Functions
template<class R , class T >
R	simdpp::bit_cast (T t)
	Casts between unrelated types. More...

basic_int16x8	simdpp::to_int16x8 (int8x16 a)
	Sign extends the first 8 values of a signed int8x16 vector to 16-bits. More...

basic_int16x16	simdpp::to_int16x16 (int8x32 a)
	Sign extends the first 16 values of a signed int8x32 vector to 16-bits. More...

basic_int16x8	simdpp::to_int16x8 (uint8x16 a)
	Extends the first 8 values of a unsigned int8x16 vector to 16-bits. More...

basic_int16x16	simdpp::to_int16x16 (uint8x32 a)
	Extends the first 16 values of a unsigned int8x32 vector to 16-bits. More...

basic_int32x4	simdpp::to_int32x4 (int16x8 a)
	Sign extends the first 4 values of a signed int16x8 vector to 32-bits. More...

basic_int32x8	simdpp::to_int32x8 (int16x16 a)
	Sign extends the first 8 values of a signed int16x16 vector to 32-bits. More...

basic_int32x4	simdpp::to_int32x4 (uint16x8 a)
	Zero-extends the values of a unsigned int16x8 vector to 32-bits. More...

basic_int32x8	simdpp::to_int32x8 (uint16x16 a)
	Zero-extends the first 8 values of a unsigned int16x16 vector to 32-bits. More...

basic_int32x4	simdpp::to_int32x4 (float32x4 a)
	Converts the values of a float32x4 vector into signed int32_t representation using truncation if only an inexact conversion can be performed. More...

basic_int32x8	simdpp::to_int32x8 (float32x8 a)
	Converts the values of a float32x4 vector into signed int32_t representation using truncation if only an inexact conversion can be performed. More...

basic_int32x4	simdpp::to_int32x4 (float64x2 a)
	Converts the values of a doublex2 vector into int32_t representation using truncation. More...

basic_int32x8	simdpp::to_int32x8 (float64x4 a)
	Converts the values of a doublex2 vector into int32_t representation using truncation. More...

basic_int64x2	simdpp::to_int64x2 (int32x4 a)
	Extends the values of a signed int32x4 vector to 64-bits. More...

basic_int64x4	simdpp::to_int64x4 (int32x8 a)
	Extends the values of a signed int32x4 vector to 64-bits. More...

basic_int64x2	simdpp::to_int64x2 (uint32x4 a)
	Extends the values of an unsigned int32x4 vector to 64-bits. More...

basic_int64x4	simdpp::to_int64x4 (uint32x8 a)
	Extends the values of a signed int32x4 vector to 64-bits. More...

float32x4	simdpp::to_float32x4 (int32x4 a)
	Converts 32-bit integer values to 32-bit float values. More...

float32x8	simdpp::to_float32x8 (int32x8 a)
	Converts 32-bit integer values to 32-bit float values. More...

float32x4	simdpp::to_float32x4 (float64x2 a)
	Converts 64-bit float values to 32-bit float values. More...

float32x8	simdpp::to_float32x8 (float64x4 a)
	Converts 64-bit float values to 32-bit float values. More...

float64x2	simdpp::to_float64x2 (int32x4 a)
	Converts the 32-bit integer values to 64-bit float values. More...

float64x4	simdpp::to_float64x4 (int32x8 a)
	Converts the 32-bit integer values to 64-bit float values. More...

float64x2	simdpp::to_float64x2 (float32x4 a)
	Converts the 32-bit float values to 64-bit float values. More...

float64x4	simdpp::to_float64x4 (float32x8 a)
	Converts the 32-bit float values to 64-bit float values. More...

Detailed Description

Function Documentation

template<class R , class T >

R simdpp::bit_cast ( T t )

Casts between unrelated types.

No changes to the stored values are performed.

float32x4 simdpp::to_float32x4 ( int32x4 a )

inline

Converts 32-bit integer values to 32-bit float values.

SSE specific:

If only inexact conversion can be performed, the current rounding mode is used.

NEON, ALTIVEC specific:

If only inexact conversion can be performed, round to nearest mode is used.

r0 = (float) a0
...
rN = (float) aN

256-bit version:

In SSE2-SSE4.1, NEON and ALTIVEC this intrinsic results in at least 2 instructions.

float32x4 simdpp::to_float32x4 ( float64x2 a )

inline

Converts 64-bit float values to 32-bit float values.

SSE specific:

If only inexact conversion can be performed, the value is rounded according to the current rounding mode.

NEON specific:

If only inexact conversion can be performed, the value is truncated.

128-bit version:: r0 = (float) a0

r1 = (float) a1

r2 = 0.0f

r3 = 0.0f

Not vectorized in NEON and .

256-bit version:: r0 = (float) a0

...

r3 = (float) a3

r4 = 0.0f

...

r7 = 0.0f

In SSE2-SSE4.1 this intrinsic results in at least 3 instructions.
Not vectorized in NEON and .

float32x8 simdpp::to_float32x8 ( int32x8 a )

inline

Converts 32-bit integer values to 32-bit float values.

SSE specific:

If only inexact conversion can be performed, the current rounding mode is used.

NEON, ALTIVEC specific:

If only inexact conversion can be performed, round to nearest mode is used.

r0 = (float) a0
...
rN = (float) aN

256-bit version:

In SSE2-SSE4.1, NEON and ALTIVEC this intrinsic results in at least 2 instructions.

float32x8 simdpp::to_float32x8 ( float64x4 a )

inline

Converts 64-bit float values to 32-bit float values.

SSE specific:

If only inexact conversion can be performed, the value is rounded according to the current rounding mode.

NEON specific:

If only inexact conversion can be performed, the value is truncated.

128-bit version:: r0 = (float) a0

r1 = (float) a1

r2 = 0.0f

r3 = 0.0f

Not vectorized in NEON and .

256-bit version:: r0 = (float) a0

...

r3 = (float) a3

r4 = 0.0f

...

r7 = 0.0f

In SSE2-SSE4.1 this intrinsic results in at least 3 instructions.
Not vectorized in NEON and .

float64x2 simdpp::to_float64x2 ( int32x4 a )

inline

Converts the 32-bit integer values to 64-bit float values.

SSE specific:

If only inexact conversion can be performed, the value is rounded according to the current rounding mode.

NEON specific:

If only inexact conversion can be performed, the value is rounded to the nearest representable value.

128-bit version:: r0 = (double) a0

r1 = (double) a1

Not vectorized in NEON and .

256-bit version:: r0 = (double) a0

...

r3 = (double) a3

In SSE2-SSE4.1 this intrinsic results in at least 3 instructions.
Not vectorized in NEON and .

float64x2 simdpp::to_float64x2 ( float32x4 a )

inline

Converts the 32-bit float values to 64-bit float values.

SSE specific:

If only inexact conversion can be performed, the value is rounded according to the current rounding mode.

NEON specific:

If only inexact conversion can be performed, the value is rounded to the nearest representable value.

128-bit version:: r0 = (double) a0

r1 = (double) a1

Not vectorized in NEON and .

256-bit version:: r0 = (double) a0

...

r3 = (double) a3

In SSE2-SSE4.1 this intrinsic results in at least 3 instructions.
Not vectorized in NEON and .

float64x4 simdpp::to_float64x4 ( int32x8 a )

inline

Converts the 32-bit integer values to 64-bit float values.

SSE specific:

If only inexact conversion can be performed, the value is rounded according to the current rounding mode.

NEON specific:

If only inexact conversion can be performed, the value is rounded to the nearest representable value.

128-bit version:: r0 = (double) a0

r1 = (double) a1

Not vectorized in NEON and .

256-bit version:: r0 = (double) a0

...

r3 = (double) a3

In SSE2-SSE4.1 this intrinsic results in at least 3 instructions.
Not vectorized in NEON and .

float64x4 simdpp::to_float64x4 ( float32x8 a )

inline

Converts the 32-bit float values to 64-bit float values.

SSE specific:

If only inexact conversion can be performed, the value is rounded according to the current rounding mode.

NEON specific:

If only inexact conversion can be performed, the value is rounded to the nearest representable value.

128-bit version:: r0 = (double) a0

r1 = (double) a1

Not vectorized in NEON and .

256-bit version:: r0 = (double) a0

...

r3 = (double) a3

In SSE2-SSE4.1 this intrinsic results in at least 3 instructions.
Not vectorized in NEON and .

basic_int16x16 simdpp::to_int16x16 ( int8x32 a )

inline

Sign extends the first 16 values of a signed int8x32 vector to 16-bits.

r0 = (int16_t) a0
...
r15 = (int16_t) a15

In SSE4.1-AVX this intrinsic results in at least 3 instructions.
In SSE2-SSSE3 this intrinsic results in at least 4 instructions.
In NEON and ALTIVEC this intrinsic results in at least 2 instructions.

basic_int16x16 simdpp::to_int16x16 ( uint8x32 a )

inline

Extends the first 16 values of a unsigned int8x32 vector to 16-bits.

r0 = (uint16_t) a0
...
r15 = (uint16_t) a15

In SSE2-AVX, NEON and ALTIVEC this intrinsic results in at least 2 instructions.

basic_int16x8 simdpp::to_int16x8 ( int8x16 a )

inline

Sign extends the first 8 values of a signed int8x16 vector to 16-bits.

r0 = (int16_t) a0
...
r7 = (int16_t) a7

In SSE2-SSSE3 this intrinsic results in at least 2 instructions.

basic_int16x8 simdpp::to_int16x8 ( uint8x16 a )

inline

Extends the first 8 values of a unsigned int8x16 vector to 16-bits.

r0 = (uint16_t) a0
...
r7 = (uint16_t) a7

basic_int32x4 simdpp::to_int32x4 ( int16x8 a )

inline

Sign extends the first 4 values of a signed int16x8 vector to 32-bits.

r0 = (int32_t) a0
...
r3 = (int32_t) a3

In SSE2-SSSE3 this intrinsic results in at least 2 instructions.

basic_int32x4 simdpp::to_int32x4 ( uint16x8 a )

inline

Zero-extends the values of a unsigned int16x8 vector to 32-bits.

r0 = (uint32_t) a0
...
r3 = (uint32_t) a3

basic_int32x4 simdpp::to_int32x4 ( float32x4 a )

inline

Converts the values of a float32x4 vector into signed int32_t representation using truncation if only an inexact conversion can be performed.

The behavior is undefined if the value can not be represented in the result type.

SSE specific: If the value can not be represented by int32_t, 0x80000000 is returned TODO: NaN handling

NEON, ALTIVEC specific: If the value can not be represented by int32_t, either 0x80000000 or 0x7fffffff is returned depending on the sign of the operand (saturation occurs). Conversion of NaNs results in 0.

r0 = (int32_t) a0
...
rN = (int32_t) aN

256-bit version:

In SSE2-SSE4.1, NEON and ALTIVEC this intrinsic results in at least 2 instructions.

basic_int32x4 simdpp::to_int32x4 ( float64x2 a )

inline

Converts the values of a doublex2 vector into int32_t representation using truncation.

The behavior is undefined if the value can not be represented in the result type.

SSE specific: If the value can not be represented by int32_t, 0x80000000 is returned

Todo:: NaN handling

NEON VFP specific: If the value can not be represented by int32_t, either 0x80000000 or 0x7fffffff is returned depending on the sign of the operand. Conversion of NaNs results in 0.

128-bit version:

Not vectorized in NEON and .

r0 = (int32_t) a0
r1 = (int32_t) a1
r2 = 0
r3 = 0

256-bit version:

r0 = (int32_t) a0
...
r3 = (int32_t) a3
r4 = 0
...
r7 = 0

Not vectorized in NEON and .
In SSE2-SSE4.1 this intrinsic results in at least 3 instructions.

basic_int32x8 simdpp::to_int32x8 ( int16x16 a )

inline

Sign extends the first 8 values of a signed int16x16 vector to 32-bits.

r0 = (int32_t) a0
...
r7 = (int32_t) a7

In SSE4.1-AVX this intrinsic results in at least 3 instructions.
In SSE2-SSSE3 this intrinsic results in at least 4 instructions.
In NEON and ALTIVEC this intrinsic results in at least 2 instructions.

basic_int32x8 simdpp::to_int32x8 ( uint16x16 a )

inline

Zero-extends the first 8 values of a unsigned int16x16 vector to 32-bits.

r0 = (uint32_t) a0
...
r7 = (uint32_t) a7

In SSE2-AVX and ALTIVEC this intrinsic results in at least 2 instructions.
In NEON this intrinsic results in at least 2 instructions.

basic_int32x8 simdpp::to_int32x8 ( float32x8 a )

inline

Converts the values of a float32x4 vector into signed int32_t representation using truncation if only an inexact conversion can be performed.

The behavior is undefined if the value can not be represented in the result type.

SSE specific: If the value can not be represented by int32_t, 0x80000000 is returned TODO: NaN handling

NEON, ALTIVEC specific: If the value can not be represented by int32_t, either 0x80000000 or 0x7fffffff is returned depending on the sign of the operand (saturation occurs). Conversion of NaNs results in 0.

r0 = (int32_t) a0
...
rN = (int32_t) aN

256-bit version:

In SSE2-SSE4.1, NEON and ALTIVEC this intrinsic results in at least 2 instructions.

basic_int32x8 simdpp::to_int32x8 ( float64x4 a )

inline

Converts the values of a doublex2 vector into int32_t representation using truncation.

The behavior is undefined if the value can not be represented in the result type.

SSE specific: If the value can not be represented by int32_t, 0x80000000 is returned

Todo:: NaN handling

NEON VFP specific: If the value can not be represented by int32_t, either 0x80000000 or 0x7fffffff is returned depending on the sign of the operand. Conversion of NaNs results in 0.

128-bit version:

Not vectorized in NEON and .

r0 = (int32_t) a0
r1 = (int32_t) a1
r2 = 0
r3 = 0

256-bit version:

r0 = (int32_t) a0
...
r3 = (int32_t) a3
r4 = 0
...
r7 = 0

Not vectorized in NEON and .
In SSE2-SSE4.1 this intrinsic results in at least 3 instructions.

basic_int64x2 simdpp::to_int64x2 ( int32x4 a )

inline

Extends the values of a signed int32x4 vector to 64-bits.

128-bit version:: r0 = (int64_t) a0

r1 = (int64_t) a1

In SSE2-SSSE3 this intrinsic results in at least 2 instructions.
In ALTIVEC this intrinsic results in at least 2-3 instructions.

256-bit version:: r0 = (int64_t) a0

...

r3 = (int64_t) a3

In SSE2-SSSE3 this intrinsic results in at least 5 instructions.
In SSE4.1-AVX this intrinsic results in at least 3 instructions.
In NEON this intrinsic results in at least 2 instructions.
In ALTIVEC this intrinsic results in at least 3-4 instructions.

basic_int64x2 simdpp::to_int64x2 ( uint32x4 a )

inline

Extends the values of an unsigned int32x4 vector to 64-bits.

128-bit version:: r0 = (uint64_t) a0

r1 = (uint64_t) a1

256-bit version:: r0 = (uint64_t) a0

...

r3 = (uint64_t) a3

In SSE2-AVX this intrinsic results in at least 3 instructions.
In NEON and ALTIVEC this intrinsic results in at least 2 instructions.

basic_int64x4 simdpp::to_int64x4 ( int32x8 a )

inline

Extends the values of a signed int32x4 vector to 64-bits.

128-bit version:: r0 = (int64_t) a0

r1 = (int64_t) a1

In SSE2-SSSE3 this intrinsic results in at least 2 instructions.
In ALTIVEC this intrinsic results in at least 2-3 instructions.

256-bit version:: r0 = (int64_t) a0

...

r3 = (int64_t) a3

In SSE2-SSSE3 this intrinsic results in at least 5 instructions.
In SSE4.1-AVX this intrinsic results in at least 3 instructions.
In NEON this intrinsic results in at least 2 instructions.
In ALTIVEC this intrinsic results in at least 3-4 instructions.

basic_int64x4 simdpp::to_int64x4 ( uint32x8 a )

inline

Extends the values of a signed int32x4 vector to 64-bits.

128-bit version:: r0 = (int64_t) a0

r1 = (int64_t) a1

In SSE2-SSSE3 this intrinsic results in at least 2 instructions.
In ALTIVEC this intrinsic results in at least 2-3 instructions.

256-bit version:: r0 = (int64_t) a0

...

r3 = (int64_t) a3

In SSE2-SSSE3 this intrinsic results in at least 5 instructions.
In SSE4.1-AVX this intrinsic results in at least 3 instructions.
In NEON this intrinsic results in at least 2 instructions.
In ALTIVEC this intrinsic results in at least 3-4 instructions.

Functions

Detailed Description

Function Documentation