/*============================================================================ This C source file is part of the SoftPosit Posit Arithmetic Package by S. H. Leong (Cerlane). Copyright 2017, 2018 A*STAR. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions, and the following disclaimer. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions, and the following disclaimer in the documentation and/or other materials provided with the distribution. 3. Neither the name of the University nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. =============================================================================*/ #include #ifdef SOFTPOSIT_QUAD #include #endif #include "platform.h" #include "internals.h" #ifdef SOFTPOSIT_QUAD void checkQuadExtraP32TwoBits(__float128 f32, __float128 temp, bool * bitsNPlusOne, bool * bitsMore ){ temp /= 2; if (temp<=f32){ *bitsNPlusOne = 1; f32-=temp; } if (f32>0) *bitsMore = 1; } uint_fast32_t convertQuadFractionP32(__float128 f32, uint_fast16_t fracLength, bool * bitNPlusOne, bool * bitsMore ){ uint_fast32_t frac=0; if(f32==0) return 0; else if(f32==INFINITY) return 0x80000000; f32 -= 1; //remove hidden bit if (fracLength==0) checkQuadExtraP32TwoBits(f32, 1.0, bitNPlusOne, bitsMore); else{ __float128 temp = 1; while (true){ temp /= 2; if (temp<=f32){ f32-=temp; fracLength--; frac = (frac<<1) + 1; //shift in one if (f32==0){ frac <<= (uint_fast32_t)fracLength; break; } if (fracLength == 0){ checkQuadExtraP32TwoBits(f32, temp, bitNPlusOne, bitsMore); break; } } else{ frac <<= 1; //shift in a zero fracLength--; if (fracLength == 0){ checkQuadExtraP32TwoBits(f32, temp, bitNPlusOne, bitsMore); break; } } } } return frac; } posit32_t convertQuadToP32(__float128 f32){ union ui32_p32 uZ; bool sign, regS; uint_fast32_t reg, frac=0; int_fast32_t exp=0; bool bitNPlusOne=0, bitsMore=0; (f32>=0) ? (sign=0) : (sign=1); if (f32 == 0 ){ uZ.ui = 0; return uZ.p; } else if(f32 == INFINITY || f32 == -INFINITY || f32 == NAN){ uZ.ui = 0x80000000; return uZ.p; } else if (f32 == 1) { uZ.ui = 0x40000000; return uZ.p; } else if (f32 == -1){ uZ.ui = 0xC0000000; return uZ.p; } else if (f32 >= 1.329227995784916e+36){ //maxpos uZ.ui = 0x7FFFFFFF; return uZ.p; } else if (f32 <= -1.329227995784916e+36){ // -maxpos uZ.ui = 0x80000001; return uZ.p; } else if(f32 <= 7.52316384526264e-37 && !sign){ //minpos uZ.ui = 0x1; return uZ.p; } else if(f32 >= -7.52316384526264e-37 && sign){ //-minpos uZ.ui = 0xFFFFFFFF; return uZ.p; } else if (f32>1 || f32<-1){ if (sign){ //Make negative numbers positive for easier computation f32 = -f32; } regS = 1; reg = 1; //because k = m-1; so need to add back 1 // minpos if (f32 <= 7.52316384526264e-37){ uZ.ui = 1; } else{ //regime while (f32>=16){ f32 *=0.0625; // f32/=16; reg++; } while (f32>=2){ f32*=0.5; exp++; } int fracLength = 28-reg; if (fracLength<0){ //remove hidden bit //in both cases, reg=29 and 30, e is n+1 bit and frac are sticky bits if(reg==29){ bitNPlusOne = exp&0x1; exp>>=1; //taken care of by the pack algo } else{//reg=30 bitNPlusOne=exp>>1; bitsMore=exp&0x1; exp=0; } if (f32!=1){ bitsMore =1; frac=0; } } else frac = convertQuadFractionP32 (f32, fracLength, &bitNPlusOne, &bitsMore); if (reg>30 ){ (regS) ? (uZ.ui= 0x7FFFFFFF): (uZ.ui=0x1); } //rounding off fraction bits else{ uint_fast32_t regime = 1; if (regS) regime = ( (1< -1 ){ if (sign){ //Make negative numbers positive for easier computation f32 = -f32; } regS = 0; reg = 0; while (f32<1){ f32 *= 16; reg++; } while (f32>=2){ f32*=0.5; exp++; } //only possible combination for reg=15 to reach here is 7FFF (maxpos) and FFFF (-minpos) //but since it should be caught on top, so no need to handle int fracLength = 28-reg; if (fracLength<0){ //in both cases, reg=29 and 30, e is n+1 bit and frac are sticky bits if(reg==29){ bitNPlusOne = exp&0x1; exp>>=1; //taken care of by the pack algo } else{//reg=30 bitNPlusOne=exp>>1; bitsMore=exp&0x1; exp=0; } if (f32!=1){//because of hidden bit bitsMore =1; frac=0; } } else frac = convertQuadFractionP32 (f32, fracLength, &bitNPlusOne, &bitsMore); if (reg>30 ){ (regS) ? (uZ.ui= 0x7FFFFFFF): (uZ.ui=0x1); } //rounding off fraction bits else{ uint_fast32_t regime = 1; if (regS) regime = ( (1<=0) ? (sign=0) : (sign=1); if (f32 == 0 ){ uZ.ui = 0; return uZ.p; } else if(f32 == INFINITY || f32 == -INFINITY || f32 == NAN){ uZ.ui = 0x80000000; return uZ.p; } else if (f32 == 1) { uZ.ui = 0x40000000; return uZ.p; } else if (f32 == -1){ uZ.ui = 0xC0000000; return uZ.p; } /*else if (f32 >= 1.329227995784916e+36){ //maxpos uZ.ui = 0x7FFFFFFF; return uZ.p; } else if (f32 <= -1.329227995784916e+36){ // -maxpos uZ.ui = 0x80000001; return uZ.p; } else if(f32 <= 7.52316384526264e-37 && !sign){ //minpos uZ.ui = 0x1; return uZ.p; } else if(f32 >= -7.52316384526264e-37 && sign){ //-minpos uZ.ui = 0xFFFFFFFF; return uZ.p; }*/ else if (f32>1 || f32<-1){ if (sign){ //Make negative numbers positive for easier computation f32 = -f32; } regS = 1; reg = 1; //because k = m-1; so need to add back 1 // minpos if (x==32 && f32 <= 7.52316384526264e-37){ uZ.ui = 1; } else{ //regime while (f32>=16){ f32 *=0.0625; // f32/=16; reg++; } while (f32>=2){ f32*=0.5; exp++; } int fracLength = x-4-reg; if (fracLength<0){ //in both cases, reg=29 and 30, e is n+1 bit and frac are sticky bits if(reg==x-3){ bitNPlusOne = exp&0x1; //exp>>=1; //taken care of by the pack algo exp&=0x2; } else{//reg=30 bitNPlusOne=exp>>1; bitsMore=exp&0x1; exp=0; } if (f32!=1){//because of hidden bit bitsMore =1; frac=0; } } else frac = convertQuadFractionP32 (f32, fracLength, &bitNPlusOne, &bitsMore); if (reg>(x-2) ){ uZ.ui=(regS) ? (0x7FFFFFFF & ((int32_t)0x80000000>>(x-1)) ): (0x1 << (32-x)); } //rounding off fraction bits else{ uint_fast32_t regime = 1; if (regS) regime = ( (1<0){ uZ.ui = 0x1 << (32-x); } if (bitNPlusOne) uZ.ui += ( ((uZ.ui>>(32-x)) & 0x1) | bitsMore ) << (32-x); } if (sign) uZ.ui = -uZ.ui & 0xFFFFFFFF; } } else if (f32 < 1 || f32 > -1 ){ if (sign){ //Make negative numbers positive for easier computation f32 = -f32; } regS = 0; reg = 0; //regime while (f32<1){ f32 *= 16; reg++; } while (f32>=2){ f32*=0.5; exp++; } int fracLength = x-4-reg; if (fracLength<0){ //in both cases, reg=29 and 30, e is n+1 bit and frac are sticky bits if(reg==x-3){ bitNPlusOne = exp&0x1; //exp>>=1; //taken care of by the pack algo exp&=0x2; } else{//reg=30 bitNPlusOne=exp>>1; bitsMore=exp&0x1; exp=0; } if (f32!=1){//because of hidden bit bitsMore =1; frac=0; } } else frac = convertQuadFractionP32 (f32, fracLength, &bitNPlusOne, &bitsMore); if (reg>(x-2) ){ uZ.ui=(regS) ? (0x7FFFFFFF & ((int32_t)0x80000000>>(x-1)) ): (0x1 << (32-x)); } //rounding off fraction bits else{ uint_fast32_t regime = 1; if (regS) regime = ( (1<0){ uZ.ui = 0x1 << (32-x); } if (bitNPlusOne){ uZ.ui += ( ((uZ.ui>>(32-x)) & 0x1) | bitsMore ) << (32-x); } } if (sign) uZ.ui = -uZ.ui & 0xFFFFFFFF; } else { //NaR - for NaN, INF and all other combinations uZ.ui = 0x80000000; } return uZ.p; } #endif void checkExtraP32TwoBits(double f32, double temp, bool * bitsNPlusOne, bool * bitsMore ){ temp /= 2; if (temp<=f32){ *bitsNPlusOne = 1; f32-=temp; } if (f32>0) *bitsMore = 1; } uint_fast32_t convertFractionP32(double f32, uint_fast16_t fracLength, bool * bitsNPlusOne, bool * bitsMore ){ uint_fast32_t frac=0; if(f32==0) return 0; else if(f32==INFINITY) return 0x80000000; f32 -= 1; //remove hidden bit if (fracLength==0) checkExtraP32TwoBits(f32, 1.0, bitsNPlusOne, bitsMore); else{ double temp = 1; while (true){ temp /= 2; if (temp<=f32){ f32-=temp; fracLength--; frac = (frac<<1) + 1; //shift in one if (f32==0){ frac <<= (uint_fast16_t)fracLength; break; } if (fracLength == 0){ checkExtraP32TwoBits(f32, temp, bitsNPlusOne, bitsMore); break; } } else{ frac <<= 1; //shift in a zero fracLength--; if (fracLength == 0){ checkExtraP32TwoBits(f32, temp, bitsNPlusOne, bitsMore); break; } } } } return frac; } posit32_t convertDoubleToP32(double f32){ union ui32_p32 uZ; bool sign, regS; uint_fast32_t reg, frac=0; int_fast32_t exp=0; bool bitNPlusOne=0, bitsMore=0; (f32>=0) ? (sign=0) : (sign=1); if (f32 == 0 ){ uZ.ui = 0; return uZ.p; } else if(f32 == INFINITY || f32 == -INFINITY || f32 == NAN){ uZ.ui = 0x80000000; return uZ.p; } else if (f32 == 1) { uZ.ui = 0x40000000; return uZ.p; } else if (f32 == -1){ uZ.ui = 0xC0000000; return uZ.p; } else if (f32 >= 1.329227995784916e+36){ //maxpos uZ.ui = 0x7FFFFFFF; return uZ.p; } else if (f32 <= -1.329227995784916e+36){ // -maxpos uZ.ui = 0x80000001; return uZ.p; } else if(f32 <= 7.52316384526264e-37 && !sign){ //minpos uZ.ui = 0x1; return uZ.p; } else if(f32 >= -7.52316384526264e-37 && sign){ //-minpos uZ.ui = 0xFFFFFFFF; return uZ.p; } else if (f32>1 || f32<-1){ if (sign){ //Make negative numbers positive for easier computation f32 = -f32; } regS = 1; reg = 1; //because k = m-1; so need to add back 1 // minpos if (f32 <= 7.52316384526264e-37){ uZ.ui = 1; } else{ //regime while (f32>=16){ f32 *=0.0625; // f32/=16; reg++; } while (f32>=2){ f32*=0.5; exp++; } int8_t fracLength = 28-reg; if (fracLength<0){ //in both cases, reg=29 and 30, e is n+1 bit and frac are sticky bits if(reg==29){ bitNPlusOne = exp&0x1; exp>>=1; //taken care of by the pack algo } else{//reg=30 bitNPlusOne=exp>>1; bitsMore=exp&0x1; exp=0; } if (f32!=1){//because of hidden bit bitsMore =1; frac=0; } } else frac = convertFractionP32 (f32, fracLength, &bitNPlusOne, &bitsMore); if (reg>30 ){ (regS) ? (uZ.ui= 0x7FFFFFFF): (uZ.ui=0x1); } //rounding off fraction bits else{ uint_fast32_t regime = 1; if (regS) regime = ( (1< -1 ){ if (sign){ //Make negative numbers positive for easier computation f32 = -f32; } regS = 0; reg = 0; //regime while (f32<1){ f32 *= 16; reg++; } while (f32>=2){ f32*=0.5; exp++; } //only possible combination for reg=15 to reach here is 7FFF (maxpos) and FFFF (-minpos) //but since it should be caught on top, so no need to handle int_fast8_t fracLength = 28-reg; if (fracLength<0){ //in both cases, reg=29 and 30, e is n+1 bit and frac are sticky bits if(reg==29){ bitNPlusOne = exp&0x1; exp>>=1; //taken care of by the pack algo } else{//reg=30 bitNPlusOne=exp>>1; bitsMore=exp&0x1; exp=0; } if (f32!=1){//because of hidden bit bitsMore =1; frac=0; } } else frac = convertFractionP32 (f32, fracLength, &bitNPlusOne, &bitsMore); if (reg>30 ){ (regS) ? (uZ.ui= 0x7FFFFFFF): (uZ.ui=0x1); } //rounding off fraction bits else{ uint_fast32_t regime = 1; if (regS) regime = ( (1<=0) ? (sign=0) : (sign=1); if (f32 == 0 ){ uZ.ui = 0; return uZ.p; } else if(f32 == INFINITY || f32 == -INFINITY || f32 == NAN){ uZ.ui = 0x80000000; return uZ.p; } else if (f32 == 1) { uZ.ui = 0x40000000; return uZ.p; } else if (f32 == -1){ uZ.ui = 0xC0000000; return uZ.p; } /*else if (f32 >= 1.329227995784916e+36){ //maxpos uZ.ui = 0x7FFFFFFF; return uZ.p; } else if (f32 <= -1.329227995784916e+36){ // -maxpos uZ.ui = 0x80000001; return uZ.p; } else if(f32 <= 7.52316384526264e-37 && !sign){ //minpos uZ.ui = 0x1; return uZ.p; } else if(f32 >= -7.52316384526264e-37 && sign){ //-minpos uZ.ui = 0xFFFFFFFF; return uZ.p; }*/ else if (f32>1 || f32<-1){ if (sign){ //Make negative numbers positive for easier computation f32 = -f32; } regS = 1; reg = 1; //because k = m-1; so need to add back 1 // minpos if (x==32 && f32 <= 7.52316384526264e-37){ uZ.ui = 1; } else{ //regime while (f32>=16){ f32 *=0.0625; // f32/=16; reg++; } while (f32>=2){ f32*=0.5; exp++; } int fracLength = x-4-reg; if (fracLength<0){ //in both cases, reg=29 and 30, e is n+1 bit and frac are sticky bits if(reg==x-3){ bitNPlusOne = exp&0x1; //exp>>=1; //taken care of by the pack algo exp&=0x2; } else{//reg=30 bitNPlusOne=exp>>1; bitsMore=exp&0x1; exp=0; } if (f32!=1){//because of hidden bit bitsMore =1; frac=0; } } else frac = convertFractionP32 (f32, fracLength, &bitNPlusOne, &bitsMore); if (reg>(x-2) ){ uZ.ui=(regS) ? (0x7FFFFFFF & ((int32_t)0x80000000>>(x-1)) ): (0x1 << (32-x)); } //rounding off fraction bits else{ uint_fast32_t regime = 1; if (regS) regime = ( (1<>=1; else if (reg<=28) exp<<= (28-reg); uZ.ui = ((uint32_t) (regime) << (30-reg)) + ((uint32_t) exp ) + ((uint32_t)(frac<<(32-x))); //minpos if (uZ.ui==0 && frac>0){ uZ.ui = 0x1 << (32-x); } if (bitNPlusOne) uZ.ui += ( ((uZ.ui>>(32-x)) & 0x1) | bitsMore ) << (32-x); } if (sign) uZ.ui = -uZ.ui & 0xFFFFFFFF; } } else if (f32 < 1 || f32 > -1 ){ if (sign){ //Make negative numbers positive for easier computation f32 = -f32; } regS = 0; reg = 0; //regime while (f32<1){ f32 *= 16; reg++; } while (f32>=2){ f32*=0.5; exp++; } int fracLength = x-4-reg; if (fracLength<0){ //in both cases, reg=29 and 30, e is n+1 bit and frac are sticky bits if(reg==x-3){ bitNPlusOne = exp&0x1; //exp>>=1; //taken care of by the pack algo exp&=0x2; } else{//reg=30 bitNPlusOne=exp>>1; bitsMore=exp&0x1; exp=0; } if (f32!=1){//because of hidden bit bitsMore =1; frac=0; } } else frac = convertFractionP32 (f32, fracLength, &bitNPlusOne, &bitsMore); if (reg>(x-2) ){ uZ.ui=(regS) ? (0x7FFFFFFF & ((int32_t)0x80000000>>(x-1)) ): (0x1 << (32-x)); } //rounding off fraction bits else{ uint_fast32_t regime = 1; if (regS) regime = ( (1<>=1; else if (reg<=28) exp<<= (28-reg); uZ.ui = ((uint32_t) (regime) << (30-reg)) + ((uint32_t) exp ) + ((uint32_t)(frac<<(32-x))); //minpos if (uZ.ui==0 && frac>0){ uZ.ui = 0x1 << (32-x); } if (bitNPlusOne){ uZ.ui += ( ((uZ.ui>>(32-x)) & 0x1) | bitsMore ) << (32-x); } } if (sign) uZ.ui = -uZ.ui & 0xFFFFFFFF; } else { //NaR - for NaN, INF and all other combinations uZ.ui = 0x80000000; } return uZ.p; }