neorv32_zfinx_extension_intrinsics.h

Go to the documentation of this file.

// ================================================================================ //
// The NEORV32 RISC-V Processor - https://github.com/stnolting/neorv32              //
// Copyright (c) NEORV32 contributors.                                              //
// Copyright (c) 2020 - 2025 Stephan Nolting. All rights reserved.                  //
// Licensed under the BSD-3-Clause license, see LICENSE for details.                //
// SPDX-License-Identifier: BSD-3-Clause                                            //
// ================================================================================ //
 
 
/**********************************************************************/
#ifndef NEORV32_ZFINX_EXTENSION_INTRINSICS_H
#define NEORV32_ZFINX_EXTENSION_INTRINSICS_H
 
#define __USE_GNU
#define _GNU_SOURCE
#include <float.h>
#include <math.h>
 
 
/**********************************************************************/
typedef union
{
  uint32_t binary_value; 
  float    float_value;  
} float_conv_t;
 
 
// ################################################################################################
// Helper functions
// ################################################################################################
 
/**********************************************************************/
float subnormal_flush(float tmp) {
 
  float res = tmp;
 
  // flush to zero if subnormal
  if (fpclassify(tmp) == FP_SUBNORMAL) {
    if (signbit(tmp) != 0) {
      res = -0.0f;
    }
    else {
      res = +0.0f;
    }
  }
 
  return res;
}
 
 
// ################################################################################################
// "Intrinsics"
// ################################################################################################
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fadds(float rs1, float rs2) {
 
  float_conv_t opa, opb, res;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  res.binary_value = CUSTOM_INSTR_R_TYPE(0b0000000, opb.binary_value, opa.binary_value, 0b000, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fsubs(float rs1, float rs2) {
 
  float_conv_t opa, opb, res;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  res.binary_value = CUSTOM_INSTR_R_TYPE(0b0000100, opb.binary_value, opa.binary_value, 0b000, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fmuls(float rs1, float rs2) {
 
  float_conv_t opa, opb, res;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  res.binary_value = CUSTOM_INSTR_R_TYPE(0b0001000, opb.binary_value, opa.binary_value, 0b000, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fmins(float rs1, float rs2) {
 
  float_conv_t opa, opb, res;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  res.binary_value = CUSTOM_INSTR_R_TYPE(0b0010100, opb.binary_value, opa.binary_value, 0b000, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fmaxs(float rs1, float rs2) {
 
  float_conv_t opa, opb, res;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  res.binary_value = CUSTOM_INSTR_R_TYPE(0b0010100, opb.binary_value, opa.binary_value, 0b001, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline uint32_t __attribute__ ((always_inline)) riscv_intrinsic_fcvt_wus(float rs1) {
 
  float_conv_t opa;
  opa.float_value = rs1;
 
  return CUSTOM_INSTR_I_TYPE(0b110000000001, opa.binary_value, 0b000, 0b1010011);
}
 
 
/**********************************************************************/
inline int32_t __attribute__ ((always_inline)) riscv_intrinsic_fcvt_ws(float rs1) {
 
  float_conv_t opa;
  opa.float_value = rs1;
 
  return (int32_t)CUSTOM_INSTR_I_TYPE(0b110000000000, opa.binary_value, 0b000, 0b1010011);
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fcvt_swu(uint32_t rs1) {
 
  float_conv_t res;
 
  res.binary_value = CUSTOM_INSTR_I_TYPE(0b110100000001, rs1, 0b000, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fcvt_sw(int32_t rs1) {
 
  float_conv_t res;
 
  res.binary_value = CUSTOM_INSTR_I_TYPE(0b110100000000, rs1, 0b000, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline uint32_t __attribute__ ((always_inline)) riscv_intrinsic_feqs(float rs1, float rs2) {
 
  float_conv_t opa, opb;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  return CUSTOM_INSTR_R_TYPE(0b1010000, opb.binary_value, opa.binary_value, 0b010, 0b1010011);
}
 
 
/**********************************************************************/
inline uint32_t __attribute__ ((always_inline)) riscv_intrinsic_flts(float rs1, float rs2) {
 
  float_conv_t opa, opb;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  return CUSTOM_INSTR_R_TYPE(0b1010000, opb.binary_value, opa.binary_value, 0b001, 0b1010011);
}
 
 
/**********************************************************************/
inline uint32_t __attribute__ ((always_inline)) riscv_intrinsic_fles(float rs1, float rs2) {
 
  float_conv_t opa, opb;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  return CUSTOM_INSTR_R_TYPE(0b1010000, opb.binary_value, opa.binary_value, 0b000, 0b1010011);
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fsgnjs(float rs1, float rs2) {
 
  float_conv_t opa, opb, res;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  res.binary_value = CUSTOM_INSTR_R_TYPE(0b0010000, opb.binary_value, opa.binary_value, 0b000, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fsgnjns(float rs1, float rs2) {
 
  float_conv_t opa, opb, res;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  res.binary_value = CUSTOM_INSTR_R_TYPE(0b0010000, opb.binary_value, opa.binary_value, 0b001, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fsgnjxs(float rs1, float rs2) {
 
  float_conv_t opa, opb, res;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  res.binary_value = CUSTOM_INSTR_R_TYPE(0b0010000, opb.binary_value, opa.binary_value, 0b010, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline uint32_t __attribute__ ((always_inline)) riscv_intrinsic_fclasss(float rs1) {
 
  float_conv_t opa;
  opa.float_value = rs1;
 
  return CUSTOM_INSTR_I_TYPE(0b111000000000, opa.binary_value, 0b001, 0b1010011);
}
 
 
// ################################################################################################
// !!! UNSUPPORTED instructions !!!
// ################################################################################################
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fdivs(float rs1, float rs2) {
 
  float_conv_t opa, opb, res;
  opa.float_value = rs1;
  opb.float_value = rs2;
 
  res.binary_value = CUSTOM_INSTR_R_TYPE(0b0001100, opb.binary_value, opa.binary_value, 0b000, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fsqrts(float rs1) {
 
  float_conv_t opa, res;
  opa.float_value = rs1;
 
  res.binary_value = CUSTOM_INSTR_I_TYPE(0b010110000000, opa.binary_value, 0b000, 0b1010011);
  return res.float_value;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fmadds(float rs1, float rs2, float rs3) {
 
  (void)rs1;
  (void)rs2;
  (void)rs3;
 
  asm volatile (".word 0x02000043");
 
  return 0;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fmsubs(float rs1, float rs2, float rs3) {
 
  (void)rs1;
  (void)rs2;
  (void)rs3;
 
  asm volatile (".word 0x02000047");
 
  return 0;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fnmsubs(float rs1, float rs2, float rs3) {
 
  (void)rs1;
  (void)rs2;
  (void)rs3;
 
  asm volatile (".word 0x0200004b");
 
  return 0;
}
 
 
/**********************************************************************/
inline float __attribute__ ((always_inline)) riscv_intrinsic_fnmadds(float rs1, float rs2, float rs3) {
 
  (void)rs1;
  (void)rs2;
  (void)rs3;
 
  asm volatile (".word 0x0200004f");
 
  return 0;
}
 
 
// ################################################################################################
// Emulation functions
// ################################################################################################
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fadds(float rs1, float rs2) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
 
  float res = opa + opb;
 
  // make NAN canonical
  if (fpclassify(res) == FP_NAN) {
    res = NAN;
  }
 
  return subnormal_flush(res);
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fsubs(float rs1, float rs2) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
 
  float res = opa - opb;
 
  // make NAN canonical
  if (fpclassify(res) == FP_NAN) {
    res = NAN;
  }
 
  return subnormal_flush(res);
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fmuls(float rs1, float rs2) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
 
  float res = opa * opb;
  return subnormal_flush(res);
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fmins(float rs1, float rs2) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
 
  union {
  uint32_t binary_value; 
  float    float_value;  
  } tmp_a, tmp_b;
 
  if ((fpclassify(opa) == FP_NAN) && (fpclassify(opb) == FP_NAN)) {
    return nanf("");
  }
 
  if (fpclassify(opa) == FP_NAN) {
    return opb;
  }
 
  if (fpclassify(opb) == FP_NAN) {
    return opa;
  }
 
  // RISC-V spec: -0 < +0
  tmp_a.float_value = opa;
  tmp_b.float_value = opb;
  if (((tmp_a.binary_value == 0x80000000) && (tmp_b.binary_value == 0x00000000)) ||
      ((tmp_a.binary_value == 0x00000000) && (tmp_b.binary_value == 0x80000000))) {
    return -0.0f;
  }
 
  return fmin(opa, opb);
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fmaxs(float rs1, float rs2) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
 
  union {
  uint32_t binary_value; 
  float    float_value;  
  } tmp_a, tmp_b;
 
 
  if ((fpclassify(opa) == FP_NAN) && (fpclassify(opb) == FP_NAN)) {
    return nanf("");
  }
 
  if (fpclassify(opa) == FP_NAN) {
    return opb;
  }
 
  if (fpclassify(opb) == FP_NAN) {
    return opa;
  }
 
  // RISC-V spec: -0 < +0
  tmp_a.float_value = opa;
  tmp_b.float_value = opb;
  if (((tmp_a.binary_value == 0x80000000) && (tmp_b.binary_value == 0x00000000)) ||
      ((tmp_a.binary_value == 0x00000000) && (tmp_b.binary_value == 0x80000000))) {
    return +0.0f;
  }
 
  return fmax(opa, opb);
}
 
 
/**********************************************************************/
uint32_t __attribute__ ((noinline)) riscv_emulate_fcvt_wus(float rs1) {
 
  float opa = subnormal_flush(rs1);
 
  return (uint32_t)rint(opa);
}
 
 
/**********************************************************************/
int32_t __attribute__ ((noinline)) riscv_emulate_fcvt_ws(float rs1) {
 
  float opa = subnormal_flush(rs1);
 
  return (int32_t)rint(opa);
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fcvt_swu(uint32_t rs1) {
 
  return (float)rs1;
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fcvt_sw(int32_t rs1) {
 
  return (float)rs1;
}
 
 
/**********************************************************************/
uint32_t __attribute__ ((noinline)) riscv_emulate_feqs(float rs1, float rs2) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
 
  if ((fpclassify(opa) == FP_NAN) || (fpclassify(opb) == FP_NAN)) {
    return 0;
  }
 
  if isless(opa, opb) {
    return 0;
  }
  else if isgreater(opa, opb) {
    return 0;
  }
  else {
    return 1;
  }
}
 
 
/**********************************************************************/
uint32_t __attribute__ ((noinline)) riscv_emulate_flts(float rs1, float rs2) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
 
  if ((fpclassify(opa) == FP_NAN) || (fpclassify(opb) == FP_NAN)) {
    return 0;
  }
 
  if isless(opa, opb) {
    return 1;
  }
  else {
    return 0;
  }
}
 
 
/**********************************************************************/
uint32_t __attribute__ ((noinline)) riscv_emulate_fles(float rs1, float rs2) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
 
  if ((fpclassify(opa) == FP_NAN) || (fpclassify(opb) == FP_NAN)) {
    return 0;
  }
 
  if islessequal(opa, opb) {
    return 1;
  }
  else {
    return 0;
  }
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fsgnjs(float rs1, float rs2) {
 
  float opa = rs1;
  float opb = rs2;
 
  int sign_1 = (int)signbit(opa);
  int sign_2 = (int)signbit(opb);
  float res = 0;
 
  if (sign_2 != 0) { // opb is negative
    if (sign_1 == 0) {
      res = -opa;
    }
    else {
      res = opa;
    }
  }
  else { // opb is positive
    if (sign_1 == 0) {
      res = opa;
    }
    else {
      res = -opa;
    }
  }
 
  return res;
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fsgnjns(float rs1, float rs2) {
 
  float opa = rs1;
  float opb = rs2;
 
  int sign_1 = (int)signbit(opa);
  int sign_2 = (int)signbit(opb);
  float res = 0;
 
  if (sign_2 != 0) { // opb is negative
    if (sign_1 == 0) {
      res = opa;
    }
    else {
      res = -opa;
    }
  }
  else { // opb is positive
    if (sign_1 == 0) {
      res = -opa;
    }
    else {
      res = opa;
    }
  }
 
  return res;
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fsgnjxs(float rs1, float rs2) {
 
  float opa = rs1;
  float opb = rs2;
 
  int sign_1 = (int)signbit(opa);
  int sign_2 = (int)signbit(opb);
  float res = 0;
 
  if (((sign_1 == 0) && (sign_2 != 0)) || ((sign_1 != 0) && (sign_2 == 0))) {
    if (sign_1 == 0) {
      res = -opa;
    }
    else {
      res = opa;
    }
  }
  else {
    if (sign_1 == 0) {
      res = opa;
    }
    else {
      res = -opa;
    }
  }
 
  return res;
}
 
 
/**********************************************************************/
uint32_t __attribute__ ((noinline)) riscv_emulate_fclasss(float rs1) {
 
  float opa = rs1;
 
  union {
    uint32_t binary_value; 
    float    float_value;  
  } aux;
 
  // RISC-V classify result layout
  const uint32_t CLASS_NEG_INF    = 1 << 0; // negative infinity
  const uint32_t CLASS_NEG_NORM   = 1 << 1; // negative normal number
  const uint32_t CLASS_NEG_DENORM = 1 << 2; // negative subnormal number
  const uint32_t CLASS_NEG_ZERO   = 1 << 3; // negative zero
  const uint32_t CLASS_POS_ZERO   = 1 << 4; // positive zero
  const uint32_t CLASS_POS_DENORM = 1 << 5; // positive subnormal number
  const uint32_t CLASS_POS_NORM   = 1 << 6; // positive normal number
  const uint32_t CLASS_POS_INF    = 1 << 7; // positive infinity
  const uint32_t CLASS_SNAN       = 1 << 8; // signaling NaN (sNaN)
  const uint32_t CLASS_QNAN       = 1 << 9; // quiet NaN (qNaN)
 
  int tmp = fpclassify(opa);
  int sgn = (int)signbit(opa);
 
  uint32_t res = 0;
 
  // infinity
  if (tmp == FP_INFINITE) {
    if (sgn) { res |= CLASS_NEG_INF; }
    else     { res |= CLASS_POS_INF; }
  }
 
  // zero
  if (tmp == FP_ZERO) {
    if (sgn) { res |= CLASS_NEG_ZERO; }
    else     { res |= CLASS_POS_ZERO; }
  }
 
  // normal
  if (tmp == FP_NORMAL) {
    if (sgn) { res |= CLASS_NEG_NORM; }
    else     { res |= CLASS_POS_NORM; }
  }
 
  // subnormal
  if (tmp == FP_SUBNORMAL) {
    if (sgn) { res |= CLASS_NEG_DENORM; }
    else     { res |= CLASS_POS_DENORM; }
  }
 
  // NaN
  if (tmp == FP_NAN) {
    aux.float_value = opa;
    if ((aux.binary_value >> 22) & 0b1) { // bit 22 (mantissa's MSB) is set -> canonical (quiet) NAN
      res |= CLASS_QNAN;
    }
    else {
      res |= CLASS_SNAN;
    }
  }
 
  return res;
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fdivs(float rs1, float rs2) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
 
  float res = opa / opb;
  return subnormal_flush(res);
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fsqrts(float rs1) {
 
  float opa = subnormal_flush(rs1);
 
  float res = sqrtf(opa);
  return subnormal_flush(res);
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fmadds(float rs1, float rs2, float rs3) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
  float opc = subnormal_flush(rs3);
 
  float res = (opa * opb) + opc;
  return subnormal_flush(res);
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fmsubs(float rs1, float rs2, float rs3) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
  float opc = subnormal_flush(rs3);
 
  float res = (opa * opb) - opc;
  return subnormal_flush(res);
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fnmsubs(float rs1, float rs2, float rs3) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
  float opc = subnormal_flush(rs3);
 
  float res = -(opa * opb) + opc;
  return subnormal_flush(res);
}
 
 
/**********************************************************************/
float __attribute__ ((noinline)) riscv_emulate_fnmadds(float rs1, float rs2, float rs3) {
 
  float opa = subnormal_flush(rs1);
  float opb = subnormal_flush(rs2);
  float opc = subnormal_flush(rs3);
 
  float res = -(opa * opb) - opc;
  return subnormal_flush(res);
}
 
 
#endif // NEORV32_ZFINX_EXTENSION_INTRINSICS_H