26381119807

Committed 25 May 2026 03:08AM UTC coverage: 84.02% (+0.01%) from 84.008%

Build # 26381119807

Build Type

Pull #993

github

Committed by

web-flow

Commit Message

Merge b7fbe7947 into 4648aaee3

Pull Request Pull Request #993: fix(erational): correct ieee754 conversion (both directions) + add tests (#986)

Coverage Stats

76 of 79 new or added lines in 1 file covered. (96.2%)

32 existing lines in 4 files now uncovered.

47426 of 56446 relevant lines covered (84.02%)

8745134.63 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

93.64

/include/sw/universal/number/cfloat/cfloat_impl.hpp

#pragma once
// cfloat_impl.hpp: implementation of an arbitrary configuration fixed-size 'classic' floating-point representation
// cfloat<> can emulate IEEE-754 floats and the new Deep Learning types, such as 
// IEEE-754 half-precision floats
// Google bfloat16
// NVIDIA TensorFloat 
// AMD FP16 and FP32
// Microsoft FP8 and FP9
// Tesla CFP8, CFP16
// 
// cfloat<> can also emulate more precise configurations, such as
// 80bit IEEE-754 extended precision floats
// true 128bit quad precision floats
//
// Copyright (C) 2017 Stillwater Supercomputing, Inc.
// SPDX-License-Identifier: MIT
//
// This file is part of the universal numbers project, which is released under an MIT Open Source license.

// compiler specific environment has been delegated to be handled by the
// number system include file <universal/number/cfloat/cfloat.hpp>
// 
// supporting types and functions
#include <cctype>
#include <limits>
#include <regex>
#include <sstream>
#include <string_view>
#include <type_traits>
#include <universal/native/ieee754.hpp>
#include <universal/native/subnormal.hpp>
#include <universal/utility/find_msb.hpp>
#include <universal/utility/decimal_to_binary.hpp>
#include <universal/number/shared/nan_encoding.hpp>
#include <universal/number/shared/infinite_encoding.hpp>
#include <universal/number/shared/specific_value_encoding.hpp>
// arithmetic tracing options
#include <universal/number/algorithm/trace_constants.hpp>
// cfloat exception structure
#include <universal/number/cfloat/exceptions.hpp>
// composition types used by cfloat
#include <universal/internal/blockbinary/blockbinary.hpp>
#include <universal/internal/blocktriple/blocktriple.hpp>
#include <universal/number/support/decimal.hpp>

#ifndef CFLOAT_THROW_ARITHMETIC_EXCEPTION
#define CFLOAT_THROW_ARITHMETIC_EXCEPTION 0
#endif

#ifndef TRACE_CONVERSION
#define TRACE_CONVERSION 0
#endif

namespace sw { namespace universal {

/*
 * classic floating-point cfloat offers subnormals, max-exponent values, and
 * saturation. The default configuration turns off subnormals, max-exponent
 * values, and projects values outside of their dynamic range to +-inf
 *
 * Behavior flags
 *   hasSubnormals   : gradual underflow: use all fraction encodings when exponent is all 0's
 *   hasMaxExpValues  : reclaim max-exponent encodings as numeric values instead of
 *                      reserving the entire all-1s exponent binade for inf/NaN;
 *                      only 4 encodings are reserved for +-inf and quiet/signalling NaN
 *   isSaturating     : saturate to maxneg or maxpos when value is out of dynamic range
 */

/// <summary>
/// decode an cfloat value into its constituent parts
/// </summary>
/// <typeparam name="bt">block type</typeparam>
/// <param name="v">cfloat value to decode (input: const ref)</param>
/// <param name="s">sign (output: bool ref)</param>
/// <param name="e">exponent (output: blockbinary ref)</param>
/// <param name="f">fraction (output: blockbinary ref)</param>
template<unsigned nbits, unsigned es, unsigned fbitsPlus1, typename bt,
        bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
void decode(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& v, bool& s, blockbinary<es, bt>& e, blockbinary<fbitsPlus1, bt>& f) {
        v.sign(s);
        v.exponent(e);
        v.fraction(f);
}

/// <summary>
/// return the binary scale of the given number
/// </summary>
/// <typeparam name="bt">Block type used for storage: derived through ADL</typeparam>
/// <param name="v">the cfloat number for which we seek to know the binary scale</param>
/// <returns>binary scale, i.e. 2^scale, of the value of the cfloat</returns>
template<unsigned nbits, unsigned es, typename bt,
        bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
int scale(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& v) {
        return v.scale();
}

/// <summary>
/// convert a blocktriple to a cfloat. blocktriples come out of the arithmetic
/// engine in the form ii.ff...ff and a scale. The conversion must take this
/// denormalized form into account during conversion.
/// 
/// The blocktriple must be in this form to round correctly, as all the bits
/// after an arithmetic operation must be taken into account.
/// 
/// Transformation:
///    ii.ff...ff  transform to    s.eee.fffff
/// All number systems that depend on blocktriple will need to have
/// the rounding decision answered, so that functionality can be
/// reused if we locate it inside blocktriple.
/// 
/// if (srcbits > fbits) // we need to round
///     if (ii.00..00 > 1) 
///         mask is at srcbits - fbits + 1
///     else 
///                    mask is at srcbits - fbits
/// }
/// else {               // no need to round
/// }
/// </summary>
/// <typeparam name="bt">type of the block used for cfloat storage</typeparam>
/// <param name="src">the blocktriple to be converted</param>
/// <param name="tgt">the resulting cfloat</param>
template<unsigned srcbits, BlockTripleOperator op, unsigned nbits, unsigned es, typename bt,
        bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline void convert(const blocktriple<srcbits, op, bt>& src, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& tgt) {
        using btType = blocktriple<srcbits, op, bt>;
        using cfloatType = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        // test special cases
        if (src.isnan()) {
                tgt.setnan(src.sign() ? NAN_TYPE_SIGNALLING : NAN_TYPE_QUIET);
        }
        else        if (src.isinf()) {
                tgt.setinf(src.sign());
        }
        else         if (src.iszero()) {
                tgt.setzero();
                tgt.setsign(src.sign()); // preserve sign
        }
        else {
                int significandScale = src.significandscale();
                int exponent = src.scale() + significandScale;
                // blocktriple keeps arithmetic results intentionally unnormalized. scale() tracks the source binade,
                // while significandScale() reports whether the exact result spilled into the integer headroom above
                // the radix. cfloat conversion has to combine both before it can classify the value as normal,
                // subnormal, or overflowing.
                // special case of underflow
                if constexpr (hasSubnormals) {
//                        std::cout << "exponent = " << exponent << " bias = " << cfloatType::EXP_BIAS << " exp subnormal = " << cfloatType::MIN_EXP_SUBNORMAL << '\n';
                        // why must exponent be less than (minExpSubnormal - 1) to be rounded to zero? 
                        // because the half-way value that would round up to minpos is at exp = (minExpSubnormal - 1)
                        if (exponent < cfloatType::MIN_EXP_SUBNORMAL) {
                                tgt.setzero();
                                if (exponent == (cfloatType::MIN_EXP_SUBNORMAL - 1)) {
                                        // -exponent because we are right shifting and exponent in this range is negative
                                        int adjustment = -(exponent + subnormal_reciprocal_shift[es]);
                                        std::pair<bool, unsigned> alignment = src.roundingDecision(adjustment);
                                        if (alignment.first) ++tgt; // we are minpos
                                }
                                tgt.setsign(src.sign());
                                return;
                        }
                }
                else {
                        if (exponent + cfloatType::EXP_BIAS <= 0) {  // value is in the subnormal range, which maps to 0
                                tgt.setzero();
                                tgt.setsign(src.sign());
                                return;
                        }
                }
                // special case of overflow
                if constexpr (hasMaxExpValues) {
                        if constexpr (isSaturating) {
                                if (exponent > cfloatType::MAX_EXP) {
                                        if (src.sign()) tgt.maxneg(); else tgt.maxpos();
                                        return;
                                }
                        }
                        else {
                                if (exponent > cfloatType::MAX_EXP) {
                                        tgt.setinf(src.sign());
                                        return;
                                }
                        }
                }
                else {  // no max-exponent values: will saturate at a different encoding: TODO can we hide it all in maxpos?
                        if constexpr (isSaturating) {
                                if (exponent > cfloatType::MAX_EXP) {
                                        if (src.sign()) tgt.maxneg(); else tgt.maxpos();
                                        return;
                                }
                        }
                        else {
                                if (exponent > cfloatType::MAX_EXP) {
                                        tgt.setinf(src.sign());
                                        return;
                                }
                        }
                }

                // our value needs to go through rounding to be correctly interpreted
                // 
                // tgt.clear();  // no need as all bits are going to be set by the code below

                // exponent construction
                int adjustment{ 0 };
                // construct exponent
                uint64_t biasedExponent = static_cast<uint64_t>(static_cast<long long>(exponent) + static_cast<long long>(cfloatType::EXP_BIAS)); // this is guaranteed to be positive if exponent in encoding range
//                        std::cout << "exponent         " << to_binary(biasedExponent) << '\n';        
                if constexpr (hasSubnormals) {
                        //if (exponent >= cfloatType::MIN_EXP_SUBNORMAL && exponent < cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>::MIN_EXP_NORMAL) {
                        if (exponent < cfloatType::MIN_EXP_NORMAL) {
                                // the value is in the subnormal range of the cfloat
                                biasedExponent = 0;
                                // -exponent because we are right shifting and exponent in this range is negative
                                adjustment = -(exponent + subnormal_reciprocal_shift[es]);
                                // this is the right shift adjustment required for subnormal representation due 
                                // to the scale of the input number, i.e. the exponent of 2^-adjustment
                        }
                        else {
                                // the value is in the normal range of the cfloat
                                biasedExponent = static_cast<uint64_t>(static_cast<long long>(exponent) + static_cast<long long>(cfloatType::EXP_BIAS)); // this is guaranteed to be positive
                        }
                }
                else {
                        if (exponent < cfloatType::MIN_EXP_NORMAL) biasedExponent = 1ull; // fixup biasedExponent if we are in the subnormal region
                }


                // get the rounding direction and the LSB right shift: 
                std::pair<bool, unsigned> alignment = src.roundingDecision(adjustment);
                unsigned rightShift = alignment.second;  // this is the shift to get the LSB of the src to the LSB of the tgt
                // rightShift is also the normalization step: it lines the blocktriple hidden bit up with the cfloat
                // hidden-bit position, while the bool tells us whether the discarded tail rounds the retained payload up.
                //std::cout << "rightShift       " << rightShift << '\n';

                if constexpr (btType::bfbits < 65) {
                        bool roundup = alignment.first;
                        //std::cout << "round-up?        " << (roundup ? "yes" : "no") << '\n';
                        // we can use a uint64_t to construct the cfloat
                        uint64_t raw = (src.sign() ? 1ull : 0ull); // process sign
                        //std::cout << "raw bits (sign)  " << to_binary(raw) << '\n';
                        // construct the fraction bits
                        uint64_t fracbits = src.significand_ull(); // get all the bits, including the integer bits
                        //std::cout << "fracbits         " << to_binary(fracbits) << '\n';
                        fracbits >>= rightShift;
                        //std::cout << "fracbits shifted " << to_binary(fracbits) << '\n';
                        fracbits &= cfloatType::ALL_ONES_FR; // remove the hidden bit
                        //std::cout << "fracbits masked  " << to_binary(fracbits) << '\n';
                        if (roundup) ++fracbits;
                        // Rounding can turn 1.111... into 10.000..., so the carry has to be reinterpreted as an exponent bump.
                        // This is the last place where the unrounded blocktriple state can still change the target binade.
                        if (fracbits == (1ull << cfloatType::fbits)) { // check for overflow
                                if (biasedExponent == cfloatType::ALL_ONES_ES) {
                                        fracbits = cfloatType::INF_ENCODING; // project to INF
                                }
                                else {
                                        ++biasedExponent;
                                        fracbits = 0;
                                }
                        }

                        raw <<= es; // shift sign to make room for the exponent bits
                        raw |= biasedExponent;
                        //std::cout << "raw bits (exp)   " << to_binary(raw) << '\n';
                        raw <<= cfloatType::fbits; // make room for the fraction bits
                        //std::cout << "raw bits (s+exp) " << to_binary(raw) << '\n';
                        raw |= fracbits;
                        //std::cout << "raw bits (final) " << to_binary(raw) << '\n';
                        tgt.setbits(raw);
//                        std::cout << "raw bits (all)   " << to_binary(raw) << '\n';
                        if constexpr (isSaturating) {
                                if (tgt.isnan()) {
                                        if (src.sign()) {
                                                tgt.maxneg();        // map back to maxneg
                                        }
                                        else {
                                                tgt.maxpos();        // map back to maxpos
                                        }
                                }
                        }
                        else {
                                // when you get too far, map it back to +-inf: 
                                // TBD: this doesn't appear to be the right algorithm to catch all overflow patterns
                                if (tgt.isnan()) tgt.setinf(src.sign());        // map back to +-inf
                        }
                }
                else {
                        // compose the segments
                        bool roundup = alignment.first;
                        auto fracbits = src.significand();
                        fracbits >>= static_cast<int>(rightShift);
                        fracbits.setbit(cfloatType::fbits, false); // remove the hidden bit (matches narrow path: fracbits &= ALL_ONES_FR)

                        // apply rounding (matches the bfbits < 65 path above)
                        if (roundup) fracbits.increment();

                        // check for rounding overflow: if the fraction carried into
                        // the hidden-bit position, clear the fraction and bump the exponent
                        if (fracbits.test(cfloatType::fbits)) {
                                fracbits.clear();
                                if (biasedExponent == cfloatType::ALL_ONES_ES) {
                                        // rounding overflowed beyond the maximum binade:
                                        // handle explicitly because the isnan() remap below
                                        // does not catch all configurations (e.g. hasMaxExpValues
                                        // with zero fraction is a valid finite encoding, not NaN)
                                        if constexpr (isSaturating) {
                                                if (src.sign()) tgt.maxneg(); else tgt.maxpos();
                                        }
                                        else {
                                                tgt.setinf(src.sign());
                                        }
                                        return;
                                }
                                else {
                                        ++biasedExponent;
                                        ++exponent;
                                }
                        }

                        // copy the blocks that contain fraction bits
                        tgt.clear();
                        for (unsigned b = 0; b < btType::nrBlocks; ++b) {
                                tgt.setblock(b, fracbits.block(b));
                        }
                        tgt.setsign(src.sign());
                        if (!tgt.setexponent(exponent)) {
                                // std::cerr is not constexpr-callable; gate the diagnostic on
                                // runtime context only. The constant-evaluator silently drops
                                // the diagnostic but the enclosing branch is rare in practice
                                // (only triggered when blocktriple bfbits >= 65 and the
                                // computed exponent doesn't fit the cfloat config).
                                if (!std::is_constant_evaluated()) {
                                        std::cerr << "exponent value is out of range: " << exponent << '\n';
                                }
                        }

                        // saturation / overflow-to-inf handling (matches the bfbits < 65 path)
                        if constexpr (isSaturating) {
                                if (tgt.isnan()) {
                                        if (src.sign()) {
                                                tgt.maxneg();
                                        }
                                        else {
                                                tgt.maxpos();
                                        }
                                }
                        }
                        else {
                                if (tgt.isnan()) tgt.setinf(src.sign());
                        }
                }
        }
}


/// <summary>
/// An arbitrary, fixed-size floating-point number with configurable gradual under/overflow and saturation/non-saturation arithmetic.
/// Default configuration offers normal encoding and non-saturating arithmetic.
/// /// </summary>
/// <typeparam name="nbits">number of bits in the encoding</typeparam>
/// <typeparam name="es">number of exponent bits in the encoding</typeparam>
/// <typeparam name="bt">the type to use as storage class: one of [uint8_t|uint16_t|uint32_t]</typeparam>
/// <typeparam name="hasSubnormals">configure gradual underflow (==subnormals)</typeparam>
/// <typeparam name="hasMaxExpValues">reclaim max-exponent encodings as numeric values</typeparam>
/// <typeparam name="isSaturating">configure saturation arithmetic</typeparam>
template<unsigned _nbits, unsigned _es, typename bt = uint8_t,
        bool _hasSubnormals = false, bool _hasMaxExpValues = false, bool _isSaturating = false>
class cfloat {
public:
        static_assert(_nbits > _es + 1ull, "nbits is too small to accomodate the requested number of exponent bits");
        static_assert(_es < 21ull, "my God that is a big number, are you trying to break the Interweb?");
        static_assert(_es > 0, "number of exponent bits must be bigger than 0 to be a classic floating point number");
        // how do you assert on the condition that if es == 1 then subnormals and max-exponent values must be true?
        static constexpr bool     subsuper = (_hasSubnormals && _hasMaxExpValues);
        static constexpr bool     special = (subsuper ? true : (_es > 1));
        static_assert(special, "when es == 1, cfloat must have both subnormals and max-exponent values");
        static constexpr unsigned bitsInByte = 8u;
        static constexpr unsigned bitsInBlock = sizeof(bt) * bitsInByte;
        static_assert(bitsInBlock <= 64, "storage unit for block arithmetic needs to be <= uint64_t"); // TODO: carry propagation on uint64_t requires assembly code

        static constexpr unsigned nbits = _nbits;
        static constexpr unsigned es = _es;
        static constexpr unsigned fbits  = nbits - 1u - es;    // number of fraction bits excluding the hidden bit
        static constexpr unsigned fhbits = nbits - es;           // number of fraction bits including the hidden bit

        static constexpr uint64_t  storageMask = (0xFFFFFFFFFFFFFFFFull >> (64ull - bitsInBlock));
        static constexpr bt       BLOCK_MASK = bt(~0);
        static constexpr bt       ALL_ONES = bt(~0); // block type specific all 1's value
        static constexpr uint32_t ALL_ONES_ES = (0xFFFF'FFFFul >> (32 - es));
        static constexpr uint64_t topfbits = fbits % 64;
        static constexpr uint64_t FR_SHIFT = (topfbits > 0 ? (64 - topfbits) : 0);
        static constexpr uint64_t ALL_ONES_FR = (topfbits > 0 ? (0xFFFF'FFFF'FFFF'FFFFull >> FR_SHIFT) : 0ull); // special case for nbits <= 64
        static constexpr uint64_t INF_ENCODING = (ALL_ONES_FR & ~1ull);

        static constexpr unsigned nrBlocks = 1u + ((nbits - 1ull) / bitsInBlock);
        static constexpr unsigned MSU = nrBlocks - 1u; // MSU == Most Significant Unit, as MSB is already taken
        static constexpr bt       MSU_MASK = (ALL_ONES >> (nrBlocks * bitsInBlock - nbits));
        static constexpr unsigned bitsInMSU = bitsInBlock - (nrBlocks * bitsInBlock - nbits);
        static constexpr unsigned fBlocks = 1ull + ((fbits - 1ull) / bitsInBlock); // nr of blocks with fraction bits
        static constexpr unsigned FSU = fBlocks - 1u;  // FSU = Fraction Significant Unit: the index of the block that contains the most significant fraction bits
        static constexpr bt       FSU_MASK = (ALL_ONES >> (fBlocks * bitsInBlock - fbits));
        static constexpr unsigned bitsInFSU = bitsInBlock - (fBlocks * bitsInBlock - fbits);

        static constexpr bt       SIGN_BIT_MASK = bt(bt(1ull) << ((nbits - 1ull) % bitsInBlock));
        static constexpr bt       LSB_BIT_MASK = bt(1ull);
        static constexpr bool     MSU_CAPTURES_EXP = (1ull + es) <= bitsInMSU;
        static constexpr unsigned EXP_SHIFT = (MSU_CAPTURES_EXP ? (1 == nrBlocks ? (nbits - 1ull - es) : (bitsInMSU - 1ull - es)) : 0);
        static constexpr bt       MSU_EXP_MASK = ((ALL_ONES << EXP_SHIFT) & ~SIGN_BIT_MASK) & MSU_MASK;
        static constexpr int      EXP_BIAS = ((1 << (es - 1u)) - 1l);
        static constexpr int      MAX_EXP = (es == 1) ? 1 : ((1 << es) - EXP_BIAS - 1);
        static constexpr int      MIN_EXP_NORMAL = 1 - EXP_BIAS;
        static constexpr int      MIN_EXP_SUBNORMAL = 1 - EXP_BIAS - int(fbits); // the scale of smallest ULP

        static constexpr bool     hasSubnormals   = _hasSubnormals;
        static constexpr bool     hasMaxExpValues = _hasMaxExpValues;
        static constexpr bool     isSaturating    = _isSaturating;
        typedef bt BlockType;

        // constructors
        cfloat() = default;
        cfloat(const cfloat&) = default;
        cfloat& operator=(const cfloat&) = default;

        // construct a cfloat from another
        template<unsigned nnbits, unsigned ees, typename bbt, bool ssub, bool ssup, bool ssat>
        cfloat(const cfloat<nnbits, ees, bbt, ssub, ssup, ssat>& rhs) noexcept : _block{} {
                if (rhs.isnan()) {
                        setnan(rhs.sign() ? NAN_TYPE_SIGNALLING : NAN_TYPE_QUIET);
                }
                else if (rhs.isinf()) {
                        setinf(rhs.sign());
                }
                else if (rhs.iszero()) {
                        setzero();
                }
                else {
                        if constexpr (std::is_same_v<bt, bbt>) {
                                // Full-precision path: create a blocktriple sized for the TARGET's
                                // fraction width, pre-align the source fraction bits into it (with
                                // proper rounding-bit collection), then let convert() handle
                                // rounding and encoding.
                                constexpr unsigned srcFbits = nnbits - 1u - ees;
                                using TgtBT = blocktriple<fbits, BlockTripleOperator::ADD, bt>;
                                constexpr unsigned addRbits = TgtBT::rbits;
                                TgtBT value;

                                bool srcSign = rhs.sign();
                                int srcScale = rhs.scale();

                                if constexpr (srcFbits < 64 && (fbits + addRbits) < 64) {
                                        // Fast path: fits in uint64_t
                                        // Guard requires (fbits + addRbits) < 64 to prevent UB from
                                        // left-shift of amount (fbits + addRbits - srcFbits) >= 64.
                                        uint64_t srcRaw = rhs.fraction_ull();
                                        if (rhs.isnormal() || rhs.ismaxexpvalue()) {
                                                // Normal and max-exponent values both have a hidden bit
                                                srcRaw |= (1ull << srcFbits);
                                        }
                                        else {
                                                // Normalize subnormal: find leading 1, shift to hidden-bit position
                                                using SrcCfloat = cfloat<nnbits, ees, bbt, ssub, ssup, ssat>;
                                                srcScale = SrcCfloat::MIN_EXP_NORMAL;
                                                for (int i = static_cast<int>(srcFbits) - 1; i >= 0; --i) {
                                                        if (srcRaw & (1ull << i)) {
                                                                unsigned shift = srcFbits - static_cast<unsigned>(i);
                                                                srcScale -= static_cast<int>(shift);
                                                                srcRaw <<= shift;
                                                                break;
                                                        }
                                                }
                                        }
                                        // Align source hidden bit (at srcFbits) to target position (fbits + addRbits)
                                        uint64_t tgtRaw;
                                        if constexpr (srcFbits >= fbits + addRbits) {
                                                constexpr unsigned rshift = srcFbits - fbits - addRbits;
                                                if constexpr (rshift > 0) {
                                                        uint64_t stickyMask = (1ull << rshift) - 1;
                                                        bool sticky = (srcRaw & stickyMask) != 0;
                                                        tgtRaw = srcRaw >> rshift;
                                                        if (sticky) tgtRaw |= 1; // fold discarded bits into sticky
                                                }
                                                else {
                                                        tgtRaw = srcRaw;
                                                }
                                        }
                                        else {
                                                constexpr unsigned lshift = fbits + addRbits - srcFbits;
                                                tgtRaw = srcRaw << lshift;
                                        }
                                        value.setsign(srcSign);
                                        value.setscale(srcScale);
                                        value.setbits(tgtRaw);
                                }
                                else {
                                        // Wide path: bit-by-bit copy with alignment
                                        int offset = static_cast<int>(fbits + addRbits) - static_cast<int>(srcFbits);
                                        // Initialize radix via setbits, then overwrite significand
                                        value.setbits(0);
                                        bool sticky = false;
                                        for (unsigned i = 0; i < srcFbits; ++i) {
                                                bool v = rhs.at(i);
                                                if (!v) continue;
                                                int tgtBit = static_cast<int>(i) + offset;
                                                if (tgtBit >= 0 && tgtBit < static_cast<int>(TgtBT::bfbits)) {
                                                        value.setbit(static_cast<unsigned>(tgtBit), true);
                                                }
                                                else if (tgtBit < 0) {
                                                        sticky = true;
                                                }
                                        }
                                        if (sticky) value.setbit(0, true);
                                        if (rhs.isnormal() || rhs.ismaxexpvalue()) {
                                                // Normal and max-exponent values both have a hidden bit
                                                value.setbit(static_cast<unsigned>(fbits + addRbits), true);
                                        }
                                        else {
                                                // Normalize subnormal for wide path.
                                                // The bit-copy loop above placed source bit i at target position
                                                // i + offset = i + (fbits + addRbits - srcFbits).  The blocktriple
                                                // scale represents the exponent such that:
                                                //   value = 2^scale * significant * 2^(-fbits-addRbits)
                                                // For the subnormal with leading 1 at source bit i, the significant
                                                // integer has its leading 1 at position i + fbits + addRbits - srcFbits,
                                                // giving value = 2^scale * 2^(i - srcFbits) * 1.xxx.
                                                // The true subnormal value is 2^MIN_EXP_NORMAL * 2^(i - srcFbits) * 1.xxx,
                                                // so scale = MIN_EXP_NORMAL with no further adjustment.
                                                using SrcCfloat = cfloat<nnbits, ees, bbt, ssub, ssup, ssat>;
                                                srcScale = SrcCfloat::MIN_EXP_NORMAL;
                                                // (No srcScale -= shift: the offset alignment already accounts for
                                                //  the bit position; subtracting shift would apply it twice.)
                                        }
                                        value.setnormal(); // clear _zero flag set by setbits(0)
                                        value.setsign(srcSign);
                                        value.setscale(srcScale);
                                }
                                convert(value, *this);
                        }
                        else {
                                // Cross-block-type: marshall through double.
                                // Safe only when the source fraction fits in double's 52-bit significand.
                                constexpr unsigned srcFbitsXB = nnbits - ees - 1u;
                                static_assert(srcFbitsXB <= 52u,
                                        "cfloat converting constructor: source and target have different "
                                        "block types and source has more fraction bits than double can represent; "
                                        "use matching block types for full-precision cross-config conversion");
                                *this = double(rhs);
                        }
                }
        }

        // converting constructors
        constexpr cfloat(const std::string& stringRep) : _block{} { assign(stringRep); }
        // specific value constructor
        constexpr cfloat(const SpecificValue code) noexcept : _block{} {
                switch (code) {
                case SpecificValue::maxpos:
                        maxpos();
                        break;
                case SpecificValue::minpos:
                        minpos();
                        break;
                case SpecificValue::zero:
                default:
                        zero();
                        break;
                case SpecificValue::minneg:
                        minneg();
                        break;
                case SpecificValue::maxneg:
                        maxneg();
                        break;
                case SpecificValue::infpos:
                        setinf(false);
                        break;
                case SpecificValue::infneg:
                        setinf(true);
                        break;
                case SpecificValue::nar: // approximation as cfloats don't have a NaR
                case SpecificValue::qnan:
                        setnan(NAN_TYPE_QUIET);
                        break;
                case SpecificValue::snan:
                        setnan(NAN_TYPE_SIGNALLING);
                        break;
                }
        }

        constexpr cfloat(signed char iv)                    noexcept : _block{} { *this = iv; }
        constexpr cfloat(short iv)                          noexcept : _block{} { *this = iv; }
        constexpr cfloat(int iv)                            noexcept : _block{} { *this = iv; }
        constexpr cfloat(long iv)                           noexcept : _block{} { *this = iv; }
        constexpr cfloat(long long iv)                      noexcept : _block{} { *this = iv; }
        constexpr cfloat(char iv)                           noexcept : _block{} { *this = iv; }
        constexpr cfloat(unsigned short iv)                 noexcept : _block{} { *this = iv; }
        constexpr cfloat(unsigned int iv)                   noexcept : _block{} { *this = iv; }
        constexpr cfloat(unsigned long iv)                  noexcept : _block{} { *this = iv; }
        constexpr cfloat(unsigned long long iv)             noexcept : _block{} { *this = iv; }
        CONSTEXPRESSION cfloat(float iv)                    noexcept : _block{} { *this = iv; }
        CONSTEXPRESSION cfloat(double iv)                   noexcept : _block{} { *this = iv; }

        // assignment operators
        constexpr cfloat& operator=(signed char rhs)        noexcept { return convert_signed_integer(rhs); }
        constexpr cfloat& operator=(short rhs)              noexcept { return convert_signed_integer(rhs); }
        constexpr cfloat& operator=(int rhs)                noexcept { return convert_signed_integer(rhs); }
        constexpr cfloat& operator=(long rhs)               noexcept { return convert_signed_integer(rhs); }
        constexpr cfloat& operator=(long long rhs)          noexcept { return convert_signed_integer(rhs); }

        constexpr cfloat& operator=(char rhs)               noexcept { return convert_unsigned_integer(rhs); }
        constexpr cfloat& operator=(unsigned short rhs)     noexcept { return convert_unsigned_integer(rhs); }
        constexpr cfloat& operator=(unsigned int rhs)       noexcept { return convert_unsigned_integer(rhs); }
        constexpr cfloat& operator=(unsigned long rhs)      noexcept { return convert_unsigned_integer(rhs); }
        constexpr cfloat& operator=(unsigned long long rhs) noexcept { return convert_unsigned_integer(rhs); }

        BIT_CAST_CONSTEXPR cfloat& operator=(float rhs)     noexcept { return convert_ieee754(rhs); }
        BIT_CAST_CONSTEXPR cfloat& operator=(double rhs)    noexcept { return convert_ieee754(rhs); }

        // make conversions to native types explicit
        explicit constexpr operator int()                       const noexcept { return to_int(); }
        explicit constexpr operator long()                      const noexcept { return to_long(); }
        explicit constexpr operator long long()                 const noexcept { return to_long_long(); }
        explicit constexpr operator float()                     const noexcept { return to_native<float>(); }
        explicit constexpr operator double()                    const noexcept { return to_native<double>(); }

        // guard long double support to enable ARM and RISC-V embedded environments
#if LONG_DOUBLE_SUPPORT
        explicit constexpr operator long double()               const noexcept { return to_native<long double>(); }
        BIT_CAST_CONSTEXPR cfloat(long double iv)           noexcept : _block{} { *this = iv; }
        BIT_CAST_CONSTEXPR cfloat& operator=(long double rhs)  noexcept { return convert_ieee754(rhs); }
#endif

        // arithmetic operators
        // prefix operator
        constexpr inline cfloat operator-() const {
                cfloat tmp(*this);
                tmp._block[MSU] ^= SIGN_BIT_MASK;
                return tmp;
        }

        constexpr cfloat& operator+=(const cfloat& rhs) CFLOAT_EXCEPT {
                if constexpr (_trace_add) std::cout << "---------------------- ADD -------------------" << std::endl;
                // special case handling of the inputs
#if CFLOAT_THROW_ARITHMETIC_EXCEPTION
                if (isnan(NAN_TYPE_SIGNALLING) || rhs.isnan(NAN_TYPE_SIGNALLING)) {
                        throw cfloat_operand_is_nan{};
                }
#else
                if (isnan(NAN_TYPE_SIGNALLING) || rhs.isnan(NAN_TYPE_SIGNALLING)) {
                        setnan(NAN_TYPE_SIGNALLING);
                        return *this;
                }
                if (isnan(NAN_TYPE_QUIET) || rhs.isnan(NAN_TYPE_QUIET)) {
                        setnan(NAN_TYPE_QUIET);
                        return *this;
                }
#endif
                // normal + inf  = inf
                // normal + -inf = -inf
                // inf + normal = inf
                // inf + inf    = inf
                // inf + -inf    = ?
                // -inf + normal = -inf
                // -inf + -inf   = -inf
                // -inf + inf    = ?
                if (isinf()) {
                        if (rhs.isinf()) {
                                if (sign() != rhs.sign()) {
                                        setnan(NAN_TYPE_SIGNALLING);
                                }
                                return *this;
                        }
                        else {
                                return *this;
                        }
                }
                else {
                        if (rhs.isinf()) {
                                *this = rhs;
                                return *this;
                        }
                }

                if (iszero()) {
                        *this = rhs;
                        return *this;
                }
                if (rhs.iszero()) return *this;

                // arithmetic operation
                blocktriple<fbits, BlockTripleOperator::ADD, bt> a, b, sum;

                // transform the inputs into (sign,scale,significant) 
                normalizeAddition(a); 
                rhs.normalizeAddition(b);
                sum.add(a, b);

                convert(sum, *this);

                return *this;
        }
        constexpr cfloat& operator+=(double rhs) CFLOAT_EXCEPT {
                return *this += cfloat(rhs);
        }
        constexpr cfloat& operator-=(const cfloat& rhs) CFLOAT_EXCEPT {
                if constexpr (_trace_sub) std::cout << "---------------------- SUB -------------------" << std::endl;
                if (rhs.isnan()) 
                        return *this += rhs;
                else 
                        return *this += -rhs;
        }
        constexpr cfloat& operator-=(double rhs) CFLOAT_EXCEPT {
                return *this -= cfloat(rhs);
        }
        constexpr cfloat& operator*=(const cfloat& rhs) CFLOAT_EXCEPT {
                if constexpr (_trace_mul) std::cout << "---------------------- MUL -------------------\n";
                // special case handling of the inputs
#if CFLOAT_THROW_ARITHMETIC_EXCEPTION
                if (isnan(NAN_TYPE_SIGNALLING) || rhs.isnan(NAN_TYPE_SIGNALLING)) {
                        throw cfloat_operand_is_nan{};
                }
#else
                if (isnan(NAN_TYPE_SIGNALLING) || rhs.isnan(NAN_TYPE_SIGNALLING)) {
                        setnan(NAN_TYPE_SIGNALLING);
                        return *this;
                }
                if (isnan(NAN_TYPE_QUIET) || rhs.isnan(NAN_TYPE_QUIET)) {
                        setnan(NAN_TYPE_QUIET);
                        return *this;
                }
#endif
                //  inf * inf = inf
                //  inf * -inf = -inf
                // -inf * inf = -inf
                // -inf * -inf = inf
                //        0 * inf = -nan(ind)
                //        inf * 0 = -nan(ind)
                bool resultSign = sign() != rhs.sign();
                if (isinf()) {
                        if (rhs.iszero()) {
                                setnan(NAN_TYPE_QUIET);
                        }
                        else {
                                setsign(resultSign);
                        }
                        return *this;
                }
                if (rhs.isinf()) {
                        if (iszero()) {
                                setnan(NAN_TYPE_QUIET);
                        }
                        else {
                                setinf(resultSign);
                        }
                        return *this;
                }

                if (iszero() || rhs.iszero()) {                        
                        setzero();
                        setsign(resultSign); // deal with negative 0
                        return *this;
                }

                // arithmetic operation
                blocktriple<fbits, BlockTripleOperator::MUL, bt> a, b, product;

                // transform the inputs into (sign,scale,significant) 
                // triples of the correct width
                normalizeMultiplication(a);
                rhs.normalizeMultiplication(b);
                product.mul(a, b);
                convert(product, *this);

                if constexpr (_trace_mul) std::cout << to_binary(a) << " : " << a << " *\n" << to_binary(b) << " : " << b << " =\n" << to_binary(product) << " : " << product << '\n';

                return *this;
        }
        constexpr cfloat& operator*=(double rhs) CFLOAT_EXCEPT {
                return *this *= cfloat(rhs);
        }
        constexpr cfloat& operator/=(const cfloat& rhs) CFLOAT_EXCEPT {
                if constexpr (_trace_div) std::cout << "---------------------- DIV -------------------" << std::endl;

                // special case handling of the inputs
                // qnan / qnan = qnan
                // qnan / snan = qnan
                // snan / qnan = snan
                // snan / snan = snan
#if CFLOAT_THROW_ARITHMETIC_EXCEPTION
                if (rhs.iszero()) throw cfloat_divide_by_zero();
                if (rhs.isnan()) throw cfloat_divide_by_nan();
                if (isnan()) throw cfloat_operand_is_nan();
#else
                if (isnan(NAN_TYPE_SIGNALLING) || rhs.isnan(NAN_TYPE_SIGNALLING)) {
                        setnan(NAN_TYPE_SIGNALLING);
                        return *this;
                }
                if (isnan(NAN_TYPE_QUIET) || rhs.isnan(NAN_TYPE_QUIET)) {
                        setnan(NAN_TYPE_QUIET);
                        return *this;
                }
                if (rhs.iszero()) {
                        if (iszero()) {
                                // zero divide by zero yields quiet NaN (in MSVC it is labeled -nan(ind) for indeterminate)
                                setnan(NAN_TYPE_QUIET);
                        }
                        else {
                                // non-zero divide by zero yields INF
                                bool resultSign = sign() != rhs.sign();
                                setinf(resultSign);
                        }
                        return *this;
                }
#endif
                //  inf /  inf = -nan(ind)
                //  inf / -inf = -nan(ind)
                // -inf /  inf = -nan(ind)
                // -inf / -inf = -nan(ind)
                //        1.0 /  inf = 0
                bool resultSign = sign() != rhs.sign();
                if (isinf()) {
                        if (rhs.isinf()) {
                                // inf divide by inf yields quiet NaN (in MSVC it is labeled -nan(ind) for indeterminate)
                                setnan(NAN_TYPE_QUIET);
                                return *this;
                        }
                        else {
                                // we stay an infinite but may change sign
                                setsign(resultSign);
                                return *this;
                        }
                }
                else {
                        if (rhs.isinf()) {
                                setzero();
                                setsign(resultSign);
                                return *this;
                        }
                }

                if (iszero()) {
                        setzero();
                        setsign(resultSign); // deal with negative 0
                        return *this;
                }

                // arithmetic operation
                using BlockTriple = blocktriple<fbits, BlockTripleOperator::DIV, bt>;
                BlockTriple a, b, quotient;

                // transform the inputs into (sign,scale,significant) 
                // triples of the correct width
                normalizeDivision(a);
                rhs.normalizeDivision(b);
                quotient.div(a, b);
                quotient.setradix(BlockTriple::radix);
                convert(quotient, *this);

                if constexpr (_trace_div) std::cout << to_binary(a) << " : " << a << " /\n" << to_binary(b) << " : " << b << " =\n" << to_binary(quotient) << " : " << quotient << '\n';

                return *this;
        }
        constexpr cfloat& operator/=(double rhs) CFLOAT_EXCEPT {
                return *this /= cfloat(rhs);
        }
        constexpr cfloat& reciprocal() CFLOAT_EXCEPT {
                cfloat c = 1.0 / *this;
                return *this = c;
        }
        /// <summary>
        /// move to the next bit encoding modulo 2^nbits
        /// </summary>
        /// <typeparam name="bt"></typeparam>
        constexpr cfloat& operator++() {
                if constexpr (0 == nrBlocks) {
                        return *this;
                }
                else if constexpr (1 == nrBlocks) {
                        if (sign()) {
                                if (_block[MSU] == (SIGN_BIT_MASK | 1ul)) { // pattern: 1.00.001 = minneg
                                        _block[MSU] = 0; // pattern: 0.00.000 = +0 
                                }
                                else {
                                        --_block[MSU];
                                }
                                if constexpr (!hasSubnormals) {
                                        if (isdenormal()) {
                                                // special case, we need to jump past all the subnormal value encodings which puts us on 0
                                                _block[MSU] = 0; // pattern: 0.00.000 = +0
                                        }
                                }
                        }
                        else {
                                if constexpr (!hasSubnormals) {
                                        if (_block[MSU] == 0) {
                                                // special case, we need to jump past all the subnormal value encodings minus 1
                                                setfraction(0xFFFF'FFFF'FFFF'FFFFull);
                                        }
                                }
                                if ((_block[MSU] & (MSU_MASK >> 1)) == (MSU_MASK >> 1)) { // pattern: 0.11.111 = nan
                                        _block[MSU] |= SIGN_BIT_MASK; // pattern: 1.11.111 = snan : wrap to the other side of the encoding
                                }
                                else {
                                        ++_block[MSU];
                                }
                        }
                }
                else {
                        if (sign()) {
                                // special case: pattern: 1.00.001 = minneg transitions to pattern: 0.00.000 = +0 
                                if (isminnegencoding()) {
                                        setzero();
                                }
                                else {
                                        //  1111 0000
                                        //  1110 1111
                                        bool borrow = true;
                                        for (unsigned i = 0; i < MSU; ++i) {
                                                if (borrow) {
                                                        if ((_block[i] & storageMask) == 0) { // block will underflow
                                                                --_block[i];
                                                                borrow = true;
                                                        }
                                                        else {
                                                                --_block[i];
                                                                borrow = false;
                                                        }
                                                }
                                        }
                                        if (borrow) {
                                                --_block[MSU];
                                        }
                                        if constexpr (!hasSubnormals) {
                                                if (isdenormal()) {
                                                        // special case, we need to jump past all the subnormal value encodings which puts us on 0
                                                        setzero(); // pattern: 0.00.000 = +0
                                                }
                                        }
                                }
                        }
                        else {
                                // special case: pattern: 0.11.111 = nan transitions to pattern: 1.11.111 = snan 
                                if (isnanencoding()) {
                                        setnan(NAN_TYPE_SIGNALLING);
                                }
                                else {
                                        if constexpr (!hasSubnormals) {
                                                if (iszero()) {
                                                        // special case, we need to jump past all the subnormal value encodings minus 1 so that the increment code below ends up on normal minpos
                                                        setfraction(0xFFFF'FFFF'FFFF'FFFFull);
                                                }
                                        }
                                        bool carry = true;
                                        for (unsigned i = 0; i < MSU; ++i) {
                                                if (carry) {
                                                        if ((_block[i] & storageMask) == storageMask) { // block will overflow
                                                                _block[i] = 0;
                                                                carry = true;
                                                        }
                                                        else {
                                                                ++_block[i];
                                                                carry = false;
                                                        }
                                                }
                                        }
                                        if (carry) {
                                                ++_block[MSU];
                                        }
                                }
                        }
                }
                return *this;
        }
        constexpr cfloat operator++(int) {
                cfloat tmp(*this);
                operator++();
                return tmp;
        }
        constexpr cfloat& operator--() {
                if constexpr (0 == nrBlocks) {
                        return *this;
                }
                else if constexpr (1 == nrBlocks) {
                        if (sign()) {
                                ++_block[MSU];
                        }
                        else {
                                // positive range
                                if (_block[MSU] == 0) { // pattern: 0.00.000 = 0
                                        if constexpr (hasSubnormals) {
                                                _block[MSU] |= SIGN_BIT_MASK | bt(1u); // pattern: 1.00.001 = minneg 
                                        }
                                        else {
                                                // special case, we need to jump past all the subnormal value encodings
                                                setfraction(0xFFFF'FFFF'FFFF'FFFFull); // set to 0.00.11...11
                                                ++_block[MSU]; // increment into 0.01.0000
                                                _block[MSU] |= SIGN_BIT_MASK; // set to 1.01.0000
                                        }
                                }
                                else {
                                        --_block[MSU];
                                }
                                if constexpr (!hasSubnormals) {
                                        if (isdenormal()) {
                                                // special case, we need to jump past all the subnormal value encodings which puts us on 0
                                                _block[MSU] = 0; // pattern: 0.00.000 = +0
                                        }
                                }
                        }
                }
                else {
                        if (sign()) {
                                bool carry = true;
                                for (unsigned i = 0; i < MSU; ++i) {
                                        if (carry) {
                                                if ((_block[i] & storageMask) == storageMask) { // block will overflow
                                                        _block[i] = 0;
                                                        carry = true;
                                                }
                                                else {
                                                        ++_block[i];
                                                        carry = false;
                                                }
                                        }
                                }
                                if (carry) {
                                        ++_block[MSU];
                                }
                        }
                        else {
                                // special case: pattern: 0.00.000 = +0 transitions to pattern: 1.00.001 = minneg
                                if (iszeroencoding()) {
                                        if constexpr (hasSubnormals) {
                                                setsign(true);
                                                setbit(0, true);
                                        }
                                        else {
                                                // special case, we need to jump past all the subnormal value encodings 1.01.0000 = minneg normal
                                                setexponent(1 - EXP_BIAS);
                                                setsign(true);
                                        }
                                }
                                else {
                                        bool borrow = true;
                                        for (unsigned i = 0; i < MSU; ++i) {
                                                if (borrow) {
                                                        if ((_block[i] & storageMask) == 0) { // block will underflow
                                                                --_block[i];
                                                                borrow = true;
                                                        }
                                                        else {
                                                                --_block[i];
                                                                borrow = false;
                                                        }
                                                }
                                        }
                                        if (borrow) {
                                                --_block[MSU];
                                        }
                                        if constexpr (!hasSubnormals) {
                                                if (isdenormal()) {
                                                        // special case, we need to jump past all the subnormal value encodings which puts us on 0
                                                        setzero(); // pattern: 0.00.000 = +0
                                                }
                                        }
                                }
                        }
                }
                return *this;
        }
        constexpr cfloat operator--(int) {
                cfloat tmp(*this);
                operator--();
                return tmp;
        }

        // modifiers        
        constexpr void clear() noexcept {
                if constexpr (0 == nrBlocks) {
                        return;
                }
                else if constexpr (1 == nrBlocks) {
                        _block[0] = bt(0);
                }
                else if constexpr (2 == nrBlocks) {
                        _block[0] = bt(0);
                        _block[1] = bt(0);
                }
                else if constexpr (3 == nrBlocks) {
                        _block[0] = bt(0);
                        _block[1] = bt(0);
                        _block[2] = bt(0);
                }
                else if constexpr (4 == nrBlocks) {
                        _block[0] = bt(0);
                        _block[1] = bt(0);
                        _block[2] = bt(0);
                        _block[3] = bt(0);
                }
                else if constexpr (5 == nrBlocks) {
                        _block[0] = bt(0);
                        _block[1] = bt(0);
                        _block[2] = bt(0);
                        _block[3] = bt(0);
                        _block[4] = bt(0);
                }
                else if constexpr (6 == nrBlocks) {
                        _block[0] = bt(0);
                        _block[1] = bt(0);
                        _block[2] = bt(0);
                        _block[3] = bt(0);
                        _block[4] = bt(0);
                        _block[5] = bt(0);
                }
                else if constexpr (7 == nrBlocks) {
                        _block[0] = bt(0);
                        _block[1] = bt(0);
                        _block[2] = bt(0);
                        _block[3] = bt(0);
                        _block[4] = bt(0);
                        _block[5] = bt(0);
                        _block[6] = bt(0);
                }
                else if constexpr (8 == nrBlocks) {
                        _block[0] = bt(0);
                        _block[1] = bt(0);
                        _block[2] = bt(0);
                        _block[3] = bt(0);
                        _block[4] = bt(0);
                        _block[5] = bt(0);
                        _block[6] = bt(0);
                        _block[7] = bt(0);
                }
                else {
                        for (unsigned i = 0; i < nrBlocks; ++i) {
                                _block[i] = bt(0);
                        }
                }
        }
        constexpr void setzero() noexcept { clear(); }
        constexpr void setinf(bool sign = true) noexcept {
                // the Inf encoding is the pattern 0b0'11...11'11...10 for a +inf, and 0b1'11...11'11...110 for a -inf
                if constexpr (0 == nrBlocks) {
                        return;
                }
                else if constexpr (1 == nrBlocks) {
                        _block[MSU] = sign ? bt(MSU_MASK ^ LSB_BIT_MASK) : bt(~SIGN_BIT_MASK & (MSU_MASK ^ LSB_BIT_MASK));
                }
                else if constexpr (2 == nrBlocks) {
                        _block[0] = BLOCK_MASK ^ LSB_BIT_MASK;
                        _block[MSU] = sign ? MSU_MASK : bt(~SIGN_BIT_MASK & MSU_MASK);
                }
                else if constexpr (3 == nrBlocks) {
                        _block[0] = BLOCK_MASK ^ LSB_BIT_MASK;
                        _block[1] = BLOCK_MASK;
                        _block[MSU] = sign ? MSU_MASK : bt(~SIGN_BIT_MASK & MSU_MASK);
                }
                else if constexpr (4 == nrBlocks) {
                        _block[0] = BLOCK_MASK ^ LSB_BIT_MASK;
                        _block[1] = BLOCK_MASK;
                        _block[2] = BLOCK_MASK;
                        _block[MSU] = sign ? MSU_MASK : bt(~SIGN_BIT_MASK & MSU_MASK);
                }
                else if constexpr (5 == nrBlocks) {
                        _block[0] = BLOCK_MASK ^ LSB_BIT_MASK;
                        _block[1] = BLOCK_MASK;
                        _block[2] = BLOCK_MASK;
                        _block[3] = BLOCK_MASK;
                        _block[MSU] = sign ? MSU_MASK : bt(~SIGN_BIT_MASK & MSU_MASK);
                }
                else if constexpr (6 == nrBlocks) {
                        _block[0] = BLOCK_MASK ^ LSB_BIT_MASK;
                        _block[1] = BLOCK_MASK;
                        _block[2] = BLOCK_MASK;
                        _block[3] = BLOCK_MASK;
                        _block[4] = BLOCK_MASK;
                        _block[MSU] = sign ? MSU_MASK : bt(~SIGN_BIT_MASK & MSU_MASK);
                }
                else if constexpr (7 == nrBlocks) {
                        _block[0] = BLOCK_MASK ^ LSB_BIT_MASK;
                        _block[1] = BLOCK_MASK;
                        _block[2] = BLOCK_MASK;
                        _block[3] = BLOCK_MASK;
                        _block[4] = BLOCK_MASK;
                        _block[5] = BLOCK_MASK;
                        _block[MSU] = sign ? MSU_MASK : bt(~SIGN_BIT_MASK & MSU_MASK);
                }
                else if constexpr (8 == nrBlocks) {
                        _block[0] = BLOCK_MASK ^ LSB_BIT_MASK;
                        _block[1] = BLOCK_MASK;
                        _block[2] = BLOCK_MASK;
                        _block[3] = BLOCK_MASK;
                        _block[4] = BLOCK_MASK;
                        _block[5] = BLOCK_MASK;
                        _block[6] = BLOCK_MASK;
                        _block[MSU] = sign ? MSU_MASK : bt(~SIGN_BIT_MASK & MSU_MASK);
                }
                else {
                        _block[0] = BLOCK_MASK ^ LSB_BIT_MASK;
                        for (unsigned i = 1; i < nrBlocks - 1; ++i) {
                                _block[i] = BLOCK_MASK;
                        }
                        _block[MSU] = sign ? MSU_MASK : bt(~SIGN_BIT_MASK & MSU_MASK);
                }        
        }
        constexpr void setnan(int NaNType = NAN_TYPE_SIGNALLING) noexcept {
                // the NaN encoding is the pattern 0b0'11...11'11...11 for a quiet Nan, and 0b1'11...11'11...111 for a signalling NaN
                if constexpr (0 == nrBlocks) {
                        return;
                }
                else if constexpr (1 == nrBlocks) {
                        // fall through
                }
                else if constexpr (2 == nrBlocks) {
                        _block[0] = BLOCK_MASK;
                }
                else if constexpr (3 == nrBlocks) {
                        _block[0] = BLOCK_MASK;
                        _block[1] = BLOCK_MASK;
                }
                else if constexpr (4 == nrBlocks) {
                        _block[0] = BLOCK_MASK;
                        _block[1] = BLOCK_MASK;
                        _block[2] = BLOCK_MASK;
                }
                else if constexpr (5 == nrBlocks) {
                        _block[0] = BLOCK_MASK;
                        _block[1] = BLOCK_MASK;
                        _block[2] = BLOCK_MASK;
                        _block[3] = BLOCK_MASK;
                }
                else if constexpr (6 == nrBlocks) {
                        _block[0] = BLOCK_MASK;
                        _block[1] = BLOCK_MASK;
                        _block[2] = BLOCK_MASK;
                        _block[3] = BLOCK_MASK;
                        _block[4] = BLOCK_MASK;
                }
                else if constexpr (7 == nrBlocks) {
                        _block[0] = BLOCK_MASK;
                        _block[1] = BLOCK_MASK;
                        _block[2] = BLOCK_MASK;
                        _block[3] = BLOCK_MASK;
                        _block[4] = BLOCK_MASK;
                        _block[5] = BLOCK_MASK;
                }
                else if constexpr (8 == nrBlocks) {
                        _block[0] = BLOCK_MASK;
                        _block[1] = BLOCK_MASK;
                        _block[2] = BLOCK_MASK;
                        _block[3] = BLOCK_MASK;
                        _block[4] = BLOCK_MASK;
                        _block[5] = BLOCK_MASK;
                        _block[6] = BLOCK_MASK;
                }
                else {
                        for (unsigned i = 0; i < nrBlocks - 1; ++i) {
                                _block[i] = BLOCK_MASK;
                        }
                }
                _block[MSU] = NaNType == NAN_TYPE_SIGNALLING ? MSU_MASK : bt(~SIGN_BIT_MASK & MSU_MASK);
        }
        constexpr void setsign(bool sign = true) {
                if (sign) {
                        _block[MSU] |= SIGN_BIT_MASK;
                }
                else {
                        _block[MSU] &= ~SIGN_BIT_MASK;
                }
        }
        constexpr bool setexponent(int scale) {
                if (scale < MIN_EXP_SUBNORMAL || scale > MAX_EXP) return false; // this scale cannot be represented
                if constexpr (nbits < 65) {
                        uint32_t exponentBits = static_cast<uint32_t>(scale + EXP_BIAS);
                        if (scale >= MIN_EXP_SUBNORMAL && scale < MIN_EXP_NORMAL) {
                                // setexponent() only writes the encoding field. In the subnormal interval the true scale is carried
                                // by a zero exponent field plus a left-shifted fraction, so callers must have already denormalized
                                // the significand before asking for this exponent pattern.
                                // we are a subnormal number: all exponent bits are 0
                                exponentBits = 0;
                        }
                        // TODO: optimize
                        uint32_t mask = (1ul << (es - 1));
                        for (unsigned i = nbits - 2; i > nbits - 2 - es; --i) {
                                setbit(i, (mask & exponentBits));
                                mask >>= 1;
                        }
                }
                else {
                        uint32_t exponentBits = static_cast<uint32_t>(scale + EXP_BIAS);
                        uint32_t mask = (1ul << (es - 1));
                        for (unsigned i = nbits - 2; i > nbits - 2 - es; --i) {
                                setbit(i, (mask & exponentBits));
                                mask >>= 1;
                        }
                }
                return true;
        }
        constexpr void setfraction(uint64_t raw_bits) {
                // unoptimized as it is not meant to be an end-user API, it is a test API
                // raw_bits is uint64_t so can have at most 64 bits of fraction data
                constexpr unsigned bitsToSet = (fbits < 64) ? fbits : 64;
                uint64_t mask{ 1ull };
                for (unsigned i = 0; i < bitsToSet; ++i) {
                        setbit(i, (mask & raw_bits));
                        mask <<= 1;
                }
        }
        constexpr void setbit(unsigned i, bool v = true) noexcept {
                unsigned blockIndex = i / bitsInBlock;
                if (blockIndex < nrBlocks) {
                        bt block = _block[blockIndex];
                        bt null = ~(1ull << (i % bitsInBlock));
                        bt bit = bt(v ? 1 : 0);
                        bt mask = bt(bit << (i % bitsInBlock));
                        _block[blockIndex] = bt((block & null) | mask);
                }
        }
        constexpr cfloat& setbits(uint64_t raw_bits) noexcept {
                if constexpr (0 == nrBlocks) {
                        return *this;
                }
                else if constexpr (1 == nrBlocks) {
                        _block[0] = raw_bits & storageMask;
                }
                else if constexpr (2 == nrBlocks) {
                        if constexpr (bitsInBlock < 64) {
                                _block[0] = raw_bits & storageMask;
                                raw_bits >>= bitsInBlock;
                                _block[1] = raw_bits & storageMask;
                        }
                        else {
                                _block[0] = raw_bits & storageMask;
                                _block[1] = 0;
                        }
                }
                else if constexpr (3 == nrBlocks) {
                        if constexpr (bitsInBlock < 64) {
                                _block[0] = raw_bits & storageMask;
                                raw_bits >>= bitsInBlock;
                                _block[1] = raw_bits & storageMask;
                                raw_bits >>= bitsInBlock;
                                _block[2] = raw_bits & storageMask;
                        }
                        else {
                                _block[0] = raw_bits & storageMask;
                                _block[1] = 0;
                                _block[2] = 0;
                        }
                }
                else if constexpr (4 == nrBlocks) {
                        if constexpr (bitsInBlock < 64) {
                                _block[0] = raw_bits & storageMask;
                                raw_bits >>= bitsInBlock;
                                _block[1] = raw_bits & storageMask;
                                raw_bits >>= bitsInBlock;
                                _block[2] = raw_bits & storageMask;
                                raw_bits >>= bitsInBlock;
                                _block[3] = raw_bits & storageMask;
                        }
                        else {
                                _block[0] = raw_bits & storageMask;
                                _block[1] = 0;
                                _block[2] = 0;
                                _block[3] = 0;
                        }
                }
                else {
                        if constexpr (bitsInBlock < 64) {
                                for (unsigned i = 0; i < nrBlocks; ++i) {
                                        _block[i] = raw_bits & storageMask;
                                        raw_bits >>= bitsInBlock;
                                }
                        }
                        else {
                                _block[0] = raw_bits & storageMask;
                                for (unsigned i = 1; i < nrBlocks; ++i) {
                                        _block[i] = 0;
                                }
                        }
                }
                _block[MSU] &= MSU_MASK; // enforce precondition for fast comparison by properly nulling bits that are outside of nbits
                return *this;
        }
        constexpr void setblock(unsigned b, bt data) noexcept {
                if (b < nrBlocks) _block[b] = data;
        }
        
        // create specific number system values of interest
        constexpr cfloat& maxpos() noexcept {
                if constexpr (isSaturating) {
                        // in a saturating encoding with max-exponent values we are removing the Inf encoding pattern 0b0'11...11'11...10 for a +inf, 
                        // and 0b1'11...11'11...110 for a -inf and using it as a value
                        if constexpr (hasMaxExpValues) {
                                // maximum positive value has this bit pattern: 0-1...1-111...110, that is, sign = 0, e = 11..11, f = 111...110
                                clear();
                                flip();
                                setbit(nbits - 1ull, false); // sign = 0
                                setbit(0ull, false); // bit0 = 0
                        }
                        else {
                                // maximum positive value has this bit pattern: 0-11...10-111...111, that is, sign = 0, e = 11..10, f = 111...111
                                clear();
                                flip();
                                setbit(fbits, false); // set least significant exponent bit to 0
                                setbit(nbits - 1ull, false); // set sign to 0
                        }
                }
                else {
                        // the Inf encoding is the pattern 0b0'11...11'11...10 for a +inf, and 0b1'11...11'11...110 for a -inf
                        // the maxpos is the encoding before that
                        if constexpr (hasMaxExpValues) {
                                // maximum positive value has this bit pattern: 0-1...1-111...101, that is, sign = 0, e = 11..11, f = 111...101
                                clear();
                                flip();
                                setbit(nbits - 1ull, false); // sign = 0
                                setbit(1ull, false); // bit1 = 0
                        }
                        else {
                                // maximum positive value has this bit pattern: 0-1...0-111...111, that is, sign = 0, e = 11..10, f = 111...111
                                clear();
                                flip();
                                setbit(fbits, false); // set least significant exponent bit to 0
                                setbit(nbits - 1ull, false); // set sign to 0
                        }
                }
                return *this;
        }
        constexpr cfloat& minpos() noexcept {
                // minpos encoding is not impacted by saturating encodings, which only affects maxpos and inf
                if constexpr (hasSubnormals) {
                        // minimum positive value has this bit pattern: 0-000-00...01, that is, sign = 0, e = 000, f = 00001
                        clear();
                        setbit(0);
                }
                else {
                        // minimum positive value has this bit pattern: 0-001-00...0, that is, sign = 0, e = 001, f = 0000
                        clear();
                        setbit(fbits);
                }
                return *this;
        }
        constexpr cfloat& zero() noexcept {
                // the zero value
                clear();
                return *this;
        }
        constexpr cfloat& minneg() noexcept {
                // minneg encoding is not impacted by saturating encodings, which only affects maxpos and inf
                if constexpr (hasSubnormals) {
                        // minimum negative value has this bit pattern: 1-000-00...01, that is, sign = 1, e = 00, f = 00001
                        clear();
                        setbit(nbits - 1ull);
                        setbit(0);
                }
                else {
                        // minimum negative value has this bit pattern: 1-001-00...0, that is, sign = 1, e = 001, f = 0000
                        clear();
                        setbit(fbits);
                        setbit(nbits - 1ull);
                }
                return *this;
        }
        constexpr cfloat& maxneg() noexcept {
                if constexpr (isSaturating) {
                        // in a saturating encoding with max-exponent values we are removing the Inf encoding pattern 0b0'11...11'11...10 for a +inf, 
                        // and 0b1'11...11'11...110 for a -inf and using it as a value
                        if constexpr (hasMaxExpValues) {
                                // maximum negative value has this bit pattern: 1-1...1-111...110, that is, sign = 1, e = 1..1, f = 111...110
                                clear();
                                flip();
                                setbit(0ull, false);
                        }
                        else {
                                // maximum negative value has this bit pattern: 1-1...0-111...111, that is, sign = 1, e = 11..10, f = 111...111
                                clear();
                                flip();
                                setbit(fbits, false);
                        }
                }
                else {
                        if constexpr (hasMaxExpValues) {
                                // maximum negative value has this bit pattern: 1-1...1-111...101, that is, sign = 1, e = 1..1, f = 111...101
                                clear();
                                flip();
                                setbit(1ull, false);
                        }
                        else {
                                // maximum negative value has this bit pattern: 1-1...0-111...111, that is, sign = 1, e = 11..10, f = 111...111
                                clear();
                                flip();
                                setbit(fbits, false);
                        }
                }
                return *this;
        }


        /// <summary>
        /// assign the value of the string representation to the cfloat
        /// </summary>
        /// <param name="stringRep">decimal scientific notation of a real number to be assigned</param>
        /// <returns>reference to this cfloat</returns>
        /// Clang doesn't support constexpr yet on string manipulations, so we need to make it conditional
        CONSTEXPRESSION cfloat& assign(const std::string& str) noexcept {
                clear();
                unsigned nrChars = static_cast<unsigned>(str.size());
                unsigned nrBits = 0;
                unsigned nrDots = 0;
                std::string bits;
                if (nrChars > 2) {
                        if (str[0] == '0' && str[1] == 'b') {
                                for (unsigned i = 2; i < nrChars; ++i) {
                                        char c = str[i];
                                        switch (c) {
                                        case '0':
                                        case '1':
                                                ++nrBits;
                                                bits += c;
                                                break;
                                        case '.':
                                                ++nrDots;
                                                bits += c;
                                                break;
                                        case '\'':
                                                // consume this delimiting character
                                                break;
                                        default:
                                                std::cerr << "string contained a non-standard character: " << c << '\n';
                                                return *this;
                                        }
                                }
                        }
                        else {
                                std::cerr << "string must start with 0b: instead input pattern was " << str << '\n';
                                return *this;
                        }
                }
                else {
                        std::cerr << "string is too short\n";
                        return *this;
                }

                if (nrBits != nbits) {
                        std::cerr << "number of bits in the string is " << nrBits << " and needs to be " << nbits << '\n';
                        return *this;
                }
                if (nrDots != 2) {
                        std::cerr << "number of segment delimiters in string is " << nrDots << " and needs to be 2 for a cfloat<>\n";
                        return *this;
                }

                // assign the bits
                int field{ 0 };  // three fields: sign, exponent, mantissa: fields are separated by a '.'
                int nrExponentBits{ -1 };
                unsigned bit = nrBits;
                for (unsigned i = 0; i < bits.size(); ++i) {
                        char c = bits[i];
                        if (c == '.') {
                                ++field;
                                if (field == 2) { // just finished parsing exponent field: we can now check the number of exponent bits
                                        if (nrExponentBits != es) {
                                                std::cerr << "provided binary string representation does not contain " << es << " exponent bits. Found " << nrExponentBits << ". Reset to 0\n";
                                                clear();
                                                return *this;
                                        }
                                }
                        }
                        else {
                                setbit(--bit, c == '1');
                        }
                        if (field == 1) { // exponent field
                                ++nrExponentBits;
                        }
                }
                if (field != 2) {
                        std::cerr << "provided binary string did not contain three fields separated by '.': Reset to 0\n";
                        clear();
                        return *this;
                }
                return *this;
        }

        // selectors
        constexpr bool sign() const noexcept { return (_block[MSU] & SIGN_BIT_MASK) == SIGN_BIT_MASK; }
        constexpr int  scale() const noexcept {
                int e{ 0 };
                if constexpr (MSU_CAPTURES_EXP) {
                        e = static_cast<int>((_block[MSU] & ~SIGN_BIT_MASK) >> EXP_SHIFT);
                        if (e == 0) {
                                // subnormal scale is determined by fraction
                                // subnormals: (-1)^s * 2^(2-2^(es-1)) * (f/2^fbits))
                                e = (2l - (1l << (es - 1ull))) - 1;
                                if constexpr (nbits > 2 + es) {
                                        for (unsigned i = nbits - 2ull - es; i > 0; --i) {
                                                if (test(i)) break;
                                                --e;
                                        }
                                }
                        }
                        else {
                                e -= EXP_BIAS;
                        }
                }
                else {
                        blockbinary<es, bt> ebits{};
                        exponent(ebits);
                        if (ebits.iszero()) {
                                // subnormal scale is determined by fraction
                                // subnormals: (-1)^s * 2^(2-2^(es-1)) * (f/2^fbits))
                                e = (2l - (1l << (es - 1ull))) - 1;
                                if constexpr (nbits > 2 + es) {
                                        for (unsigned i = nbits - 2ull - es; i > 0; --i) {
                                                if (test(i)) break;
                                                --e;
                                        }
                                }
                        }
                        else {
                                e = static_cast<int>(unsigned(ebits) - EXP_BIAS);
                        }
                }
                return e;
        }
        constexpr bool isneg() const noexcept {
                if (isnan()) return false;
                return sign(); 
        }
        constexpr bool ispos() const noexcept { 
                if (isnan()) return false;
                return !sign(); 
        }
        constexpr bool iszero() const noexcept {
                // NOTE: this is a very specific design that makes the decsion that
                // for subnormal encodings found in a configuration that doesn't
                // support them, we assume that these values map to 0.
                if constexpr (hasSubnormals) {
                        return iszeroencoding();
                }
                else { // all subnormals round to 0
                        blockbinary<es, bt> ebits{};
                        exponent(ebits);
                        if (ebits.iszero()) return true; else return false;
                }
        }
        constexpr bool isone() const noexcept {
                // unbiased exponent = scale = 0, fraction = 0
                int s = scale();
                if (s == 0) {
                        blockbinary<fbits, bt> f{};
                        fraction(f);
                        return f.iszero();
                }
                return false;
        }
        constexpr bool isinf(int InfType = INF_TYPE_EITHER) const noexcept {
                // the bit pattern encoding of Inf is independent of the max-exponent value configuration
                bool isNegInf = false;
                bool isPosInf = false;
                if constexpr (0 == nrBlocks) {
                        return false;
                }
                else if constexpr (1 == nrBlocks) {
                        isNegInf = (_block[MSU] & MSU_MASK) == (MSU_MASK ^ LSB_BIT_MASK);
                        isPosInf = (_block[MSU] & MSU_MASK) == ((MSU_MASK ^ SIGN_BIT_MASK) ^ LSB_BIT_MASK);
                }
                else if constexpr (2 == nrBlocks) {
                        bool isInf = (_block[0] == (BLOCK_MASK ^ LSB_BIT_MASK));
                        isNegInf = isInf && ((_block[MSU] & MSU_MASK) == MSU_MASK);
                        isPosInf = isInf && (_block[MSU] & MSU_MASK) == (MSU_MASK ^ SIGN_BIT_MASK);
                }
                else if constexpr (3 == nrBlocks) {
                        bool isInf = (_block[0] == (BLOCK_MASK ^ LSB_BIT_MASK)) && (_block[1] == BLOCK_MASK);
                        isNegInf = isInf && ((_block[MSU] & MSU_MASK) == MSU_MASK);
                        isPosInf = isInf && (_block[MSU] & MSU_MASK) == (MSU_MASK ^ SIGN_BIT_MASK);
                }
                else if constexpr (4 == nrBlocks) {
                        bool isInf = (_block[0] == (BLOCK_MASK ^ LSB_BIT_MASK)) && (_block[1] == BLOCK_MASK) && (_block[2] == BLOCK_MASK);
                        isNegInf = isInf && ((_block[MSU] & MSU_MASK) == MSU_MASK);
                        isPosInf = isInf && (_block[MSU] & MSU_MASK) == (MSU_MASK ^ SIGN_BIT_MASK);
                }
                else {
                        bool isInf = (_block[0] == (BLOCK_MASK ^ LSB_BIT_MASK));
                        for (unsigned i = 1; i < nrBlocks - 1; ++i) {
                                if (_block[i] != BLOCK_MASK) {
                                        isInf = false;
                                        break;
                                }
                        }
                        isNegInf = isInf && ((_block[MSU] & MSU_MASK) == MSU_MASK);
                        isPosInf = isInf && (_block[MSU] & MSU_MASK) == (MSU_MASK ^ SIGN_BIT_MASK);
                }

                return (InfType == INF_TYPE_EITHER ? (isNegInf || isPosInf) :
                        (InfType == INF_TYPE_NEGATIVE ? isNegInf :
                                (InfType == INF_TYPE_POSITIVE ? isPosInf : false)));
        }
        constexpr bool isnan(int NaNType = NAN_TYPE_EITHER) const noexcept {
                if constexpr (hasMaxExpValues) {
                        return isnanencoding(NaNType);
                }
                else {
                        if (ismaxexpvalue()) {
                                // all these max-exponent encodings are NANs, except for the encoding representing INF
                                bool isNaN = isinf() ? false : true;
                                bool isNegNaN = isNaN && sign();
                                bool isPosNaN = isNaN && !sign();
                                return (NaNType == NAN_TYPE_EITHER ? (isNaN) :
                                        (NaNType == NAN_TYPE_SIGNALLING ? isNegNaN :
                                                (NaNType == NAN_TYPE_QUIET ? isPosNaN : false)));
                        }
                        else {
                                return false;
                        }
                }
        }
        // iszeroencoding returns true if it finds a pure -0 or +0 pattern and returns false otherwise
        constexpr bool iszeroencoding() const noexcept {
                if constexpr (0 == nrBlocks) {
                        return true;
                }
                else if constexpr (1 == nrBlocks) {
                        return (_block[MSU] & ~SIGN_BIT_MASK) == 0;
                }
                else if constexpr (2 == nrBlocks) {
                        return (_block[0] == 0) && (_block[MSU] & ~SIGN_BIT_MASK) == 0;
                }
                else if constexpr (3 == nrBlocks) {
                        return (_block[0] == 0) && _block[1] == 0 && (_block[MSU] & ~SIGN_BIT_MASK) == 0;
                }
                else if constexpr (4 == nrBlocks) {
                        return (_block[0] == 0) && _block[1] == 0 && _block[2] == 0 && (_block[MSU] & ~SIGN_BIT_MASK) == 0;
                }
                else {
                        for (unsigned i = 0; i < nrBlocks - 1; ++i) if (_block[i] != 0) return false;
                        return (_block[MSU] & ~SIGN_BIT_MASK) == 0;
                }
        }
        // isminnegencoding returns true if it find the pattern 1.00.00001 and returns false otherwise
        constexpr bool isminnegencoding() const noexcept {
                if constexpr (0 == nrBlocks) {
                        return false;
                }
                else if constexpr (1 == nrBlocks) {
                        return (_block[MSU] & (SIGN_BIT_MASK | 1ul));
                }
                else if constexpr (2 == nrBlocks) {
                        return ((_block[0] == 1ul) && (_block[1] == SIGN_BIT_MASK));
                }
                else if constexpr (3 == nrBlocks) {
                        return ((_block[0] == 1ul) && (_block[1] == 0) && (_block[2] == SIGN_BIT_MASK));
                }
                else if constexpr (4 == nrBlocks) {
                        return ((_block[0] == 1ul) && (_block[1] == 0) && (_block[2] == 0) && (_block[3] == SIGN_BIT_MASK));
                }
                else {
                        if (_block[0] != 1ul) return false;
                        for (unsigned i = 1; i < nrBlocks - 2; ++i) if (_block[i] != 0) return false;
                        return (_block[MSU] == SIGN_BIT_MASK);
                }
        }
        constexpr bool isnanencoding(int NaNType = NAN_TYPE_EITHER) const noexcept {
                // the bit encoding of NaN is independent of the gradual overflow configuration
                bool isNaN = true;
                bool isNegNaN = false;
                bool isPosNaN = false;

                if constexpr (0 == nrBlocks) {
                        return false;
                }
                else if constexpr (1 == nrBlocks) {
                }
                else if constexpr (2 == nrBlocks) {
                        isNaN = (_block[0] == BLOCK_MASK);
                }
                else if constexpr (3 == nrBlocks) {
                        isNaN = (_block[0] == BLOCK_MASK) && (_block[1] == BLOCK_MASK);
                }
                else if constexpr (4 == nrBlocks) {
                        isNaN = (_block[0] == BLOCK_MASK) && (_block[1] == BLOCK_MASK) && (_block[2] == BLOCK_MASK);
                }
                else {
                        for (unsigned i = 0; i < nrBlocks - 1; ++i) {
                                if (_block[i] != BLOCK_MASK) {
                                        isNaN = false;
                                        break;
                                }
                        }
                }
                isNegNaN = isNaN && ((_block[MSU] & MSU_MASK) == MSU_MASK);
                isPosNaN = isNaN && (_block[MSU] & MSU_MASK) == (MSU_MASK ^ SIGN_BIT_MASK);

                return (NaNType == NAN_TYPE_EITHER ? (isNegNaN || isPosNaN) :
                        (NaNType == NAN_TYPE_SIGNALLING ? isNegNaN :
                                (NaNType == NAN_TYPE_QUIET ? isPosNaN : false)));
        }
        // isnormal returns true if 0 or exponent bits are not all zero or one, false otherwise
        constexpr bool isnormal() const noexcept {
                if (iszeroencoding()) return true; // filter out the one special case
                blockbinary<es, bt> e{};
                exponent(e);
                return !e.iszero() && !e.all();
        }
        // isdenormal returns true if exponent bits are all zero, false otherwise
        constexpr bool isdenormal() const noexcept {
                if (iszeroencoding()) return false; // filter out the one special case
                blockbinary<es, bt> e{};
                exponent(e);
                return e.iszero(); 
        }
        // ismaxexpvalue returns true if exponent bits are all one, false otherwise
        constexpr bool ismaxexpvalue() const noexcept {
                blockbinary<es, bt> e{};
                exponent(e);
                return e.all();
        }
        // isinteger is TBD
        constexpr bool isinteger() const noexcept { return false; } // return (floor(*this) == *this) ? true : false; }
        
        template<typename NativeReal>
        constexpr bool inrange(NativeReal v) const noexcept {
                // the valid range for this cfloat includes the interval between 
                // maxpos and the value that would round down to maxpos
                bool bIsInRange = true;                
                if (v > 0) {
                        cfloat c(SpecificValue::maxpos);
                        cfloat<nbits + 1, es, BlockType, hasSubnormals, hasMaxExpValues, isSaturating> d{};
                        d = NativeReal(c);
                        ++d;
                        if (v >= NativeReal(d)) bIsInRange = false;
                }
                else {
                        cfloat c(SpecificValue::maxneg);
                        cfloat<nbits + 1, es, BlockType, hasSubnormals, hasMaxExpValues, isSaturating> d{};
                        d = NativeReal(c);
                        --d;
                        if (v <= NativeReal(d)) bIsInRange = false;
                }

                return bIsInRange;
        }
        constexpr bool test(unsigned bitIndex) const noexcept {
                return at(bitIndex);
        }
        constexpr bool at(unsigned bitIndex) const noexcept {
                if (bitIndex < nbits) {
                        bt word = _block[bitIndex / bitsInBlock];
                        bt mask = bt(1ull << (bitIndex % bitsInBlock));
                        return (word & mask);
                }
                return false;
        }
        constexpr uint8_t nibble(unsigned n) const noexcept {
                if (n < (1 + ((nbits - 1) >> 2))) {
                        bt word = _block[(n * 4) / bitsInBlock];
                        int nibbleIndexInWord = int(n % (bitsInBlock >> 2ull));
                        bt mask = bt(0xF << (nibbleIndexInWord * 4));
                        bt nibblebits = bt(mask & word);
                        return uint8_t(nibblebits >> (nibbleIndexInWord * 4));
                }
                return 0;
        }
        constexpr bt block(unsigned b) const noexcept {
                if (b < nrBlocks) {
                        return _block[b];
                }
                return 0;
        }

        constexpr void sign(bool& s) const {
                s = sign();
        }
        constexpr void exponent(blockbinary<es, bt>& e) const {
                e.clear();
                if constexpr (0 == nrBlocks) return;
                else if constexpr (1 == nrBlocks) {
                        bt ebits = bt(_block[MSU] & ~SIGN_BIT_MASK);
                        e.setbits(uint64_t(ebits >> EXP_SHIFT));
                }
                else if constexpr (nrBlocks > 1) {
                        if (MSU_CAPTURES_EXP) {
                                bt ebits = bt(_block[MSU] & ~SIGN_BIT_MASK);
                                e.setbits(uint64_t(ebits >> ((nbits - 1ull - es) % bitsInBlock)));
                        }
                        else {
                                for (unsigned i = 0; i < es; ++i) { e.setbit(i, at(nbits - 1ull - es + i)); }
                        }
                }
        }
        template<unsigned targetFractionBits>
        constexpr blockbinary<targetFractionBits, bt>& fraction(blockbinary<targetFractionBits, bt>& f) const {
                static_assert(targetFractionBits >= fbits, "target blockbinary is too small and can't receive all fraction bits");
                f.clear();
                if constexpr (0 == nrBlocks) return f;
                else if constexpr (1 == nrBlocks) {
                        bt fraction = bt(_block[MSU] & ~MSU_EXP_MASK);
                        f.setbits(fraction);
                }
                else if constexpr (nrBlocks > 1) {
                        for (unsigned i = 0; i < fbits; ++i) { f.setbit(i, at(i)); }
                }
                return f;
        }
        constexpr uint64_t fraction_ull() const {
                uint64_t raw{ 0 };
                if constexpr (nbits - es - 1ull < 65ull) { // no-op if precondition doesn't hold
                        if constexpr (1 == nrBlocks) {
                                uint64_t fbitMask = 0xFFFF'FFFF'FFFF'FFFF >> (64 - fbits);
                                raw = fbitMask & uint64_t(_block[0]);
                        }
                        else if constexpr (2 == nrBlocks) {
                                uint64_t fbitMask = 0xFFFF'FFFF'FFFF'FFFF >> (64 - fbits);
                                raw = fbitMask & ((uint64_t(_block[1]) << bitsInBlock) | uint64_t(_block[0]));
                        }
                        else if constexpr (3 == nrBlocks) {
                                uint64_t fbitMask = 0xFFFF'FFFF'FFFF'FFFF >> (64 - fbits);
                                raw = fbitMask & ((uint64_t(_block[2]) << (2 * bitsInBlock)) | (uint64_t(_block[1]) << bitsInBlock) | uint64_t(_block[0]));
                        }
                        else if constexpr (4 == nrBlocks) {
                                uint64_t fbitMask = 0xFFFF'FFFF'FFFF'FFFF >> (64 - fbits);
                                raw = fbitMask & ((uint64_t(_block[3]) << (3 * bitsInBlock)) | (uint64_t(_block[2]) << (2 * bitsInBlock)) | (uint64_t(_block[1]) << bitsInBlock) | uint64_t(_block[0]));
                        }
                        else {
                                uint64_t mask{ 1 };
                                for (unsigned i = 0; i < fbits; ++i) { 
                                        if (test(i)) {
                                                raw |= mask;
                                        }
                                        mask <<= 1;
                                }
                        }
                }
                return raw;
        }
        // construct the significant from the encoding, returns normalization offset
        constexpr unsigned significant(blockbinary<fhbits, bt>& s, bool isNormal = true) const {
                unsigned shift = 0;
                if (iszero()) return 0;
                if constexpr (0 == nrBlocks) return 0;
                else if constexpr (1 == nrBlocks) {
                        bt significant = bt(_block[MSU] & ~MSU_EXP_MASK & ~SIGN_BIT_MASK);
                        if (isNormal) {
                                significant |= (bt(0x1ul) << fbits);
                        }
                        else {
                                unsigned msb = find_msb(significant);
//                                std::cout << "msb : " << msb << " : fhbits : " << fhbits << " : " << to_binary(significant, true) << std::endl;
                                shift = fhbits - msb;
                                significant <<= shift;
                        }
                        s.setbits(significant);
                }
                else if constexpr (nrBlocks > 1) {
                        s.clear();
                        // TODO: design and implement a block-oriented algorithm, this sequential algorithm is super slow
                        if (isNormal) {
                                s.setbit(fbits);
                                for (unsigned i = 0; i < fbits; ++i) { s.setbit(i, at(i)); }
                        }
                        else {
                                // Find the MSB of the subnormal: 
                                unsigned msb = 0;
                                for (unsigned i = 0; i < fbits; ++i) { // msb protected from not being assigned through iszero test at prelude of function
                                        msb = fbits - 1ull - i;
                                        if (test(msb)) break;
                                }
                                //      m-----lsb
                                // h00001010101
                                // 101010100000
                                for (unsigned i = 0; i <= msb; ++i) {
                                        s.setbit(fbits - msb + i, at(i));
                                }
                                shift = fhbits - msb;
                        }
                }
                return shift;
        }
        template<unsigned targetbits>
        constexpr void bits(blockbinary<targetbits, bt>& b) const {
                unsigned upperbound = (nbits > targetbits ? targetbits : nbits);
                b.clear();
                for (unsigned i = 0; i < upperbound; ++i) { b.setbit(i, at(i)); }
        }

        // casts to native types
        // Float-to-integer conversion is undefined when the (truncated) source value
        // is outside the destination type's representable range [conv.fpint]. Clamp
        // to min/max before the cast. Note: float(INT_MAX) typically rounds up to
        // 2^31, so we use >= (not >) to catch the boundary.
        constexpr int to_int() const {
                if (isnan()) return 0;
                if (isinf()) return sign() ? std::numeric_limits<int>::min() : std::numeric_limits<int>::max();
                float f = to_native<float>();
                if (f >= static_cast<float>(std::numeric_limits<int>::max())) return std::numeric_limits<int>::max();
                if (f <  static_cast<float>(std::numeric_limits<int>::min())) return std::numeric_limits<int>::min();
                return int(f);
        }
        constexpr long to_long() const {
                if (isnan()) return 0;
                if (isinf()) return sign() ? std::numeric_limits<long>::min() : std::numeric_limits<long>::max();
                double d = to_native<double>();
                if (d >= static_cast<double>(std::numeric_limits<long>::max())) return std::numeric_limits<long>::max();
                if (d <  static_cast<double>(std::numeric_limits<long>::min())) return std::numeric_limits<long>::min();
                return long(d);
        }
        constexpr long long to_long_long() const {
                if (isnan()) return 0;
                if (isinf()) return sign() ? std::numeric_limits<long long>::min() : std::numeric_limits<long long>::max();
                double d = to_native<double>();
                if (d >= static_cast<double>(std::numeric_limits<long long>::max())) return std::numeric_limits<long long>::max();
                if (d <  static_cast<double>(std::numeric_limits<long long>::min())) return std::numeric_limits<long long>::min();
                return (long long)(d);
        }

        // transform an cfloat to a native C++ floating-point. We are using the native
        // precision to compute, which means that all sub-values need to be representable 
        // by the native precision.
        // A more accurate approximation would require an adaptive precision algorithm
        // with a final rounding step.
        template<typename TargetFloat>
        constexpr
        TargetFloat to_native() const { 
                TargetFloat v{ 0.0 };
                if (iszero()) {
                        // the optimizer might destroy the sign
                        return sign() ? -TargetFloat(0) : TargetFloat(0);
                }
                else if (isnan()) {
                        v = sign() ? std::numeric_limits<TargetFloat>::signaling_NaN() : std::numeric_limits<TargetFloat>::quiet_NaN();
                }
                else if (isinf()) {
                        v = sign() ? -std::numeric_limits<TargetFloat>::infinity() : std::numeric_limits<TargetFloat>::infinity();
                }
                else { // TODO: this approach has catastrophic cancellation when nbits is large and native target float is too small
                        TargetFloat f{ 0 };
                        TargetFloat fbit{ 0.5 };
                        for (int i = static_cast<int>(nbits - 2ull - es); i >= 0; --i) {
                                f += at(static_cast<unsigned>(i)) ? fbit : TargetFloat(0);
                                fbit *= TargetFloat(0.5);
                        }
                        blockbinary<es, bt> ebits;
                        exponent(ebits);
                        if constexpr (hasSubnormals) {
                                if (ebits.iszero()) {
                                        // subnormals: (-1)^s * 2^(2-2^(es-1)) * (f/2^fbits))
                                        TargetFloat exponentiation = TargetFloat(subnormal_exponent[es]); // precomputed values for 2^(2-2^(es-1))
                                        v = exponentiation * f;  // f is already f/2^fbits
                                        return sign() ? -v : v;
                                }
                        }
                        else {
                                if (ebits.iszero()) { // underflow to 0
                                        // compiler fast float optimization might destroy the sign
                                        return sign() ? -TargetFloat(0) : TargetFloat(0);
                                }
                        }
                        if constexpr (hasMaxExpValues) {
                                // regular: (-1)^s * 2^(e+1-2^(es-1)) * (1 + f/2^fbits))
                                int exponent = static_cast<int>(unsigned(ebits) - EXP_BIAS);
                                if (-64 < exponent && exponent < 64) {
                                        // NB: the negative branch uses std::ldexp instead of
                                        // 1.0f / TargetFloat(1ull << -exponent) because clang -O2
                                        // miscompiles the latter for TargetFloat == long double via
                                        // a buggy direct-bit-pattern optimisation (see issue #937).
                                        // gcc handles either form correctly; ldexp is portable.
                                        TargetFloat exponentiation = (exponent >= 0
                                                ? TargetFloat(1ull << exponent)
                                                : std::ldexp(TargetFloat(1), exponent));
                                        v = exponentiation * (TargetFloat(1.0) + f);
                                }
                                else {
                                        double exponentiation = ipow(exponent);
                                        v = TargetFloat(exponentiation * (1.0 + f));
                                }
                        }
                        else {
                                if (ebits.all()) {
                                        // max-exponent values are mapped to quiet NaNs
                                        v = std::numeric_limits<TargetFloat>::quiet_NaN();
                                        return v;
                                }
                                else {
                                        // regular: (-1)^s * 2^(e+1-2^(es-1)) * (1 + f/2^fbits))
                                        int exponent = static_cast<int>(unsigned(ebits) - EXP_BIAS);
                                        if (-64 < exponent && exponent < 64) {
                                                // See above re: clang -O2 miscompile, issue #937.
                                                TargetFloat exponentiation = (exponent >= 0
                                                        ? TargetFloat(1ull << exponent)
                                                        : std::ldexp(TargetFloat(1), exponent));
                                                v = exponentiation * (TargetFloat(1.0) + f);
                                        }
                                        else {
                                                double exponentiation = ipow(exponent);
                                                v = TargetFloat(exponentiation * (1.0 + f));
                                        }
                                }
                        }
                        v = sign() ? -v : v;
                }
                return v;
        }

        // convert a cfloat to a blocktriple with the fraction format 1.ffff
        // we are using the same block type so that we can use block copies to move bits around.
        // Since we tend to have at least two exponent bits, this will lead to
        // most cfloat<->blocktriple cases being efficient as the block types are aligned.
        // The relationship between the source cfloat and target blocktriple is not
        // arbitrary, enforce it by setting the blocktriple fbits to the cfloat's (nbits - es - 1)
        template<typename TargetBlockType = bt>
        constexpr void normalize(blocktriple<fbits, BlockTripleOperator::REP, TargetBlockType>& tgt) const {
                // test special cases
                if (isnan()) {
                        tgt.setnan();
                }
                else if (isinf()) {
                        tgt.setinf();
                }
                else if (iszero()) {
                        tgt.setzero();
                }
                else {
                        tgt.setnormal(); // a blocktriple is always normalized
                        int scale = this->scale();
                        tgt.setsign(sign());
                        tgt.setscale(scale);
                        // set significant
                        // we are going to unify to the format 01.ffffeeee
                        // where 'f' is a fraction bit, and 'e' is an extension bit
                        // so that normalize can be used to generate blocktriples for add/sub/mul/div/sqrt
                        if (isnormal() || ismaxexpvalue()) {
                                // normal encoding or max-exponent encoding (hasMaxExpValues=true):
                                // significand has a hidden bit at position fbits
                                if constexpr (fbits < 64) { // max 63 bits of fraction to yield 64bit of raw significant bits
                                        uint64_t raw = fraction_ull();
                                        raw |= (1ull << fbits);
                                        tgt.setbits(raw);
                                }
                                else {
                                        blockcopy(tgt);
                                        tgt.setbit(fbits);
                                }
                        }
                        else { // it is a subnormal encoding in this target cfloat
                                int shift = MIN_EXP_NORMAL - scale;
                                if (shift < 0) shift = 0; // guard against negative shifts
                                // Stored subnormals have no hidden bit, but blocktriple REP expects a normalized 1.ffff payload.
                                // Shift the encoded fraction up until the leading 1 reaches the hidden-bit position, while leaving
                                // scale at the value returned by cfloat::scale(); that preserves the original real value in canonical form.
                                if constexpr (fbits < 64) {
                                        uint64_t raw = fraction_ull();
                                        raw <<= shift;
                                        raw |= (1ull << fbits);
                                        tgt.setbits(raw);
                                }
                                else {
                                        blockcopy(tgt);
                                        tgt <<= shift;
                                        tgt.setbit(fbits);
                                }
                        }
                }
        }

        // normalize a cfloat to a blocktriple used in add/sub, which has the form 00h.fffff
        // that is 3 + fbits, the 3 extra bits are required to be able to use 2's complement 
        // and capture the largest value of an addition/subtraction.
        // TODO: currently abits = 2*fhbits as the worst case input argument size to
        // capture the smallest normal value in aligned form. There is a faster/smaller
        // implementation where the input is constrainted to just the round, guard, and sticky bits.
        constexpr void normalizeAddition(blocktriple<fbits, BlockTripleOperator::ADD, bt>& tgt) const {
                using BlockTripleConfiguration = blocktriple<fbits, BlockTripleOperator::ADD, bt>;
                // test special cases
                if (isnan()) {
                        tgt.setnan();
                }
                else if (isinf()) {
                        tgt.setinf();
                }
                else if (iszero()) {
                        tgt.setzero();
                }
                else {
                        tgt.setnormal(); // a blocktriple is always normalized
                        int scale = this->scale();
                        tgt.setsign(sign());
                        tgt.setscale(scale);
                        // set significant
                        // we are going to unify to the format 001.ffffeeee
                        // where 'f' is a fraction bit, and 'e' is an extension bit
                        // so that normalize can be used to generate blocktriples for add/sub/mul/div/sqrt
                        if (isnormal()) {
                                if constexpr (fbits < 64 && BlockTripleConfiguration::rbits < (64 - fbits)) {
                                        uint64_t raw = fraction_ull();
                                        raw |= (1ull << fbits); // add the hidden bit
                                        //std::cout << "normalize      : " << *this << '\n';
                                        //std::cout << "significant    : " << to_binary(raw, fbits + 2) << '\n';
                                        raw <<= BlockTripleConfiguration::rbits;  // rounding bits required for correct rounding
                                        //std::cout << "rounding shift : " << to_binary(raw, fbits + 2 + BlockTripleConfiguration::rbits) << '\n';
                                        tgt.setbits(raw);
                                }
                                else {
                                        blockcopy(tgt);
                                        tgt.setradix();
                                        tgt.setbit(fbits); // add the hidden bit
                                        tgt.bitShift(BlockTripleConfiguration::rbits);  // rounding bits required for correct rounding
                                }
                        }
                        else {
                                if (isdenormal()) { // it is a subnormal encoding in this target cfloat
                                        if constexpr (hasSubnormals) {
                                                if constexpr (BlockTripleConfiguration::rbits < (64 - fbits)) {
                                                        uint64_t raw = fraction_ull();
                                                        int shift = MIN_EXP_NORMAL - scale;
                                                        raw <<= shift; // shift but do NOT add a hidden bit as the MSB of the subnormal is shifted in the hidden bit position
                                                        raw <<= BlockTripleConfiguration::rbits;  // rounding bits required for correct rounding
                                                        tgt.setbits(raw);
                                                }
                                                else {
                                                        blockcopy(tgt);
                                                        tgt.setradix();
                                                        int shift = MIN_EXP_NORMAL - scale;
                                                        tgt.bitShift(shift + BlockTripleConfiguration::rbits);  // rounding bits required for correct rounding
                                                }
                                        }
                                        else {  // this cfloat has no subnormals
                                                tgt.setzero(tgt.sign()); // preserve the sign
                                        }
                                }
                                else {
                                        // by design, a cfloat is either normal, subnormal, or max-exponent value, so this else clause is by deduction covering a max-exponent value
                                        if constexpr (hasMaxExpValues) {
                                                if constexpr (fbits < 64 && BlockTripleConfiguration::rbits < (64 - fbits)) {
                                                        uint64_t raw = fraction_ull();
                                                        raw |= (1ull << fbits); // add the hidden bit
                                                        raw <<= BlockTripleConfiguration::rbits;  // rounding bits required for correct rounding
                                                        tgt.setbits(raw);
                                                }
                                                else {
                                                        blockcopy(tgt);
                                                        tgt.setradix();
                                                        tgt.setbit(fbits); // add the hidden bit
                                                        tgt.bitShift(BlockTripleConfiguration::rbits);  // rounding bits required for correct rounding
                                                }
                                        }
                                        else {  // this cfloat has no max-exponent values and thus this represents a nan, signalling or quiet determined by the sign
                                                tgt.setnan(tgt.sign());
                                        }                        
                                }
                        }
                }
                // tgt.setradix(radix);
        }

        // Normalize a cfloat to a blocktriple used in mul, which has the form 0'00001.fffff
        // that is 2*fbits, plus 1 overflow bit, and the radix set at <fbits>.
        // The result radix will go to 2*fbits after multiplication.
        constexpr void normalizeMultiplication(blocktriple<fbits, BlockTripleOperator::MUL, bt>& tgt) const {
                // test special cases
                if (isnan()) {
                        tgt.setnan();
                }
                else if (isinf()) {
                        tgt.setinf();
                }
                else if (iszero()) {
                        tgt.setzero();
                }
                else {
                        tgt.setnormal(); // a blocktriple is always normalized
                        int scale = this->scale();
                        tgt.setsign(sign());
                        tgt.setscale(scale);

                        // set significant
                        // we are going to unify to the format 01.ffffeeee
                        // where 'f' is a fraction bit, and 'e' is an extension bit
                        // so that normalize can be used to generate blocktriples for add/sub/mul/div/sqrt
                        if (isnormal() || ismaxexpvalue()) {
                                if constexpr (fbits < 64) { // max 63 bits of fraction to yield 64bit of raw significant bits
                                        uint64_t raw = fraction_ull();
                                        raw |= (1ull << fbits);
                                        tgt.setbits(raw);
                                }
                                else {
                                        blockcopy(tgt);
                                        tgt.setradix();
                                        tgt.setbit(fbits); // add the hidden bit
                                }
                        }
                        else { 
                                // it is a subnormal encoding in this target cfloat
                                if constexpr (hasSubnormals) {
                                        if constexpr (fbits < 64) {
                                                uint64_t raw = fraction_ull();
                                                int shift = MIN_EXP_NORMAL - scale;
                                                raw <<= shift;
                                                raw |= (1ull << fbits);
                                                tgt.setbits(raw);
                                        }
                                        else {
                                                blockcopy(tgt);
                                                int shift = MIN_EXP_NORMAL - scale;
                                                tgt.bitShift(shift);
                                                tgt.setbit(fbits);
                                        }
                                }
                                else { // this cfloat has no subnormals
                                        tgt.setzero(tgt.sign()); // preserve the sign
                                }
                        }
                }
                tgt.setradix(fbits); // override the radix with the input scale for accurate value printing
        }

        // normalize a cfloat to a blocktriple used in div, which has the form 0'00000'00001.fffff
        // that is 3*fbits, plus 1 overflow bit, and the radix set at <fbits>.
        // the result radix will go to 2*fbits after multiplication.
        // TODO: needs implementation
        constexpr void normalizeDivision(blocktriple<fbits, BlockTripleOperator::DIV, bt>& tgt) const {
                constexpr unsigned divshift = blocktriple<fbits, BlockTripleOperator::DIV, bt>::divshift;
                // test special cases
                if (isnan()) {
                        tgt.setnan();
                }
                else if (isinf()) {
                        tgt.setinf();
                }
                else if (iszero()) {
                        tgt.setzero();
                }
                else {
                        tgt.setnormal(); // a blocktriple is always normalized
                        int scale = this->scale();
                        tgt.setsign(sign());
                        tgt.setscale(scale);
                        // set significant
                        // we are going to unify to the format 01.ffffeeee
                        // where 'f' is a fraction bit, and 'e' is an extension bit
                        // so that normalize can be used to generate blocktriples for add/sub/mul/div/sqrt
                        if (isnormal() || ismaxexpvalue()) {
                                if constexpr (fbits < 64 && divshift < (64 - fbits)) {
                                        uint64_t raw = fraction_ull();
                                        raw |= (1ull << fbits);
                                        raw <<= divshift; // shift the input value to the output radix
                                        tgt.setbits(raw);
                                }
                                else {
                                        // brute force copy of blocks
                                        blockcopy(tgt);
                                        tgt.setbit(fbits);
                                        tgt.bitShift(divshift); // shift the input value to the output radix
                                }
                        }
                        else { // it is a subnormal encoding in this target cfloat
                                if constexpr (fbits < 64 && divshift < (64 - fbits)) {
                                        uint64_t raw = fraction_ull();
                                        int shift = MIN_EXP_NORMAL - scale;
                                        raw <<= shift;
                                        raw |= (1ull << fbits);
                                        raw <<= divshift; // shift the input value to the output radix
                                        tgt.setbits(raw);
                                }
                                else {
                                        blockcopy(tgt);
                                        int shift = MIN_EXP_NORMAL - scale;
                                        tgt.bitShift(shift);
                                        tgt.setbit(fbits);
                                        tgt.bitShift(divshift); // shift the input value to the output radix
                                }
                        }
                }
                tgt.setradix(blocktriple<fbits, BlockTripleOperator::DIV, bt>::radix);
        }

        // helper debug function
        void constexprClassParameters() const noexcept {
                std::cout << "-------------------------------------------------------------\n";
                std::cout << "type              : " << typeid(*this).name() << '\n';
                std::cout << "nbits             : " << nbits << '\n';
                std::cout << "es                : " << es << std::endl;
                std::cout << "hasSubnormals     : " << (hasSubnormals ? "true" : "false") << '\n';
                std::cout << "hasMaxExpValues   : " << (hasMaxExpValues ? "true" : "false") << '\n';
                std::cout << "isSaturating      : " << (isSaturating ? "true" : "false") << '\n';
                std::cout << "ALL_ONES          : " << to_binary(ALL_ONES, 0, true) << '\n';
                std::cout << "BLOCK_MASK        : " << to_binary(BLOCK_MASK, 0, true) << '\n';
                std::cout << "nrBlocks          : " << nrBlocks << '\n';
                std::cout << "bits in MSU       : " << bitsInMSU << '\n';
                std::cout << "MSU               : " << MSU << '\n';
                std::cout << "MSU MASK          : " << to_binary(MSU_MASK, 0, true) << '\n';
                std::cout << "SIGN_BIT_MASK     : " << to_binary(SIGN_BIT_MASK, 0, true) << '\n';
                std::cout << "LSB_BIT_MASK      : " << to_binary(LSB_BIT_MASK, 0, true) << '\n';
                std::cout << "MSU CAPTURES_EXP  : " << (MSU_CAPTURES_EXP ? "yes\n" : "no\n");
                std::cout << "EXP_SHIFT         : " << EXP_SHIFT << '\n';
                std::cout << "MSU EXP MASK      : " << to_binary(MSU_EXP_MASK, 0, true) << '\n';
                std::cout << "ALL_ONE_MASK_ES   : " << to_binary(ALL_ONES_ES) << '\n';
                std::cout << "EXP_BIAS          : " << EXP_BIAS << '\n';
                std::cout << "MAX_EXP           : " << MAX_EXP << '\n';
                std::cout << "MIN_EXP_NORMAL    : " << MIN_EXP_NORMAL << '\n';
                std::cout << "MIN_EXP_SUBNORMAL : " << MIN_EXP_SUBNORMAL << '\n';
                std::cout << "fraction Blocks   : " << fBlocks << '\n';
                std::cout << "bits in FSU       : " << bitsInFSU << '\n';
                std::cout << "FSU               : " << FSU << '\n';
                std::cout << "FSU MASK          : " << to_binary(FSU_MASK, 0, true) << '\n';
                std::cout << "topfbits          : " << topfbits << '\n';
                std::cout << "ALL_ONE_MASK_FR   : " << to_binary(ALL_ONES_FR) << '\n';
        }
        void showLimbs() const {
                for (unsigned b = 0; b < nrBlocks; ++b) {
                        std::cout << to_binary(_block[nrBlocks - b - 1], sizeof(bt) * 8) << ' ';
                }
                std::cout << '\n';
        }

protected:
        // HELPER methods

        /// <summary>
        /// 1's complement of the encoding used to set up specific encoding patterns.
        /// This is not an arithmetic operator that makes sense for floating-point numbers.
        /// </summary>
        /// <returns>reference to this cfloat object</returns>
        constexpr cfloat& flip() noexcept { // in-place one's complement
                for (unsigned i = 0; i < nrBlocks; ++i) {
                        _block[i] = bt(~_block[i]);
                }
                _block[MSU] &= MSU_MASK; // assert precondition of properly nulled leading non-bits
                return *this;
        }

        /// <summary>
        /// shift left is a bit level encoding helper for fast limb-based conversions between different cfloats
        /// </summary>
        /// <param name="bitsToShift"></param>
        void shiftLeft(unsigned bitsToShift) {
                if (bitsToShift == 0) return;
                if (bitsToShift > nbits) {
                        setzero();
                }
                if (bitsToShift >= bitsInBlock) {
                        int blockShift = static_cast<int>(bitsToShift / bitsInBlock);
                        for (int i = static_cast<int>(MSU); i >= blockShift; --i) {
                                _block[i] = _block[i - blockShift];
                        }
                        for (int i = blockShift - 1; i >= 0; --i) {
                                _block[i] = bt(0);
                        }
                        // adjust the shift
                        bitsToShift -= blockShift * bitsInBlock;
                        if (bitsToShift == 0) return;
                }
                if constexpr (MSU > 0) {
                        // construct the mask for the upper bits in the block that need to move to the higher word
                        bt mask = 0xFFFFFFFFFFFFFFFF << (bitsInBlock - bitsToShift);
                        for (unsigned i = MSU; i > 0; --i) {
                                _block[i] <<= bitsToShift;
                                // mix in the bits from the right
                                bt bits = bt(mask & _block[i - 1]);
                                _block[i] |= (bits >> (bitsInBlock - bitsToShift));
                        }
                }
                _block[0] <<= bitsToShift;
        }

        // convert an unsigned integer into a cfloat
        // TODO: this method does not protect against being called with a signed integer
        template<typename Ty>
        constexpr cfloat& convert_unsigned_integer(const Ty& rhs) noexcept {
                clear();
                if (0 == rhs) return *this;

                uint64_t raw = static_cast<uint64_t>(rhs);
                int msb = static_cast<int>(find_msb(raw)) - 1; // msb > 0 due to zero test above
                int exponent = msb;
                // remove the MSB as it represents the hidden bit in the cfloat representation
                uint64_t hmask = ~(1ull << msb);
                raw &= hmask;

                constexpr uint32_t sizeInBits = 8 * sizeof(Ty);
                uint32_t shift = sizeInBits - exponent - 1;
                raw <<= shift;
                raw = round<sizeInBits, uint64_t>(raw, exponent);

                // check for exponent overflow after rounding
                if (exponent > MAX_EXP) {
                        if constexpr (isSaturating) {
                                this->maxpos();
                        }
                        else {
                                setinf(false);
                        }
                        return *this;
                }

                // construct the target cfloat
                if constexpr (fbits < (64 - es)) {
                        uint64_t biasedExponent = static_cast<uint64_t>(static_cast<long long>(exponent) + static_cast<long long>(EXP_BIAS));
                        uint64_t bits = 0;
                        bits <<= es;
                        bits |= biasedExponent;
                        bits <<= fbits;
                        bits |= raw;
                        setbits(bits);
                }
                else {
                        setsign(false);
                        setexponent(exponent);
                        for (int i = 0; i < exponent; ++i) {
                                bool bit = (raw >> (sizeInBits - 2 - i)) & 1;
                                setbit(static_cast<unsigned>(fbits - 1 - i), bit);
                        }
                }
                // post-rounding cleanup: rounding at the maxpos boundary can
                // produce a NaN encoding; project back to inf or maxpos
                if (isnan()) {
                        if constexpr (isSaturating) {
                                this->maxpos();
                        }
                        else {
                                setinf(false);
                        }
                }
                return *this;
        }
        // convert a signed integer into a cfloat
        // TODO: this method does not protect against being called with a signed integer
        template<typename Ty>
        constexpr cfloat& convert_signed_integer(const Ty& rhs) noexcept {
                clear();
                if (0 == rhs) return *this;
                bool s = (rhs < 0);
                using UnsignedTy = std::make_unsigned_t<Ty>;
                UnsignedTy urhs = static_cast<UnsignedTy>(rhs);
                uint64_t raw = static_cast<uint64_t>(s ? (UnsignedTy(0) - urhs) : urhs);

                int msb = static_cast<int>(find_msb(raw)) - 1; // msb > 0 due to zero test above 
                int exponent = msb;
                // remove the MSB as it represents the hidden bit in the cfloat representation
                uint64_t hmask = ~(1ull << msb);
                raw &= hmask;

                // shift the msb to the msb of the fraction
                constexpr uint32_t sizeInBits = 8 * sizeof(Ty);
                uint32_t shift = sizeInBits - exponent - 1;
                raw <<= shift;
                raw = round<sizeInBits, uint64_t>(raw, exponent);

                // check for exponent overflow after rounding
                if (exponent > MAX_EXP) {
                        if constexpr (isSaturating) {
                                if (s) this->maxneg(); else this->maxpos();
                        }
                        else {
                                setinf(s);
                        }
                        return *this;
                }

                // construct the target cfloat
                if constexpr (fbits < (64 - es)) {
                        uint64_t biasedExponent = static_cast<uint64_t>(static_cast<long long>(exponent) + static_cast<long long>(EXP_BIAS));
                        uint64_t bits = (s ? 1ull : 0ull);
                        bits <<= es;
                        bits |= biasedExponent;
                        bits <<= fbits;
                        bits |= raw;
                        setbits(bits);
                }
                else {
                        setsign(s);
                        setexponent(exponent);
                        for (int i = 0; i < exponent; ++i) {
                                bool bit = (raw >> (sizeInBits - 2 - i)) & 1;
                                setbit(static_cast<unsigned>(fbits - 1 - i), bit);
                        }
                }
                // post-rounding cleanup: rounding at the maxpos boundary can
                // produce a NaN encoding; project back to inf or maxpos/maxneg
                if (isnan()) {
                        if constexpr (isSaturating) {
                                if (s) this->maxneg(); else this->maxpos();
                        }
                        else {
                                setinf(s);
                        }
                }
                return *this;
        }

public:
        template<typename Real>
        CONSTEXPRESSION cfloat& convert_ieee754(Real rhs) noexcept {
                if constexpr (nbits == 32 && es == 8 && sizeof(Real) == 4) {
                        // we CANNOT use the native conversion to float as cfloats have max-exponent values
                        // which IEEE-754 does not have and thus a native conversion would destroy
                        // only if the Real type is a float can we use the direct conversion

                        // when our cfloat is a perfect match to single precision IEEE-754
                        bool s{ false };
                        uint64_t rawExponent{ 0 };
                        uint64_t rawFraction{ 0 };
                        uint64_t bits{ 0 };
                        extractFields(rhs, s, rawExponent, rawFraction, bits);
                        if (rawExponent == ieee754_parameter<Real>::eallset) { // nan and inf need to be remapped
                                if (rawFraction == (ieee754_parameter<Real>::fmask & ieee754_parameter<Real>::snanmask) ||
                                        rawFraction == (ieee754_parameter<Real>::fmask & (ieee754_parameter<Real>::qnanmask | ieee754_parameter<Real>::snanmask))) {
                                        // 1.11111111.00000000.......00000001 signalling nan
                                        // 0.11111111.00000000000000000000001 signalling nan
                                        // MSVC
                                        // 1.11111111.10000000.......00000001 signalling nan
                                        // 0.11111111.10000000.......00000001 signalling nan
                                        setnan(NAN_TYPE_SIGNALLING);
                                        //setsign(s);  a cfloat encodes a signalling nan with sign = 1, and a quiet nan with sign = 0
                                        return *this;
                                }
                                if (rawFraction == (ieee754_parameter<Real>::fmask & ieee754_parameter<Real>::qnanmask)) {
                                        // 1.11111111.10000000.......00000000 quiet nan
                                        // 0.11111111.10000000.......00000000 quiet nan
                                        setnan(NAN_TYPE_QUIET);
                                        //setsign(s);  a cfloat encodes a signalling nan with sign = 1, and a quiet nan with sign = 0
                                        return *this;
                                }
                                if (rawFraction == 0ull) {
                                        // 1.11111111.0000000.......000000000 -inf
                                        // 0.11111111.0000000.......000000000 +inf
                                        setinf(s);
                                        return *this;
                                }
                        }
                        uint64_t raw{ s ? 1ull : 0ull };
                        raw <<= 31;
                        raw |= (rawExponent << fbits);
                        raw |= rawFraction;
                        setbits(raw);
                        return *this;
                }
                else if constexpr (nbits == 64 && es == 11 && sizeof(Real) == 8) {
                        // when our cfloat is a perfect match to double precision IEEE-754
                        bool s{ false };
                        uint64_t rawExponent{ 0 };
                        uint64_t rawFraction{ 0 };
                        uint64_t bits{ 0 };
                        extractFields(rhs, s, rawExponent, rawFraction, bits);
                        if (rawExponent == ieee754_parameter<Real>::eallset) { // nan and inf need to be remapped
                                if (rawFraction == (ieee754_parameter<Real>::fmask & ieee754_parameter<Real>::snanmask) ||
                                        rawFraction == (ieee754_parameter<Real>::fmask & (ieee754_parameter<Real>::qnanmask | ieee754_parameter<Real>::snanmask))) {
                                        // 1.11111111.00000000.......00000001 signalling nan
                                        // 0.11111111.00000000000000000000001 signalling nan
                                        // MSVC
                                        // 1.11111111.10000000.......00000001 signalling nan
                                        // 0.11111111.10000000.......00000001 signalling nan
                                        setnan(NAN_TYPE_SIGNALLING);
                                        //setsign(s);  a cfloat encodes a signalling nan with sign = 1, and a quiet nan with sign = 0
                                        return *this;
                                }
                                if (rawFraction == (ieee754_parameter<Real>::fmask & ieee754_parameter<Real>::qnanmask)) {
                                        // 1.11111111.10000000.......00000000 quiet nan
                                        // 0.11111111.10000000.......00000000 quiet nan
                                        setnan(NAN_TYPE_QUIET);
                                        //setsign(s);  a cfloat encodes a signalling nan with sign = 1, and a quiet nan with sign = 0
                                        return *this;
                                }
                                if (rawFraction == 0ull) {
                                        // 1.11111111.0000000.......000000000 -inf
                                        // 0.11111111.0000000.......000000000 +inf
                                        setinf(s);
                                        return *this;
                                }
                        }
                        // normal and subnormal handling
                        uint64_t raw{ s ? 1ull : 0ull };
                        raw <<= 63;
                        raw |= (rawExponent << fbits);
                        raw |= rawFraction;
                        setbits(raw);
                        return *this;
                }
                else {
                        clear();
                        // extract raw IEEE-754 bits
                        bool s{ false };
                        uint64_t rawExponent{ 0 };
                        uint64_t rawFraction{ 0 };
                        uint64_t bits{ 0 };
                        extractFields(rhs, s, rawExponent, rawFraction, bits);
                        // special case handling
                        if (rawExponent == ieee754_parameter<Real>::eallset) { // nan and inf
                                if (rawFraction == (ieee754_parameter<Real>::fmask & ieee754_parameter<Real>::snanmask) ||
                                        rawFraction == (ieee754_parameter<Real>::fmask & (ieee754_parameter<Real>::qnanmask | ieee754_parameter<Real>::snanmask))) {
                                        // 1.11111111.00000000.......00000001 signalling nan
                                        // 0.11111111.00000000000000000000001 signalling nan
                                        // MSVC
                                        // 1.11111111.10000000.......00000001 signalling nan
                                        // 0.11111111.10000000.......00000001 signalling nan
                                        setnan(NAN_TYPE_SIGNALLING);
                                        //setsign(s);  a cfloat encodes a signalling nan with sign = 1, and a quiet nan with sign = 0
                                        return *this;
                                }
                                if (rawFraction == (ieee754_parameter<Real>::fmask & ieee754_parameter<Real>::qnanmask)) {
                                        // 1.11111111.10000000.......00000000 quiet nan
                                        // 0.11111111.10000000.......00000000 quiet nan
                                        setnan(NAN_TYPE_QUIET);
                                        //setsign(s);  a cfloat encodes a signalling nan with sign = 1, and a quiet nan with sign = 0
                                        return *this;
                                }
                                if (rawFraction == 0ull) {
                                        // 1.11111111.0000000.......000000000 -inf
                                        // 0.11111111.0000000.......000000000 +inf
                                        setinf(s);
                                        return *this;
                                }
                        }
                        if (rhs == 0.0) { // IEEE rule: this is valid for + and - 0.0
                                setbit(nbits - 1ull, s);
                                return *this;
                        }
        
                        // normal number consists of fbits fraction bits and one hidden bit
                        // subnormal number has no hidden bit
                        int exponent = static_cast<int>(rawExponent) - ieee754_parameter<Real>::bias;  // unbias the exponent

                        // normalize subnormal source values so downstream code can treat them as normal
                        // A subnormal has rawExponent == 0 and value = 0.fraction * 2^(1-bias).
                        // We find the MSB of the fraction, shift it into the hidden bit position,
                        // mask it off (like a normal's implicit 1), and adjust the exponent.
                        // Setting rawExponent = 1 causes all downstream if(rawExponent != 0) branches
                        // to take the normal-number path, which already handles normal-to-normal and
                        // normal-to-subnormal target conversions correctly.
                        if (rawExponent == 0 && rawFraction != 0) {
                                exponent = 1 - ieee754_parameter<Real>::bias; // effective exponent for subnormals
                                unsigned msb_pos = find_msb(rawFraction); // 1-indexed position of MSB
                                unsigned normalizeShift = static_cast<unsigned>(ieee754_parameter<Real>::fbits) + 1u - msb_pos;
                                rawFraction = (rawFraction << normalizeShift) & static_cast<uint64_t>(ieee754_parameter<Real>::fmask);
                                exponent -= static_cast<int>(normalizeShift);
                                rawExponent = 1; // mark as normalized so downstream paths treat this as a normal number
                        }

                        // check special case of
                        //  1- saturating to maxpos/maxneg, or 
                        //  2- projecting to +-inf 
                        // if the value is out of range.
                        // 
                        // One problem here is that at the rounding cusps of maxpos <-> inf <-> nan
                        // you need to go through the rounding logic to know which encoding you end up
                        // with. 
                        // For each specific cfloat configuration, you can work out these rounding cusps
                        // but they need to go through the value transformation to map them back to native
                        // IEEE-754. That is a complex computation to do as a static constexpr as you need
                        // to construct the value, then evaluate it, and store it.
                        // 
                        // The algorithm used here is to check for the obvious out of range values by
                        // comparing their scale to the max scale this cfloat encoding can represent.
                        // For the rounding cusps, we go through the rounding logic, and then clean up
                        // after rounding using the observation that no conversion from a value can ever
                        // yield the NaN encoding.
                        //
                        // The rounding logic will correctly sort between maxpos and inf, and we clean
                        // up any NaN encodings by projecting back to the configuration's saturation rule.
                        //
                        // We could improve on this by creating the database of rounding cusps and
                        // referencing them with a direct value comparison with the input. That would be
                        // the most performant implementation.
                        if (exponent > MAX_EXP) {
                                if constexpr (isSaturating) {
                                        if (s) this->maxneg(); else this->maxpos(); // saturate to maxpos or maxneg
                                }
                                else {
                                        setinf(s);
                                }
                                return *this;
                        }
                        if constexpr (hasSubnormals) {
                                if (exponent < MIN_EXP_SUBNORMAL - 1) { 
                                        // map to +-0 any values that have a scale less than (MIN_EXP_SUBMORNAL - 1)
                                        this->setbit(nbits - 1, s);
                                        return *this;
                                }
                        }
                        else {
                                if (exponent < MIN_EXP_NORMAL - 1) {
                                        // map to +-0 any values that have a scale less than (MIN_EXP_MORNAL - 1)
                                        this->setbit(nbits - 1, s);
                                        return *this;
                                }
                        }

                        /////////////////  
                        /// end of special case processing, move on to value sampling and rounding

#if TRACE_CONVERSION
                        std::cout << '\n';
                        std::cout << "value             : " << rhs << '\n';
                        std::cout << "segments          : " << to_binary(rhs) << '\n';
                        std::cout << "sign     bit      : " << (s ? '1' : '0') << '\n';
                        std::cout << "exponent bits     : " << to_binary(rawExponent, ieee754_parameter<Real>::ebits, true) << '\n';
                        std::cout << "fraction bits     : " << to_binary(rawFraction, ieee754_parameter<Real>::fbits, true) << '\n';
                        std::cout << "exponent value    : " << exponent << '\n';
#endif

                        // do the following scenarios have different rounding bits?
                        // input is normal, cfloat is normal           <-- rounding can happen with native ieee-754 bits
                        // input is normal, cfloat is subnormal
                        // input is subnormal, cfloat is normal
                        // input is subnormal, cfloat is subnormal

                        // The first condition is the relationship between the number 
                        // of fraction bits from the source and the number of fraction bits 
                        // in the target cfloat: these are constexpressions and guard the shifts
                        // input fbits >= cfloat fbits                 <-- need to round
                        // input fbits < cfloat fbits                  <-- no need to round

                        // quick check if we are truncating to 0 for all subnormal values
                        if constexpr (!hasSubnormals) {
                                if (exponent < MIN_EXP_NORMAL) {
                                        setsign(s); // rest of the bits, exponent and fraction, are already set correctly
                                        return *this;
                                }
                        }
                        if constexpr (fbits < ieee754_parameter<Real>::fbits) {
                                // this is the common case for cfloats that are smaller in size compared to single and double precision IEEE-754
                                constexpr int rightShift = ieee754_parameter<Real>::fbits - fbits; // this is the bit shift to get the MSB of the src to the MSB of the tgt
                                uint32_t biasedExponent{ 0 };
                                int adjustment{ 0 }; // right shift adjustment for subnormal representation
                                uint64_t mask;
                                if (rawExponent != 0) {
                                        // the source real is a normal number, 
//                                        if (exponent >= (MIN_EXP_SUBNORMAL - 1) && exponent < MIN_EXP_NORMAL) {
                                        if (exponent < MIN_EXP_NORMAL) {
//                                                exponent = (exponent < MIN_EXP_SUBNORMAL ? MIN_EXP_SUBNORMAL : exponent); // clip to the smallest subnormal exponent, otherwise the adjustment is off
                                                // the value is a subnormal number in this representation: biasedExponent = 0
                                                // add the hidden bit to the fraction bits so the denormalization has the correct MSB
                                                rawFraction |= ieee754_parameter<Real>::hmask;

                                                // fraction processing: we have 1 hidden + 23 explicit fraction bits 
                                                // f = 1.ffff 2^exponent * 2^fbits * 2^-(2-2^(es-1)) = 1.ff...ff >> (23 - (-exponent + fbits - (2 -2^(es-1))))
                                                // -exponent because we are right shifting and exponent in this range is negative
                                                adjustment = -(exponent + subnormal_reciprocal_shift[es]);
                                                // this is the right shift adjustment required for subnormal representation due 
                                                // to the scale of the input number, i.e. the exponent of 2^-adjustment
                                        }
                                        else {
                                                // the value is a normal number in this representation: common case
                                                biasedExponent = static_cast<uint32_t>(exponent + EXP_BIAS); // project the exponent into the target 
                                                // fraction processing
                                                // float structure is: seee'eeee'efff'ffff'ffff'ffff'ffff'ffff, s = sign, e - exponent bit, f = fraction bit
                                                // target structure is for example cfloat<8,2>: seef'ffff
                                                // since both are normals, we can shift the incoming fraction to the target structure bits, and round
                                                // MSB of source = 23 - 1, MSB of target = fbits - 1: shift = MSB of src - MSB of tgt => 23 - fbits
                                                adjustment = 0;
                                        }
                                        if constexpr (rightShift > 0) {
                                                // if true we need to round
                                                // round-to-even logic
                                                //  ... lsb | guard  round sticky   round
                                                //       x     0       x     x       down
                                                //       0     1       0     0       down  round to even
                                                //       1     1       0     0        up   round to even
                                                //       x     1       0     1        up
                                                //       x     1       1     0        up
                                                //       x     1       1     1        up
                                                // collect lsb, guard, round, and sticky bits


#if TRACE_CONVERSION
                                                std::cout << "fraction bits     : " << to_binary(rawFraction, ieee754_parameter<Real>::nbits, true) << '\n';
                                                std::cout << "lsb mask bits     : " << to_binary(mask, ieee754_parameter<Real>::nbits, true) << '\n';
#endif
                                                mask = (1ull << (rightShift + adjustment)); // bit mask for the lsb bit
                                                bool lsb = (mask & rawFraction);
                                                mask >>= 1;
                                                bool guard = (mask & rawFraction);
                                                mask >>= 1;
                                                bool round = (mask & rawFraction);
                                                if ((rightShift + adjustment) > 1) {
                                                        mask = (0xFFFF'FFFF'FFFF'FFFFull << (rightShift + adjustment - 2));
                                                        mask = ~mask;
                                                }
                                                else {
                                                        mask = 0;
                                                }
#if TRACE_CONVERSION
                                                std::cout << "right shift       : " << rightShift << '\n';
                                                std::cout << "adjustment        : " << adjustment << '\n';
                                                std::cout << "shift to LSB      : " << (rightShift + adjustment) << '\n';
                                                std::cout << "fraction bits     : " << to_binary(rawFraction, ieee754_parameter<Real>::nbits, true) << '\n';
                                                std::cout << "sticky mask bits  : " << to_binary(mask, ieee754_parameter<Real>::nbits, true) << '\n';
#endif
                                                bool sticky = (mask & rawFraction);
                                                rawFraction >>= (static_cast<int64_t>(rightShift) + static_cast<int64_t>(adjustment));

                                                // execute rounding operation
                                                if (guard) {
                                                        if (lsb && (!round && !sticky)) ++rawFraction; // round to even
                                                        if (round || sticky) ++rawFraction;
                                                        if (rawFraction == (1ull << fbits)) { // overflow
                                                                if (biasedExponent == ALL_ONES_ES) { // overflow to INF == .111..01
                                                                        rawFraction = INF_ENCODING;
                                                                }
                                                                else {
                                                                        ++biasedExponent;
                                                                        rawFraction = 0;
                                                                }
                                                        }
                                                }
#if TRACE_CONVERSION
                                                std::cout << "lsb               : " << (lsb ? "1\n" : "0\n");
                                                std::cout << "guard             : " << (guard ? "1\n" : "0\n");
                                                std::cout << "round             : " << (round ? "1\n" : "0\n");
                                                std::cout << "sticky            : " << (sticky ? "1\n" : "0\n");
                                                std::cout << "rounding decision : " << (lsb && (!round && !sticky) ? "round to even\n" : "-\n");
                                                std::cout << "rounding direction: " << (round || sticky ? "round up\n" : "round down\n");
#endif
                                        }
                                        else { // all bits of the float go into this representation and need to be shifted up, no rounding necessary
                                                int shiftLeft = fbits - ieee754_parameter<Real>::fbits;
                                                rawFraction <<= shiftLeft;
                                        }
#if TRACE_CONVERSION
                                        std::cout << "biased exponent   : " << biasedExponent << " : 0x" << std::hex << biasedExponent << std::dec << '\n';
                                        std::cout << "right shift       : " << rightShift << '\n';
                                        std::cout << "adjustment shift  : " << adjustment << '\n';
                                        std::cout << "sticky bit mask   : " << to_binary(mask, 32, true) << '\n';
                                        std::cout << "fraction bits     : " << to_binary(rawFraction, 32, true) << '\n';
#endif
                                        // construct the target cfloat
                                        bits = (s ? 1ull : 0ull);
                                        bits <<= es;
                                        bits |= biasedExponent;
                                        bits <<= fbits;
                                        bits |= rawFraction;
#if TRACE_CONVERSION
                                        std::cout << "sign bit          : " << (s ? '1' : '0') << '\n';
                                        std::cout << "biased exponent   : " << biasedExponent << " : 0x" << std::hex << biasedExponent << std::dec << '\n';
                                        std::cout << "fraction bits     : " << to_binary(rawFraction, 32, true) << '\n';
                                        std::cout << "cfloat bits       : " << to_binary(bits, nbits, true) << '\n';
#endif
                                        setbits(bits);
                                }
                                else {
                                        // the source real is a subnormal number                                
                                        mask = 0x00FF'FFFFu >> (fbits + exponent + subnormal_reciprocal_shift[es] + 1); // mask for sticky bit 

                                        // fraction processing: we have fbits+1 bits = 1 hidden + fbits explicit fraction bits 
                                        // f = 1.ffff  2^exponent * 2^fbits * 2^-(2-2^(es-1)) = 1.ff...ff >> (23 - (-exponent + fbits - (2 -2^(es-1))))
                                        // -exponent because we are right shifting and exponent in this range is negative
                                        adjustment = -(exponent + subnormal_reciprocal_shift[es]); // this is the right shift adjustment due to the scale of the input number, i.e. the exponent of 2^-adjustment
#if TRACE_CONVERSION                                        
                                        std::cout << "source is subnormal: TBD\n";
                                        std::cout << "shift to LSB    " << (rightShift + adjustment) << '\n';
                                        std::cout << "adjustment      " << adjustment << '\n';
                                        std::cout << "exponent        " << exponent << '\n';
                                        std::cout << "subnormal shift " << subnormal_reciprocal_shift[es] << '\n';
#endif
                                        if (exponent >= (MIN_EXP_SUBNORMAL - 1) && exponent < MIN_EXP_NORMAL) {
                                                // the value is a subnormal number in this representation
                                        }
                                        else {
                                                // the value is a normal number in this representation
                                        }
                                }
                        }
                        else {
                                // no need to round, but we need to shift left to deliver the bits
                                // cfloat<40,  8> = float
                                // cfloat<48,  9> = float
                                // cfloat<56, 10> = float
                                // cfloat<64, 11> = float
                                // cfloat<64, 10> = double 
                                // can we go from an input subnormal to a cfloat normal? 
                                // yes, for example a cfloat<64,11> assigned to a subnormal float

                                // map exponent into target cfloat encoding
                                constexpr int upshift = fbits - ieee754_parameter<Real>::fbits;
                                uint64_t biasedExponent{ 0 };
                                bool subnormalInTarget = (exponent < MIN_EXP_NORMAL);
                                int denormShift = 0;
                                if (subnormalInTarget) {
                                        biasedExponent = 0;
                                        denormShift = MIN_EXP_NORMAL - exponent;
                                }
                                else {
                                        biasedExponent = static_cast<uint64_t>(static_cast<int64_t>(exponent) + EXP_BIAS);
                                }
                                // output processing
                                if constexpr (nbits < 65) {
                                        // we can compose the bits in a native 64-bit unsigned integer
                                        // common case: normal to normal
                                        // reference example: nbits = 40, es = 8, fbits = 31: rhs = float fbits = 23; shift left by (31 - 23) = 8

                                        if (rawExponent != 0) {
                                                // rhs is a normal encoding (or a normalized former-subnormal)
                                                uint64_t raw{ s ? 1ull : 0ull };
                                                raw <<= es;
                                                raw |= biasedExponent;
                                                raw <<= fbits;
                                                if (subnormalInTarget) {
                                                        // add the hidden bit and denormalize the fraction
                                                        uint64_t significand = rawFraction | ieee754_parameter<Real>::hmask;
                                                        int netShift = upshift - denormShift;
                                                        if (netShift >= 0) {
                                                                rawFraction = significand << netShift;
                                                        }
                                                        else {
                                                                // right shift with round-to-nearest-even
                                                                int rshift = -netShift;
                                                                uint64_t lsbMask = (1ull << rshift);
                                                                bool lsb    = (significand & lsbMask) != 0;
                                                                bool guard  = (rshift >= 1) && ((significand & (lsbMask >> 1)) != 0);
                                                                bool round  = (rshift >= 2) && ((significand & (lsbMask >> 2)) != 0);
                                                                bool sticky = (rshift >= 3) && ((significand & ((lsbMask >> 2) - 1)) != 0);
                                                                rawFraction = significand >> rshift;
                                                                if (guard) {
                                                                        if (lsb && (!round && !sticky)) ++rawFraction; // round to even
                                                                        if (round || sticky) ++rawFraction;
                                                                }
                                                        }
                                                }
                                                else {
                                                        rawFraction <<= upshift;
                                                }
                                                raw |= rawFraction;
                                                setbits(raw);
                                        }
                                        else {
                                                // rhs is a subnormal (unreachable after normalization above, kept for safety)
                                        }
                                }
                                else {
                                        // we need to write and shift bits into place
                                        // use cases are cfloats like cfloat<80, 11, bt>
                                        // even though the bits that come in are single or double precision
                                        // we need to write the fields and then shifting them in place
                                        // 
                                        // common case: normal to normal
                                        if constexpr (bitsInBlock < 64) {
                                                if (rawExponent != 0) {
                                                        // reference example: nbits = 128, es = 15, fbits = 112: rhs = float: shift left by (112 - 23) = 89
                                                        setbits(biasedExponent);
                                                        shiftLeft(fbits);
                                                        // if subnormal in target, add hidden bit before copying
                                                        uint64_t fractionToCopy = subnormalInTarget
                                                                ? (rawFraction | ieee754_parameter<Real>::hmask)
                                                                : rawFraction;
                                                        // shift fraction bits
                                                        int bitsToShift = subnormalInTarget ? (upshift - denormShift) : upshift;
                                                        // for subnormal targets near minpos, bitsToShift can be negative (right shift)
                                                        // apply rounding on the 64-bit significand before splitting into blocks
                                                        if (bitsToShift < 0) {
                                                                int rshift = -bitsToShift;
                                                                uint64_t lsbMask = (1ull << rshift);
                                                                bool lsb    = (fractionToCopy & lsbMask) != 0;
                                                                bool guard  = (rshift >= 1) && ((fractionToCopy & (lsbMask >> 1)) != 0);
                                                                bool round  = (rshift >= 2) && ((fractionToCopy & (lsbMask >> 2)) != 0);
                                                                bool sticky = (rshift >= 3) && ((fractionToCopy & ((lsbMask >> 2) - 1)) != 0);
                                                                fractionToCopy >>= rshift;
                                                                if (guard) {
                                                                        if (lsb && (!round && !sticky)) ++fractionToCopy; // round to even
                                                                        if (round || sticky) ++fractionToCopy;
                                                                }
                                                                bitsToShift = 0;
                                                        }
                                                        bt fractionBlock[nrBlocks]{ 0 };
                                                        // copy fraction bits
                                                        unsigned blocksRequired = (8 * sizeof(fractionToCopy) + 1) / bitsInBlock;
                                                        unsigned maxBlockNr = (blocksRequired < nrBlocks ? blocksRequired : nrBlocks);
                                                        uint64_t mask = static_cast<uint64_t>(ALL_ONES); // set up the block mask
                                                        unsigned shift = 0;
                                                        for (unsigned i = 0; i < maxBlockNr; ++i) {
                                                                fractionBlock[i] = bt((mask & fractionToCopy) >> shift);
                                                                mask <<= bitsInBlock;
                                                                shift += bitsInBlock;
                                                        }
                                                        if (bitsToShift >= static_cast<int>(bitsInBlock)) {
                                                                int blockShift = static_cast<int>(bitsToShift / bitsInBlock);
                                                                for (int i = MSU; i >= blockShift; --i) {
                                                                        fractionBlock[i] = fractionBlock[i - blockShift];
                                                                }
                                                                for (int i = blockShift - 1; i >= 0; --i) {
                                                                        fractionBlock[i] = bt(0);
                                                                }
                                                                // adjust the shift
                                                                bitsToShift -= blockShift * bitsInBlock;
                                                        }
                                                        if (bitsToShift > 0) {
                                                                // construct the mask for the upper bits in the block that need to move to the higher word
                                                                bt bitsToMoveMask = bt(ALL_ONES << (bitsInBlock - bitsToShift));
                                                                for (unsigned i = MSU; i > 0; --i) {
                                                                        fractionBlock[i] <<= bitsToShift;
                                                                        // mix in the bits from the right
                                                                        bt fracbits = static_cast<bt>(bitsToMoveMask & fractionBlock[i - 1]); // operator & yields an int
                                                                        fractionBlock[i] |= (fracbits >> (bitsInBlock - bitsToShift));
                                                                }
                                                                fractionBlock[0] <<= bitsToShift;
                                                        }
                                                        // OR the bits in
                                                        for (unsigned i = 0; i <= MSU; ++i) {
                                                                _block[i] |= fractionBlock[i];
                                                        }
                                                        // enforce precondition for fast comparison by properly nulling bits that are outside of nbits
                                                        _block[MSU] &= MSU_MASK;
                                                        // finally, set the sign bit
                                                        setsign(s);
                                                }
                                                else {
                                                        // rhs is a subnormal
                //                                        std::cerr << "rhs is a subnormal : " << to_binary(rhs) << " : " << rhs << '\n';
                                                }
                                        }
                                        else {
                                                // BlockType is incorrect
                                        }
                                }
                        }
                        // post-processing results to implement saturation and projection after rounding logic
                        if constexpr (isSaturating) {
                                if (isinf(INF_TYPE_POSITIVE) || isnan(NAN_TYPE_QUIET)) {
                                        maxpos();
                                }
                                else if (isinf(INF_TYPE_NEGATIVE) || isnan(NAN_TYPE_SIGNALLING)) {
                                        maxneg();
                                }
                        }
                        else {
                                if (isnan(NAN_TYPE_QUIET)) {
                                        setinf(false);
                                }
                                else if (isnan(NAN_TYPE_SIGNALLING)) {
                                        setinf(true);
                                }
                        }
                        return *this;  // TODO: unreachable in some configurations        
                }
        }

        // post-processing results to implement saturation and projection after rounding logic
        // arithmetic bit operations can't produce NaN encodings, so we need to re-interpret
        // these encodings and 'project' them to the proper values.
        void constexpr post_process() noexcept {
                if constexpr (isSaturating) {
                        if (isinf(INF_TYPE_POSITIVE) || isnan(NAN_TYPE_QUIET)) {
                                maxpos();
                        }
                        else if (isinf(INF_TYPE_NEGATIVE) || isnan(NAN_TYPE_SIGNALLING)) {
                                maxneg();
                        }
                }
                else {
                        if (isnan(NAN_TYPE_QUIET)) {
                                setinf(false);
                        }
                        else if (isnan(NAN_TYPE_SIGNALLING)) {
                                setinf(true);
                        }
                }
        }

protected:

        /// <summary>
        /// round a set of source bits to the present representation.
        /// srcbits is the number of bits of significant in the source representation
        /// </summary>
        /// <typeparam name="StorageType"></typeparam>
        /// <param name="raw"></param>
        /// <returns></returns>
        template<unsigned srcbits, typename StorageType>
        constexpr uint64_t round(StorageType raw, int& exponent) noexcept {
                if constexpr (fhbits < srcbits) {
                        // round to even: lsb guard round sticky
                    // collect guard, round, and sticky bits
                    // this same logic will work for the case where
                    // we only have a guard bit and no round and sticky bits
                    // because the mask logic will make round and sticky both 0
                        constexpr uint32_t shift = srcbits - fhbits - 1ull;
                        StorageType mask = (StorageType(1ull) << shift);
                        bool guard = (mask & raw);
                        mask >>= 1;
                        bool round = (mask & raw);
                        if constexpr (shift > 1u) { // protect against a negative shift
                                StorageType allones(StorageType(~0));
                                mask = StorageType(allones << (shift - 1));
                                mask = ~mask;
                        }
                        else {
                                mask = 0;
                        }
                        bool sticky = (mask & raw);

                        raw >>= (shift + 1);  // shift out the bits we are rounding away
                        bool lsb = (raw & 0x1u);
                        //  ... lsb | guard  round sticky   round
                        //       x     0       x     x       down
                        //       0     1       0     0       down  round to even
                        //       1     1       0     0        up   round to even
                        //       x     1       0     1        up
                        //       x     1       1     0        up
                        //       x     1       1     1        up
                        if (guard) {
                                if (lsb && (!round && !sticky)) ++raw; // round to even
                                if (round || sticky) ++raw;
                                if (raw == (1ull << fbits)) { // overflow into next power of two
                                        ++exponent;
                                        raw = 0; // fraction is zero after carry into hidden bit
                                }
                        }
                }
                else {
                        // Target has more precision than source - need to left-shift to align
                        // For large types where fhbits > 64, the fraction bits cannot fit in
                        // a 64-bit raw after shifting. In this case, skip the shift and let
                        // convert_signed/unsigned_integer place bits using setbit().
                        // The caller positions fraction bits at the top of srcbits, and we
                        // need to ensure they don't overflow 64 bits after our shift.
                        constexpr unsigned shift = fhbits - srcbits;
                        if constexpr (fhbits <= 64 && shift < 64) {
                                raw <<= shift;
                        }
                        // else: raw stays as-is; caller will extract bits and place them
                }
                uint64_t significant = raw;
                return significant;
        }

        template<typename ArgumentBlockType>
        constexpr void copyBits(ArgumentBlockType v) {
                unsigned blocksRequired = (8 * sizeof(v) + 1 ) / bitsInBlock;
                unsigned maxBlockNr = (blocksRequired < nrBlocks ? blocksRequired : nrBlocks);
                bt b{ 0ul }; b = bt(~b);
                ArgumentBlockType mask = ArgumentBlockType(b);
                unsigned shift = 0;
                for (unsigned i = 0; i < maxBlockNr; ++i) {
                        _block[i] = bt((mask & v) >> shift);
                        mask <<= bitsInBlock;
                        shift += bitsInBlock;
                }
        }
        void shiftLeft(int leftShift) {
                if (leftShift == 0) return;
                if (leftShift < 0) return shiftRight(-leftShift);
                if (leftShift > long(nbits)) leftShift = nbits; // clip to max
                if (leftShift >= long(bitsInBlock)) {
                        int blockShift = leftShift / static_cast<int>(bitsInBlock);
                        for (signed i = signed(MSU); i >= blockShift; --i) {
                                _block[i] = _block[i - blockShift];
                        }
                        for (signed i = blockShift - 1; i >= 0; --i) {
                                _block[i] = bt(0);
                        }
                        // adjust the shift
                        leftShift -= (long)(blockShift * bitsInBlock);
                        if (leftShift == 0) return;
                }
                // construct the mask for the upper bits in the block that need to move to the higher word
//                bt mask = static_cast<bt>(0xFFFFFFFFFFFFFFFFull << (bitsInBlock - leftShift));
                bt mask = ALL_ONES;
                mask <<= (bitsInBlock - leftShift);
                for (unsigned i = MSU; i > 0; --i) {
                        _block[i] <<= leftShift;
                        // mix in the bits from the right
                        bt bits = static_cast<bt>(mask & _block[i - 1]);
                        _block[i] |= (bits >> (bitsInBlock - leftShift));
                }
                _block[0] <<= leftShift;
        }
        void shiftRight(int rightShift) {
                if (rightShift == 0) return;
                if (rightShift < 0) return shiftLeft(-rightShift);
                if (rightShift >= long(nbits)) {
                        setzero();
                        return;
                }
                bool signext = sign();
                unsigned blockShift = 0;
                if (rightShift >= long(bitsInBlock)) {
                        blockShift = rightShift / bitsInBlock;
                        if (MSU >= blockShift) {
                                // shift by blocks
                                for (unsigned i = 0; i <= MSU - blockShift; ++i) {
                                        _block[i] = _block[i + blockShift];
                                }
                        }
                        // adjust the shift
                        rightShift -= (long)(blockShift * bitsInBlock);
                        if (rightShift == 0) {
                                // fix up the leading zeros if we have a negative number
                                if (signext) {
                                        // rightShift is guaranteed to be less than nbits
                                        rightShift += (long)(blockShift * bitsInBlock);
                                        for (unsigned i = nbits - rightShift; i < nbits; ++i) {
                                                this->setbit(i);
                                        }
                                }
                                else {
                                        // clean up the blocks we have shifted clean
                                        rightShift += (long)(blockShift * bitsInBlock);
                                        for (unsigned i = nbits - rightShift; i < nbits; ++i) {
                                                this->setbit(i, false);
                                        }
                                }
                                return;  // shift was aligned to block boundary, no per-bit shift needed
                        }
                }

                bt mask = ALL_ONES;
                mask >>= (bitsInBlock - rightShift); // this is a mask for the lower bits in the block that need to move to the lower word
                for (unsigned i = 0; i < MSU; ++i) {  // TODO: can this be improved? we should not have to work on the upper blocks in case we block shifted
                        _block[i] >>= rightShift;
                        // mix in the bits from the left
                        bt bits = static_cast<bt>(mask & _block[i + 1]); // & operator returns an int
                        _block[i] |= (bits << (bitsInBlock - rightShift));
                }
                _block[MSU] >>= rightShift;

                // fix up the leading zeros if we have a negative number
                if (signext) {
                        // bitsToShift is guaranteed to be less than nbits
                        rightShift += (long)(blockShift * bitsInBlock);
                        for (unsigned i = nbits - rightShift; i < nbits; ++i) {
                                this->setbit(i);
                        }
                }
                else {
                        // clean up the blocks we have shifted clean
                        rightShift += (long)(blockShift * bitsInBlock);
                        for (unsigned i = nbits - rightShift; i < nbits; ++i) {
                                this->setbit(i, false);
                        }
                }

                // enforce precondition for fast comparison by properly nulling bits that are outside of nbits
                _block[MSU] &= MSU_MASK;
        }

        // calculate the integer power 2 ^ b using exponentiation by squaring
        constexpr double ipow(int exponent) const {
                bool negative = (exponent < 0);
                exponent = negative ? -exponent : exponent;
                double result(1.0);
                double base = 2.0;
                for (;;) {
                        if (exponent % 2) result *= base;
                        exponent >>= 1;
                        if (exponent == 0) break;
                        base *= base;
                }
                return (negative ? (1.0 / result) : result);
        }

        template<BlockTripleOperator btop, typename TargetBlockType = bt>
        constexpr void blockcopy(blocktriple<fbits, btop, TargetBlockType>& tgt) const {
                // brute force copy of blocks
                if constexpr (1 == fBlocks) {
                        tgt.setblock(0, static_cast<TargetBlockType>(_block[0] & FSU_MASK));
                }
                else if constexpr (2 == fBlocks) {
                        tgt.setblock(0, static_cast<TargetBlockType>(_block[0]));
                        tgt.setblock(1, static_cast<TargetBlockType>(_block[1] & FSU_MASK));
                }
                else if constexpr (3 == fBlocks) {
                        tgt.setblock(0, static_cast<TargetBlockType>(_block[0]));
                        tgt.setblock(1, static_cast<TargetBlockType>(_block[1]));
                        tgt.setblock(2, static_cast<TargetBlockType>(_block[2] & FSU_MASK));
                }
                else if constexpr (4 == fBlocks) {
                        tgt.setblock(0, static_cast<TargetBlockType>(_block[0]));
                        tgt.setblock(1, static_cast<TargetBlockType>(_block[1]));
                        tgt.setblock(2, static_cast<TargetBlockType>(_block[2]));
                        tgt.setblock(3, static_cast<TargetBlockType>(_block[3] & FSU_MASK));
                }
                else if constexpr (5 == fBlocks) {
                        tgt.setblock(0, static_cast<TargetBlockType>(_block[0]));
                        tgt.setblock(1, static_cast<TargetBlockType>(_block[1]));
                        tgt.setblock(2, static_cast<TargetBlockType>(_block[2]));
                        tgt.setblock(3, static_cast<TargetBlockType>(_block[3]));
                        tgt.setblock(4, static_cast<TargetBlockType>(_block[4] & FSU_MASK));
                }
                else if constexpr (6 == fBlocks) {
                        tgt.setblock(0, static_cast<TargetBlockType>(_block[0]));
                        tgt.setblock(1, static_cast<TargetBlockType>(_block[1]));
                        tgt.setblock(2, static_cast<TargetBlockType>(_block[2]));
                        tgt.setblock(3, static_cast<TargetBlockType>(_block[3]));
                        tgt.setblock(4, static_cast<TargetBlockType>(_block[4]));
                        tgt.setblock(5, static_cast<TargetBlockType>(_block[5] & FSU_MASK));
                }
                else if constexpr (7 == fBlocks) {
                        tgt.setblock(0, static_cast<TargetBlockType>(_block[0]));
                        tgt.setblock(1, static_cast<TargetBlockType>(_block[1]));
                        tgt.setblock(2, static_cast<TargetBlockType>(_block[2]));
                        tgt.setblock(3, static_cast<TargetBlockType>(_block[3]));
                        tgt.setblock(4, static_cast<TargetBlockType>(_block[4]));
                        tgt.setblock(5, static_cast<TargetBlockType>(_block[5]));
                        tgt.setblock(6, static_cast<TargetBlockType>(_block[6] & FSU_MASK));
                }
                else if constexpr (8 == fBlocks) {
                        tgt.setblock(0, static_cast<TargetBlockType>(_block[0]));
                        tgt.setblock(1, static_cast<TargetBlockType>(_block[1]));
                        tgt.setblock(2, static_cast<TargetBlockType>(_block[2]));
                        tgt.setblock(3, static_cast<TargetBlockType>(_block[3]));
                        tgt.setblock(4, static_cast<TargetBlockType>(_block[4]));
                        tgt.setblock(5, static_cast<TargetBlockType>(_block[5]));
                        tgt.setblock(6, static_cast<TargetBlockType>(_block[6]));
                        tgt.setblock(7, static_cast<TargetBlockType>(_block[7] & FSU_MASK));
                }
                else {
                        for (unsigned i = 0; i < FSU; ++i) {
                                tgt.setblock(i, static_cast<TargetBlockType>(_block[i]));
                        }
                        tgt.setblock(FSU, static_cast<TargetBlockType>(_block[FSU] & FSU_MASK));
                }
        }

private:
        bt _block[nrBlocks];

        //////////////////////////////////////////////////////////////////////////////
        // friend functions

        // template parameters need names different from class template parameters (for gcc and clang)
        template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
        friend std::ostream& operator<< (std::ostream& ostr, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& r);
        template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
        friend std::istream& operator>> (std::istream& istr, cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& r);

        template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
        friend constexpr bool operator==(const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& lhs, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& rhs);
        template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
        friend constexpr bool operator!=(const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& lhs, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& rhs);
        template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
        friend constexpr bool operator< (const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& lhs, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& rhs);
        template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
        friend constexpr bool operator> (const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& lhs, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& rhs);
        template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
        friend constexpr bool operator<=(const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& lhs, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& rhs);
        template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
        friend constexpr bool operator>=(const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& lhs, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& rhs);
};

///////////////////////////// IOSTREAM operators ///////////////////////////////////////////////

// convert cfloat to decimal fixpnt string, i.e. "-1234.5678"
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
std::string to_decimal_fixpnt_string(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& value, long long precision) {
        constexpr unsigned fbits = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>::fbits;
        constexpr unsigned bias = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>::EXP_BIAS;
        std::stringstream str;
        if (value.iszero()) {
                str << '0';
                return str.str();
        }
        if (value.sign()) str << '-';

        // construct the discretization levels of the fraction part
        support::decimal range, discretizationLevels, step;
        // create the decimal range we are discretizing
        range.setdigit(1);
        range.shiftLeft(fbits); // the decimal range of the fraction
        discretizationLevels.powerOf2(fbits); // calculate the discretization levels of this range
        step = div(range, discretizationLevels);
        // now construct the value of this range by adding the fraction samples
        support::decimal partial, multiplier;
        partial.setzero();  // if you just want the fraction
        multiplier.setdigit(1);
        // convert the fraction part
        for (unsigned i = 0; i < fbits; ++i) {
                if (value.at(i)) {
                        support::add(partial, multiplier);
                }
                support::add(multiplier, multiplier);
        }
        if (value.isdenormal()) {
                support::mul(partial, step);
                support::decimal scale;
                scale.powerOf2(bias - 1ull);
                partial = support::div(partial, scale);
        } 
        else {
                support::add(partial, multiplier); // add the hidden bit
                support::mul(partial, step);
                support::decimal scale;
                int exponent = value.scale();
                if (exponent < 0) {
                        scale.powerOf2(static_cast<unsigned>(-exponent));
                        partial = support::div(partial, scale);
                }
                else {
                        scale.powerOf2(static_cast<unsigned>(exponent));
                        support::mul(partial, scale);
                }
        }

        // the radix is at fbits
        // The partial represents the parts in the range, so we can deduce
        // the number of leading zeros by comparing to the length of range
        int nrLeadingZeros = static_cast<int>(range.size()) - static_cast<int>(partial.size()) - 1;
        if (nrLeadingZeros >= 0) str << "0.";
        for (int i = 0; i < nrLeadingZeros; ++i) str << '0';
        int digitsWritten = (nrLeadingZeros > 0) ? nrLeadingZeros : 0;
        int position = static_cast<int>(partial.size()) - 1;
        for (support::decimal::const_reverse_iterator rit = partial.rbegin(); rit != partial.rend(); ++rit) {
                str << (int)*rit;
                ++digitsWritten;
                if (position == fbits) str << '.';
                --position;
        }
        if (digitsWritten < precision) { // deal with trailing 0s
                for (unsigned i = static_cast<unsigned>(digitsWritten); i < fbits; ++i) {
                        str << '0';
                }
        }

        return str.str();
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
std::string to_string(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& value, long long precision) {
        constexpr unsigned fbits = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>::fbits;
        std::stringstream str;
        if (value.iszero()) {
                str << '0';
                return str.str();
        }
        if (value.sign()) str << '-';

        // denormalize the number to gain access to the most sigificant digits
        // 1.ffff^e
        // scale is e
        // lsbScale is e - fbits
        // shift to get lsb to position 2^0 = (e - fbits)
        std::int64_t scale = value.scale();
//        std::int64_t shift = scale + fbits; // we want the lsb at 2^0
        std::int64_t lsbScale = scale - fbits;  // scale of the lsb
        support::decimal partial, multiplier;
        partial.setzero();

        multiplier.powerOf2(lsbScale);

        // convert the fraction bits 
        for (unsigned i = 0; i < fbits; ++i) {
                if (value.at(i)) {
                        support::add(partial, multiplier);
                }
                support::add(multiplier, multiplier);
        }
        if (!value.isdenormal()) {
                support::add(partial, multiplier); // add the hidden bit
        }
        str << partial;
        return str.str();
}

//////////////////////////////////////////////////////////////////////////////////////////////
/// stream operators

// ostream output generates an ASCII format for the floating-point argument
// Uses native binary-to-decimal conversion via blocktriple::to_string()
// to produce exact output for all cfloat sizes without double conversion.
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
inline std::ostream& operator<<(std::ostream& ostr, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& v) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        constexpr unsigned cfbits = Cfloat::fbits;

        std::streamsize prec  = ostr.precision();
        std::streamsize width = ostr.width();
        std::ios_base::fmtflags ff = ostr.flags();
        bool bFixed      = (ff & std::ios_base::fixed) == std::ios_base::fixed;
        bool bScientific = (ff & std::ios_base::scientific) == std::ios_base::scientific;
        bool bShowpos    = (ff & std::ios_base::showpos) != 0;
        bool bUppercase  = (ff & std::ios_base::uppercase) != 0;
        bool bInternal   = (ff & std::ios_base::internal) != 0;
        bool bLeft       = (ff & std::ios_base::left) != 0;
        char fillChar    = ostr.fill();

        if constexpr (cfbits == 0) {
                // degenerate cfloat with no fraction bits: fall back to double
                std::ostringstream oss;
                oss.precision(prec);
                if (bFixed) oss << std::fixed;
                if (bScientific) oss << std::scientific;
                if (bUppercase) oss << std::uppercase;
                if (bShowpos) oss << std::showpos;
                oss << static_cast<double>(v);
                std::string s = oss.str();
                if (width > 0 && s.length() < static_cast<size_t>(width)) {
                        size_t pad = static_cast<size_t>(width) - s.length();
                        if (bLeft) { s.append(pad, fillChar); }
                        else { s.insert(0u, pad, fillChar); }
                }
                return ostr << s;
        } else {
                blocktriple<cfbits, BlockTripleOperator::REP, bt> a;
                v.normalize(a);
                return ostr << a.to_string(prec, width, bFixed, bScientific,
                                           bInternal, bLeft, bShowpos, bUppercase, fillChar);
        }
}

// parse a cfloat from a string in either cfloat hex format (nbits.esxHEXVALUEc)
// or a decimal floating-point representation
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
bool parse(const std::string& txt, cfloat<nbits,es,bt,hasSubnormals,hasMaxExpValues,isSaturating>& v) {
        // check if the txt is of the native cfloat form: nbits.esX[0x]hexvaluec
        std::regex cfloat_regex(R"(^[0-9]+\.[0-9]+[xX](0[xX])?[0-9A-Fa-f]+c?$)");
        if (std::regex_match(txt, cfloat_regex)) {
                // found a cfloat representation: parse nbits.esxHEXVALUEc
                std::string nbitsStr, esStr, bitStr;
                auto it = txt.begin();
                for (; it != txt.end(); ++it) {
                        if (*it == '.') break;
                        nbitsStr.append(1, *it);
                }
                for (++it; it != txt.end(); ++it) {
                        if (*it == 'x' || *it == 'X') break;
                        esStr.append(1, *it);
                }
                for (++it; it != txt.end(); ++it) {
                        if (*it == 'c') break;
                        bitStr.append(1, *it);
                }
                unsigned nbits_in = 0;
                unsigned es_in = 0;
                {
                        std::istringstream ss(nbitsStr);
                        ss >> nbits_in;
                        if (ss.fail()) return false;
                }
                {
                        std::istringstream ss(esStr);
                        ss >> es_in;
                        if (ss.fail()) return false;
                }
                // native cfloat form must match target configuration
                if (nbits_in != nbits || es_in != es) return false;
                uint64_t raw = 0;
                std::istringstream ss(bitStr);
                ss >> std::hex >> raw;
                if (ss.fail()) return false;
                ss >> std::ws;
                if (!ss.eof()) return false;
                v.setbits(raw);
                return true;
        }
        else {
                // Decimal floating-point representation.
                // Route through the high-precision decimal_to_binary utility so that
                // wide cfloat configurations (nbits > 64, including IEEE quad and
                // posit-killer formats) don't lose precision through an intermediate
                // double. The utility delivers a normalized mantissa with
                // target_mantissa_bits bits plus guard/sticky; we package that as a
                // blocktriple and hand it to convert(blocktriple, cfloat), which
                // handles all IEEE-754 edge cases (subnormals, saturation, etc.).
                // Special-value tokens (nan / inf in any common spelling) are
                // recognised directly so callers don't depend on locale-sensitive
                // std::istringstream parsing of those literals.
                using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
                {
                        std::string t; t.reserve(txt.size());
                        for (char c : txt) t.push_back(static_cast<char>(std::tolower(static_cast<unsigned char>(c))));
                        bool negative = !t.empty() && t.front() == '-';
                        std::string body = t;
                        if (!body.empty() && (body.front() == '+' || body.front() == '-')) body.erase(0, 1);
                        if (body == "nan") {
                                v.setnan(NAN_TYPE_QUIET);
                                return true;
                        }
                        if (body == "inf" || body == "infinity") {
                                v.setinf(negative);
                                return true;
                        }
                }
                // We pack the d2b result into a blocktriple<fbits, MUL, bt>. The MUL
                // layout has bfbits = 2*fbits + 2 and radix = 2*fbits, which gives us
                // exactly fbits of headroom below the cfloat fraction for guard/
                // sticky -- enough room for cfloat's convert() to make a correct
                // round-to-nearest-even decision. (The REP layout has no such
                // headroom: its fraction is exactly cfloat::fbits wide.)
                using BT = blocktriple<Cfloat::fbits, BlockTripleOperator::MUL, bt>;
                constexpr unsigned radix_pos          = static_cast<unsigned>(BT::radix);
                constexpr unsigned target_mantissa_bits = radix_pos + 1u;
                auto d = ::sw::universal::decimal_to_binary::convert(
                        std::string_view{txt}, target_mantissa_bits);
                if (!d.valid) return false;
                if (d.is_zero) {
                        v.setzero();
                        v.setsign(d.negative);
                        return true;
                }
                BT bt_val;
                bt_val.setnormal();
                bt_val.setsign(d.negative);
                bt_val.setscale(static_cast<int>(d.binary_scale));
                // d2b mantissa has its MSB at position radix_pos (the hidden bit);
                // below it are radix_pos extra precision bits. Copy bit-for-bit
                // into the blocktriple significand aligned to the same radix.
                for (unsigned i = 0; i <= radix_pos; ++i) {
                        if (d.mantissa.at(i)) bt_val.setbit(i, true);
                }
                // Fold d2b's residual guard/sticky into the lowest bit so cfloat's
                // own rounding decision picks them up as sticky tail.
                if (d.guard_bit || d.sticky_bit) bt_val.setbit(0, true);
                convert(bt_val, v);
                return true;
        }
}

// read an ASCII float or cfloat format: nbits.esxNN...NNc, for example: 16.5x7C00c
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
inline std::istream& operator>>(std::istream& istr, cfloat<nbits,es,bt,hasSubnormals,hasMaxExpValues,isSaturating>& v) {
        std::string txt;
        istr >> txt;
        if (!parse(txt, v)) {
                std::cerr << "unable to parse -" << txt << "- into a cfloat value\n";
                istr.setstate(std::ios::failbit);
        }
        return istr;
}

// encoding helpers

// return the Unit in the Last Position
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> ulp(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& a) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> b(a);
        return ++b - a;
}

// transform cfloat to a binary representation
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
inline std::string to_binary(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& number, bool nibbleMarker = false) {
        std::stringstream s;
        s << "0b";
        unsigned index = nbits;
        s << (number.at(--index) ? '1' : '0') << '.';

        for (int i = int(es) - 1; i >= 0; --i) {
                s << (number.at(--index) ? '1' : '0');
                if (i > 0 && (i % 4) == 0 && nibbleMarker) s << '\'';
        }

        s << '.';

        constexpr int fbits = nbits - 1ull - es;
        for (int i = fbits - 1; i >= 0; --i) {
                s << (number.at(--index) ? '1' : '0');
                if (i > 0 && (i % 4) == 0 && nibbleMarker) s << '\'';
        }

        return s.str();
}

// native semantic representation: radix-2, delegates to to_binary
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
inline std::string to_native(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& number, bool nibbleMarker = false) {
        return to_binary(number, nibbleMarker);
}

// transform a cfloat into a triple representation
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
inline std::string to_triple(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& number, bool nibbleMarker = true) {
        std::stringstream s;
        blocktriple<cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>::fbits, BlockTripleOperator::REP, bt> triple;
        number.normalize(triple);
        s << to_triple(triple, nibbleMarker);
        return s.str();
}

// Magnitude of a cfloat (equivalent to turning the sign bit off).
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>
abs(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& v) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> a(v);
        a.setsign(false);
        return a;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>
fabs(cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> v) {
        return abs(v);
}

////////////////////// debug helpers

// convenience method to gain access to the values of the constexpr variables that govern the cfloat behavior
template<unsigned nbits, unsigned es, typename bt = uint8_t, bool hasSubnormals = false, bool hasMaxExpValues = false, bool isSaturating = false>
void ReportCfloatClassParameters() {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> a;
        a.constexprClassParameters();
}

//////////////////////////////////////////////////////
/// cfloat - cfloat binary logic operators

template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
constexpr inline bool operator==(const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& lhs, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& rhs) {
        if (lhs.isnan() || rhs.isnan()) return false;
        if (lhs.iszero() && rhs.iszero()) return true; // +0 == -0 per IEEE-754 §5.11
        for (unsigned i = 0; i < lhs.nrBlocks; ++i) {
                if (lhs._block[i] != rhs._block[i]) {
                        return false;
                }
        }
        return true;
}
template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
constexpr inline bool operator!=(const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& lhs, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& rhs) { return !operator==(lhs, rhs); }
template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
constexpr inline bool operator< (const cfloat<nnbits, nes, nbt, nsub, nsup, nsat>& lhs, const cfloat<nnbits, nes, nbt, nsub, nsup, nsat>& rhs) {
        if (lhs.isnan() || rhs.isnan()) return false;
        // need this as arithmetic difference is defined as snan(indeterminate)
        if (lhs.isinf(INF_TYPE_NEGATIVE) && rhs.isinf(INF_TYPE_NEGATIVE)) return false;
        if (lhs.isinf(INF_TYPE_POSITIVE) && rhs.isinf(INF_TYPE_POSITIVE)) return false;
        if constexpr (nsub) {
                cfloat<nnbits, nes, nbt, nsub, nsup, nsat> diff = (lhs - rhs);
                return (!diff.iszero() && diff.sign()) ? true : false;  // got to guard against -0
        }
        else {
                if (lhs.iszero() && rhs.iszero()) return false;  // we need to 'collapse' all zero encodings
                if (lhs.sign() && !rhs.sign()) return true;
                if (!lhs.sign() && rhs.sign()) return false;
                bool positive = lhs.ispos();
                if (positive) {
                        if (lhs.scale() < rhs.scale()) return true;
                        if (lhs.scale() > rhs.scale()) return false;
                }
                else {
                        if (lhs.scale() > rhs.scale()) return true;
                        if (lhs.scale() < rhs.scale()) return false;
                }
                // sign and scale are the same
                if (lhs.scale() == rhs.scale()) {
                        // compare fractions: we do not have subnormals, so we can ignore the hidden bit
                        blockbinary<nnbits - 1ull - nes, nbt> l, r;
                        lhs.fraction(l);
                        rhs.fraction(r);
                        blockbinary<nnbits - nes, nbt> ll, rr; // fbits + 1 so we can 0 extend to honor 2's complement encoding of blockbinary
                        ll.assignWithoutSignExtend(l);
                        rr.assignWithoutSignExtend(r);
                        return (positive ? (ll < rr) : (ll > rr));
                }
                return false;
        }
}
template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
constexpr inline bool operator> (const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& lhs, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& rhs) { 
        if (lhs.isnan() || rhs.isnan()) return false;
        // need this as arithmetic difference is defined as snan(indeterminate)
        if (lhs.isinf(INF_TYPE_NEGATIVE) && rhs.isinf(INF_TYPE_NEGATIVE)) return false;
        if (lhs.isinf(INF_TYPE_POSITIVE) && rhs.isinf(INF_TYPE_POSITIVE)) return false;
        return  operator< (rhs, lhs); 
}
template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
constexpr inline bool operator<=(const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& lhs, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& rhs) { 
        if (lhs.isnan() || rhs.isnan()) return false;
        return !operator> (lhs, rhs); 
}
template<unsigned nnbits, unsigned nes, typename nbt, bool nsub, bool nsup, bool nsat>
constexpr inline bool operator>=(const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& lhs, const cfloat<nnbits,nes,nbt,nsub,nsup,nsat>& rhs) {
        if (lhs.isnan() || rhs.isnan()) return false;
        return !operator< (lhs, rhs); 
}

//////////////////////////////////////////////////////
/// cfloat - cfloat binary arithmetic operators

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> sum(lhs);
        sum += rhs;
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> mul(lhs);
        mul *= rhs;
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> ratio(lhs);
        ratio /= rhs;
        return ratio;
}

/// binary cfloat - literal arithmetic operators

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(float lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> sum(lhs);
        sum += rhs;
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(float lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(float lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> mul(lhs);
        mul *= rhs;
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(float lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> ratio(lhs);
        ratio /= rhs;
        return ratio;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(double lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> sum(lhs);
        sum += rhs;
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(double lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(double lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> mul(lhs);
        mul *= rhs;
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(double lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> ratio(lhs);
        ratio /= rhs;
        return ratio;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(int lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> sum(lhs);
        sum += rhs;
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(int lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(int lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> mul(lhs);
        mul *= rhs;
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(int lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> ratio(lhs);
        ratio /= rhs;
        return ratio;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(unsigned int lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> sum(lhs);
        sum += rhs;
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(unsigned int lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(unsigned int lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> mul(lhs);
        mul *= rhs;
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(unsigned int lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> ratio(lhs);
        ratio /= rhs;
        return ratio;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(long long lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> sum(lhs);
        sum += rhs;
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(long long lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(long long lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> mul(lhs);
        mul *= rhs;
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(long long lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> ratio(lhs);
        ratio /= rhs;
        return ratio;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(unsigned long long lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> sum(lhs);
        sum += rhs;
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(unsigned long long lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(unsigned long long lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> mul(lhs);
        mul *= rhs;
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(unsigned long long lhs, const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& rhs) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> ratio(lhs);
        ratio /= rhs;
        return ratio;
}

///  binary cfloat - literal arithmetic operators

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, float rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat sum(lhs);
        sum += Cfloat(rhs);
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, float rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, float rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat mul(lhs);
        mul *= Cfloat(rhs);
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, float rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat ratio(lhs);
        ratio /= Cfloat(rhs);
        return ratio;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, double rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat sum(lhs);
        sum += Cfloat(rhs);
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, double rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, double rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat mul(lhs);
        mul *= Cfloat(rhs);
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, double rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat ratio(lhs);
        ratio /= Cfloat(rhs);
        return ratio;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, int rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat sum(lhs);
        sum += Cfloat(rhs);
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, int rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, int rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat mul(lhs);
        mul *= Cfloat(rhs);
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, int rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat ratio(lhs);
        ratio /= Cfloat(rhs);
        return ratio;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, unsigned int rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat sum(lhs);
        sum += Cfloat(rhs);
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, unsigned int rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, unsigned int rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat mul(lhs);
        mul *= Cfloat(rhs);
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, unsigned int rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat ratio(lhs);
        ratio /= Cfloat(rhs);
        return ratio;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long long rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat sum(lhs);
        sum += Cfloat(rhs);
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long long rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long long rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat mul(lhs);
        mul *= Cfloat(rhs);
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long long rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat ratio(lhs);
        ratio /= Cfloat(rhs);
        return ratio;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator+(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, unsigned long long rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat sum(lhs);
        sum += Cfloat(rhs);
        return sum;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator-(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, unsigned long long rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat diff(lhs);
        diff -= rhs;
        return diff;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator*(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, unsigned long long rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat mul(lhs);
        mul *= Cfloat(rhs);
        return mul;
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> operator/(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, unsigned long long rhs) {
        using Cfloat = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>;
        Cfloat ratio(lhs);
        ratio /= Cfloat(rhs);
        return ratio;
}

///////////////////////////////////////////////////////////////////////
///   binary logic literal comparisons

// cfloat - literal float logic operators
// Promote rhs into the cfloat domain rather than narrowing lhs to native float;
// otherwise cfloat<80, 11>(1.0 + 1ulp) == 1.0f reports true because float(lhs)
// rounds the wider value into single precision. Comparisons must happen in the
// shared cfloat domain to preserve precision.
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator==(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, float rhs) {
        return operator==(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator!=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, float rhs) {
        return operator!=(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator< (const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, float rhs) {
        return operator< (lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator> (const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, float rhs) {
        return operator> (lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator<=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, float rhs) {
        return operator<=(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator>=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, float rhs) {
        return operator>=(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
// cfloat - literal double logic operators (same promotion strategy)
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator==(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, double rhs) {
        return operator==(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator!=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, double rhs) {
        return operator!=(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator< (const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, double rhs) {
        return operator< (lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator> (const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, double rhs) {
        return operator> (lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator<=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, double rhs) {
        return operator<=(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator>=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, double rhs) {
        return operator>=(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}

#if LONG_DOUBLE_SUPPORT
// cfloat - literal long double logic operators (same promotion strategy)
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator==(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long double rhs) {
        return operator==(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator!=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long double rhs) {
        return operator!=(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator< (const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long double rhs) {
        return operator< (lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator> (const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long double rhs) {
        return operator> (lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator<=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long double rhs) {
        return operator<=(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator>=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long double rhs) {
        return operator>=(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
#endif

// cfloat - literal int logic operators
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator==(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, int rhs) {
        return operator==(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator!=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, int rhs) {
        return operator!=(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator< (const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, int rhs) {
        return operator<(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator> (const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, int rhs) {
        return operator<(cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs), lhs);
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator<=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, int rhs) {
        return operator<(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs)) || operator==(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator>=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, int rhs) {
        return !operator<(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}

// cfloat - long long logic operators
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator==(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long long rhs) {
        return operator==(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator!=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long long rhs) {
        return operator!=(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator< (const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long long rhs) {
        return operator<(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator> (const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long long rhs) {
        return operator<(cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs), lhs);
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator<=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long long rhs) {
        return operator<(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs)) || operator==(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline bool operator>=(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& lhs, long long rhs) {
        return !operator<(lhs, cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>(rhs));
}

// standard library functions for floating point

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> frexp(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& x, int* exp) {
        *exp = x.scale();
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> fraction(x);
        fraction.setexponent(0);
        return fraction;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> ldexp(const cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& x, int exp) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> result(x);
        int xexp = x.scale();
        result.setexponent(xexp + exp);  // TODO: this does not work for subnormals
        return result;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
constexpr inline cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> 
fma(cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> x,
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> y,
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> z) {
        cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> fused{ 0 };
        constexpr unsigned FBITS = cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>::fbits;
        constexpr unsigned EXTRA_FBITS = FBITS+2;
        constexpr unsigned EXTENDED_PRECISION = nbits + EXTRA_FBITS;
        // the C++ fma spec indicates that the x*y+z is evaluated in 'infinite' precision
        // with only a single rounding event. The minimum finite precision that would behave like this
        // is the precision where the product x*y does not need to be rounded, which will
        // need at least 2*(fbits+1) mantissa bits to capture all bits that can be
        // generated by the product.
        cfloat<EXTENDED_PRECISION, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> preciseX(x), preciseY(y), preciseZ(z);
//        ReportValue(preciseX, "extended precision x");
//        ReportValue(preciseY, "extended precision y");
//        ReportValue(preciseZ, "extended precision z");
        cfloat<EXTENDED_PRECISION, es, bt, hasSubnormals, hasMaxExpValues, isSaturating> product = preciseX * preciseY;
//        ReportValue(product, "extended precision p");
        fused = product + preciseZ;
        return fused;
}

template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>&
minpos(cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& c) {
        return c.minpos();
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>&
maxpos(cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& c) {
        return c.maxpos();
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>&
minneg(cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& c) {
        return c.minneg();
}
template<unsigned nbits, unsigned es, typename bt, bool hasSubnormals, bool hasMaxExpValues, bool isSaturating>
cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>&
maxneg(cfloat<nbits, es, bt, hasSubnormals, hasMaxExpValues, isSaturating>& c) {
        return c.maxneg();
}

}} // namespace sw::universal

stillwater-sc / universal / 26381119807

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous