/*
 * Copyright (c) 2020 Raspberry Pi (Trading) Ltd.
 *
 * SPDX-License-Identifier: BSD-3-Clause
 */

#ifndef _PICO_FLOAT_H
#define _PICO_FLOAT_H

#include <math.h>
#include <float.h>
#include "pico.h"
#include "pico/bootrom/sf_table.h"

#ifdef __cplusplus
extern "C" {
#endif

/** \file float.h
* \defgroup pico_float pico_float
*
* \brief Optimized single-precision floating point functions
*
* An application can take control of the floating point routines used in the application over and above what is provided by the compiler,
* by depending on the pico_float library. A user might want to do this
*
* 1. To use optimized software implementations provided by the RP2-series device's bootrom or the SDK
* 2. To use optimized combined software/hardware implementations utilizing custom RP2-series hardware for acceleration
* 3. To control the amount of C compiler/library code bloat
* 4. To make sure no floating point is called at all
*
* The pico_float library comes in three main flavors:
*
* 1. `pico_float_none` - all floating point operations cause a \ref panic - no single-precision floating point code is included
* 2. `pico_float_compiler` - no custom functions are provided; all single-precision floating point is handled by the C compiler/library
* 3. `pico_float_pico` - the smallest and fastest available for the platform, along with additional functionality (e.g. fixed point conversions) which are detailed below
*
* The user can control which version they want (e.g. **pico_float_xxx** by either setting the CMake global variable
* `PICO_DEFAULT_FLOAT_IMPL=xxx`, or by using the CMake function `pico_set_float_implementation(<TARGET> xxx)`. Note that in the absence
* of either, pico_float_pico is used by default.
*
* \if rp2040_specific
* On RP2040, `pico_float_pico` uses optimized hand coded implementations from the bootrom and the SDK for both
* basic single-precision floating point operations and floating point math library functions. These implementations
* are generally faster and smaller than those provided by the C compiler/library, though they don't support all the features of a fully compliant
* floating point implementation; they are however usually fine for the majority of cases
* \endif
*
* \if rp2350_specific
* On Arm on RP2350, there are multiple options for `pico_float_pico`:
*
* 1. `pico_float_pico_vfp` - this library leaves basic C single-precision floating point operations to the compiler
* which can use inlined VFP (Arm FPU) code. Custom optimized versions of trigonometric and scientific functions are provided.
* No DCP (RP2350 Double co-processor) instructions are used.
* 2. `pico_float_pico_dcp` - this library prevents the compiler injecting inlined VFP code, and also implements
* all single-precision floating point operations in optimized DCP or M33 code. This option is not quite as fast
* as pico_float_pico_vfp, however it allows floating point operations without enabling the floating point co-processor
* on the CPU; this can be beneficial in certain circumstances, e.g. where leaving stack in tasks or interrupts
* for the floating point state is undesirable.
*
* Note: `pico_float_pico` is equivalent to `pico_float_pico_vfp` on RP2350, as this is the most sensible default
* \endif
*
* On Arm, (replacement) optimized implementations are provided for the following compiler built-ins
* and math library functions when using `_pico` variants of `pico_float`:
*
* - basic arithmetic: (except `pico_float_pico_vfp`)
*
*   __aeabi_fadd, __aeabi_fdiv, __aeabi_fmul, __aeabi_frsub, __aeabi_fsub
*
* - comparison: (except `pico_float_pico_vfp`)
*
*   __aeabi_cfcmpeq, __aeabi_cfrcmple, __aeabi_cfcmple, __aeabi_fcmpeq, __aeabi_fcmplt, __aeabi_fcmple, __aeabi_fcmpge, __aeabi_fcmpgt, __aeabi_fcmpun
*
* - (u)int32 <-> float: (except `pico_float_pico_vfp`)
*
*    __aeabi_i2f, __aeabi_ui2f, __aeabi_f2iz, __aeabi_f2uiz
*
* - (u)int64 <-> float: (except `pico_float_pico_vfp`)
*
*   __aeabi_l2f, __aeabi_ul2f, __aeabi_f2lz, __aeabi_f2ulz
*
* - float -> double: (except `pico_float_pico_vfp`)
*
*   __aeabi_f2d
*
* - basic trigonometric:
*
*   sqrtf, cosf, sinf, tanf, atan2f, expf, logf
*
* - trigonometric and scientific
*
*   ldexpf, copysignf, truncf, floorf, ceilf, roundf, asinf, acosf, atanf, sinhf, coshf, tanhf, asinhf, acoshf, atanhf, exp2f, log2f, exp10f, log10f, powf, hypotf, cbrtf, fmodf, dremf, remainderf, remquof, expm1f, log1pf, fmaf
*
* - GNU exetnsions:
*
*   powintf, sincosf
*
* On Arm, the following additional optimized functions are also provided (when using `_pico` variants of `pico_float`):
*
* - Conversions to/from integer types:
*
*   - (u)int -> float (round to nearest):
*
*     int2float, uint2float, int642float, uint642float
*
*     note: on `pico_float_pico_vfp` the 32-bit functions are also provided as C macros since they map to inline VFP code
*
*   - (u)float -> int (round towards zero):
*
*     float2int_z, float2uint_z, float2int64_z, float2uint64_z
*
*     note: on `pico_float_pico_vfp` the 32-bit functions are also provided as C macros since they map to inline VFP code
*
*   - (u)float -> int (round towards -infinity):
*
*     float2int, float2uint, float2int64, float2uint64
*
* - Conversions to/from fixed point integers:
*
*   - (u)fix -> float (round to nearest):
*
*       fix2float, ufix2float, fix642float, ufix642float
*
*   - float -> (u)fix (round towards zero):
*
*       float2fix_z, float2ufix_z, float2fix64_z, float2ufix64_z
*
*     note: on `pico_float_pico_vfp` the 32-bit functions are also provided as C macros since they can map to inline VFP code
*     when the number of fractional bits is a compile time constant between 1 and 32
*
*   - float -> (u)fix (round towards -infinity):
*
*       float2fix, float2ufix, float2fix64, float2ufix64
*
*     note: on `pico_float_pico_vfp` the 32-bit functions are also provided as C macros since they can map to inline VFP code
*     when the number of fractional bits is a compile time constant between 1 and 32
*
* - Even faster versions of divide and square-root functions that do not round correctly: (`pico_float_pico_dcp` only)
*
*   fdiv_fast, sqrtf_fast
*
* \if rp2350_specific
* On RISC-V, (replacement) optimized implementations are provided for the following compiler built-ins when using the `pico_float_pico`
* library (note that there are no variants of this library like there are on Arm):
*
* - basic arithmetic:
*
*   __addsf3, __subsf3, __mulsf3
* \endif
*/
#if !defined(__riscv) || PICO_COMBINED_DOCS

#if PICO_COMBINED_DOCS || !LIB_PICO_FLOAT_COMPILER
float int2float(int32_t i);
float uint2float(uint32_t i);
float int642float(int64_t i);
float uint642float(uint64_t i);
float fix2float(int32_t m, int e);
float ufix2float(uint32_t m, int e);
float fix642float(int64_t m, int e);
float ufix642float(uint64_t m, int e);

// These methods round towards 0, which IS the C way
int32_t float2int_z(float f);
int64_t float2int64_z(float f);
int32_t float2uint_z(float f);
int64_t float2uint64_z(float f);
int32_t float2fix_z(float f, int e);
uint32_t float2ufix_z(float f, int e);
int64_t float2fix64_z(float f, int e);
uint64_t float2ufix64_z(float f, int e);

// These methods round towards -Infinity - which IS NOT the C way for negative numbers;
// as such the naming is not ideal, however is kept for backwards compatibility
int32_t float2int(float f);
uint32_t float2uint(float f);
int64_t float2int64(float f);
uint64_t float2uint64(float f);
int32_t float2fix(float f, int e);
uint32_t float2ufix(float f, int e);
int64_t float2fix64(float f, int e);
uint64_t float2ufix64(float f, int e);

#if LIB_PICO_FLOAT_PICO_VFP
// a bit of a hack to inline VFP fixed point conversion when exponent is constant and in range 1-32
#define fix2float(m, e) __builtin_choose_expr(__builtin_constant_p(e), (e) >= 1 && (e) <= 32 ? _fix2float_inline(m, e) : fix2 ## float(m, e), fix2 ## float(m, e))
#define ufix2float(m, e) __builtin_choose_expr(__builtin_constant_p(e), (e) >= 1 && (e) <= 32 ? _ufix2float_inline(m, e) : ufix2 ## float(m, e), ufix2 ## float(m, e))
#define float2fix_z(f, e) __builtin_choose_expr(__builtin_constant_p(e), (e) >= 1 && (e) <= 32 ? _float2fix_z_inline(f, e) : float2 ## fix_z(f, e), float2 ## fix_z(f, e))
#define float2ufix_z(f, e) __builtin_choose_expr(__builtin_constant_p(e), (e) >= 1 && (e) <= 32 ? _float2ufix_z_inline(f, e) : float2 ## ufix_z(f, e), float2 ## ufix_z(f, e))
#define float2fix(f, e) __builtin_choose_expr(__builtin_constant_p(e), (e) >= 1 && (e) <= 32 ? _float2fix_inline(f, e) : float2 ## fix(f, e), float2 ## fix(f, e))
#define float2ufix(f, e) __builtin_choose_expr(__builtin_constant_p(e), (e) >= 1 && (e) <= 32 ? _float2ufix_inline(f, e) : float2 ## ufix(f, e), float2 ## ufix(f, e))

#define _fix2float_inline(m, e) ({ \
    int32_t _m = m; \
    float f; \
    pico_default_asm( \
        "vmov %0, %1\n" \
        "vcvt.f32.s32 %0, %0, %2\n" \
        : "=t" (f) \
        : "r" (_m), "i" (e) \
    ); \
    f; \
})
#define _ufix2float_inline(m, e) ({ \
    uint32_t _m = m; \
    float f; \
    pico_default_asm( \
        "vmov %0, %1\n" \
        "vcvt.f32.u32 %0, %0, %2\n" \
        : "=t" (f) \
        : "r" (_m), "i" (e) \
    ); \
    f; \
})
#define _float2fix_z_inline(f, e) ({ \
    int32_t _m; \
    float _f = (f); \
    pico_default_asm( \
        "vcvt.s32.f32 %0, %0, %2\n" \
        "vmov %1, %0\n" \
        : "+t" (_f), "=r" (_m) \
        : "i" (e) \
    ); \
    _m; \
})
#define _float2ufix_z_inline(f, e) ({ \
    uint32_t _m; \
    float _f = (f); \
    pico_default_asm( \
        "vcvt.u32.f32 %0, %0, %2\n" \
        "vmov %1, %0\n" \
        : "+t" (_f), "=r" (_m) \
        : "i" (e) \
    ); \
    _m; \
})
#define _float2fix_z_inline(f, e) ({ \
    int32_t _m; \
    float _f = (f); \
    pico_default_asm( \
        "vcvt.s32.f32 %0, %0, %2\n" \
        "vmov %1, %0\n" \
        : "+t" (_f), "=r" (_m) \
        : "i" (e) \
    ); \
    _m; \
})
#define _float2fix_inline(f, e) ({ \
    union { float _f; int32_t _i; } _u; \
    _u._f = (f); \
    uint rc, tmp; \
    pico_default_asm( \
        "vcvt.s32.f32 %0, %0, %4\n" \
        "vmov %2, %0\n" \
        "lsls %1, #1\n" \
        "bls 2f\n" /* positive or zero or -zero are ok with the result we have */ \
        "lsrs %3, %1, #24\n" \
        "subs %3, #0x7f - %c4\n" \
        "bcc 1f\n" /* 0 < abs(f) < 1 ^ e, so need to round down */ \
        /* mask off all but fractional bits */ \
        "lsls %1, %3\n" \
        "lsls %1, #8\n" \
        "beq 2f\n" /* integers can round towards zero */ \
        "1:\n" \
        /* need to subtract 1 from the result to round towards -infinity... */ \
        /* this will never cause an overflow, because to get here we must have had a non integer/infinite value which */ \
        /* therefore cannot have been equal to INT64_MIN when rounded towards zero */ \
        "subs %2, #1\n" \
        "2:\n" \
        : "+t" (_u._f), "+r" (_u._i), "=r" (rc), "=r" (tmp) \
        : "i" (e) \
    ); \
    rc; \
})
#define _float2ufix_inline(f, e) _float2ufix_z_inline((f), (e))
#endif

#if LIB_PICO_FLOAT_PICO_VFP
// may as well provide inline macros for VFP
#define int2float(i) ((float)(int32_t)(i))
#define uint2float(i) ((float)(uint32_t)(i))
#define float2int_z(f) ((int32_t)(f))
#define float2uint_z(f) ((uint32_t)(f))
#endif

#endif

float exp10f(float x);
void sincosf(float x, float *sinx, float *cosx);
float powintf(float x, int y);

#if !PICO_RP2040 || PICO_COMBINED_DOCS
float fdiv_fast(float n, float d);
float sqrtf_fast(float f);
#endif

#endif

#if defined(__riscv) || LIB_PICO_FLOAT_COMPILER
// when using the compiler or RISC-V, we provide as many functions as we trivially can - these will be efficient
// when using hard-float on Arm
static inline float int2float(int32_t i) { return (float)i; }
static inline float uint2float(uint32_t i) { return (float)i; }
static inline float int642float(int64_t i) { return (float)i; }
static inline float uint642float(uint64_t i) { return (float)i; }

static inline int32_t float2int_z(float f) { return (int32_t)f; }
static inline int64_t float2int64_z(float f) { return (int64_t)f; }
static inline int32_t float2uint_z(float f) { return (uint32_t)f; }
static inline int64_t float2uint64_z(float f) { return (uint64_t)f; }
#endif

#ifdef __cplusplus
}
#endif

#endif