TMS320C6000产品是美国
TI公司于1997年推出的dsp芯片,该DSP芯片定点、浮点兼容,其中,定点系列是TMS320C62xx系列,浮点系列是TMS320C67xx系列,2000年3月,TI发布新的C64xx内核,主频为1.1GHz,处理速度9000MIPS,在图像处理和流媒体领域得到了广泛的应用。
C6000片内有8个并行的处理单元,分为相同的两组。DSP的体系结构采用超长指令字(vliw)结构,单指令字长为32位,指令包里有8条指令,总字长达到256位。执行指令的功能单元已经在编译时分配好,程序运行时通过专门的指令分配模块,可以将每个256为的指令包同时分配到8个处理单元,并有8个单元同时运行。芯片最高
时钟频率为300MHz(67xx系列),且内部8个处理单元并行运行时,其最大处理能力可达到1600MIPS。
DSP320C6000系列
DSP320C6000的指令列表汇集
内联指令 汇编指令 简要描述
int _abs (int src);
int _labs (__int40_t src)ABS返回src的绝对值
int _add2 (int src1, int src2)ADD2把src1的高、低16位和src2的高、低16位分别相加,放入结果的高、低16位
ushort & _amem2 (void *ptr);LDHU
STHU从内存中加载一个halfword到dst里,必须2by
te对齐(读或存)
const ushort & _amem2_const (const void *ptr);LDHU必须2byte对齐(读)
unsigned & _amem4 (void *ptr);LDW
STW必须4byte对齐(读或存)
const unsigned & _amem4_const (const void *ptr);LDW必须4byte对齐(读)
double & _amemd8 (void *ptr);LDW/LDW
STW/STW必须8byte对齐(读或存)
const double & _amemd8_const (const void *ptr);LDDW必须8byte对齐(读)
unsigned _clr (unsigned src2, unsigned csta,unsigned cstb);CLR指定了从需要清0的首位和末位
unsigned _clrr (unsigned src2, int src1);CLR将src2中指定位清0,清0的首位和末位由src1的低10位指定
__int40_t _dtol (double src); 将一个double
寄存器重新解释成一个__int40_t
long long _dtoll (double src); 将一个double寄存器重新解释成一个long long
int _ext (int src2, unsigned csta, unsigned cstb);EXT从src2里提取csta和cstb指定的区域且符号扩展到32位。提取出的区域先符号左移再右移。
int _extr (int src2, int src1);EXT同上,区别:左右移的位数由src1的低10位指定
unsigned _extu (unsigned src2, unsigned csta , unsigned cstb);EXTU同上上,区别最后是0扩展到32位。
unsigned _extur (unsigned src2, intsrc1);EXTU同上,区别:左右移的位数由src1的低10位指定例:
_ftoi (1.0) == 1065353216U
unsigned _ftoi (float src); 将float的比特位解释成unsigned
unsigned _hi (double src); 返回double寄存器的高位(奇数位)
unsigned _hill (long long src); 返回longlong寄存器的高位(奇数位)
double _itod (unsigned src2, unsignedsrc1); 创建一个新的double寄存器为了解释2个unsigned的值,其中src2是高(奇数)寄存器,src1是低(偶数)寄存器
float _itof (unsigned src); 将unsigned中的比特位解释成float例:
_itof (0x3f800000) = 1.0
long long _itoll (unsigned src2, unsignedsrc1); 创建一个新的longlong寄存器为了解释2个unsigned的值,其中src2是高(奇数)寄存器,src1是低(偶数)
unsigned _lmbd (unsigned src1, unsignedsrc2);LMBD搜索src2里面的1或0,1或0是由src1的LSB决定的,返回比特位变化的位数
unsigned _lo (double src); 返回double寄存器的低(奇数)寄存器
unsigned _loll (long long src); 返回longlong寄存器的低(奇数)寄存器
double _ltod (__int40_t src); 把一个__int40_t寄存器解释成一个double寄存器
double _lltod (long long src); 把一个longlong寄存器解释成一个double寄存器
int _mpy (int src1, int src2);MPYSrc1和src2相乘,操作数默认为有符号的
int _mpyus (unsigned src1, int src2);MPYUS无符号src1和有符号src2相乘,S是用来那个是有符号的操作数,当两个操作数都是有符号的或者无符号的
int _mpysu (int src1, unsigned src2);MPYSU同上
unsigned _mpyu (unsigned src1,unsigned src2);MPYU同上上上,默认为无符号
int _mpyh (int src1, int src2);MPYH同上,区别见图示
int _mpyhus (unsigned src1, int src2);MPYHUS
int _mpyhsu (int src1, unsigned src2);MPYHSU
unsigned _mpyhu (unsigned src1,unsigned src2);MPYHU
int _mpyhl (int src1, int src2);MPYHL同上,区别见图示
int _mpyhuls (unsigned src1, int src2);MPYHULS
int _mpyhslu (int src1, unsigned src2);MPYHSLU
unsigned _mpyhlu (unsigned src1,unsigned src2);MPYHLU
int _mpylh (int src1, int src2);MPYLH
int _mpyluhs (unsigned src1, int src2);MPYLUHS
int _mpylshu (int src1, unsigned src2);MPYLSHU
unsigned _mpylhu (unsigned src1,unsigned src2);MPYLHU
void _
nassert (int src); 不生成代码,告诉优化器一些事情
unsigned _norm (int src);
unsigned _lnorm (__int40_t src);NORM
返回src2的冗余的符号比特位的个数,具体见图示
int _sadd (int src1, int src2);
long _lsadd (int src1, __int40_t src2);SADD将src1和src2相加,且饱和其结果
int _sat (__int40_t src2);SAT将一个40比特的long转换为一个32比特的有符号int,如有需要,对结果进行饱和
unsigned _set (unsigned src2, unsignedcsta , unsigned cstb);SET将src2中指定的区域置位1,指定的区域由csta和cstb指定
unsigned _setr (unit src2, int src1);SET
int _smpy (int src1, int src2);SMPY把src1的低16位和src2的低16位相乘
int _smpyh (int src1, int src2);SMPYH高16位
int _smpyhl (int src1, int src2);SMPYHL
int _smpylh (int src1, int src2);SMPYLH
int _sshl (int src2, unsigned src1);SSHL以src1操作数将src2左移,并且将结果饱和在32位
int _ssub (int src1, int src2);
__int40_t _lssub (int src1, __int40_tsrc2);SSUB从src1中减去src2,并饱和结果(src1-src2)
unsigned _subc (unsigned src1, unsignedsrc2);SUBC有条件的减和左移(常用于除法)
int _sub2 (int src1, int src2);SUB2把src1的高低16位分别减去src2的高低16位。任何低16位的借位不会影响高16位。
int _abs2 (int src);ABS2计算16位的绝对值
int _add4 (int src1, int src2);ADD4把src1和src2的4对8位数相加。不会进行饱和,进位不会影响其他的8位数
long long & _amem8 (void *ptr);LDDW
STDW加载和存储8bytes,指针必须8byte对齐
const long long & _amem8_const (const void *ptr);LDDW加载8bytes,指针必须8byte对齐
__float2_t & _amem8_f2(void * ptr);LDDW
STDW加载和存储8bytes,指针必须8byte对齐,必须包含c6x.h
const __float2_t & _amem8_f2_const(void * ptr);LDDW加载8bytes,指针必须8byte对齐,必须包含c6x.h
double & _amemd8 (void *ptr);LDDW
STDW
const double & _amemd8_const (const void *ptr);LDDW
int _avg2 (int src1, int src2);AVG2计算每对有符号16位置的平均值
unsigned _avgu4 (unsigned, unsigned);AVGU4计算每对有符号8位数的平均值
unsigned _bitc4 (unsigned src);BITC4统计每个8位的比特位是1的个数,写入结果对应位置
unsigned _bitr (unsigned src);BITR翻转比特位的顺序
int _cmpeq2 (int src1, int src2);CMPEQ2比较每16位的值是否相等,结果放入dst的最低2位
int _cmpeq4 (int src1, int src2);CMPEQ4比较每8位的值是否相等,结果放入dst的最低4位,相等置1,否则为0
int _cmpgt2 (int src1, int src2);CMPGT2每16位有符号比较,src1》src2,置为1;否则置为0。结果放入dst的最低2位
unsigned _cmpgtu4 (unsigned src1,unsigned src2);CMPGTU4每8位无符号比较,src1》src2,置为1;否则置为0。结果放入dst的最低4位
unsigned _deal (unsigned src );DEAL将src中的比特位的奇数位和偶数位抽出来进行重组,偶数位放在低的16位,奇数位放在高的16位
int _dotp2 (int src1, int src2);
__int40_t _
ldotp2 (int src1, int src2);DOTP2
DOTP2将src1中的和src2中的16位有符号对进行点积,结果被写成有符号32位int或者符号扩展为64位
int _dotpn2 (int src1, int src2);DOTPN2将src1和src2中的16位有符号数进行点积相减
int _dotpn
rsu2 (int src1, unsigned src2);DOTPNRSU2Src1和src2的高16位的点积减去低16位的点积。Src1中的数被当做有符号,src2中的数被当做无符号,再加上2^15,结果再符号右移16位
int _dotprsu2 (int src1, unsigned src2);DOTPRSU2Src1和src2的高16位的点积加上低16位的点积。Src1中的数被当做有符号,src2中的数被当做无符号,再加上2^15,结果再符号右移16位
int _dotpsu4 (int src1, unsigned src2);DOTPSU4将src1和src2的每8位进行相乘再求和,src1的每8位数被当做有符号,src2的每8位数被当做无符号
unsigned _dotpu4 (unsigned src1,unsigned src2);DOTPU4都被当做无符号的
int _gmpy4 (int src1, int src2);GMPY4将src1和src2的4个无符号进行伽罗瓦域的乘法
int _max2 (int src1, int src2);MAX2将src1和src2的2个有符号16位整数比较,取较大值
int _min2 (int src1, int src2);MIN2将src1和src2的2个有符号16位整数比较,取较小值
unsigned _maxu4 (unsigned src1,unsigned src2);MAXU4将src1和src2的4个无符号8位整数比较,取较大值
unsigned _minu4 (unsigned src1,unsigned src2);MINU4将src1和src2的4个无符号8位整数比较,取较小值
ushort & _mem2 (void * ptr);LDB/LDB
STB/STB加载和存储2byte,不需要对齐
const ushort & _mem2_const (const void * ptr);LDB/LDB加载2byte,不需要对齐
unsigned & _mem4 (void * ptr);LDNW
STNW加载和存储4byte,不需要对齐
const unsigned & _mem4_const (const void * ptr);LDNW加载4byte,不需要对齐
long long & _mem8 (void * ptr);LDNDW
STNDW加载和存储8byte,不需要对齐
const long long & _mem8_const (const void * ptr);LDNDW加载8byte,不需要对齐
double & _memd8 (void * ptr);LDNDW
STNDW加载和存储8byte,不需要对齐
const double & _memd8_const (const void * ptr);LDNDW加载8byte,不需要对齐
long long _mpy2ll (int src1, int src2);MPY2将src1和src2中的2个有符号16位分别相乘,将2个32位的结果写入longlong中
long long _mpyhill (int src1, int src2);MPYHI将src1中高16位作为1个有符号16位乘以src2的有符号32位,结果写入longlong的低48位
long long _mpylill (int src1, int src2);MPYLI将src1中低16位作为1个有符号16位乘以src2的有符号32位,结果写入longlong的低48位
int _mpyhir (int src1, int src2);MPYHIR将src1的高16位作为一个16位有符号乘以src2的有符号32位。乘积利用round模式通过加2^14转成32位,最后再右移15位
int _mpylir (int src1, int src2);MPYLIR将src1的低16位作为一个16位有符号乘以src2的有符号32位。乘积利用round模式通过加2^14转成32位,最后再右移15位
long long _mpysu4ll (int src1, unsignedsrc2);MPYSU4将src1的4个8位有符号乘src2的4个8位无符号,得到4个16位有符号,组成一个64位
long long _mpyu4ll (unsigned src1,unsigned src2);MPYU4将src1和src2的4个无符号8位相乘,得到4个无符号16位组成一个64位的数
int _mvd (int src2 );MVD将src2的数据移入返回值中,利用了乘法流水线(延迟)
unsigned _pack2 (unsigned src1,unsigned src2);PACK2
unsigned _packh2 (unsigned src1,unsigned src2);PACKH2
unsigned _packh4 (unsigned src1,unsigned src2);PACKH4
unsigned _packl4 (unsigned src1,unsigned src2);PACKL4
unsigned _packhl2 (unsigned src1,unsigned src2);PACKHL2
unsigned _packlh2 (unsigned src1,unsigned src2);PACKLH2
unsigned _rotl (unsigned src1, unsignedsrc2);ROTL按照src1的最低5位的数去左移src2的32位,src1中剩下的高的5-31位被忽略
int _sadd2 (int src1, int src2);SADD2将src1和src2中的2个16位有符号数相加,生成2个16有符号数并且是饱和过的。
int _saddus2 (unsigned src1, int src2);SADDUS2将src1中的2个无符号16位数和src中的2个16位有符号数相加,得到2个无符号16位数
unsigned _saddu4 (unsigned src1,unsigned src2);SADDU4将src1和src2中的4个无符号8位数相加
unsigned _shfl (unsigned src2);SHFL将src2的高16和低16位进行交织
unsigned _shlmb (unsigned src1,unsigned src2);SHLMB将src2左移1byte,然后将src1的最高位充入src2左移后多出来的位置
unsigned _shrmb (unsigned src1,unsigned src2);SHRMB将src2右移1byte,然后将src1的最低位充入src2右移后多出来的位置
int _shr2 (int src1, unsigned src2);SHR2将src2的2个16位有符号数分别右移,右移的位数由src1的低5位决定,多出的位置由符号位扩展
unsigned shru2 (unsigned src1, unsignedsrc2);SHRU2将src2的2个16位无符号数分别右移,右移的位数由src1的低5位决定,多出的位置由0扩展
long long _smpy2ll (int src1, int src2);SMPY2将src1和src2中的2个有符号16位数相乘,然后左移1位,再进行饱和。
int _spack2 (int src1, int src2);SPACK2将src1和src2中的1个有符号32位数进行饱和到有符号16位,然后把src1的饱和结果放入dst的高16位,src2的饱和结果放入dst的低16位
unsigned _spacku4 (int src1 , int src2);SPACKU4将src1和src2中的4个有符号16位数饱和成无符号8位数,
int _sshvl (int src2, int src1);SSHVL将src2中的有符号32位数左移或右移,移位的数量由src1指定的比特数确定。
src1在[-31,31]之间,如果src1为正,src2则左移;如果src1为负,src2右移|src1|且符号位扩展
int _sshvr (int src2, int src1);SSHVR将src2中的有符号32位数左移或右移,移位的数量由src1指定的比特数确定。
src1在[-31,31]之间,如果src1为正,src2则右移且是符号扩展;如果src1为负,src2左移|src1|
int _sub4 (int src1, int src2);SUB4将src1和src2中的4个8位数相减,不进行饱和
int _subabs4 (int src1, int src2);SUBABS4将src1和src2中的4个无符号8位相减求绝对值
unsigned _swap4 (unsigned src);SWAP4将src的4个8位无符号数按图示换位置
unsigned _unpkhu4 (unsigned src);UNPKHU4扩展0
unsigned _unpklu4 (unsigned src);UNPKLU4扩0
unsigned _xpnd2 (unsigned src);XPND2按src的最低2位进行扩展,bit1扩展高16位,bit0扩展低16位
unsigned _xpnd4 (unsigned src);XPND4按src的最低4位进行扩展
long long _addsub (int src1, int src2);ADDSUB平行做2步:
1、src2+src1-》dst_o
2、src1-src2-》dst_e
long long _addsub2 (int src1, int src2);ADDSUB216位有符号
ADD2:src2的高、低16位+src1的高、低16位-》dst_o
SUB2: src1的高、低16位-src2的高、低16位-》dst_e
long long _cmpy (unsigned src1,unsigned src2);CMPY有符号16位
Src1和src2的高16位的点积-src1和src2的低16位点积-》dst_o
饱和(src1和src2的高16位的点积+src1和src2的低16位点积)-》dst_e
unsigned _cmpyr (unsigned src1,unsigned src2);CMPYR
unsigned _cmpyr1 (unsigned src1,unsigned src2 );CMPYR1
long long _ddotp4 (unsigned src1,unsigned src2);DDOTP4没有饱和
long long _ddotph2 (long long src1,unsigned src2);DDOTPH2
long long _ddotpl2 (long long src1,unsigned src2);DDOTPL2
unsigned _ddotph2r (long long src1,unsigned src2);DDOTPH2R
unsigned _ddotpl2r (long long src1,unsigned src2);DDOTPL2R
long long _dmv (int src1, int src2);DMV将两个寄存器移入一个寄存器一次性的
long long _dpack2 (unsigned src1,unsigned src2);DPACK2
long long _dpackx2 (unsigned src1,unsigned src2);DPACKX2
__float2_t _fmdv_f2(float src1, floatsrc2)DMV
unsigned _gmpy (unsigned src1,unsigned src2);GMPY伽罗瓦域上的乘法
long long _mpy2ir (int src1, int src2);MPY2IR进行16位乘32位。
将src1的高16位和低16位当做有符号16位;将src2的值当做有符号32位。
乘积通过加上2^14round到32位,然后结果右移15位。
2个结果的低32位写入dst_o:dst_e
int _mpy32 (int src1, int src2);MPY32进行32位乘32位。都是有符号的,64位结果中的低32位写入dst
long long _mpy32ll (int src1, int src2);MPY3232位有符号数×32位有符号数,有符号的64位结果被写入dst
long long _mpy32su (int src1, int src2);MPY32SUsrc1有符号32位×src2无符号32位=dst有符号64位
long long _mpy32us (unsigned src1, intsrc2);MPY32USsrc1无符号32位×src2有符号32位=dst有符号64位
long long _mpy32u (unsigned src1,unsigned src2);MPY32Usrc1无符号32位×src2无符号32位=dst无符号64位
int _rpack2 (int src1, int src2);RPACK2
long long _saddsub (unsigned src1,unsigned src2);SADDSUB并行进行:
1、饱和(src1+src2)-》dst_o
2、饱和(src1-src2)-》dst_e
long long _saddsub2 (unsigned src1,unsigned src2);SADDSUB2并行进行SADD2和SSUB2指令
long long _shfl3 (unsigned src1, unsignedsrc2);SHFL3如图,生成一个longlong
int _smpy32 (int src1, int src2);SMPY3232位有符号×32位有符号,64位的结果左移1位然后饱和,然后将之后的结果的高32位写入dst
int _ssub2 (unsigned src1, unsignedsrc2);SSUB2Src1中的2个16位有符号-src2中的2个有符号16位,结果进行饱和
unsigned _xormpy (unsigned src1,unsigned src2);XORMPY加瓦罗域乘法
int _d
pint (double src);DPINT将double转成int(round)
__int40_t _f2tol(__float2_t src); 将一个__float2_t解释成一个__int40
__float2_t _f2toll(__float2_t src); 将一个__float2_t解释成一个longlong
double _fabs (double src);ABSDP将src的绝对值放入dst。
float _fabsf (float src);ABSSP
__float2_t _lltof2(long long src); 将一个longlong解释成一个__float2_t
__float2_t _ltof2(__int40_t src); 将一个__int40解释成一个__float2_t
__float2_t & _mem8_f2(void * ptr);LDNDW
STNDW从内存里加载一个64位值
const __float2_t & _mem8_f2_const(void * ptr);LDNDW
STNDW
long long _mpyidll (int src1, int src2);MPYIDSrc1×src2-》dst
double_mpysp2dp (float src1, float src2);MPYSP2DPSrc1×src2-》dst
double_mpyspdp (float src1, doublesrc2);MPYSPDPSrc1×src2-》dst
double _rcpdp (double src);RCPDP64位double倒数近似值放入dst
float _rcpsp (float src);RCPSP32位float的倒数近似值
double _rsqrdp (double src);RSQRDP64位double的平方根倒数近似值
float _rsqrsp (float src);RSQRSP32位float的平方根倒数近似值
int _spint (float);SPINTFloat转为int
ADDDP2个double相加
ADDSP2个float相加
AND位与
ANDN与后取反
MPYSP2个float相乘
OR位或
SUBDP2个double相减
SUBSP2和float相减
XOR异或
__x128_t _ccmatmpy (long long src1,__x128_t src2);CMATMPY
long long _ccmatmpyr1 (long long src1,__x128_t src2);CCMATMPYR1
long long _ccmpy32r1 (long long src1,long long src2);CCMPY32R1
__x128_t _cmatmpy (long long src1,__x128_t src2);CMATMPY
long long _cmatmpyr1 (long long src1,__x128_t src2);CMATMPYR1
long long _cmpy32r1 (long long src1,long long src2);CMPY32R1
__x128_t _cmpysp (__float2_t src1,__float2_t src2);CMPYSP
double _complex_conjugate_mpysp (double src1, double src2);CMPYSP
DSUBSP
double _complex_mpysp (double src1,double src2);CMPYSP
DADDSP
int _crot90 (int src);CROT90复数的90度旋转
int _crot270 (int src);CROT270复数的270度旋转
long long _dadd (long long src1, long longsrc2);DADDSrc1的2个32位有符号数+src2的2个32位有符号数
long long _dadd2 (long long src1, long long src2);DADD24路有符号16位相加
__float2_t _daddsp (__float2_t src1,__float2_t src2);DADDSP
long long _dadd_c (scst5 immediate src1,long long src2);DADD2路float加法
long long _dapys2 (long long src1, long long src2);DAPYS2
long long _davg2 (long long src1, long long src2);DAVG2有符号16位
long long _davgnr2 (long long src1, long long src2);DAVGNR2有符号16位,无round模式
long long _davgnru4 (long long src1,long long src2);DAVGNRU4无符号8位,无round模式
long long _davgu4 (long long src1, long long src2);DAVGU4无符号8位
long long _dccmpyr1 (long long src1,long long src2);DCCMPYR1
unsigned _dcmpeq2 (long long src1, long long src2);DCMPEQ216位比较,相等返回1,不等返回0
unsigned _dcmpeq4 (long long src1, long long src2);DCMPEQ48位比较,相等返回1,不等返回0
unsigned _dcmpgt2 (long long src1, long long src2);DCMPGT216位比较,src1》src-》1,否则返回0
unsigned _dcmpgtu4 (long long src1,long long src2);DCMPGTU48位比较,src1》src-》1,否则返回0
__x128_t _dccmpy (long long src1, long long src2);DCCMPY
__x128_t _dcmpy (long long src1, long long src2);DCMPY
long long _dcmpyr1 (long long src1, long long src2);DCMPYR1
long long _dcrot90 (long long src);DCROT90
long long _dcrot270 (long long src);DCROT270
long long _ddotp4h (__x128_t src1,__x128_t src2 );DDOTP4H执行2个dotp4h,都是有符号的
long long _ddotpsu4h (__x128_t src1,__x128_t src2 );DDOTPSU4H执行2个dotpsu4h,一个有符号,一个无符号
__float2_t _dinthsp (int src);DINTHSPSrc中的16位有符号数转成单精度浮点放入dst_e和dst_o中
__float2_t _dinthspu (unsigned src);DINTHSPUSrc中的16位无符号数转成单精度浮点放入dst_e和dst_o中
__float2_t _dintsp(long long src);DINTSPSrc中的有符号32位转成单精度浮点,放入dst_e和dst_o中
__float2_t _dintspu(long long src);DINTSPUSrc中的无符号32位转成单精度浮点,放入dst_e和dst_o中
long long _
dmax2 (long long src1, long long src2);DMAX2对src1和src2中的16位有符号数比大小,将大的放入dst中
long long _dmaxu4 (long long src1, long long src2);DMAXU4对src1和src2中的8位有符号数比大小,将大的放入dst中
long long _dmin2 (long long src1, long long src2);DMIN2对src1和src2中的16位有符号数比大小,将小的放入dst中
long long _dminu4 (long long src1, long long src2);DMINU4对src1和src2中的8位有符号数比大小,将小的放入dst中
__x128_t _dmpy2 (long long src1, long long src2);DMPY2将src1和src2中的16位有符号数相乘,得到32位有符号数放入128位寄存器中
__float2_t _dmpysp (__float2_t src1,__float2_t src2);DMPYSP
__x128_t _dmpysu4 (long long src1,long long src2);DMPYSU4将src1中的8位有符号数乘以src2中的无符号8位,等到有符号16位
__x128_t _dmpyu2 (long long src1, long long src2);DMPYU216位无符号数相乘,得到32位数放入128位寄存器中
__x128_t _dmpyu4 (long long src1, long long src2);DMPYU48位无符号数相乘,得到有符号16位结果
long long _dmvd (long long src1,unsigned src2 );DMVD将2个寄存器移入一个寄存器中。依次进行2次移动,当处理很多的double word时很有用。减轻寄存器压力
int _dotp4h (long long src1, long longsrc2 );DOTP4H进行两个系列的16位值的点积
long long _dotp4hll (long long src1, long long src2 );DOTP4H返回值不同
int _dotpsu4h (long long src1, long longsrc2);DOTPSU4HSrc1中被当做有符号16位,src2被当做无符号16位,得到32位结果
long long _dotspu4hll (long long src1,long long src2);DOTPSU4HSrc1中被当做有符号16位,src2被当做无符号16位,得到64位结果
long long _dpackh2 (long long src1, long long src2);DPACKH2
long long _dpackh4 (long long src1, long long src2);DPACKH4并行执行2个PACKH4
long long _dpacklh2 (long long src1, long long src2);DPACKLH2
long long _dpacklh4 (unsigned src1,unsigned src2);DPACKLH4并行执行PACKH4和PACKL4
long long _dpackl2 (long long src1, long long src2);DPACKL2
long long _dpackl4 (long long src1, long long src2);DPACKL4并行执行2个PACKL4
long long _dsadd (long long src1, long long src2);DSADD将src1中的2个有符号32位数加上src2中的2个有符号32位数,结果进行饱和
long long _dsadd2 (long long src1, long long src2);DSADD2结果饱和到[-2^15 2^15]
long long _dshl (long long src1, unsignedsrc2);DSHL将longlong中的2个32位左移,用0补位(有符号32位)
long long _dshl2 (long long src1,unsigned src2);DSHL2将longlong中的4个16位左移,用0补位(有符号16位)
long long _dshr (long long src1, unsignedsrc2);DSHR右移,符号位补位(有符号32位)
long long _dshr2 (long long src1,unsigned src2);DSHR2右移,符号位补位(有符号16位)
long long _dshru (long long src1,unsigned src2);DSHRU右移,0补位(无符号32位)
long long _dshru2 (long long src1,unsigned src2);DSHRU2右移,0补位(无符号16位)
__x128_t _dsmpy2 (long long src1, long long src2);DSMPY2见图示
long long _dspacku4 (long long src1, long long src2);DSPACKU4并行进行2个SPACK4
long long _dspint (__float2_t src);DSPINT将src中的2个单精度数转成2个整型
unsigned _dspinth (__float2_t src);DSPINTH将src_e和src_o的两个单精度浮点数转陈高个有符号的16位整数
long long _dssub (long long src1, long long src2);DSSUB将src1中的2个32位有符号数减src2中的2个32位有符号数,得到的结果进行饱和[-2^31 (2^31)-1]
long long _dssub2 (long long src1, long long src2);DSSUB24个16位有符号数相减,结果进行饱和[-2^15 (2^15)-1]
long long _dsub (long long src1, long longsrc2);DSUB不饱和
long long _dsub2 (long long src1, long long src2);DSUB2不饱和
__float2_t _dsubsp (__float2_t src1,__float2_t src2);DSUBSP32位单精度数相减
long long _dxpnd2 (unsigned src);DXPND2
long long _dxpnd4 (unsigned src);DXPND4
__float2_t _fdmvd_f2(float src1, floatsrc2);DMVD见MVD
int _land (int src1, int src2);LAND逻辑与
int _landn (int src1, int src2);LANDN
int _lor (int src1, int src2);LOR逻辑或
void _mfence();MFENCE延迟取指令流水线一直到内存系统的busy标志降低
double_mpysp2dp (float src1, float src2);MPYSP2DP将2个float相乘得到1个double结果
double_mpyspdp (float src1, doublesrc2);MPYSPDP1个float×1个double得到1个double
long long _mpyu2 (unsigned src1,unsigned src2 );MPYU22个无符号16位数×2个无符号16位数得到2个无符号32位数
__x128_t _qmpy32 (__x128_t src1,__x128_t src2);QMPY324路:32位有符号×32位有符号,结果的低32位放入dst
__x128_t _qmpysp (__x128_t src1,__x128_t src2);QMPYSP
__x128_t _qsmpy32r1 (__x128_t src1,__x128_t src2);QSMPY32R14路:有符号32位×有符号32位,得到32位。和QMOY32的区别是饱和round
unsigned _shl2 (unsigned src1, unsignedsrc2);SHL22个有符号16位,左移。Src2的低4位是移动的位数。结果也是当做有符号16位
long long _unpkbu4 (unsigned src);UNPKBU4将无符号8位扩成无符号16位
long long _unpkh2 (unsigned src);UNPKH2有符号16位符号扩展
long long _unpkhu2 (unsigned src);UNPKHU2无符号16位进行0扩展
long long _xorll_c (scst5 immediate src1,long long src2);XOR逻辑异或