Project Ne10
An open, optimized software library for the ARM architecture.
Macros
NE10_fft.neonintrinsic.h File Reference
#include "NE10_fft.h"
#include <arm_neon.h>

Go to the source code of this file.

Macros

#define NE10_CPX_ADD_NEON_F32(Z, A, B)
 
#define NE10_CPX_SUB_NEON_F32(Z, A, B)
 
#define NE10_CPX_MUL_NEON_F32(Z, A, B)
 
#define NE10_CPX_MUL_INV_NEON_F32(Z, A, B)
 
#define NE10_BUTTERFLY_NEON_F32(O1, O2, I1, I2)
 
#define NE10_DECLARE_2(TYPE, NAME)
 
#define NE10_DECLARE_3(TYPE, NAME)
 
#define NE10_DECLARE_4(TYPE, NAME)
 
#define NE10_DECLARE_8(TYPE, NAME)
 
#define NE10_REVERSE_FLOAT32X4(VECTOR4F)
 
#define NE10_REVERSE_OUT_FLOAT32X4(VECTOR4F_OUT, VECTOR4F)
 
#define NE10_RADIX4X4C_TRANSPOSE_NEON(Q2_OUT, Q2_IN)
 
#define VDUPQ_N_F32(VAR)   { VAR, VAR, VAR, VAR }
 
#define CONST_TW_81   0.70710678
 
#define CONST_TW_81N   -0.70710678
 
#define DIV_TW81   1.4142136f
 
#define DIV_TW81N   -1.4142136f
 
#define NE10_RADIX8x4_R2C_NEON_KERNEL_S1(Q_OUT, Q_IN)
 
#define NE10_RADIX8x4_R2C_NEON_KERNEL_S2(Q_OUT, Q_IN)
 
#define NE10_RADIX8x4_C2R_NEON_KERNEL_S1(Q_OUT, Q_IN)
 
#define NE10_RADIX8x4_C2R_NEON_KERNEL_S2(Q_OUT, Q_IN)
 
#define NE10_RADIX8x4_C2R_NEON_KERNEL_SCALE(Q_OUT)
 
#define NE10_RADIX4x4_C2R_NEON_KERNEL_SCALE(Q_OUT)
 
#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL_SCALE(Q2_OUT)
 
#define NE10_RADIX8x4_R2C_NEON_KERNEL(Q_OUT, Q_IN)
 
#define NE10_RADIX4x4_R2C_NEON_KERNEL(Q_OUT, Q_IN)
 
#define NE10_RADIX4x4_C2R_NEON_KERNEL(Q_OUT, Q_IN)
 
#define NE10_RADIX8x4_C2R_NEON_KERNEL(Q_OUT, Q_IN)
 
#define NE10_RADIX8x4_R2C_NEON_LOAD(PTR_IN, Q_IN, IN_STEP)
 
#define NE10_RADIX4x4_R2C_NEON_LOAD(PTR_IN, Q_IN, IN_STEP)
 
#define NE10_RADIX8x4_R2C_NEON_STORE(PTR_OUT, Q_OUT, OUT_STEP)
 
#define NE10_RADIX4x4_R2C_NEON_STORE(PTR_OUT, Q_OUT, OUT_STEP)
 
#define NE10_RADIX4x4_R2C_TW_MUL_NEON(Q2_OUT, Q2_IN, Q2_TW)
 
#define NE10_RADIX4x4_C2R_TW_MUL_NEON(Q2_OUT, Q2_IN, Q2_TW)
 
#define NE10_RADIX4x4_R2C_TW_NEON_KERNEL_S1(Q2_OUT, Q2_IN)
 
#define NE10_RADIX4x4_R2C_TW_NEON_KERNEL_S2(Q2_OUT, Q2_IN)
 
#define NE10_RADIX4x4_R2C_TW_NEON_KERNEL_LAST(Q_OUT, Q_IN)
 
#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL_LAST(Q_OUT, Q_IN)
 
#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL_S1(Q2_OUT, Q2_IN)
 
#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL_S2(Q2_OUT, Q2_IN)
 
#define NE10_RADIX4x4_R2C_TW_NEON_KERNEL(Q2_OUT, Q2_IN, Q2_TW)
 
#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL(Q2_OUT, Q2_IN, Q2_TW)
 
#define NE10_PRINT_Qx8_VECTOR(Q_VECTOR)   ;
 
#define NE10_PRINT_Qx4_VECTOR(Q_VECTOR)   ;
 
#define NE10_PRINT_Q2x4_VECTOR(Q2_VECTOR)   ;
 

Macro Definition Documentation

#define CONST_TW_81   0.70710678

Definition at line 112 of file NE10_fft.neonintrinsic.h.

#define CONST_TW_81N   -0.70710678

Definition at line 113 of file NE10_fft.neonintrinsic.h.

#define DIV_TW81   1.4142136f

Definition at line 118 of file NE10_fft.neonintrinsic.h.

#define DIV_TW81N   -1.4142136f

Definition at line 119 of file NE10_fft.neonintrinsic.h.

#define NE10_BUTTERFLY_NEON_F32 (   O1,
  O2,
  I1,
  I2 
)
Value:
do { \
NE10_CPX_ADD_NEON(O1,I1,I2); \
NE10_CPX_SUB_NEON(O2,I1,I2); \
} while(0);
O2
#define O2
I1
#define I1
I2
#define I2
O1
#define O1

Definition at line 64 of file NE10_fft.neonintrinsic.h.

#define NE10_CPX_ADD_NEON_F32 (   Z,
  A,
  B 
)
Value:
do { \
Z.val[0] = A.val[0] + B.val[0]; \
Z.val[1] = A.val[1] + B.val[1]; \
} while (0);

Definition at line 38 of file NE10_fft.neonintrinsic.h.

#define NE10_CPX_MUL_INV_NEON_F32 (   Z,
  A,
  B 
)
Value:
do { \
float32x4_t ARBR = vmulq_f32( A.val[0], B.val[0] ); \
float32x4_t AIBI = vmulq_f32( A.val[1], B.val[1] ); \
float32x4_t ARBI = vmulq_f32( A.val[0], B.val[1] ); \
float32x4_t AIBR = vmulq_f32( A.val[1], B.val[0] ); \
Z.val[0] = vaddq_f32(ARBR,AIBI); \
Z.val[1] = vsubq_f32(AIBR,ARBI); \
} while (0);

Definition at line 55 of file NE10_fft.neonintrinsic.h.

#define NE10_CPX_MUL_NEON_F32 (   Z,
  A,
  B 
)
Value:
do { \
float32x4_t ARBR = vmulq_f32( A.val[0], B.val[0] ); \
float32x4_t ARBI = vmulq_f32( A.val[0], B.val[1] ); \
Z.val[0] = vmlsq_f32(ARBR, A.val[1], B.val[1]); \
Z.val[1] = vmlaq_f32(ARBI, A.val[1], B.val[0]); \
} while (0);

Definition at line 48 of file NE10_fft.neonintrinsic.h.

#define NE10_CPX_SUB_NEON_F32 (   Z,
  A,
  B 
)
Value:
do { \
Z.val[0] = A.val[0] - B.val[0]; \
Z.val[1] = A.val[1] - B.val[1]; \
} while (0);

Definition at line 43 of file NE10_fft.neonintrinsic.h.

#define NE10_DECLARE_2 (   TYPE,
  NAME 
)
Value:
TYPE NAME ## 0; \
TYPE NAME ## 1;

Definition at line 69 of file NE10_fft.neonintrinsic.h.

#define NE10_DECLARE_3 (   TYPE,
  NAME 
)
Value:
NE10_DECLARE_2(TYPE,NAME); \
TYPE NAME ## 2;
NE10_DECLARE_2
#define NE10_DECLARE_2(TYPE, NAME)
Definition: NE10_fft.neonintrinsic.h:69

Definition at line 72 of file NE10_fft.neonintrinsic.h.

#define NE10_DECLARE_4 (   TYPE,
  NAME 
)
Value:
NE10_DECLARE_3(TYPE,NAME); \
TYPE NAME ## 3;
NE10_DECLARE_3
#define NE10_DECLARE_3(TYPE, NAME)
Definition: NE10_fft.neonintrinsic.h:72

Definition at line 75 of file NE10_fft.neonintrinsic.h.

#define NE10_DECLARE_8 (   TYPE,
  NAME 
)
Value:
NE10_DECLARE_4(TYPE,NAME); \
TYPE NAME ## 4; \
TYPE NAME ## 5; \
TYPE NAME ## 6; \
TYPE NAME ## 7;
NE10_DECLARE_4
#define NE10_DECLARE_4(TYPE, NAME)
Definition: NE10_fft.neonintrinsic.h:75

Definition at line 78 of file NE10_fft.neonintrinsic.h.

#define NE10_PRINT_Q2x4_VECTOR (   Q2_VECTOR)    ;

Definition at line 427 of file NE10_fft.neonintrinsic.h.

#define NE10_PRINT_Qx4_VECTOR (   Q_VECTOR)    ;

Definition at line 426 of file NE10_fft.neonintrinsic.h.

#define NE10_PRINT_Qx8_VECTOR (   Q_VECTOR)    ;

Definition at line 425 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_C2R_NEON_KERNEL (   Q_OUT,
  Q_IN 
)
Value:
do { \
NE10_DECLARE_4(float32x4_t,Q_S_IN); \
Q_S_IN0 = vaddq_f32 (Q_IN##0, Q_IN##3); \
Q_S_IN1 = vsubq_f32 (Q_IN##0, Q_IN##3); \
Q_S_IN2 = vaddq_f32 (Q_IN##1, Q_IN##1); \
Q_S_IN3 = vaddq_f32 (Q_IN##2, Q_IN##2); \
Q_OUT ## 0 = vaddq_f32 (Q_S_IN0, Q_S_IN2); \
Q_OUT ## 1 = vsubq_f32 (Q_S_IN1, Q_S_IN3); \
Q_OUT ## 2 = vsubq_f32 (Q_S_IN0, Q_S_IN2); \
Q_OUT ## 3 = vaddq_f32 (Q_S_IN1, Q_S_IN3); \
} while(0);
NE10_DECLARE_4
#define NE10_DECLARE_4(TYPE, NAME)
Definition: NE10_fft.neonintrinsic.h:75

Definition at line 231 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_C2R_NEON_KERNEL_SCALE (   Q_OUT)
Value:
do { \
Q_OUT ## 0 = vmulq_f32( Q_OUT ## 0, QUAD_NEON); \
Q_OUT ## 1 = vmulq_f32( Q_OUT ## 1, QUAD_NEON); \
Q_OUT ## 2 = vmulq_f32( Q_OUT ## 2, QUAD_NEON); \
Q_OUT ## 3 = vmulq_f32( Q_OUT ## 3, QUAD_NEON); \
} while(0);

Definition at line 197 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_C2R_TW_MUL_NEON (   Q2_OUT,
  Q2_IN,
  Q2_TW 
)
Value:
do { \
Q2_OUT ## 0 = Q2_IN ## 0; \
NE10_CPX_MUL_INV_NEON_F32(Q2_OUT ## 1,Q2_IN ## 1,Q2_TW ## 0); \
NE10_CPX_MUL_INV_NEON_F32(Q2_OUT ## 2,Q2_IN ## 2,Q2_TW ## 1); \
NE10_CPX_MUL_INV_NEON_F32(Q2_OUT ## 3,Q2_IN ## 3,Q2_TW ## 2); \
} while(0);
NE10_CPX_MUL_INV_NEON_F32
#define NE10_CPX_MUL_INV_NEON_F32(Z, A, B)
Definition: NE10_fft.neonintrinsic.h:55

Definition at line 304 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL (   Q2_OUT,
  Q2_IN,
  Q2_TW 
)
Value:
do { \
NE10_RADIX4x4_C2R_TW_NEON_KERNEL_S1(Q2_OUT,Q2_IN); \
NE10_RADIX4x4_C2R_TW_NEON_KERNEL_S2(Q2_IN,Q2_OUT); \
NE10_RADIX4x4_C2R_TW_MUL_NEON(Q2_OUT,Q2_IN,Q2_TW); \
} while(0);
NE10_RADIX4x4_C2R_TW_NEON_KERNEL_S1
#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL_S1(Q2_OUT, Q2_IN)
Definition: NE10_fft.neonintrinsic.h:360
NE10_RADIX4x4_C2R_TW_MUL_NEON
#define NE10_RADIX4x4_C2R_TW_MUL_NEON(Q2_OUT, Q2_IN, Q2_TW)
Definition: NE10_fft.neonintrinsic.h:304
NE10_RADIX4x4_C2R_TW_NEON_KERNEL_S2
#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL_S2(Q2_OUT, Q2_IN)
Definition: NE10_fft.neonintrinsic.h:372

Definition at line 385 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL_LAST (   Q_OUT,
  Q_IN 
)
Value:
do { \
float32x4_t Q_TMP; \
Q_IN ## 1 = vnegq_f32(Q_IN ## 1 ); \
Q_OUT ## 0 = vaddq_f32(Q_IN ## 0, Q_IN ## 2); \
Q_OUT ## 1 = vsubq_f32(Q_IN ## 0, Q_IN ## 2); \
Q_OUT ## 2 = vaddq_f32(Q_IN ## 1, Q_IN ## 3); \
Q_OUT ## 3 = vsubq_f32(Q_IN ## 1, Q_IN ## 3); \
Q_TMP = vaddq_f32(Q_OUT ## 1, Q_OUT ## 3); \
Q_OUT ## 3 = vsubq_f32(Q_OUT ## 3, Q_OUT ## 1); \
Q_OUT ## 1 = Q_TMP; \
Q_OUT ## 1 = vmulq_f32( Q_OUT ## 1, DIV_TW81_NEON); \
Q_OUT ## 3 = vmulq_f32( Q_OUT ## 3, DIV_TW81_NEON); \
Q_OUT ## 0 = vaddq_f32( Q_OUT ## 0, Q_OUT ## 0 ); \
Q_OUT ## 2 = vaddq_f32( Q_OUT ## 2, Q_OUT ## 2 ); \
} while(0);

Definition at line 344 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL_S1 (   Q2_OUT,
  Q2_IN 
)
Value:
do { \
Q2_IN ## 3 .val[1] = vnegq_f32(Q2_IN ## 3 .val[1]); \
Q2_OUT ## 0 .val[0] = vaddq_f32(Q2_IN ## 0 .val[0] , Q2_IN ## 2 .val[0]); \
Q2_OUT ## 0 .val[1] = vsubq_f32(Q2_IN ## 0 .val[1] , Q2_IN ## 2 .val[1]); \
Q2_OUT ## 2 .val[0] = vsubq_f32(Q2_IN ## 0 .val[0] , Q2_IN ## 2 .val[0]); \
Q2_OUT ## 2 .val[1] = vaddq_f32(Q2_IN ## 2 .val[1] , Q2_IN ## 0 .val[1]); \
Q2_OUT ## 1 .val[0] = vaddq_f32(Q2_IN ## 1 .val[0] , Q2_IN ## 3 .val[0]); \
Q2_OUT ## 1 .val[1] = vaddq_f32(Q2_IN ## 1 .val[1] , Q2_IN ## 3 .val[1]); \
Q2_OUT ## 3 .val[0] = vsubq_f32(Q2_IN ## 3 .val[1] , Q2_IN ## 1 .val[1]); \
Q2_OUT ## 3 .val[1] = vsubq_f32(Q2_IN ## 1 .val[0] , Q2_IN ## 3 .val[0]); \
} while(0);

Definition at line 360 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL_S2 (   Q2_OUT,
  Q2_IN 
)
Value:
do { \
NE10_CPX_ADD_NEON_F32(Q2_OUT ## 0,Q2_IN ## 0,Q2_IN ## 1); \
NE10_CPX_SUB_NEON_F32(Q2_OUT ## 2,Q2_IN ## 0,Q2_IN ## 1); \
NE10_CPX_ADD_NEON_F32(Q2_OUT ## 1,Q2_IN ## 2,Q2_IN ## 3); \
NE10_CPX_SUB_NEON_F32(Q2_OUT ## 3,Q2_IN ## 2,Q2_IN ## 3); \
} while(0);
NE10_CPX_ADD_NEON_F32
#define NE10_CPX_ADD_NEON_F32(Z, A, B)
Definition: NE10_fft.neonintrinsic.h:38
NE10_CPX_SUB_NEON_F32
#define NE10_CPX_SUB_NEON_F32(Z, A, B)
Definition: NE10_fft.neonintrinsic.h:43

Definition at line 372 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_C2R_TW_NEON_KERNEL_SCALE (   Q2_OUT)
Value:
do { \
Q2_OUT ## 0 .val[0] = vmulq_f32( Q2_OUT ## 0 .val[0], QUAD_NEON); \
Q2_OUT ## 1 .val[0] = vmulq_f32( Q2_OUT ## 1 .val[0], QUAD_NEON); \
Q2_OUT ## 2 .val[0] = vmulq_f32( Q2_OUT ## 2 .val[0], QUAD_NEON); \
Q2_OUT ## 3 .val[0] = vmulq_f32( Q2_OUT ## 3 .val[0], QUAD_NEON); \
Q2_OUT ## 0 .val[1] = vmulq_f32( Q2_OUT ## 0 .val[1], QUAD_NEON); \
Q2_OUT ## 1 .val[1] = vmulq_f32( Q2_OUT ## 1 .val[1], QUAD_NEON); \
Q2_OUT ## 2 .val[1] = vmulq_f32( Q2_OUT ## 2 .val[1], QUAD_NEON); \
Q2_OUT ## 3 .val[1] = vmulq_f32( Q2_OUT ## 3 .val[1], QUAD_NEON); \
} while(0);

Definition at line 204 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_R2C_NEON_KERNEL (   Q_OUT,
  Q_IN 
)
Value:
do { \
NE10_DECLARE_4(float32x4_t,Q_S_IN); \
Q_S_IN0 = vaddq_f32 (Q_IN ## 0, Q_IN ## 2); \
Q_S_IN1 = vaddq_f32 (Q_IN ## 1, Q_IN ## 3); \
Q_OUT ## 0 = vaddq_f32 (Q_S_IN0, Q_S_IN1); \
Q_OUT ## 1 = vsubq_f32 (Q_IN##0, Q_IN##2); \
Q_OUT ## 2 = vsubq_f32 (Q_IN##3, Q_IN##1); \
Q_OUT ## 3 = vsubq_f32 (Q_S_IN0, Q_S_IN1); \
} while(0);
NE10_DECLARE_4
#define NE10_DECLARE_4(TYPE, NAME)
Definition: NE10_fft.neonintrinsic.h:75

Definition at line 221 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_R2C_NEON_LOAD (   PTR_IN,
  Q_IN,
  IN_STEP 
)
Value:
do {\
Q_IN ## 0 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
Q_IN ## 1 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
Q_IN ## 2 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
Q_IN ## 3 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
} while(0);
ne10_float32_t
float ne10_float32_t
Definition: NE10_types.h:80

Definition at line 268 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_R2C_NEON_STORE (   PTR_OUT,
  Q_OUT,
  OUT_STEP 
)
Value:
do { \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 0 * OUT_STEP ), Q_OUT ## 0); \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 1 * OUT_STEP ), Q_OUT ## 1); \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 2 * OUT_STEP ), Q_OUT ## 2); \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 3 * OUT_STEP ), Q_OUT ## 3); \
} while(0);
ne10_float32_t
float ne10_float32_t
Definition: NE10_types.h:80

Definition at line 290 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_R2C_TW_MUL_NEON (   Q2_OUT,
  Q2_IN,
  Q2_TW 
)
Value:
do { \
Q2_OUT ## 0 = Q2_IN ## 0; \
NE10_CPX_MUL_NEON_F32(Q2_OUT ## 1,Q2_IN ## 1,Q2_TW ## 0); \
NE10_CPX_MUL_NEON_F32(Q2_OUT ## 2,Q2_IN ## 2,Q2_TW ## 1); \
NE10_CPX_MUL_NEON_F32(Q2_OUT ## 3,Q2_IN ## 3,Q2_TW ## 2); \
} while(0);
NE10_CPX_MUL_NEON_F32
#define NE10_CPX_MUL_NEON_F32(Z, A, B)
Definition: NE10_fft.neonintrinsic.h:48

Definition at line 297 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_R2C_TW_NEON_KERNEL (   Q2_OUT,
  Q2_IN,
  Q2_TW 
)
Value:
do { \
NE10_RADIX4x4_R2C_TW_MUL_NEON(Q2_OUT,Q2_IN,Q2_TW); \
NE10_RADIX4x4_R2C_TW_NEON_KERNEL_S1(Q2_IN,Q2_OUT); \
NE10_RADIX4x4_R2C_TW_NEON_KERNEL_S2(Q2_OUT,Q2_IN); \
} while(0);
NE10_RADIX4x4_R2C_TW_NEON_KERNEL_S1
#define NE10_RADIX4x4_R2C_TW_NEON_KERNEL_S1(Q2_OUT, Q2_IN)
Definition: NE10_fft.neonintrinsic.h:311
NE10_RADIX4x4_R2C_TW_NEON_KERNEL_S2
#define NE10_RADIX4x4_R2C_TW_NEON_KERNEL_S2(Q2_OUT, Q2_IN)
Definition: NE10_fft.neonintrinsic.h:318
NE10_RADIX4x4_R2C_TW_MUL_NEON
#define NE10_RADIX4x4_R2C_TW_MUL_NEON(Q2_OUT, Q2_IN, Q2_TW)
Definition: NE10_fft.neonintrinsic.h:297

Definition at line 379 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_R2C_TW_NEON_KERNEL_LAST (   Q_OUT,
  Q_IN 
)
Value:
do { \
float32x4_t Q_TMP; \
Q_IN ## 1 = vmulq_f32(Q_IN ## 1, Q_TW_81); \
Q_IN ## 3 = vmulq_f32(Q_IN ## 3, Q_TW_81); \
Q_TMP = vsubq_f32(Q_IN ## 1, Q_IN ## 3); \
Q_IN ## 3 = vaddq_f32(Q_IN ## 1, Q_IN ## 3); \
Q_IN ## 1 = Q_TMP; \
Q_OUT ## 0 = vaddq_f32(Q_IN ## 0, Q_IN ## 1); \
Q_OUT ## 1 = vaddq_f32(Q_IN ## 2, Q_IN ## 3); \
Q_OUT ## 2 = vsubq_f32(Q_IN ## 0, Q_IN ## 1); \
Q_OUT ## 3 = vsubq_f32(Q_IN ## 2, Q_IN ## 3); \
Q_OUT ## 1 = vnegq_f32(Q_OUT ## 1); \
} while(0);

Definition at line 330 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_R2C_TW_NEON_KERNEL_S1 (   Q2_OUT,
  Q2_IN 
)
Value:
do { \
NE10_CPX_ADD_NEON_F32(Q2_OUT ## 0,Q2_IN ## 0,Q2_IN ## 2); \
NE10_CPX_SUB_NEON_F32(Q2_OUT ## 1,Q2_IN ## 0,Q2_IN ## 2); \
NE10_CPX_ADD_NEON_F32(Q2_OUT ## 2,Q2_IN ## 1,Q2_IN ## 3); \
NE10_CPX_SUB_NEON_F32(Q2_OUT ## 3,Q2_IN ## 1,Q2_IN ## 3); \
} while(0);
NE10_CPX_ADD_NEON_F32
#define NE10_CPX_ADD_NEON_F32(Z, A, B)
Definition: NE10_fft.neonintrinsic.h:38
NE10_CPX_SUB_NEON_F32
#define NE10_CPX_SUB_NEON_F32(Z, A, B)
Definition: NE10_fft.neonintrinsic.h:43

Definition at line 311 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4x4_R2C_TW_NEON_KERNEL_S2 (   Q2_OUT,
  Q2_IN 
)
Value:
do { \
Q2_OUT ## 0 .val[0] = vaddq_f32(Q2_IN ## 0 .val[0] , Q2_IN ## 2 .val[0]); \
Q2_OUT ## 0 .val[1] = vaddq_f32(Q2_IN ## 0 .val[1] , Q2_IN ## 2 .val[1]); \
Q2_OUT ## 2 .val[0] = vsubq_f32(Q2_IN ## 0 .val[0] , Q2_IN ## 2 .val[0]); \
Q2_OUT ## 2 .val[1] = vsubq_f32(Q2_IN ## 2 .val[1] , Q2_IN ## 0 .val[1]); \
Q2_OUT ## 1 .val[0] = vaddq_f32(Q2_IN ## 1 .val[0] , Q2_IN ## 3 .val[1]); \
Q2_OUT ## 1 .val[1] = vsubq_f32(Q2_IN ## 1 .val[1] , Q2_IN ## 3 .val[0]); \
Q2_OUT ## 3 .val[0] = vsubq_f32(Q2_IN ## 1 .val[0] , Q2_IN ## 3 .val[1]); \
Q2_OUT ## 3 .val[1] = vaddq_f32(Q2_IN ## 3 .val[0] , Q2_IN ## 1 .val[1]); \
Q2_OUT ## 3 .val[1] = vnegq_f32(Q2_OUT ## 3 .val[1]); \
} while(0);

Definition at line 318 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX4X4C_TRANSPOSE_NEON (   Q2_OUT,
  Q2_IN 
)
Value:
do { \
NE10_DECLARE_4(float32x4x2_t,q2_tmp); \
q2_tmp0 = vtrnq_f32 (Q2_IN ## 0 .val[0], Q2_IN ## 1 .val[0]); \
q2_tmp1 = vtrnq_f32 (Q2_IN ## 0 .val[1], Q2_IN ## 1 .val[1]); \
q2_tmp2 = vtrnq_f32 (Q2_IN ## 2 .val[0], Q2_IN ## 3 .val[0]); \
q2_tmp3 = vtrnq_f32 (Q2_IN ## 2 .val[1], Q2_IN ## 3 .val[1]); \
Q2_OUT ## 0 .val[0] = vcombine_f32 (vget_low_f32 (q2_tmp0.val[0]), vget_low_f32 (q2_tmp2.val[0])); \
Q2_OUT ## 0 .val[1] = vcombine_f32 (vget_low_f32 (q2_tmp1.val[0]), vget_low_f32 (q2_tmp3.val[0])); \
Q2_OUT ## 1 .val[0] = vcombine_f32 (vget_low_f32 (q2_tmp0.val[1]), vget_low_f32 (q2_tmp2.val[1])); \
Q2_OUT ## 1 .val[1] = vcombine_f32 (vget_low_f32 (q2_tmp1.val[1]), vget_low_f32 (q2_tmp3.val[1])); \
Q2_OUT ## 2 .val[0] = vcombine_f32 (vget_high_f32 (q2_tmp0.val[0]), vget_high_f32 (q2_tmp2.val[0])); \
Q2_OUT ## 2 .val[1] = vcombine_f32 (vget_high_f32 (q2_tmp1.val[0]), vget_high_f32 (q2_tmp3.val[0])); \
Q2_OUT ## 3 .val[0] = vcombine_f32 (vget_high_f32 (q2_tmp0.val[1]), vget_high_f32 (q2_tmp2.val[1])); \
Q2_OUT ## 3 .val[1] = vcombine_f32 (vget_high_f32 (q2_tmp1.val[1]), vget_high_f32 (q2_tmp3.val[1])); \
} while(0);
NE10_DECLARE_4
#define NE10_DECLARE_4(TYPE, NAME)
Definition: NE10_fft.neonintrinsic.h:75

Definition at line 94 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX8x4_C2R_NEON_KERNEL (   Q_OUT,
  Q_IN 
)
Value:
do { \
NE10_DECLARE_8(float32x4_t,Q_S_IN_C2R_KERNEL); \
NE10_RADIX8x4_C2R_NEON_KERNEL_S1(Q_S_IN_C2R_KERNEL,Q_IN); \
NE10_RADIX8x4_C2R_NEON_KERNEL_S2(Q_OUT,Q_S_IN_C2R_KERNEL); \
} while(0);
NE10_RADIX8x4_C2R_NEON_KERNEL_S1
#define NE10_RADIX8x4_C2R_NEON_KERNEL_S1(Q_OUT, Q_IN)
Definition: NE10_fft.neonintrinsic.h:153
NE10_DECLARE_8
#define NE10_DECLARE_8(TYPE, NAME)
Definition: NE10_fft.neonintrinsic.h:78
NE10_RADIX8x4_C2R_NEON_KERNEL_S2
#define NE10_RADIX8x4_C2R_NEON_KERNEL_S2(Q_OUT, Q_IN)
Definition: NE10_fft.neonintrinsic.h:173

Definition at line 243 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX8x4_C2R_NEON_KERNEL_S1 (   Q_OUT,
  Q_IN 
)
Value:
do { \
NE10_DECLARE_8(float32x4_t,Q_S_IN); \
Q_S_IN0 = vaddq_f32(Q_IN ## 0, Q_IN ## 7); \
Q_S_IN1 = vsubq_f32(Q_IN ## 0, Q_IN ## 7); \
Q_S_IN2 = vaddq_f32(Q_IN ## 1, Q_IN ## 5); \
Q_S_IN3 = vsubq_f32(Q_IN ## 1, Q_IN ## 5); \
Q_S_IN4 = vaddq_f32(Q_IN ## 6, Q_IN ## 2); \
Q_S_IN5 = vsubq_f32(Q_IN ## 6, Q_IN ## 2); \
Q_S_IN6 = vaddq_f32(Q_IN ## 3, Q_IN ## 3); \
Q_S_IN7 = vaddq_f32(Q_IN ## 4, Q_IN ## 4); \
Q_OUT ## 0 = vaddq_f32(Q_S_IN0, Q_S_IN6); \
Q_OUT ## 1 = vaddq_f32(Q_S_IN2, Q_S_IN2); \
Q_OUT ## 2 = vsubq_f32(Q_S_IN1, Q_S_IN7); \
Q_OUT ## 3 = vsubq_f32(Q_S_IN3, Q_S_IN4); \
Q_OUT ## 4 = vsubq_f32(Q_S_IN0, Q_S_IN6); \
Q_OUT ## 5 = vaddq_f32(Q_S_IN5, Q_S_IN5); \
Q_OUT ## 6 = vaddq_f32(Q_S_IN1, Q_S_IN7); \
Q_OUT ## 7 = vaddq_f32(Q_S_IN4, Q_S_IN3); \
} while (0);
NE10_DECLARE_8
#define NE10_DECLARE_8(TYPE, NAME)
Definition: NE10_fft.neonintrinsic.h:78

Definition at line 153 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX8x4_C2R_NEON_KERNEL_S2 (   Q_OUT,
  Q_IN 
)
Value:
do { \
Q_IN ## 3 = vmulq_f32(Q_IN ## 3,DIV_TW81_NEON); \
Q_IN ## 7 = vmulq_f32(Q_IN ## 7,DIV_TW81N_NEON); \
Q_OUT ## 0 = vaddq_f32(Q_IN ## 0, Q_IN ## 1); \
Q_OUT ## 4 = vsubq_f32(Q_IN ## 0, Q_IN ## 1); \
Q_OUT ## 1 = vaddq_f32(Q_IN ## 2, Q_IN ## 3); \
Q_OUT ## 5 = vsubq_f32(Q_IN ## 2, Q_IN ## 3); \
Q_OUT ## 2 = vaddq_f32(Q_IN ## 4, Q_IN ## 5); \
Q_OUT ## 6 = vsubq_f32(Q_IN ## 4, Q_IN ## 5); \
Q_OUT ## 3 = vaddq_f32(Q_IN ## 6, Q_IN ## 7); \
Q_OUT ## 7 = vsubq_f32(Q_IN ## 6, Q_IN ## 7); \
} while(0);

Definition at line 173 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX8x4_C2R_NEON_KERNEL_SCALE (   Q_OUT)
Value:
do { \
Q_OUT ## 0 = vmulq_f32( Q_OUT ## 0, EIGH_NEON); \
Q_OUT ## 1 = vmulq_f32( Q_OUT ## 1, EIGH_NEON); \
Q_OUT ## 2 = vmulq_f32( Q_OUT ## 2, EIGH_NEON); \
Q_OUT ## 3 = vmulq_f32( Q_OUT ## 3, EIGH_NEON); \
Q_OUT ## 4 = vmulq_f32( Q_OUT ## 4, EIGH_NEON); \
Q_OUT ## 5 = vmulq_f32( Q_OUT ## 5, EIGH_NEON); \
Q_OUT ## 6 = vmulq_f32( Q_OUT ## 6, EIGH_NEON); \
Q_OUT ## 7 = vmulq_f32( Q_OUT ## 7, EIGH_NEON); \
} while(0);

Definition at line 186 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX8x4_R2C_NEON_KERNEL (   Q_OUT,
  Q_IN 
)
Value:
do { \
NE10_DECLARE_8(float32x4_t,Q_S_IN); \
NE10_RADIX8x4_R2C_NEON_KERNEL_S1(Q_S_IN,Q_IN); \
NE10_RADIX8x4_R2C_NEON_KERNEL_S2(Q_OUT,Q_S_IN); \
} while(0);
NE10_RADIX8x4_R2C_NEON_KERNEL_S1
#define NE10_RADIX8x4_R2C_NEON_KERNEL_S1(Q_OUT, Q_IN)
Definition: NE10_fft.neonintrinsic.h:124
NE10_DECLARE_8
#define NE10_DECLARE_8(TYPE, NAME)
Definition: NE10_fft.neonintrinsic.h:78
NE10_RADIX8x4_R2C_NEON_KERNEL_S2
#define NE10_RADIX8x4_R2C_NEON_KERNEL_S2(Q_OUT, Q_IN)
Definition: NE10_fft.neonintrinsic.h:137

Definition at line 215 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX8x4_R2C_NEON_KERNEL_S1 (   Q_OUT,
  Q_IN 
)
Value:
do { \
Q_OUT ## 0 = vaddq_f32 (Q_IN ## 0, Q_IN ## 4); \
Q_OUT ## 1 = vsubq_f32 (Q_IN ## 0, Q_IN ## 4); \
Q_OUT ## 2 = vaddq_f32 (Q_IN ## 1, Q_IN ## 5); \
Q_OUT ## 3 = vsubq_f32 (Q_IN ## 1, Q_IN ## 5); \
Q_OUT ## 4 = vaddq_f32 (Q_IN ## 2, Q_IN ## 6); \
Q_OUT ## 5 = vsubq_f32 (Q_IN ## 2, Q_IN ## 6); \
Q_OUT ## 6 = vaddq_f32 (Q_IN ## 3, Q_IN ## 7); \
Q_OUT ## 7 = vsubq_f32 (Q_IN ## 3, Q_IN ## 7); \
Q_OUT ## 3 = vmulq_f32 (Q_OUT ## 3, Q_TW_81 ); \
Q_OUT ## 7 = vmulq_f32 (Q_OUT ## 7, Q_TW_81N); \
} while(0);

Definition at line 124 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX8x4_R2C_NEON_KERNEL_S2 (   Q_OUT,
  Q_IN 
)
Value:
do { \
NE10_DECLARE_4(float32x4_t,Q_S); \
Q_S0 = vaddq_f32 (Q_IN ## 0, Q_IN ## 4); \
Q_S1 = vaddq_f32 (Q_IN ## 2, Q_IN ## 6); \
Q_S2 = vsubq_f32 (Q_IN ## 7, Q_IN ## 3); \
Q_S3 = vaddq_f32 (Q_IN ## 3, Q_IN ## 7); \
Q_OUT ## 0 = vaddq_f32 ( Q_S0, Q_S1 ); \
Q_OUT ## 1 = vaddq_f32 ( Q_IN ## 1, Q_S3 ); \
Q_OUT ## 2 = vsubq_f32 ( Q_S2, Q_IN ## 5 ); \
Q_OUT ## 3 = vsubq_f32 ( Q_IN ## 0, Q_IN ## 4 ); \
Q_OUT ## 4 = vsubq_f32 ( Q_IN ## 6, Q_IN ## 2 ); \
Q_OUT ## 5 = vsubq_f32 ( Q_IN ## 1, Q_S3 ); \
Q_OUT ## 6 = vaddq_f32 ( Q_IN ## 5, Q_S2 ); \
Q_OUT ## 7 = vsubq_f32 ( Q_S0, Q_S1 ); \
} while(0);
NE10_DECLARE_4
#define NE10_DECLARE_4(TYPE, NAME)
Definition: NE10_fft.neonintrinsic.h:75

Definition at line 137 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX8x4_R2C_NEON_LOAD (   PTR_IN,
  Q_IN,
  IN_STEP 
)
Value:
do { \
Q_IN ## 0 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
Q_IN ## 1 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
Q_IN ## 2 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
Q_IN ## 3 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
Q_IN ## 4 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
Q_IN ## 5 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
Q_IN ## 6 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
Q_IN ## 7 = vld1q_f32( (ne10_float32_t*) ( PTR_IN ) ); \
PTR_IN += IN_STEP; \
} while(0);
ne10_float32_t
float ne10_float32_t
Definition: NE10_types.h:80

Definition at line 249 of file NE10_fft.neonintrinsic.h.

#define NE10_RADIX8x4_R2C_NEON_STORE (   PTR_OUT,
  Q_OUT,
  OUT_STEP 
)
Value:
do { \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 0 * OUT_STEP ), Q_OUT ## 0); \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 1 * OUT_STEP ), Q_OUT ## 1); \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 2 * OUT_STEP ), Q_OUT ## 2); \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 3 * OUT_STEP ), Q_OUT ## 3); \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 4 * OUT_STEP ), Q_OUT ## 4); \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 5 * OUT_STEP ), Q_OUT ## 5); \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 6 * OUT_STEP ), Q_OUT ## 6); \
vst1q_f32( (ne10_float32_t*) ( PTR_OUT + 7 * OUT_STEP ), Q_OUT ## 7); \
} while(0);
ne10_float32_t
float ne10_float32_t
Definition: NE10_types.h:80

Definition at line 279 of file NE10_fft.neonintrinsic.h.

#define NE10_REVERSE_FLOAT32X4 (   VECTOR4F)
Value:
do { \
VECTOR4F = vrev64q_f32(VECTOR4F); \
VECTOR4F = vcombine_f32( vget_high_f32( VECTOR4F ), vget_low_f32( VECTOR4F ) ); \
} while (0);

Definition at line 84 of file NE10_fft.neonintrinsic.h.

#define NE10_REVERSE_OUT_FLOAT32X4 (   VECTOR4F_OUT,
  VECTOR4F 
)
Value:
do { \
float32x4_t Q_TMP = vrev64q_f32(VECTOR4F); \
VECTOR4F_OUT = vcombine_f32( vget_high_f32( Q_TMP ), vget_low_f32( Q_TMP ) ); \
} while (0);

Definition at line 89 of file NE10_fft.neonintrinsic.h.

#define VDUPQ_N_F32 (   VAR)    { VAR, VAR, VAR, VAR }

Definition at line 110 of file NE10_fft.neonintrinsic.h.

Generated on Fri Jun 30 2017 10:50:54 for Project Ne10 by   doxygen 1.8.11