triton.language.extra.cann.libdevice

Libdevice Ops

abs	Computes the element-wise absolute value of x.
acos	Computes the element-wise arccosine (inverse cosine) of the input tensor.
acosh	Computes the element-wise inverse hyperbolic cosine of the input tensor.
add_rd	Computes x + y rounded down.
add_rn	Computes x + y rounded to nearest even.
add_ru	Computes x + y rounded up.
add_rz	Computes x + y rounded toward zero.
asin	Computes the element-wise arcsine (inverse sine) of the input tensor.
asinh	Computes the element-wise inverse hyperbolic sine of the input tensor.
atan	Computes the element-wise arctangent (inverse tangent) of the input tensor.
atan2	Computes the arctangent of arg0/arg1, using signs to determine the quadrant.
atanh	Computes the element-wise inverse hyperbolic tangent of the input tensor.
brev	Reverses the bit order of a 32-bit integer.
byte_perm	Selects bytes from two 32-bit integers x and y according to the selector s and returns the combined integer.
cbrt	Computes the cube root of x.
ceil	Rounds x up to the nearest integer.
clz	Counts the number of leading zero bits in a 32-bit integer.
copysign	Creates a value with the magnitude of arg0 and the sign of arg1.
cos	Computes the element-wise cosine of x.
cosh	Computes the element-wise hyperbolic cosine of the input tensor.
cospi	Computes cos(pi * x).
cyl_bessel_i0	Computes the modified Bessel function of the first kind, order 0.
cyl_bessel_i1	Computes the modified Bessel function of the first kind of order 1.
div_rd	Computes x / y rounded down.
div_rn	Computes x / y rounded to nearest even.
div_ru	Computes x / y rounded up.
div_rz	Computes the division with round-toward-zero mode.
erf	Computes the error function of x.
erfc	Computes the complementary error function of x.
erfcinv	Computes the inverse complementary error function of x.
erfcx	Computes the scaled complementary error function of x.
erfinv	Computes the inverse error function.
exp	Computes e raised to the power x.
exp10	Computes 10 raised to the power x.
exp2	Computes 2 raised to the power x.
expm1	Computes e^x - 1 with better precision for small x.
fast_cosf	Fast approximation of the cosine of x.
fast_dividef	Computes a fast floating-point division.
fast_exp10f	Fast approximation of 10 raised to the power x.
fast_expf	Computes a fast exponential (e^x) of the input tensor.
fast_log10f	Fast approximation of the base-10 logarithm of x.
fast_log2f	Fast approximation of the base-2 logarithm of x.
fast_logf	Fast approximation of the natural logarithm of x.
fast_powf	Fast approximation of x raised to the power y.
fast_sinf	Fast approximation of the sine of x.
fast_tanf	Fast approximation of the tangent of x.
fdim	Computes the positive difference max(x - y, 0).
ffs	Finds the index of the first (least significant) bit set to 1, in the range [0, 32].
finitef	Determines whether x is a finite value.
float2int_rd	Converts x to int32, rounding down.
float2int_rn	Converts x to int32, rounding to nearest even.
float2int_ru	Converts x to int32, rounding up.
float2int_rz	Converts x to int32, rounding toward zero.
float2ll_rd	Converts x to int64, rounding down.
float2ll_rn	Converts x to int64, rounding to nearest even.
float2ll_ru	Converts x to int64, rounding up.
float2ll_rz	Converts x to int64, rounding toward zero.
float2uint_rd	Converts x to uint32, rounding down.
float2uint_rn	Converts x to uint32, rounding to nearest even.
float2uint_ru	Converts x to uint32, rounding up.
float2uint_rz	Converts x to uint32, rounding toward zero.
float2ull_rd	Converts x to uint64, rounding down.
float2ull_rn	Converts x to uint64, rounding to nearest even.
float2ull_ru	Converts x to uint64, rounding up.
float2ull_rz	Converts x to uint64, rounding toward zero.
float_as_int	Reinterprets the bits of a float32 value as an int32.
float_as_uint	Reinterprets the bits of a float32 value as uint32.
floor	Rounds x down to the nearest integer.
fma	Computes x * y + z with a single rounding.
fma_rd	Computes x * y + z with a single rounding, rounded down.
fma_rn	Computes x * y + z with a single rounding, rounded to nearest even.
fma_ru	Computes x * y + z with a single rounding, rounded up.
fma_rz	Computes x * y + z with a single rounding, rounded toward zero.
fmod	Computes the floating-point remainder of arg0 / arg1.
gamma	Computes the Gamma function using the Lanczos approximation.
hadd	Computes the average of x and y.
hypot	Computes the Euclidean distance: sqrt(arg0^2 + arg1^2).
ilogb	Returns the integer binary exponent of the input tensor.
int2float_rd	Converts an int32 value to float32.
int2float_rn	Converts an int32 value to float32.
int2float_ru	Converts an int32 value to float32.
int2float_rz	Converts an int32 value to float32.
int_as_float	Reinterprets the bits of an int32 value as float32.
isinf	Tests whether each element of the input tensor is infinity.
isnan	Tests whether each element of the input tensor is NaN.
j0	Computes the Bessel function of the first kind of order 0.
j1	Computes the Bessel function of the first kind of order 1.
jn	Computes the Bessel function of the first kind of integer order n.
ldexp	Computes x * 2^exp from a mantissa and an exponent.
lgamma	Computes the natural logarithm of the absolute value of the Gamma function.
ll2float_rd	Converts an int64 value to float32.
ll2float_rn	Converts an int64 value to float32.
ll2float_ru	Converts an int64 value to float32.
ll2float_rz	Converts an int64 value to float32.
llrint	Rounds x to the nearest int64 value.
llround	Rounds x to the nearest int64 value.
log	Computes the natural logarithm of x.
log10	Computes the element-wise base-10 logarithm of the input tensor.
log1p	Computes the element-wise natural logarithm of (1 + x).
log2	Computes the base-2 logarithm of x.
logb	Computes the unbiased exponent of x, i.e. floor(log2(|x|)).
mul24	Computes the low 24-bit multiplication result of x and y.
mul_rd	Computes x * y rounded down.
mul_rn	Computes x * y rounded to nearest even.
mul_ru	Computes x * y rounded up.
mul_rz	Computes x * y rounded toward zero.
mulhi	Computes the high 32 bits of the product x * y.
nearbyint	Rounds the input tensor to the nearest integer using round-to-nearest-even.
nextafter	Returns the next representable floating-point value after arg0 toward arg1.
norm3d	Computes the Euclidean norm of (x, y, z).
norm4d	Computes the Euclidean norm of (x, y, z, w).
normcdf	Computes the standard normal cumulative distribution function.
normcdfinv	Computes the inverse of the standard normal cumulative distribution function.
popc	Counts the number of bits set to 1 in x, in the range [0, 32].
pow	Computes arg0 raised to the power of arg1.
rcbrt	Computes the reciprocal cube root of x.
rcp_rd	Computes 1 / x rounded down.
rcp_rn	Computes 1 / x rounded to nearest even.
rcp_ru	Computes 1 / x rounded up.
rcp_rz	Computes 1 / x rounded toward zero.
reciprocal	Computes the element-wise reciprocal (1/x) of the input tensor.
relu	Computes the element-wise ReLU activation: max(0, x).
remainder	Computes the IEEE remainder of x / y.
rhadd	Computes the rounded average of x and y.
rhypot	Computes 1 / sqrt(x^2 + y^2).
rint	Rounds the input tensor to the nearest integer using round-to-nearest-even.
rnorm3d	Computes 1 / sqrt(x^2 + y^2 + z^2).
rnorm4d	Computes 1 / sqrt(x^2 + y^2 + z^2 + w^2).
round	Rounds the input tensor to the nearest integer.
rsqrt	Computes the reciprocal square root of x.
rsqrt_rn	Computes the reciprocal square root of x, rounded to nearest even.
sad	Computes |x - y| + z, where x and y are signed integers and z is an unsigned integer.
saturatef	Clamps x to the range [0, 1].
scalbn	Computes x * 2^n.
signbit	Returns the sign bit of the input tensor.
sin	Computes the element-wise sine of x.
sinh	Computes the element-wise hyperbolic sine of the input tensor.
sinpi	Computes sin(pi * x).
sqrt	Computes the square root of x.
sqrt_rd	Computes the square root of x, rounded down.
sqrt_rn	Computes the square root of x, rounded to nearest even.
sqrt_ru	Computes the square root of x, rounded up.
sqrt_rz	Computes the square root of x, rounded toward zero.
sub_rd	Computes x - y rounded down.
sub_rn	Computes x - y rounded to nearest even.
sub_ru	Computes x - y rounded up.
sub_rz	Computes x - y rounded toward zero.
tan	Computes the element-wise tangent of the input tensor.
tanh	Computes the element-wise hyperbolic tangent of the input tensor.
tgamma	Computes the gamma function of x.
trunc	Truncates the input tensor to the nearest integer toward zero.
uint2float_rd	Converts a uint32 value to float32.
uint2float_rn	Converts a uint32 value to float32.
uint2float_ru	Converts a uint32 value to float32.
uint2float_rz	Converts a uint32 value to float32.
uint_as_float	Reinterprets the bits of a uint32 value as float32.
ull2float_rd	Converts a uint64 value to float32.
ull2float_rn	Converts a uint64 value to float32.
ull2float_ru	Converts a uint64 value to float32.
ull2float_rz	Converts a uint64 value to float32.
y0	Computes the Bessel function of the second kind of order 0.
y1	Computes the Bessel function of the second kind of order 1.
yn	Computes the Bessel function of the second kind of integer order n.