Message ID | 1387660099-22390-1-git-send-email-peter.maydell@linaro.org |
---|---|
State | New |
Headers | show |
On Sat, Dec 21, 2013 at 09:08:19PM +0000, Peter Maydell wrote: > If the input to float*_scalbn() is denormal then it represents > a number 0.[mantissabits] * 2^(1-exponentbias) (and the actual > exponent field is all zeroes). This means that when we convert > it to our unpacked encoding the unpacked exponent must be one > greater than for a normal number, which represents > 1.[mantissabits] * 2^(e-exponentbias) for an exponent field e. > > This meant we were giving answers too small by a factor of 2 for > all denormal inputs. > > Note that the float-to-int routines also have this behaviour > of not adjusting the exponent for denormals; however there it is > harmless because denormals will all convert to integer zero anyway. > > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > --- > Changes v1->v2: propagated fix to 80 bit and 128 bit float functions. > > These function names remind me of Ken Thompson's reply to a question > about what he'd do differently if he were redesigning UNIX: > "I'd spell creat with an e." > > fpu/softfloat.c | 29 +++++++++++++++++++++-------- > 1 file changed, 21 insertions(+), 8 deletions(-) > > diff --git a/fpu/softfloat.c b/fpu/softfloat.c > index dbda61b..7f5e8c7 100644 > --- a/fpu/softfloat.c > +++ b/fpu/softfloat.c > @@ -6795,10 +6795,13 @@ float32 float32_scalbn( float32 a, int n STATUS_PARAM ) > } > return a; > } > - if ( aExp != 0 ) > + if (aExp != 0) { > aSig |= 0x00800000; > - else if ( aSig == 0 ) > + } else if (aSig == 0) { > return a; > + } else { > + aExp++; > + } > > if (n > 0x200) { > n = 0x200; > @@ -6828,10 +6831,13 @@ float64 float64_scalbn( float64 a, int n STATUS_PARAM ) > } > return a; > } > - if ( aExp != 0 ) > + if (aExp != 0) { > aSig |= LIT64( 0x0010000000000000 ); > - else if ( aSig == 0 ) > + } else if (aSig == 0) { > return a; > + } else { > + aExp++; > + } > > if (n > 0x1000) { > n = 0x1000; > @@ -6861,8 +6867,12 @@ floatx80 floatx80_scalbn( floatx80 a, int n STATUS_PARAM ) > return a; > } > > - if (aExp == 0 && aSig == 0) > - return a; > + if (aExp == 0) { > + if (aSig == 0) { > + return a; > + } > + aExp++; > + } > > if (n > 0x10000) { > n = 0x10000; > @@ -6891,10 +6901,13 @@ float128 float128_scalbn( float128 a, int n STATUS_PARAM ) > } > return a; > } > - if ( aExp != 0 ) > + if (aExp != 0) { > aSig0 |= LIT64( 0x0001000000000000 ); > - else if ( aSig0 == 0 && aSig1 == 0 ) > + } else if (aSig0 == 0 && aSig1 == 0) { > return a; > + } else { > + aExp++; > + } > > if (n > 0x10000) { > n = 0x10000; Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
On 12/21/2013 01:08 PM, Peter Maydell wrote: > If the input to float*_scalbn() is denormal then it represents > a number 0.[mantissabits] * 2^(1-exponentbias) (and the actual > exponent field is all zeroes). This means that when we convert > it to our unpacked encoding the unpacked exponent must be one > greater than for a normal number, which represents > 1.[mantissabits] * 2^(e-exponentbias) for an exponent field e. > > This meant we were giving answers too small by a factor of 2 for > all denormal inputs. > > Note that the float-to-int routines also have this behaviour > of not adjusting the exponent for denormals; however there it is > harmless because denormals will all convert to integer zero anyway. > > Signed-off-by: Peter Maydell <peter.maydell@linaro.org> > --- > Changes v1->v2: propagated fix to 80 bit and 128 bit float functions. > > These function names remind me of Ken Thompson's reply to a question > about what he'd do differently if he were redesigning UNIX: > "I'd spell creat with an e." > > fpu/softfloat.c | 29 +++++++++++++++++++++-------- > 1 file changed, 21 insertions(+), 8 deletions(-) Reviewed-by: Richard Henderson <rth@twiddle.net> r~
diff --git a/fpu/softfloat.c b/fpu/softfloat.c index dbda61b..7f5e8c7 100644 --- a/fpu/softfloat.c +++ b/fpu/softfloat.c @@ -6795,10 +6795,13 @@ float32 float32_scalbn( float32 a, int n STATUS_PARAM ) } return a; } - if ( aExp != 0 ) + if (aExp != 0) { aSig |= 0x00800000; - else if ( aSig == 0 ) + } else if (aSig == 0) { return a; + } else { + aExp++; + } if (n > 0x200) { n = 0x200; @@ -6828,10 +6831,13 @@ float64 float64_scalbn( float64 a, int n STATUS_PARAM ) } return a; } - if ( aExp != 0 ) + if (aExp != 0) { aSig |= LIT64( 0x0010000000000000 ); - else if ( aSig == 0 ) + } else if (aSig == 0) { return a; + } else { + aExp++; + } if (n > 0x1000) { n = 0x1000; @@ -6861,8 +6867,12 @@ floatx80 floatx80_scalbn( floatx80 a, int n STATUS_PARAM ) return a; } - if (aExp == 0 && aSig == 0) - return a; + if (aExp == 0) { + if (aSig == 0) { + return a; + } + aExp++; + } if (n > 0x10000) { n = 0x10000; @@ -6891,10 +6901,13 @@ float128 float128_scalbn( float128 a, int n STATUS_PARAM ) } return a; } - if ( aExp != 0 ) + if (aExp != 0) { aSig0 |= LIT64( 0x0001000000000000 ); - else if ( aSig0 == 0 && aSig1 == 0 ) + } else if (aSig0 == 0 && aSig1 == 0) { return a; + } else { + aExp++; + } if (n > 0x10000) { n = 0x10000;
If the input to float*_scalbn() is denormal then it represents a number 0.[mantissabits] * 2^(1-exponentbias) (and the actual exponent field is all zeroes). This means that when we convert it to our unpacked encoding the unpacked exponent must be one greater than for a normal number, which represents 1.[mantissabits] * 2^(e-exponentbias) for an exponent field e. This meant we were giving answers too small by a factor of 2 for all denormal inputs. Note that the float-to-int routines also have this behaviour of not adjusting the exponent for denormals; however there it is harmless because denormals will all convert to integer zero anyway. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- Changes v1->v2: propagated fix to 80 bit and 128 bit float functions. These function names remind me of Ken Thompson's reply to a question about what he'd do differently if he were redesigning UNIX: "I'd spell creat with an e." fpu/softfloat.c | 29 +++++++++++++++++++++-------- 1 file changed, 21 insertions(+), 8 deletions(-)