == 0) - type = VFP_INFINITY; - else if (s->significand & VFP_DOUBLE_SIGNIFICAND_QNAN) - type = VFP_QNAN; - else - type = VFP_SNAN; - } else if (s->exponent == 0) { - if (s->significand == 0) - type |= VFP_ZERO; - else - type |= VFP_DENORMAL; - } - return type; + int type = VFP_NUMBER; + if (s->exponent == 2047) { + if (s->significand == 0) + type = VFP_INFINITY; + else if (s->significand & VFP_DOUBLE_SIGNIFICAND_QNAN) + type = VFP_QNAN; + else + type = VFP_SNAN; + } else if (s->exponent == 0) { + if (s->significand == 0) + type |= VFP_ZERO; + else + type |= VFP_DENORMAL; + } + return type; } -u32 vfp_double_normaliseround(ARMul_State* state, int dd, struct vfp_double *vd, u32 fpscr, u32 exceptions, const char *func); - u32 vfp_estimate_sqrt_significand(u32 exponent, u32 significand); -/* - * A special flag to tell the normalisation code not to normalise. - */ -#define VFP_NAN_FLAG 0x100 - -/* - * A bit pattern used to indicate the initial (unset) value of the - * exception mask, in case nothing handles an instruction. This - * doesn't include the NAN flag, which get masked out before - * we check for an error. - */ -#define VFP_EXCEPTION_ERROR ((u32)-1 & ~VFP_NAN_FLAG) - -/* - * A flag to tell vfp instruction type. - * OP_SCALAR - this operation always operates in scalar mode - * OP_SD - the instruction exceptionally writes to a single precision result. - * OP_DD - the instruction exceptionally writes to a double precision result. - * OP_SM - the instruction exceptionally reads from a single precision operand. - */ -#define OP_SCALAR (1 << 0) -#define OP_SD (1 << 1) -#define OP_DD (1 << 1) -#define OP_SM (1 << 2) +// A special flag to tell the normalisation code not to normalise. +#define VFP_NAN_FLAG 0x100 + +// A bit pattern used to indicate the initial (unset) value of the +// exception mask, in case nothing handles an instruction. This +// doesn't include the NAN flag, which get masked out before +// we check for an error. +#define VFP_EXCEPTION_ERROR ((u32)-1 & ~VFP_NAN_FLAG) + +// A flag to tell vfp instruction type. +// OP_SCALAR - This operation always operates in scalar mode +// OP_SD - The instruction exceptionally writes to a single precision result. +// OP_DD - The instruction exceptionally writes to a double precision result. +// OP_SM - The instruction exceptionally reads from a single precision operand. +enum : u32 { + OP_SCALAR = (1 << 0), + OP_SD = (1 << 1), + OP_DD = (1 << 1), + OP_SM = (1 << 2) +}; struct op { - u32 (* const fn)(ARMul_State* state, int dd, int dn, int dm, u32 fpscr); - u32 flags; + u32 (* const fn)(ARMul_State* state, int dd, int dn, int dm, u32 fpscr); + u32 flags; }; -static u32 fls(ARMword x) +static inline u32 fls(ARMword x) { - int r = 32; - - if (!x) - return 0; - if (!(x & 0xffff0000u)) { - x <<= 16; - r -= 16; - } - if (!(x & 0xff000000u)) { - x <<= 8; - r -= 8; - } - if (!(x & 0xf0000000u)) { - x <<= 4; - r -= 4; - } - if (!(x & 0xc0000000u)) { - x <<= 2; - r -= 2; - } - if (!(x & 0x80000000u)) { - x <<= 1; - r -= 1; - } - return r; + int r = 32; + + if (!x) + return 0; + if (!(x & 0xffff0000u)) { + x <<= 16; + r -= 16; + } + if (!(x & 0xff000000u)) { + x <<= 8; + r -= 8; + } + if (!(x & 0xf0000000u)) { + x <<= 4; + r -= 4; + } + if (!(x & 0xc0000000u)) { + x <<= 2; + r -= 2; + } + if (!(x & 0x80000000u)) { + x <<= 1; + r -= 1; + } + return r; } -u32 vfp_double_normaliseroundintern(ARMul_State* state, struct vfp_double *vd, u32 fpscr, u32 exceptions, const char *func); -u32 vfp_double_multiply(struct vfp_double *vdd, struct vfp_double *vdn, struct vfp_double *vdm, u32 fpscr); -u32 vfp_double_add(struct vfp_double *vdd, struct vfp_double *vdn, struct vfp_double *vdm, u32 fpscr); -u32 vfp_double_fcvtsinterncutting(ARMul_State* state, int sd, struct vfp_double* dm, u32 fpscr); +u32 vfp_double_multiply(vfp_double* vdd, vfp_double* vdn, vfp_double* vdm, u32 fpscr); +u32 vfp_double_add(vfp_double* vdd, vfp_double* vdn, vfp_double *vdm, u32 fpscr); +u32 vfp_double_normaliseround(ARMul_State* state, int dd, vfp_double* vd, u32 fpscr, u32 exceptions, const char* func); diff --git a/src/core/arm/skyeye_common/vfp/vfpdouble.cpp b/src/core/arm/skyeye_common/vfp/vfpdouble.cpp index 15a5d1265..9a7088088 100644 --- a/src/core/arm/skyeye_common/vfp/vfpdouble.cpp +++ b/src/core/arm/skyeye_common/vfp/vfpdouble.cpp @@ -83,134 +83,6 @@ static void vfp_double_normalise_denormal(struct vfp_double *vd) vfp_double_dump("normalise_denormal: out", vd); } -u32 vfp_double_normaliseroundintern(ARMul_State* state, struct vfp_double *vd, u32 fpscr, u32 exceptions, const char *func) -{ - u64 significand, incr; - int exponent, shift, underflow; - u32 rmode; - - vfp_double_dump("pack: in", vd); - - /* - * Infinities and NaNs are a special case. - */ - if (vd->exponent == 2047 && (vd->significand == 0 || exceptions)) - goto pack; - - /* - * Special-case zero. - */ - if (vd->significand == 0) { - vd->exponent = 0; - goto pack; - } - - exponent = vd->exponent; - significand = vd->significand; - - shift = 32 - fls((ARMword)(significand >> 32)); - if (shift == 32) - shift = 64 - fls((ARMword)significand); - if (shift) { - exponent -= shift; - significand <<= shift; - } - -#if 1 - vd->exponent = exponent; - vd->significand = significand; - vfp_double_dump("pack: normalised", vd); -#endif - - /* - * Tiny number? - */ - underflow = exponent < 0; - if (underflow) { - significand = vfp_shiftright64jamming(significand, -exponent); - exponent = 0; -#if 1 - vd->exponent = exponent; - vd->significand = significand; - vfp_double_dump("pack: tiny number", vd); -#endif - if (!(significand & ((1ULL << (VFP_DOUBLE_LOW_BITS + 1)) - 1))) - underflow = 0; - } - - /* - * Select rounding increment. - */ - incr = 0; - rmode = fpscr & FPSCR_RMODE_MASK; - - if (rmode == FPSCR_ROUND_NEAREST) { - incr = 1ULL << VFP_DOUBLE_LOW_BITS; - if ((significand & (1ULL << (VFP_DOUBLE_LOW_BITS + 1))) == 0) - incr -= 1; - } - else if (rmode == FPSCR_ROUND_TOZERO) { - incr = 0; - } - else if ((rmode == FPSCR_ROUND_PLUSINF) ^ (vd->sign != 0)) - incr = (1ULL << (VFP_DOUBLE_LOW_BITS + 1)) - 1; - - LOG_TRACE(Core_ARM11, "VFP: rounding increment = 0x%08llx\n", incr); - - /* - * Is our rounding going to overflow? - */ - if ((significand + incr) < significand) { - exponent += 1; - significand = (significand >> 1) | (significand & 1); - incr >>= 1; -#if 1 - vd->exponent = exponent; - vd->significand = significand; - vfp_double_dump("pack: overflow", vd); -#endif - } - - /* - * If any of the low bits (which will be shifted out of the - * number) are non-zero, the result is inexact. - */ - if (significand & ((1 << (VFP_DOUBLE_LOW_BITS + 1)) - 1)) - exceptions |= FPSCR_IXC; - - /* - * Do our rounding. - */ - significand += incr; - - /* - * Infinity? - */ - if (exponent >= 2046) { - exceptions |= FPSCR_OFC | FPSCR_IXC; - if (incr == 0) { - vd->exponent = 2045; - vd->significand = 0x7fffffffffffffffULL; - } - else { - vd->exponent = 2047; /* infinity */ - vd->significand = 0; - } - } - else { - if (significand >> (VFP_DOUBLE_LOW_BITS + 1) == 0) - exponent = 0; - if (exponent || significand > 0x8000000000000000ULL) - underflow = 0; - if (underflow) - exceptions |= FPSCR_UFC; - vd->exponent = exponent; - vd->significand = significand >> 1; - } - pack: - return 0; -} - u32 vfp_double_normaliseround(ARMul_State* state, int dd, struct vfp_double *vd, u32 fpscr, u32 exceptions, const char *func) { u64 significand, incr; @@ -511,7 +383,7 @@ static u32 vfp_compare(ARMul_State* state, int dd, int signal_on_qnan, int dm, u LOG_TRACE(Core_ARM11, "In %s, state=0x%x, fpscr=0x%x\n", __FUNCTION__, state, fpscr); m = vfp_get_double(state, dm); if (vfp_double_packed_exponent(m) == 2047 && vfp_double_packed_mantissa(m)) { - ret |= FPSCR_C | FPSCR_V; + ret |= FPSCR_CFLAG | FPSCR_VFLAG; if (signal_on_qnan || !(vfp_double_packed_mantissa(m) & (1ULL << (VFP_DOUBLE_MANTISSA_BITS - 1)))) /* * Signalling NaN, or signalling on quiet NaN @@ -521,7 +393,7 @@ static u32 vfp_compare(ARMul_State* state, int dd, int signal_on_qnan, int dm, u d = vfp_get_double(state, dd); if (vfp_double_packed_exponent(d) == 2047 && vfp_double_packed_mantissa(d)) { - ret |= FPSCR_C | FPSCR_V; + ret |= FPSCR_CFLAG | FPSCR_VFLAG; if (signal_on_qnan || !(vfp_double_packed_mantissa(d) & (1ULL << (VFP_DOUBLE_MANTISSA_BITS - 1)))) /* * Signalling NaN, or signalling on quiet NaN @@ -535,7 +407,7 @@ static u32 vfp_compare(ARMul_State* state, int dd, int signal_on_qnan, int dm, u /* * equal */ - ret |= FPSCR_Z | FPSCR_C; + ret |= FPSCR_ZFLAG | FPSCR_CFLAG; //printf("In %s,1 ret=0x%x\n", __FUNCTION__, ret); } else if (vfp_double_packed_sign(d ^ m)) { /* @@ -545,22 +417,22 @@ static u32 vfp_compare(ARMul_State* state, int dd, int signal_on_qnan, int dm, u /* * d is negative, so d < m */ - ret |= FPSCR_N; + ret |= FPSCR_NFLAG; else /* * d is positive, so d > m */ - ret |= FPSCR_C; + ret |= FPSCR_CFLAG; } else if ((vfp_double_packed_sign(d) != 0) ^ (d < m)) { /* * d < m */ - ret |= FPSCR_N; + ret |= FPSCR_NFLAG; } else if ((vfp_double_packed_sign(d) != 0) ^ (d > m)) { /* * d > m */ - ret |= FPSCR_C; + ret |= FPSCR_CFLAG; } } LOG_TRACE(Core_ARM11, "In %s, state=0x%x, ret=0x%x\n", __FUNCTION__, state, ret); @@ -592,49 +464,6 @@ static u32 vfp_double_fcmpez(ARMul_State* state, int dd, int unused, int dm, u32 return vfp_compare(state, dd, 1, VFP_REG_ZERO, fpscr); } -u32 vfp_double_fcvtsinterncutting(ARMul_State* state, int sd, struct vfp_double* dm, u32 fpscr) //ichfly for internal use only -{ - struct vfp_single vsd; - int tm; - u32 exceptions = 0; - - LOG_TRACE(Core_ARM11, "In %s\n", __FUNCTION__); - - tm = vfp_double_type(dm); - - /* - * If we have a signalling NaN, signal invalid operation. - */ - if (tm == VFP_SNAN) - exceptions = FPSCR_IOC; - - if (tm & VFP_DENORMAL) - vfp_double_normalise_denormal(dm); - - vsd.sign = dm->sign; - vsd.significand = vfp_hi64to32jamming(dm->significand); - - /* - * If we have an infinity or a NaN, the exponent must be 255 - */ - if (tm & (VFP_INFINITY | VFP_NAN)) { - vsd.exponent = 255; - if (tm == VFP_QNAN) - vsd.significand |= VFP_SINGLE_SIGNIFICAND_QNAN; - goto pack_nan; - } - else if (tm & VFP_ZERO) - vsd.exponent = 0; - else - vsd.exponent = dm->exponent - (1023 - 127); - - return vfp_single_normaliseround(state, sd, &vsd, fpscr, exceptions, "fcvts"); - -pack_nan: - vfp_put_float(state, vfp_single_pack(&vsd), sd); - return exceptions; -} - static u32 vfp_double_fcvts(ARMul_State* state, int sd, int unused, int dm, u32 fpscr) { struct vfp_double vdm; @@ -723,7 +552,7 @@ static u32 vfp_double_ftoui(ARMul_State* state, int sd, int unused, int dm, u32 exceptions |= FPSCR_IDC; if (tm & VFP_NAN) - vdm.sign = 0; + vdm.sign = 1; if (vdm.exponent >= 1023 + 32) { d = vdm.sign ? 0 : 0xffffffff; @@ -880,18 +709,18 @@ static struct op fops_ext[] = { { NULL, 0 }, { NULL, 0 }, { vfp_double_fcvts, OP_SCALAR|OP_DD }, //0x0000000F - FEXT_FCVT - { vfp_double_fuito, OP_SCALAR }, //0x00000010 - FEXT_FUITO - { vfp_double_fsito, OP_SCALAR }, //0x00000011 - FEXT_FSITO + { vfp_double_fuito, OP_SCALAR|OP_SM }, //0x00000010 - FEXT_FUITO + { vfp_double_fsito, OP_SCALAR|OP_SM }, //0x00000011 - FEXT_FSITO { NULL, 0 }, { NULL, 0 }, { NULL, 0 }, { NULL, 0 }, { NULL, 0 }, { NULL, 0 }, - { vfp_double_ftoui, OP_SCALAR }, //0x00000018 - FEXT_FTOUI - { vfp_double_ftouiz, OP_SCALAR }, //0x00000019 - FEXT_FTOUIZ - { vfp_double_ftosi, OP_SCALAR }, //0x0000001A - FEXT_FTOSI - { vfp_double_ftosiz, OP_SCALAR }, //0x0000001B - FEXT_FTOSIZ + { vfp_double_ftoui, OP_SCALAR|OP_SD }, //0x00000018 - FEXT_FTOUI + { vfp_double_ftouiz, OP_SCALAR|OP_SD }, //0x00000019 - FEXT_FTOUIZ + { vfp_double_ftosi, OP_SCALAR|OP_SD }, //0x0000001A - FEXT_FTOSI + { vfp_double_ftosiz, OP_SCALAR|OP_SD }, //0x0000001B - FEXT_FTOSIZ }; static u32 @@ -1079,6 +908,9 @@ vfp_double_multiply_accumulate(ARMul_State* state, int dd, int dn, int dm, u32 f vdp.sign = vfp_sign_negate(vdp.sign); vfp_double_unpack(&vdn, vfp_get_double(state, dd)); + if (vdn.exponent == 0 && vdn.significand != 0) + vfp_double_normalise_denormal(&vdn); + if (negate & NEG_SUBTRACT) vdn.sign = vfp_sign_negate(vdn.sign); diff --git a/src/core/arm/skyeye_common/vfp/vfpinstr.cpp b/src/core/arm/skyeye_common/vfp/vfpinstr.cpp index 1bdbfec8e..0aa2d5089 100644 --- a/src/core/arm/skyeye_common/vfp/vfpinstr.cpp +++ b/src/core/arm/skyeye_common/vfp/vfpinstr.cpp @@ -59,73 +59,6 @@ VMLA_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmla), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmla)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmla)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - //arch_arm_undef(cpu, bb, instr); - int m; - int n; - int d ; - int add = (BIT(6) == 0); - int s = BIT(8) == 0; - Value *mm; - Value *nn; - Value *tmp; - if(s){ - m = BIT(5) | BITS(0,3) << 1; - n = BIT(7) | BITS(16,19) << 1; - d = BIT(22) | BITS(12,15) << 1; - mm = FR32(m); - nn = FR32(n); - tmp = FPMUL(nn,mm); - if(!add) - tmp = FPNEG32(tmp); - mm = FR32(d); - tmp = FPADD(mm,tmp); - //LETS(d,tmp); - LETFPS(d,tmp); - }else { - m = BITS(0,3) | BIT(5) << 4; - n = BITS(16,19) | BIT(7) << 4; - d = BIT(22) << 4 | BITS(12,15); - //mm = SITOFP(32,RSPR(m)); - //LETS(d,tmp); - mm = ZEXT64(IBITCAST32(FR32(2 * m))); - nn = ZEXT64(IBITCAST32(FR32(2 * m + 1))); - tmp = OR(SHL(nn,CONST64(32)),mm); - mm = FPBITCAST64(tmp); - tmp = ZEXT64(IBITCAST32(FR32(2 * n))); - nn = ZEXT64(IBITCAST32(FR32(2 * n + 1))); - nn = OR(SHL(nn,CONST64(32)),tmp); - nn = FPBITCAST64(nn); - tmp = FPMUL(nn,mm); - if(!add) - tmp = FPNEG64(tmp); - mm = ZEXT64(IBITCAST32(FR32(2 * d))); - nn = ZEXT64(IBITCAST32(FR32(2 * d + 1))); - mm = OR(SHL(nn,CONST64(32)),mm); - mm = FPBITCAST64(mm); - tmp = FPADD(mm,tmp); - mm = TRUNC32(LSHR(IBITCAST64(tmp),CONST64(32))); - nn = TRUNC32(AND(IBITCAST64(tmp),CONST64(0xffffffff))); - LETFPS(2*d ,FPBITCAST32(nn)); - LETFPS(d*2 + 1 , FPBITCAST32(mm)); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VNMLS */ /* cond 1110 0D00 Vn-- Vd-- 101X N1M0 Vm-- */ @@ -176,74 +109,6 @@ VMLS_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmls), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmls)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmls)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s VMLS instruction is executed out of here.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int m; - int n; - int d ; - int add = (BIT(6) == 0); - int s = BIT(8) == 0; - Value *mm; - Value *nn; - Value *tmp; - if(s){ - m = BIT(5) | BITS(0,3) << 1; - n = BIT(7) | BITS(16,19) << 1; - d = BIT(22) | BITS(12,15) << 1; - mm = FR32(m); - nn = FR32(n); - tmp = FPMUL(nn,mm); - if(!add) - tmp = FPNEG32(tmp); - mm = FR32(d); - tmp = FPADD(mm,tmp); - //LETS(d,tmp); - LETFPS(d,tmp); - }else { - m = BITS(0,3) | BIT(5) << 4; - n = BITS(16,19) | BIT(7) << 4; - d = BIT(22) << 4 | BITS(12,15); - //mm = SITOFP(32,RSPR(m)); - //LETS(d,tmp); - mm = ZEXT64(IBITCAST32(FR32(2 * m))); - nn = ZEXT64(IBITCAST32(FR32(2 * m + 1))); - tmp = OR(SHL(nn,CONST64(32)),mm); - mm = FPBITCAST64(tmp); - tmp = ZEXT64(IBITCAST32(FR32(2 * n))); - nn = ZEXT64(IBITCAST32(FR32(2 * n + 1))); - nn = OR(SHL(nn,CONST64(32)),tmp); - nn = FPBITCAST64(nn); - tmp = FPMUL(nn,mm); - if(!add) - tmp = FPNEG64(tmp); - mm = ZEXT64(IBITCAST32(FR32(2 * d))); - nn = ZEXT64(IBITCAST32(FR32(2 * d + 1))); - mm = OR(SHL(nn,CONST64(32)),mm); - mm = FPBITCAST64(mm); - tmp = FPADD(mm,tmp); - mm = TRUNC32(LSHR(IBITCAST64(tmp),CONST64(32))); - nn = TRUNC32(AND(IBITCAST64(tmp),CONST64(0xffffffff))); - LETFPS(2*d ,FPBITCAST32(nn)); - LETFPS(d*2 + 1 , FPBITCAST32(mm)); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VNMLA */ /* cond 1110 0D01 Vn-- Vd-- 101X N1M0 Vm-- */ @@ -294,75 +159,6 @@ VNMLA_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vnmla), -DYNCOM_FILL_ACTION(vnmla), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vnmla)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vnmla)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s VNMLA instruction is executed out of here.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int m; - int n; - int d ; - int add = (BIT(6) == 0); - int s = BIT(8) == 0; - Value *mm; - Value *nn; - Value *tmp; - if(s){ - m = BIT(5) | BITS(0,3) << 1; - n = BIT(7) | BITS(16,19) << 1; - d = BIT(22) | BITS(12,15) << 1; - mm = FR32(m); - nn = FR32(n); - tmp = FPMUL(nn,mm); - if(!add) - tmp = FPNEG32(tmp); - mm = FR32(d); - tmp = FPADD(FPNEG32(mm),tmp); - //LETS(d,tmp); - LETFPS(d,tmp); - }else { - m = BITS(0,3) | BIT(5) << 4; - n = BITS(16,19) | BIT(7) << 4; - d = BIT(22) << 4 | BITS(12,15); - //mm = SITOFP(32,RSPR(m)); - //LETS(d,tmp); - mm = ZEXT64(IBITCAST32(FR32(2 * m))); - nn = ZEXT64(IBITCAST32(FR32(2 * m + 1))); - tmp = OR(SHL(nn,CONST64(32)),mm); - mm = FPBITCAST64(tmp); - tmp = ZEXT64(IBITCAST32(FR32(2 * n))); - nn = ZEXT64(IBITCAST32(FR32(2 * n + 1))); - nn = OR(SHL(nn,CONST64(32)),tmp); - nn = FPBITCAST64(nn); - tmp = FPMUL(nn,mm); - if(!add) - tmp = FPNEG64(tmp); - mm = ZEXT64(IBITCAST32(FR32(2 * d))); - nn = ZEXT64(IBITCAST32(FR32(2 * d + 1))); - mm = OR(SHL(nn,CONST64(32)),mm); - mm = FPBITCAST64(mm); - tmp = FPADD(FPNEG64(mm),tmp); - mm = TRUNC32(LSHR(IBITCAST64(tmp),CONST64(32))); - nn = TRUNC32(AND(IBITCAST64(tmp),CONST64(0xffffffff))); - LETFPS(2*d ,FPBITCAST32(nn)); - LETFPS(d*2 + 1 , FPBITCAST32(mm)); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VNMLS */ /* cond 1110 0D01 Vn-- Vd-- 101X N0M0 Vm-- */ @@ -414,75 +210,6 @@ VNMLS_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vnmls), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vnmls)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vnmls)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int m; - int n; - int d ; - int add = (BIT(6) == 0); - int s = BIT(8) == 0; - Value *mm; - Value *nn; - Value *tmp; - if(s){ - m = BIT(5) | BITS(0,3) << 1; - n = BIT(7) | BITS(16,19) << 1; - d = BIT(22) | BITS(12,15) << 1; - mm = FR32(m); - nn = FR32(n); - tmp = FPMUL(nn,mm); - if(!add) - tmp = FPNEG32(tmp); - mm = FR32(d); - tmp = FPADD(FPNEG32(mm),tmp); - //LETS(d,tmp); - LETFPS(d,tmp); - }else { - m = BITS(0,3) | BIT(5) << 4; - n = BITS(16,19) | BIT(7) << 4; - d = BIT(22) << 4 | BITS(12,15); - //mm = SITOFP(32,RSPR(m)); - //LETS(d,tmp); - mm = ZEXT64(IBITCAST32(FR32(2 * m))); - nn = ZEXT64(IBITCAST32(FR32(2 * m + 1))); - tmp = OR(SHL(nn,CONST64(32)),mm); - mm = FPBITCAST64(tmp); - tmp = ZEXT64(IBITCAST32(FR32(2 * n))); - nn = ZEXT64(IBITCAST32(FR32(2 * n + 1))); - nn = OR(SHL(nn,CONST64(32)),tmp); - nn = FPBITCAST64(nn); - tmp = FPMUL(nn,mm); - if(!add) - tmp = FPNEG64(tmp); - mm = ZEXT64(IBITCAST32(FR32(2 * d))); - nn = ZEXT64(IBITCAST32(FR32(2 * d + 1))); - mm = OR(SHL(nn,CONST64(32)),mm); - mm = FPBITCAST64(mm); - tmp = FPADD(FPNEG64(mm),tmp); - mm = TRUNC32(LSHR(IBITCAST64(tmp),CONST64(32))); - nn = TRUNC32(AND(IBITCAST64(tmp),CONST64(0xffffffff))); - LETFPS(2*d ,FPBITCAST32(nn)); - LETFPS(d*2 + 1 , FPBITCAST32(mm)); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VNMUL */ /* cond 1110 0D10 Vn-- Vd-- 101X N0M0 Vm-- */ @@ -533,65 +260,6 @@ VNMUL_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vnmul), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vnmul)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vnmul)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int m; - int n; - int d ; - int add = (BIT(6) == 0); - int s = BIT(8) == 0; - Value *mm; - Value *nn; - Value *tmp; - if(s){ - m = BIT(5) | BITS(0,3) << 1; - n = BIT(7) | BITS(16,19) << 1; - d = BIT(22) | BITS(12,15) << 1; - mm = FR32(m); - nn = FR32(n); - tmp = FPMUL(nn,mm); - //LETS(d,tmp); - LETFPS(d,FPNEG32(tmp)); - }else { - m = BITS(0,3) | BIT(5) << 4; - n = BITS(16,19) | BIT(7) << 4; - d = BIT(22) << 4 | BITS(12,15); - //mm = SITOFP(32,RSPR(m)); - //LETS(d,tmp); - mm = ZEXT64(IBITCAST32(FR32(2 * m))); - nn = ZEXT64(IBITCAST32(FR32(2 * m + 1))); - tmp = OR(SHL(nn,CONST64(32)),mm); - mm = FPBITCAST64(tmp); - tmp = ZEXT64(IBITCAST32(FR32(2 * n))); - nn = ZEXT64(IBITCAST32(FR32(2 * n + 1))); - nn = OR(SHL(nn,CONST64(32)),tmp); - nn = FPBITCAST64(nn); - tmp = FPMUL(nn,mm); - tmp = FPNEG64(tmp); - mm = TRUNC32(LSHR(IBITCAST64(tmp),CONST64(32))); - nn = TRUNC32(AND(IBITCAST64(tmp),CONST64(0xffffffff))); - LETFPS(2*d ,FPBITCAST32(nn)); - LETFPS(d*2 + 1 , FPBITCAST32(mm)); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VMUL */ /* cond 1110 0D10 Vn-- Vd-- 101X N0M0 Vm-- */ @@ -642,77 +310,6 @@ VMUL_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmul), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmul)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmul)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int m; - int n; - int d ; - int s = BIT(8) == 0; - Value *mm; - Value *nn; - Value *tmp; - if(s){ - m = BIT(5) | BITS(0,3) << 1; - n = BIT(7) | BITS(16,19) << 1; - d = BIT(22) | BITS(12,15) << 1; - //mm = SITOFP(32,FR(m)); - //nn = SITOFP(32,FRn)); - mm = FR32(m); - nn = FR32(n); - tmp = FPMUL(nn,mm); - //LETS(d,tmp); - LETFPS(d,tmp); - }else { - m = BITS(0,3) | BIT(5) << 4; - n = BITS(16,19) | BIT(7) << 4; - d = BIT(22) << 4 | BITS(12,15); - //mm = SITOFP(32,RSPR(m)); - //LETS(d,tmp); - Value *lo = FR32(2 * m); - Value *hi = FR32(2 * m + 1); - hi = IBITCAST32(hi); - lo = IBITCAST32(lo); - Value *hi64 = ZEXT64(hi); - Value* lo64 = ZEXT64(lo); - Value* v64 = OR(SHL(hi64,CONST64(32)),lo64); - Value* m0 = FPBITCAST64(v64); - lo = FR32(2 * n); - hi = FR32(2 * n + 1); - hi = IBITCAST32(hi); - lo = IBITCAST32(lo); - hi64 = ZEXT64(hi); - lo64 = ZEXT64(lo); - v64 = OR(SHL(hi64,CONST64(32)),lo64); - Value *n0 = FPBITCAST64(v64); - tmp = FPMUL(n0,m0); - Value *val64 = IBITCAST64(tmp); - hi = LSHR(val64,CONST64(32)); - lo = AND(val64,CONST64(0xffffffff)); - hi = TRUNC32(hi); - lo = TRUNC32(lo); - hi = FPBITCAST32(hi); - lo = FPBITCAST32(lo); - LETFPS(2*d ,lo); - LETFPS(d*2 + 1 , hi); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VADD */ /* cond 1110 0D11 Vn-- Vd-- 101X N0M0 Vm-- */ @@ -763,73 +360,6 @@ VADD_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vadd), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vadd)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vadd)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction will implement out of JIT.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int m; - int n; - int d ; - int s = BIT(8) == 0; - Value *mm; - Value *nn; - Value *tmp; - if(s){ - m = BIT(5) | BITS(0,3) << 1; - n = BIT(7) | BITS(16,19) << 1; - d = BIT(22) | BITS(12,15) << 1; - mm = FR32(m); - nn = FR32(n); - tmp = FPADD(nn,mm); - LETFPS(d,tmp); - }else { - m = BITS(0,3) | BIT(5) << 4; - n = BITS(16,19) | BIT(7) << 4; - d = BIT(22) << 4 | BITS(12,15); - Value *lo = FR32(2 * m); - Value *hi = FR32(2 * m + 1); - hi = IBITCAST32(hi); - lo = IBITCAST32(lo); - Value *hi64 = ZEXT64(hi); - Value* lo64 = ZEXT64(lo); - Value* v64 = OR(SHL(hi64,CONST64(32)),lo64); - Value* m0 = FPBITCAST64(v64); - lo = FR32(2 * n); - hi = FR32(2 * n + 1); - hi = IBITCAST32(hi); - lo = IBITCAST32(lo); - hi64 = ZEXT64(hi); - lo64 = ZEXT64(lo); - v64 = OR(SHL(hi64,CONST64(32)),lo64); - Value *n0 = FPBITCAST64(v64); - tmp = FPADD(n0,m0); - Value *val64 = IBITCAST64(tmp); - hi = LSHR(val64,CONST64(32)); - lo = AND(val64,CONST64(0xffffffff)); - hi = TRUNC32(hi); - lo = TRUNC32(lo); - hi = FPBITCAST32(hi); - lo = FPBITCAST32(lo); - LETFPS(2*d ,lo); - LETFPS(d*2 + 1 , hi); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VSUB */ /* cond 1110 0D11 Vn-- Vd-- 101X N1M0 Vm-- */ @@ -879,71 +409,6 @@ VSUB_INST: GOTO_NEXT_INST; } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vsub), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vsub)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vsub)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instr=0x%x, instruction is executed out of JIT.\n", __FUNCTION__, instr); - //arch_arm_undef(cpu, bb, instr); - int m; - int n; - int d ; - int s = BIT(8) == 0; - Value *mm; - Value *nn; - Value *tmp; - if(s){ - m = BIT(5) | BITS(0,3) << 1; - n = BIT(7) | BITS(16,19) << 1; - d = BIT(22) | BITS(12,15) << 1; - mm = FR32(m); - nn = FR32(n); - tmp = FPSUB(nn,mm); - LETFPS(d,tmp); - }else { - m = BITS(0,3) | BIT(5) << 4; - n = BITS(16,19) | BIT(7) << 4; - d = BIT(22) << 4 | BITS(12,15); - Value *lo = FR32(2 * m); - Value *hi = FR32(2 * m + 1); - hi = IBITCAST32(hi); - lo = IBITCAST32(lo); - Value *hi64 = ZEXT64(hi); - Value* lo64 = ZEXT64(lo); - Value* v64 = OR(SHL(hi64,CONST64(32)),lo64); - Value* m0 = FPBITCAST64(v64); - lo = FR32(2 * n); - hi = FR32(2 * n + 1); - hi = IBITCAST32(hi); - lo = IBITCAST32(lo); - hi64 = ZEXT64(hi); - lo64 = ZEXT64(lo); - v64 = OR(SHL(hi64,CONST64(32)),lo64); - Value *n0 = FPBITCAST64(v64); - tmp = FPSUB(n0,m0); - Value *val64 = IBITCAST64(tmp); - hi = LSHR(val64,CONST64(32)); - lo = AND(val64,CONST64(0xffffffff)); - hi = TRUNC32(hi); - lo = TRUNC32(lo); - hi = FPBITCAST32(hi); - lo = FPBITCAST32(lo); - LETFPS(2*d ,lo); - LETFPS(d*2 + 1 , hi); - } - return No_exp; -} -#endif /* ----------------------------------------------------------------------- */ /* VDIV */ @@ -995,73 +460,6 @@ VDIV_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vdiv), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vdiv)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vdiv)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int m; - int n; - int d ; - int s = BIT(8) == 0; - Value *mm; - Value *nn; - Value *tmp; - if(s){ - m = BIT(5) | BITS(0,3) << 1; - n = BIT(7) | BITS(16,19) << 1; - d = BIT(22) | BITS(12,15) << 1; - mm = FR32(m); - nn = FR32(n); - tmp = FPDIV(nn,mm); - LETFPS(d,tmp); - }else { - m = BITS(0,3) | BIT(5) << 4; - n = BITS(16,19) | BIT(7) << 4; - d = BIT(22) << 4 | BITS(12,15); - Value *lo = FR32(2 * m); - Value *hi = FR32(2 * m + 1); - hi = IBITCAST32(hi); - lo = IBITCAST32(lo); - Value *hi64 = ZEXT64(hi); - Value* lo64 = ZEXT64(lo); - Value* v64 = OR(SHL(hi64,CONST64(32)),lo64); - Value* m0 = FPBITCAST64(v64); - lo = FR32(2 * n); - hi = FR32(2 * n + 1); - hi = IBITCAST32(hi); - lo = IBITCAST32(lo); - hi64 = ZEXT64(hi); - lo64 = ZEXT64(lo); - v64 = OR(SHL(hi64,CONST64(32)),lo64); - Value *n0 = FPBITCAST64(v64); - tmp = FPDIV(n0,m0); - Value *val64 = IBITCAST64(tmp); - hi = LSHR(val64,CONST64(32)); - lo = AND(val64,CONST64(0xffffffff)); - hi = TRUNC32(hi); - lo = TRUNC32(lo); - hi = FPBITCAST32(hi); - lo = FPBITCAST32(lo); - LETFPS(2*d ,lo); - LETFPS(d*2 + 1 , hi); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VMOVI move immediate */ /* cond 1110 1D11 im4H Vd-- 101X 0000 im4L */ @@ -1111,46 +509,6 @@ VMOVI_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmovi), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmovi)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmovi)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int single = (BIT(8) == 0); - int d; - int imm32; - Value *v; - Value *tmp; - v = CONST32(BITS(0,3) | BITS(16,19) << 4); - //v = CONST64(0x3ff0000000000000); - if(single){ - d = BIT(22) | BITS(12,15) << 1; - }else { - d = BITS(12,15) | BIT(22) << 4; - } - if(single){ - LETFPS(d,FPBITCAST32(v)); - }else { - //v = UITOFP(64,v); - //tmp = IBITCAST64(v); - LETFPS(d*2 ,FPBITCAST32(TRUNC32(AND(v,CONST64(0xffffffff))))); - LETFPS(d * 2 + 1,FPBITCAST32(TRUNC32(LSHR(v,CONST64(32))))); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VMOVR move register */ /* cond 1110 1D11 0000 Vd-- 101X 01M0 Vm-- */ @@ -1196,40 +554,6 @@ VMOVR_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmovr), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmovr)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - if(instr >> 28 != 0xe) - *tag |= TAG_CONDITIONAL; - - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmovr)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - int single = BIT(8) == 0; - int d = (single ? BITS(12,15)<<1 | BIT(22) : BIT(22) << 4 | BITS(12,15)); - int m = (single ? BITS(0, 3)<<1 | BIT(5) : BITS(0, 3) | BIT(5)<<4); - - if (single) - { - LETFPS(d, FR32(m)); - } - else - { - /* Check endian please */ - LETFPS((d*2 + 1), FR32(m*2 + 1)); - LETFPS((d * 2), FR32(m * 2)); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VABS */ /* cond 1110 1D11 0000 Vd-- 101X 11M0 Vm-- */ @@ -1280,57 +604,6 @@ VABS_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vabs), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vabs)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vabs)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - //arch_arm_undef(cpu, bb, instr); - int single = BIT(8) == 0; - int d = (single ? BITS(12,15)<<1 | BIT(22) : BIT(22) << 4 | BITS(12,15)); - int m = (single ? BITS(0, 3)<<1 | BIT(5) : BITS(0, 3) | BIT(5)<<4); - Value* m0; - if (single) - { - m0 = FR32(m); - m0 = SELECT(FPCMP_OLT(m0,FPCONST32(0.0)),FPNEG32(m0),m0); - LETFPS(d,m0); - } - else - { - /* Check endian please */ - Value *lo = FR32(2 * m); - Value *hi = FR32(2 * m + 1); - hi = IBITCAST32(hi); - lo = IBITCAST32(lo); - Value *hi64 = ZEXT64(hi); - Value* lo64 = ZEXT64(lo); - Value* v64 = OR(SHL(hi64,CONST64(32)),lo64); - m0 = FPBITCAST64(v64); - m0 = SELECT(FPCMP_OLT(m0,FPCONST64(0.0)),FPNEG64(m0),m0); - Value *val64 = IBITCAST64(m0); - hi = LSHR(val64,CONST64(32)); - lo = AND(val64,CONST64(0xffffffff)); - hi = TRUNC32(hi); - lo = TRUNC32(lo); - hi = FPBITCAST32(hi); - lo = FPBITCAST32(lo); - LETFPS(2*d ,lo); - LETFPS(d*2 + 1 , hi); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VNEG */ /* cond 1110 1D11 0001 Vd-- 101X 11M0 Vm-- */ @@ -1382,59 +655,6 @@ VNEG_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vneg), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vneg)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vneg)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int single = BIT(8) == 0; - int d = (single ? BITS(12,15)<<1 | BIT(22) : BIT(22) << 4 | BITS(12,15)); - int m = (single ? BITS(0, 3)<<1 | BIT(5) : BITS(0, 3) | BIT(5)<<4); - Value* m0; - if (single) - { - m0 = FR32(m); - m0 = FPNEG32(m0); - LETFPS(d,m0); - } - else - { - /* Check endian please */ - Value *lo = FR32(2 * m); - Value *hi = FR32(2 * m + 1); - hi = IBITCAST32(hi); - lo = IBITCAST32(lo); - Value *hi64 = ZEXT64(hi); - Value* lo64 = ZEXT64(lo); - Value* v64 = OR(SHL(hi64,CONST64(32)),lo64); - m0 = FPBITCAST64(v64); - m0 = FPNEG64(m0); - Value *val64 = IBITCAST64(m0); - hi = LSHR(val64,CONST64(32)); - lo = AND(val64,CONST64(0xffffffff)); - hi = TRUNC32(hi); - lo = TRUNC32(lo); - hi = FPBITCAST32(hi); - lo = FPBITCAST32(lo); - LETFPS(2*d ,lo); - LETFPS(d*2 + 1 , hi); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VSQRT */ /* cond 1110 1D11 0001 Vd-- 101X 11M0 Vm-- */ @@ -1485,47 +705,6 @@ VSQRT_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vsqrt), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vsqrt)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vsqrt)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int dp_op = (BIT(8) == 1); - int d = dp_op ? BITS(12,15) | BIT(22) << 4 : BIT(22) | BITS(12,15) << 1; - int m = dp_op ? BITS(0,3) | BIT(5) << 4 : BIT(5) | BITS(0,3) << 1; - Value* v; - Value* tmp; - if(dp_op){ - v = SHL(ZEXT64(IBITCAST32(FR32(2 * m + 1))),CONST64(32)); - tmp = ZEXT64(IBITCAST32(FR32(2 * m))); - v = OR(v,tmp); - v = FPSQRT(FPBITCAST64(v)); - tmp = TRUNC32(LSHR(IBITCAST64(v),CONST64(32))); - v = TRUNC32(AND(IBITCAST64(v),CONST64( 0xffffffff))); - LETFPS(2 * d , FPBITCAST32(v)); - LETFPS(2 * d + 1, FPBITCAST32(tmp)); - }else { - v = FR32(m); - v = FPSQRT(FPEXT(64,v)); - v = FPTRUNC(32,v); - LETFPS(d,v); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VCMP VCMPE */ /* cond 1110 1D11 0100 Vd-- 101X E1M0 Vm-- Encoding 1 */ @@ -1576,74 +755,6 @@ VCMP_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vcmp), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vcmp)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vcmp)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is executed out of JIT.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int dp_op = (BIT(8) == 1); - int d = dp_op ? BITS(12,15) | BIT(22) << 4 : BIT(22) | BITS(12,15) << 1; - int m = dp_op ? BITS(0,3) | BIT(5) << 4 : BIT(5) | BITS(0,3) << 1; - Value* v; - Value* tmp; - Value* n; - Value* z; - Value* c; - Value* vt; - Value* v1; - Value* nzcv; - if(dp_op){ - v = SHL(ZEXT64(IBITCAST32(FR32(2 * m + 1))),CONST64(32)); - tmp = ZEXT64(IBITCAST32(FR32(2 * m))); - v1 = OR(v,tmp); - v = SHL(ZEXT64(IBITCAST32(FR32(2 * d + 1))),CONST64(32)); - tmp = ZEXT64(IBITCAST32(FR32(2 * d))); - v = OR(v,tmp); - z = FPCMP_OEQ(FPBITCAST64(v),FPBITCAST64(v1)); - n = FPCMP_OLT(FPBITCAST64(v),FPBITCAST64(v1)); - c = FPCMP_OGE(FPBITCAST64(v),FPBITCAST64(v1)); - tmp = FPCMP_UNO(FPBITCAST64(v),FPBITCAST64(v1)); - v1 = tmp; - c = OR(c,tmp); - n = SHL(ZEXT32(n),CONST32(31)); - z = SHL(ZEXT32(z),CONST32(30)); - c = SHL(ZEXT32(c),CONST32(29)); - v1 = SHL(ZEXT32(v1),CONST(28)); - nzcv = OR(OR(OR(n,z),c),v1); - v = R(VFP_FPSCR); - tmp = OR(nzcv,AND(v,CONST32(0x0fffffff))); - LET(VFP_FPSCR,tmp); - }else { - z = FPCMP_OEQ(FR32(d),FR32(m)); - n = FPCMP_OLT(FR32(d),FR32(m)); - c = FPCMP_OGE(FR32(d),FR32(m)); - tmp = FPCMP_UNO(FR32(d),FR32(m)); - c = OR(c,tmp); - v1 = tmp; - n = SHL(ZEXT32(n),CONST32(31)); - z = SHL(ZEXT32(z),CONST32(30)); - c = SHL(ZEXT32(c),CONST32(29)); - v1 = SHL(ZEXT32(v1),CONST(28)); - nzcv = OR(OR(OR(n,z),c),v1); - v = R(VFP_FPSCR); - tmp = OR(nzcv,AND(v,CONST32(0x0fffffff))); - LET(VFP_FPSCR,tmp); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VCMP VCMPE */ /* cond 1110 1D11 0100 Vd-- 101X E1M0 Vm-- Encoding 2 */ @@ -1694,74 +805,6 @@ VCMP2_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vcmp2), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vcmp2)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vcmp2)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction will executed out of JIT.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int dp_op = (BIT(8) == 1); - int d = dp_op ? BITS(12,15) | BIT(22) << 4 : BIT(22) | BITS(12,15) << 1; - //int m = dp_op ? BITS(0,3) | BIT(5) << 4 : BIT(5) | BITS(0,3) << 1; - Value* v; - Value* tmp; - Value* n; - Value* z; - Value* c; - Value* vt; - Value* v1; - Value* nzcv; - if(dp_op){ - v1 = CONST64(0); - v = SHL(ZEXT64(IBITCAST32(FR32(2 * d + 1))),CONST64(32)); - tmp = ZEXT64(IBITCAST32(FR32(2 * d))); - v = OR(v,tmp); - z = FPCMP_OEQ(FPBITCAST64(v),FPBITCAST64(v1)); - n = FPCMP_OLT(FPBITCAST64(v),FPBITCAST64(v1)); - c = FPCMP_OGE(FPBITCAST64(v),FPBITCAST64(v1)); - tmp = FPCMP_UNO(FPBITCAST64(v),FPBITCAST64(v1)); - v1 = tmp; - c = OR(c,tmp); - n = SHL(ZEXT32(n),CONST32(31)); - z = SHL(ZEXT32(z),CONST32(30)); - c = SHL(ZEXT32(c),CONST32(29)); - v1 = SHL(ZEXT32(v1),CONST(28)); - nzcv = OR(OR(OR(n,z),c),v1); - v = R(VFP_FPSCR); - tmp = OR(nzcv,AND(v,CONST32(0x0fffffff))); - LET(VFP_FPSCR,tmp); - }else { - v1 = CONST(0); - v1 = FPBITCAST32(v1); - z = FPCMP_OEQ(FR32(d),v1); - n = FPCMP_OLT(FR32(d),v1); - c = FPCMP_OGE(FR32(d),v1); - tmp = FPCMP_UNO(FR32(d),v1); - c = OR(c,tmp); - v1 = tmp; - n = SHL(ZEXT32(n),CONST32(31)); - z = SHL(ZEXT32(z),CONST32(30)); - c = SHL(ZEXT32(c),CONST32(29)); - v1 = SHL(ZEXT32(v1),CONST(28)); - nzcv = OR(OR(OR(n,z),c),v1); - v = R(VFP_FPSCR); - tmp = OR(nzcv,AND(v,CONST32(0x0fffffff))); - LET(VFP_FPSCR,tmp); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VCVTBDS between double and single */ /* cond 1110 1D11 0111 Vd-- 101X 11M0 Vm-- */ @@ -1812,48 +855,6 @@ VCVTBDS_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vcvtbds), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vcvtbds)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vcvtbds)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is executed out.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - int dp_op = (BIT(8) == 1); - int d = dp_op ? BITS(12,15) << 1 | BIT(22) : BIT(22) << 4 | BITS(12,15); - int m = dp_op ? BITS(0,3) | BIT(5) << 4 : BIT(5) | BITS(0,3) << 1; - int d2s = dp_op; - Value* v; - Value* tmp; - Value* v1; - if(d2s){ - v = SHL(ZEXT64(IBITCAST32(FR32(2 * m + 1))),CONST64(32)); - tmp = ZEXT64(IBITCAST32(FR32(2 * m))); - v1 = OR(v,tmp); - tmp = FPTRUNC(32,FPBITCAST64(v1)); - LETFPS(d,tmp); - }else { - v = FR32(m); - tmp = FPEXT(64,v); - v = IBITCAST64(tmp); - tmp = TRUNC32(AND(v,CONST64(0xffffffff))); - v1 = TRUNC32(LSHR(v,CONST64(32))); - LETFPS(2 * d, FPBITCAST32(tmp) ); - LETFPS(2 * d + 1, FPBITCAST32(v1)); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VCVTBFF between floating point and fixed point */ /* cond 1110 1D11 1op2 Vd-- 101X X1M0 Vm-- */ @@ -1906,26 +907,6 @@ VCVTBFF_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vcvtbff), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vcvtbff)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vcvtbff)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - arch_arm_undef(cpu, bb, instr); - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VCVTBFI between floating point and integer */ /* cond 1110 1D11 1op2 Vd-- 101X X1M0 Vm-- */ @@ -1976,114 +957,6 @@ VCVTBFI_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vcvtbfi), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vcvtbfi)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - LOG_TRACE(Core_ARM11, "\t\tin %s, instruction will be executed out of JIT.\n", __FUNCTION__); - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vcvtbfi)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s, instruction will be executed out of JIT.\n", __FUNCTION__); - //arch_arm_undef(cpu, bb, instr); - unsigned int opc2 = BITS(16,18); - int to_integer = ((opc2 >> 2) == 1); - int dp_op = (BIT(8) == 1); - unsigned int op = BIT(7); - int m,d; - Value* v; - Value* hi; - Value* lo; - Value* v64; - if(to_integer){ - d = BIT(22) | (BITS(12,15) << 1); - if(dp_op) - m = BITS(0,3) | BIT(5) << 4; - else - m = BIT(5) | BITS(0,3) << 1; - }else { - m = BIT(5) | BITS(0,3) << 1; - if(dp_op) - d = BITS(12,15) | BIT(22) << 4; - else - d = BIT(22) | BITS(12,15) << 1; - } - if(to_integer){ - if(dp_op){ - lo = FR32(m * 2); - hi = FR32(m * 2 + 1); - hi = ZEXT64(IBITCAST32(hi)); - lo = ZEXT64(IBITCAST32(lo)); - v64 = OR(SHL(hi,CONST64(32)),lo); - if(BIT(16)){ - v = FPTOSI(32,FPBITCAST64(v64)); - } - else - v = FPTOUI(32,FPBITCAST64(v64)); - - v = FPBITCAST32(v); - LETFPS(d,v); - }else { - v = FR32(m); - if(BIT(16)){ - v = FPTOSI(32,v); - } - else - v = FPTOUI(32,v); - LETFPS(d,FPBITCAST32(v)); - } - }else { - if(dp_op){ - v = IBITCAST32(FR32(m)); - if(BIT(7)) - v64 = SITOFP(64,v); - else - v64 = UITOFP(64,v); - v = IBITCAST64(v64); - hi = FPBITCAST32(TRUNC32(LSHR(v,CONST64(32)))); - lo = FPBITCAST32(TRUNC32(AND(v,CONST64(0xffffffff)))); - LETFPS(2 * d , lo); - LETFPS(2 * d + 1, hi); - }else { - v = IBITCAST32(FR32(m)); - if(BIT(7)) - v = SITOFP(32,v); - else - v = UITOFP(32,v); - LETFPS(d,v); - } - } - return No_exp; -} - -/** -* @brief The implementation of c language for vcvtbfi instruction of dyncom -* -* @param cpu -* @param instr -* -* @return -*/ -int vcvtbfi_instr_impl(arm_core_t* cpu, uint32 instr){ - int dp_operation = BIT(8); - int ret; - if (dp_operation) - ret = vfp_double_cpdo(cpu, instr, cpu->VFP[VFP_OFFSET(VFP_FPSCR)]); - else - ret = vfp_single_cpdo(cpu, instr, cpu->VFP[VFP_OFFSET(VFP_FPSCR)]); - - vfp_raise_exceptions(cpu, ret, instr, cpu->VFP[VFP_OFFSET(VFP_FPSCR)]); - return 0; -} -#endif - /* ----------------------------------------------------------------------- */ /* MRC / MCR instructions */ /* cond 1110 AAAL XXXX XXXX 101C XBB1 XXXX */ @@ -2135,35 +1008,6 @@ VMOVBRS_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmovbrs), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmovbrs)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmovbrs)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - int to_arm = BIT(20) == 1; - int t = BITS(12, 15); - int n = BIT(7) | BITS(16, 19)<<1; - - if (to_arm) - { - LET(t, IBITCAST32(FR32(n))); - } - else - { - LETFPS(n, FPBITCAST32(R(t))); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VMSR */ /* cond 1110 1110 reg- Rt-- 1010 0001 0000 */ @@ -2211,48 +1055,6 @@ VMSR_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmsr), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmsr)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmsr)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - //arch_arm_undef(cpu, bb, instr); - if(RD == 15) { - LOG_ERROR(Core_ARM11, "in %s is not implementation.\n", __FUNCTION__); - exit(-1); - } - - Value *data = NULL; - int reg = RN; - int Rt = RD; - if (reg == 1) - { - LET(VFP_FPSCR, R(Rt)); - } - else - { - switch (reg) - { - case 8: - LET(VFP_FPEXC, R(Rt)); - break; - default: - break; - } - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VMOVBRC register to scalar */ /* cond 1110 0XX0 Vd-- Rt-- 1011 DXX1 0000 */ @@ -2302,26 +1104,6 @@ VMOVBRC_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmovbrc), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmovbrc)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmovbrc)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - arch_arm_undef(cpu, bb, instr); - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VMRS */ /* cond 1110 1111 CRn- Rt-- 1010 0001 0000 */ @@ -2404,64 +1186,6 @@ VMRS_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmrs), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmrs)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmrs)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - //arch_arm_undef(cpu, bb, instr); - - Value *data = NULL; - int reg = BITS(16, 19);; - int Rt = BITS(12, 15); - if (reg == 1) - { - if (Rt != 15) - { - LET(Rt, R(VFP_FPSCR)); - } - else - { - //LET(Rt, R(VFP_FPSCR)); - update_cond_from_fpscr(cpu, instr, bb, pc); - } - } - else - { - switch (reg) - { - case 0: - LET(Rt, R(VFP_FPSID)); - break; - case 6: - /* MVFR1, VFPv3 only ? */ - LOG_TRACE(Core_ARM11, "\tr%d <= MVFR1 unimplemented\n", Rt); - break; - case 7: - /* MVFR0, VFPv3 only? */ - LOG_TRACE(Core_ARM11, "\tr%d <= MVFR0 unimplemented\n", Rt); - break; - case 8: - LET(Rt, R(VFP_FPEXC)); - break; - default: - break; - } - } - - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VMOVBCR scalar to register */ /* cond 1110 XXX1 Vd-- Rt-- 1011 NXX1 0000 */ @@ -2511,26 +1235,6 @@ VMOVBCR_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmovbcr), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmovbcr)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmovbcr)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - LOG_TRACE(Core_ARM11, "\t\tin %s instruction is not implemented.\n", __FUNCTION__); - arch_arm_undef(cpu, bb, instr); - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* MRRC / MCRR instructions */ /* cond 1100 0101 Rt2- Rt-- copr opc1 CRm- MRRC */ @@ -2584,39 +1288,6 @@ VMOVBRRSS_INST: GOTO_NEXT_INST; } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmovbrrss), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmovbrrss)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - if (instr >> 28 != 0xE) - *tag |= TAG_CONDITIONAL; - - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmovbrrss)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc) -{ - int to_arm = BIT(20) == 1; - int t = BITS(12, 15); - int t2 = BITS(16, 19); - int n = BIT(5)<<4 | BITS(0, 3); - if (to_arm) { - LET(t, IBITCAST32(FR32(n + 0))); - LET(t2, IBITCAST32(FR32(n + 1))); - } - else { - LETFPS(n + 0, FPBITCAST32(R(t))); - LETFPS(n + 1, FPBITCAST32(R(t2))); - } - return No_exp; -} -#endif /* ----------------------------------------------------------------------- */ /* VMOVBRRD between 2 registers and 1 double */ @@ -2667,38 +1338,6 @@ VMOVBRRD_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vmovbrrd), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vmovbrrd)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - if(instr >> 28 != 0xe) - *tag |= TAG_CONDITIONAL; - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vmovbrrd)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - //arch_arm_undef(cpu, bb, instr); - int to_arm = BIT(20) == 1; - int t = BITS(12, 15); - int t2 = BITS(16, 19); - int n = BIT(5)<<4 | BITS(0, 3); - if(to_arm){ - LET(t, IBITCAST32(FR32(n * 2))); - LET(t2, IBITCAST32(FR32(n * 2 + 1))); - } - else{ - LETFPS(n * 2, FPBITCAST32(R(t))); - LETFPS(n * 2 + 1, FPBITCAST32(R(t2))); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* LDC/STC between 2 registers and 1 double */ /* cond 110X XXX1 Rn-- CRd- copr imm- imm- LDC */ @@ -2764,53 +1403,6 @@ VSTR_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vstr), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vstr)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - *tag |= TAG_NEW_BB; - if(instr >> 28 != 0xe) - *tag |= TAG_CONDITIONAL; - - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vstr)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - int single = BIT(8) == 0; - int add = BIT(23); - int imm32 = BITS(0,7) << 2; - int d = (single ? BITS(12, 15)<<1|BIT(22) : BITS(12, 15)|(BIT(22)<<4)); - int n = BITS(16, 19); - - Value* base = (n == 15) ? ADD(AND(R(n), CONST(0xFFFFFFFC)), CONST(8)): R(n); - Value* Addr = add ? ADD(base, CONST(imm32)) : SUB(base, CONST(imm32)); - //if(single) - // bb = arch_check_mm(cpu, bb, Addr, 4, 0, cpu->dyncom_engine->bb_trap); - //else - // bb = arch_check_mm(cpu, bb, Addr, 8, 0, cpu->dyncom_engine->bb_trap); - //Value* phys_addr; - if(single){ - //memory_write(cpu, bb, Addr, RSPR(d), 32); - memory_write(cpu, bb, Addr, IBITCAST32(FR32(d)), 32); - bb = cpu->dyncom_engine->bb; - } - else{ - //memory_write(cpu, bb, Addr, RSPR(d * 2), 32); - memory_write(cpu, bb, Addr, IBITCAST32(FR32(d * 2)), 32); - bb = cpu->dyncom_engine->bb; - //memory_write(cpu, bb, ADD(Addr, CONST(4)), RSPR(d * 2 + 1), 32); - memory_write(cpu, bb, ADD(Addr, CONST(4)), IBITCAST32(FR32(d * 2 + 1)), 32); - bb = cpu->dyncom_engine->bb; - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VPUSH */ /* cond 1101 0D10 1101 Vd-- 101X imm8 imm8 */ @@ -2873,63 +1465,6 @@ VPUSH_INST: GOTO_NEXT_INST; } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vpush), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vpush)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - *tag |= TAG_NEW_BB; - if(instr >> 28 != 0xe) - *tag |= TAG_CONDITIONAL; - - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vpush)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - int single = BIT(8) == 0; - int d = (single ? BITS(12, 15)<<1|BIT(22) : BITS(12, 15)|(BIT(22)<<4)); - int imm32 = BITS(0, 7)<<2; - int regs = (single ? BITS(0, 7) : BITS(1, 7)); - - Value* Addr = SUB(R(13), CONST(imm32)); - //if(single) - // bb = arch_check_mm(cpu, bb, Addr, regs * 4, 0, cpu->dyncom_engine->bb_trap); - //else - // bb = arch_check_mm(cpu, bb, Addr, regs * 8, 0, cpu->dyncom_engine->bb_trap); - //Value* phys_addr; - - for (int i = 0; i < regs; i++) - { - if (single) - { - //Memory::Write32(addr, cpu->ExtReg[inst_cream->d+i]); - //memory_write(cpu, bb, Addr, RSPR(d + i), 32); - memory_write(cpu, bb, Addr, IBITCAST32(FR32(d + i)), 32); - bb = cpu->dyncom_engine->bb; - Addr = ADD(Addr, CONST(4)); - } - else - { - /* Careful of endianness, little by default */ - //memory_write(cpu, bb, Addr, RSPR((d + i) * 2), 32); - memory_write(cpu, bb, Addr, IBITCAST32(FR32((d + i) * 2)), 32); - bb = cpu->dyncom_engine->bb; - //memory_write(cpu, bb, ADD(Addr, CONST(4)), RSPR((d + i) * 2 + 1), 32); - memory_write(cpu, bb, ADD(Addr, CONST(4)), IBITCAST32(FR32((d + i) * 2 + 1)), 32); - bb = cpu->dyncom_engine->bb; - - Addr = ADD(Addr, CONST(8)); - } - } - LET(13, SUB(R(13), CONST(imm32))); - - return No_exp; -} -#endif /* ----------------------------------------------------------------------- */ /* VSTM */ @@ -3004,76 +1539,6 @@ VSTM_INST: /* encoding 1 */ } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vstm), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vstm)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - *tag |= TAG_NEW_BB; - if(instr >> 28 != 0xe) - *tag |= TAG_CONDITIONAL; - - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vstm)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - //arch_arm_undef(cpu, bb, instr); - int single = BIT(8) == 0; - int add = BIT(23); - int wback = BIT(21); - int d = single ? BITS(12, 15)<<1|BIT(22) : BITS(12, 15)|(BIT(22)<<4); - int n = BITS(16, 19); - int imm32 = BITS(0, 7)<<2; - int regs = single ? BITS(0, 7) : BITS(1, 7); - - Value* Addr = SELECT(CONST1(add), R(n), SUB(R(n), CONST(imm32))); - //if(single) - // bb = arch_check_mm(cpu, bb, Addr, regs * 4, 0, cpu->dyncom_engine->bb_trap); - //else - // bb = arch_check_mm(cpu, bb, Addr, regs * 8, 0, cpu->dyncom_engine->bb_trap); - - int i; - Value* phys_addr; - for (i = 0; i < regs; i++) - { - if (single) - { - //Memory::Write32(addr, cpu->ExtReg[inst_cream->d+i]); - /* if R(i) is R15? */ - //memory_write(cpu, bb, Addr, RSPR(d + i), 32); - memory_write(cpu, bb, Addr, IBITCAST32(FR32(d + i)),32); - bb = cpu->dyncom_engine->bb; - Addr = ADD(Addr, CONST(4)); - } - else - { - //Memory::Write32(addr, cpu->ExtReg[(inst_cream->d+i)*2]); - //memory_write(cpu, bb, Addr, RSPR((d + i) * 2), 32); - memory_write(cpu, bb, Addr, IBITCAST32(FR32((d + i) * 2)),32); - bb = cpu->dyncom_engine->bb; - - //Memory::Write32(addr + 4, cpu->ExtReg[(inst_cream->d+i)*2 + 1]); - //memory_write(cpu, bb, ADD(Addr, CONST(4)), RSPR((d + i) * 2 + 1), 32); - memory_write(cpu, bb, ADD(Addr, CONST(4)), IBITCAST32(FR32((d + i) * 2 + 1)), 32); - bb = cpu->dyncom_engine->bb; - //addr += 8; - Addr = ADD(Addr, CONST(8)); - } - } - if (wback){ - //cpu->Reg[n] = (add ? cpu->Reg[n] + imm32 : - // cpu->Reg[n] - imm32); - LET(n, SELECT(CONST1(add), ADD(R(n), CONST(imm32)), SUB(R(n), CONST(imm32)))); - } - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VPOP */ /* cond 1100 1D11 1101 Vd-- 101X imm8 imm8 */ @@ -3142,70 +1607,6 @@ VPOP_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vpop), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vpop)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - /* Should check if PC is destination register */ - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - *tag |= TAG_NEW_BB; - if(instr >> 28 != 0xe) - *tag |= TAG_CONDITIONAL; - - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vpop)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - //arch_arm_undef(cpu, bb, instr); - int single = BIT(8) == 0; - int d = (single ? BITS(12, 15)<<1|BIT(22) : BITS(12, 15)|(BIT(22)<<4)); - int imm32 = BITS(0, 7)<<2; - int regs = (single ? BITS(0, 7) : BITS(1, 7)); - - int i; - unsigned int value1, value2; - - Value* Addr = R(13); - Value* val; - //if(single) - // bb = arch_check_mm(cpu, bb, Addr, regs * 4, 1, cpu->dyncom_engine->bb_trap); - //else - // bb = arch_check_mm(cpu, bb, Addr, regs * 4, 1, cpu->dyncom_engine->bb_trap); - //Value* phys_addr; - for (i = 0; i < regs; i++) - { - if (single) - { - memory_read(cpu, bb, Addr, 0, 32); - bb = cpu->dyncom_engine->bb; - val = new LoadInst(cpu->dyncom_engine->read_value, "", false, bb); - LETFPS(d + i, FPBITCAST32(val)); - Addr = ADD(Addr, CONST(4)); - } - else - { - /* Careful of endianness, little by default */ - memory_read(cpu, bb, Addr, 0, 32); - bb = cpu->dyncom_engine->bb; - val = new LoadInst(cpu->dyncom_engine->read_value, "", false, bb); - LETFPS((d + i) * 2, FPBITCAST32(val)); - memory_read(cpu, bb, ADD(Addr, CONST(4)), 0, 32); - bb = cpu->dyncom_engine->bb; - val = new LoadInst(cpu->dyncom_engine->read_value, "", false, bb); - LETFPS((d + i) * 2 + 1, FPBITCAST32(val)); - - Addr = ADD(Addr, CONST(8)); - } - } - LET(13, ADD(R(13), CONST(imm32))); - return No_exp; -} -#endif /* ----------------------------------------------------------------------- */ /* VLDR */ @@ -3271,67 +1672,6 @@ VLDR_INST: } #endif -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vldr), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vldr)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - /* Should check if PC is destination register */ - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - *tag |= TAG_NEW_BB; - if(instr >> 28 != 0xe) - *tag |= TAG_CONDITIONAL; - - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vldr)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - int single = BIT(8) == 0; - int add = BIT(23); - int wback = BIT(21); - int d = (single ? BITS(12, 15)<<1|BIT(22) : BITS(12, 15)|(BIT(22)<<4)); - int n = BITS(16, 19); - int imm32 = BITS(0, 7)<<2; - int regs = (single ? BITS(0, 7) : BITS(1, 7)); - Value* base = R(n); - if(n == 15){ - base = ADD(AND(base, CONST(0xFFFFFFFC)), CONST(8)); - } - Value* Addr = add ? (ADD(base, CONST(imm32))) : (SUB(base, CONST(imm32))); - //if(single) - // bb = arch_check_mm(cpu, bb, Addr, 4, 1, cpu->dyncom_engine->bb_trap); - //else - // bb = arch_check_mm(cpu, bb, Addr, 8, 1, cpu->dyncom_engine->bb_trap); - //Value* phys_addr; - Value* val; - if(single){ - memory_read(cpu, bb, Addr, 0, 32); - bb = cpu->dyncom_engine->bb; - val = new LoadInst(cpu->dyncom_engine->read_value, "", false, bb); - //LETS(d, val); - LETFPS(d,FPBITCAST32(val)); - } - else{ - memory_read(cpu, bb, Addr, 0, 32); - bb = cpu->dyncom_engine->bb; - val = new LoadInst(cpu->dyncom_engine->read_value, "", false, bb); - //LETS(d * 2, val); - LETFPS(d * 2,FPBITCAST32(val)); - memory_read(cpu, bb, ADD(Addr, CONST(4)), 0,32); - bb = cpu->dyncom_engine->bb; - val = new LoadInst(cpu->dyncom_engine->read_value, "", false, bb); - //LETS(d * 2 + 1, val); - LETFPS( d * 2 + 1,FPBITCAST32(val)); - } - - return No_exp; -} -#endif - /* ----------------------------------------------------------------------- */ /* VLDM */ /* cond 110P UDW1 Rn-- Vd-- 101X imm8 imm8 */ @@ -3403,76 +1743,3 @@ VLDM_INST: GOTO_NEXT_INST; } #endif - -#ifdef VFP_DYNCOM_TABLE -DYNCOM_FILL_ACTION(vldm), -#endif -#ifdef VFP_DYNCOM_TAG -int DYNCOM_TAG(vldm)(cpu_t *cpu, addr_t pc, uint32_t instr, tag_t *tag, addr_t *new_pc, addr_t *next_pc) -{ - int instr_size = INSTR_SIZE; - //arm_tag_trap(cpu, pc, instr, tag, new_pc, next_pc); - arm_tag_continue(cpu, pc, instr, tag, new_pc, next_pc); - *tag |= TAG_NEW_BB; - if(instr >> 28 != 0xe) - *tag |= TAG_CONDITIONAL; - - return instr_size; -} -#endif -#ifdef VFP_DYNCOM_TRANS -int DYNCOM_TRANS(vldm)(cpu_t *cpu, uint32_t instr, BasicBlock *bb, addr_t pc){ - int single = BIT(8) == 0; - int add = BIT(23); - int wback = BIT(21); - int d = single ? BITS(12, 15)<<1|BIT(22) : BITS(12, 15)|BIT(22)<<4; - int n = BITS(16, 19); - int imm32 = BITS(0, 7)<<2; - int regs = single ? BITS(0, 7) : BITS(1, 7); - - Value* Addr = SELECT(CONST1(add), R(n), SUB(R(n), CONST(imm32))); - //if(single) - // bb = arch_check_mm(cpu, bb, Addr, regs * 4, 1, cpu->dyncom_engine->bb_trap); - //else - // bb = arch_check_mm(cpu, bb, Addr, regs * 4, 1, cpu->dyncom_engine->bb_trap); - - int i; - //Value* phys_addr; - Value* val; - for (i = 0; i < regs; i++) - { - if (single) - { - //Memory::Write32(addr, cpu->ExtReg[inst_cream->d+i]); - /* if R(i) is R15? */ - memory_read(cpu, bb, Addr, 0, 32); - bb = cpu->dyncom_engine->bb; - val = new LoadInst(cpu->dyncom_engine->read_value, "", false, bb); - //LETS(d + i, val); - LETFPS(d + i, FPBITCAST32(val)); - Addr = ADD(Addr, CONST(4)); - } - else - { - memory_read(cpu, bb, Addr, 0, 32); - bb = cpu->dyncom_engine->bb; - val = new LoadInst(cpu->dyncom_engine->read_value, "", false, bb); - LETFPS((d + i) * 2, FPBITCAST32(val)); - memory_read(cpu, bb, Addr, 0, 32); - bb = cpu->dyncom_engine->bb; - val = new LoadInst(cpu->dyncom_engine->read_value, "", false, bb); - LETFPS((d + i) * 2 + 1, FPBITCAST32(val)); - - //Memory::Write(addr + 4, phys_addr, cpu->ExtReg[(inst_cream->d+i)*2 + 1], 32); - //addr += 8; - Addr = ADD(Addr, CONST(8)); - } - } - if (wback){ - //cpu->Reg[n] = (add ? cpu->Reg[n] + imm32 : - // cpu->Reg[n] - imm32); - LET(n, SELECT(CONST1(add), ADD(R(n), CONST(imm32)), SUB(R(n), CONST(imm32)))); - } - return No_exp; -} -#endif diff --git a/src/core/arm/skyeye_common/vfp/vfpsingle.cpp b/src/core/arm/skyeye_common/vfp/vfpsingle.cpp index b7872bdc4..8b2dfa388 100644 --- a/src/core/arm/skyeye_common/vfp/vfpsingle.cpp +++ b/src/core/arm/skyeye_common/vfp/vfpsingle.cpp @@ -419,7 +419,7 @@ static u32 vfp_compare(ARMul_State* state, int sd, int signal_on_qnan, s32 m, u3 d = vfp_get_float(state, sd); if (vfp_single_packed_exponent(m) == 255 && vfp_single_packed_mantissa(m)) { - ret |= FPSCR_C | FPSCR_V; + ret |= FPSCR_CFLAG | FPSCR_VFLAG; if (signal_on_qnan || !(vfp_single_packed_mantissa(m) & (1 << (VFP_SINGLE_MANTISSA_BITS - 1)))) /* * Signalling NaN, or signalling on quiet NaN @@ -428,7 +428,7 @@ static u32 vfp_compare(ARMul_State* state, int sd, int signal_on_qnan, s32 m, u3 } if (vfp_single_packed_exponent(d) == 255 && vfp_single_packed_mantissa(d)) { - ret |= FPSCR_C | FPSCR_V; + ret |= FPSCR_CFLAG | FPSCR_VFLAG; if (signal_on_qnan || !(vfp_single_packed_mantissa(d) & (1 << (VFP_SINGLE_MANTISSA_BITS - 1)))) /* * Signalling NaN, or signalling on quiet NaN @@ -441,7 +441,7 @@ static u32 vfp_compare(ARMul_State* state, int sd, int signal_on_qnan, s32 m, u3 /* * equal */ - ret |= FPSCR_Z | FPSCR_C; + ret |= FPSCR_ZFLAG | FPSCR_CFLAG; } else if (vfp_single_packed_sign(d ^ m)) { /* * different signs @@ -450,22 +450,22 @@ static u32 vfp_compare(ARMul_State* state, int sd, int signal_on_qnan, s32 m, u3 /* * d is negative, so d < m */ - ret |= FPSCR_N; + ret |= FPSCR_NFLAG; else /* * d is positive, so d > m */ - ret |= FPSCR_C; + ret |= FPSCR_CFLAG; } else if ((vfp_single_packed_sign(d) != 0) ^ (d < m)) { /* * d < m */ - ret |= FPSCR_N; + ret |= FPSCR_NFLAG; } else if ((vfp_single_packed_sign(d) != 0) ^ (d > m)) { /* * d > m */ - ret |= FPSCR_C; + ret |= FPSCR_CFLAG; } } return ret; @@ -491,46 +491,6 @@ static u32 vfp_single_fcmpez(ARMul_State* state, int sd, int unused, s32 m, u32 return vfp_compare(state, sd, 1, 0, fpscr); } -static s64 vfp_single_to_doubleintern(ARMul_State* state, s32 m, u32 fpscr) //ichfly for internal use only -{ - struct vfp_single vsm; - struct vfp_double vdd; - int tm; - u32 exceptions = 0; - - vfp_single_unpack(&vsm, m); - - tm = vfp_single_type(&vsm); - - /* - * If we have a signalling NaN, signal invalid operation. - */ - if (tm == VFP_SNAN) - exceptions = FPSCR_IOC; - - if (tm & VFP_DENORMAL) - vfp_single_normalise_denormal(&vsm); - - vdd.sign = vsm.sign; - vdd.significand = (u64)vsm.significand << 32; - - /* - * If we have an infinity or NaN, the exponent must be 2047. - */ - if (tm & (VFP_INFINITY | VFP_NAN)) { - vdd.exponent = 2047; - if (tm == VFP_QNAN) - vdd.significand |= VFP_DOUBLE_SIGNIFICAND_QNAN; - goto pack_nan; - } else if (tm & VFP_ZERO) - vdd.exponent = 0; - else - vdd.exponent = vsm.exponent + (1023 - 127); -pack_nan: - vfp_double_normaliseroundintern(state, &vdd, fpscr, exceptions, "fcvtd"); - return vfp_double_pack(&vdd); -} - static u32 vfp_single_fcvtd(ARMul_State* state, int dd, int unused, s32 m, u32 fpscr) { struct vfp_single vsm; @@ -981,6 +941,9 @@ vfp_single_multiply_accumulate(ARMul_State* state, int sd, int sn, s32 m, u32 fp v = vfp_get_float(state, sd); pr_debug("VFP: s%u = %08x\n", sd, v); vfp_single_unpack(&vsn, v); + if (vsn.exponent == 0 && vsn.significand != 0) + vfp_single_normalise_denormal(&vsn); + if (negate & NEG_SUBTRACT) vsn.sign = vfp_sign_negate(vsn.sign); |