sp
/
cln_mirror
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
Browse Source

Fix scale_float for large factors on x86. 

The routine scale_float(cl_LF, cl_I) in file cl_LF_scale_I.cc had not
been adapted to 64-bit exponents on systems where uintD is 32 bit.

Reported by Michael Miller <millermj@lemoyne.edu>.
master
Richard Kreckel 14 years ago
parent
b338e09d3b
commit
9b55b28892
1 changed files with 21 additions and 51 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 62
      src/float/lfloat/elem/cl_LF_scale_I.cc

62
src/float/lfloat/elem/cl_LF_scale_I.cc
View File

@ -27,7 +27,7 @@ const cl_LF scale_float (const cl_LF& x, const cl_I& delta)
var uintE uexp = TheLfloat(x)->expo;
if (uexp==0) { return x; }
var uintE udelta;
// |delta| muß <= LF_exp_high-LF_exp_low < 2^32 sein. Wie bei I_to_UL:
// |delta| muß <= LF_exp_high-LF_exp_low < 2^intEsize sein.
if (fixnump(delta)) {
// Fixnum
var sintV sdelta = FN_to_V(delta);
@ -39,51 +39,21 @@ const cl_LF scale_float (const cl_LF& x, const cl_I& delta)
// Bignum
var cl_heap_bignum* bn = TheBignum(delta);
if ((sintD)mspref(arrayMSDptr(bn->data,bn->length),0) >= 0) {
#define IF_LENGTH(i) \
if (bn_minlength <= i) /* genau i Digits überhaupt möglich? */\
if (bn->length == i) /* genau i Digits? */ \
/* 2^((i-1)*intDsize-1) <= delta < 2^(i*intDsize-1) */ \
if ( (i*intDsize-1 > 32) \
&& ( ((i-1)*intDsize-1 >= 32) \
|| (mspref(arrayMSDptr(bn->data,i),0) >= (uintD)bitc(32-(i-1)*intDsize)) \
) ) \
goto overflow; \
else
IF_LENGTH(1)
{ udelta = get_uint1D_Dptr(arrayLSDptr(bn->data,1)); goto pos; }
IF_LENGTH(2)
{ udelta = get_uint2D_Dptr(arrayLSDptr(bn->data,2)); goto pos; }
IF_LENGTH(3)
{ udelta = get_uint3D_Dptr(arrayLSDptr(bn->data,3)); goto pos; }
IF_LENGTH(4)
{ udelta = get_uint4D_Dptr(arrayLSDptr(bn->data,4)); goto pos; }
IF_LENGTH(5)
{ udelta = get_uint4D_Dptr(arrayLSDptr(bn->data,5)); goto pos; }
#undef IF_LENGTH
goto overflow; // delta zu groß
// delta >= 0
try {
udelta = cl_I_to_UE(delta);
goto pos;
} catch (const runtime_exception&) {
goto overflow;
}
} else {
#define IF_LENGTH(i) \
if (bn_minlength <= i) /* genau i Digits überhaupt möglich? */\
if (bn->length == i) /* genau i Digits? */ \
/* - 2^((i-1)*intDsize-1) > delta >= - 2^(i*intDsize-1) */\
if ( (i*intDsize-1 > 32) \
&& ( ((i-1)*intDsize-1 >= 32) \
|| (mspref(arrayMSDptr(bn->data,i),0) < (uintD)(-bitc(32-(i-1)*intDsize))) \
) ) \
goto underflow; \
else
IF_LENGTH(1)
{ udelta = get_sint1D_Dptr(arrayLSDptr(bn->data,1)); goto pos; }
IF_LENGTH(2)
{ udelta = get_sint2D_Dptr(arrayLSDptr(bn->data,2)); goto pos; }
IF_LENGTH(3)
{ udelta = get_sint3D_Dptr(arrayLSDptr(bn->data,3)); goto pos; }
IF_LENGTH(4)
{ udelta = get_sint4D_Dptr(arrayLSDptr(bn->data,4)); goto pos; }
IF_LENGTH(5)
{ udelta = get_sint4D_Dptr(arrayLSDptr(bn->data,5)); goto pos; }
#undef IF_LENGTH
goto underflow; // delta zu klein
// delta < 0
try {
udelta = cl_I_to_E(delta);
goto neg;
} catch (const runtime_exception&) {
goto underflow;
}
}
}
@ -95,7 +65,7 @@ const cl_LF scale_float (const cl_LF& x, const cl_I& delta)
{ throw floating_point_overflow_exception(); }
goto ok;
neg: // delta <0, udelta = 2^32+delta
neg: // delta <0, udelta = 2^intEsize+delta
if ( ((uexp = uexp+udelta) >= udelta) // oder Exponent-Unterlauf?
|| (uexp < LF_exp_low) // oder Exponent zu klein?
)

Write Preview
Loading…
Cancel
Save