diff --git a/.gitattributes b/.gitattributes
index b97122598b..97f741126f 100644
--- a/.gitattributes
+++ b/.gitattributes
@@ -7052,6 +7052,7 @@ packages/rtl-extra/src/inc/mvecimp.inc svneol=native#text/plain
 packages/rtl-extra/src/inc/objects.pp svneol=native#text/plain
 packages/rtl-extra/src/inc/printer.inc svneol=native#text/plain
 packages/rtl-extra/src/inc/printerh.inc svneol=native#text/plain
+packages/rtl-extra/src/inc/real48utils.pp svneol=native#text/plain
 packages/rtl-extra/src/inc/sockets.inc svneol=native#text/plain
 packages/rtl-extra/src/inc/socketsh.inc svneol=native#text/plain
 packages/rtl-extra/src/inc/sockovl.inc svneol=native#text/plain
diff --git a/packages/rtl-extra/fpmake.pp b/packages/rtl-extra/fpmake.pp
index d6cdba19f7..0785264245 100644
--- a/packages/rtl-extra/fpmake.pp
+++ b/packages/rtl-extra/fpmake.pp
@@ -83,6 +83,7 @@ begin
 
     // Add clocale for Android first in order to compile the source file
     // from the 'android' dir, not the 'unix' dir.
+    T:=P.Targets.AddUnit('real48utils.pp');
     T:=P.Targets.AddUnit('clocale.pp',[android]);
 
     T:=P.Targets.AddUnit('ucomplex.pp',UComplexOSes);
diff --git a/packages/rtl-extra/src/inc/real48utils.pp b/packages/rtl-extra/src/inc/real48utils.pp
new file mode 100644
index 0000000000..5901ce57f1
--- /dev/null
+++ b/packages/rtl-extra/src/inc/real48utils.pp
@@ -0,0 +1,155 @@
+unit Real48Utils;
+
+{$mode objfpc}{$H+}
+
+interface
+
+type
+  { Over 32 bits does not work }
+  //TBit52 = 0..$FFFFFFFFFFFFF; { (1 shl 52) - 1 }
+  //TBit40 = 0..$FFFFFFFFFF;    { (1 shl 40) - 1 }
+  //TBit39 = 0..$7FFFFFFFFF;    { (1 shl 39) - 1 }
+  TBit32 = 0..$FFFFFFFF;      { (1 shl 32) - 1 }
+  TBit20 = 0..(1 shl 20) - 1;
+  TBit11 = 0..(1 shl 11) - 1;
+  TBit07 = 0..(1 shl 07) - 1;
+  TBit01 = 0..(1 shl 01) - 1;
+
+  //Double
+  //S1 E11[Bias $3FF] F52
+  TDoubleRec = bitpacked record
+    { F:TBit52; }
+    F2:TBit20;
+    F1:TBit32;
+    E:TBit11;
+    S:TBit01;
+  end;
+  PDoubleRec = ^TDoubleRec;
+
+  //Real48
+  //S1 F39 E8[Bias 129]
+  TReal48Rec = bitpacked record
+    E:Byte;
+    { F:TBit39; }
+    F2:TBit07;
+    F1:TBit32;
+    S:TBit01;
+  end;
+  PReal48Rec = ^TReal48Rec;
+
+function Double2Real(d : double) : real48;
+
+operator explicit (r:Real48) d:double; inline;
+operator explicit (d:double) r:Real48; inline;
+operator := (d:double) r:real48; inline;
+operator := (r:real48) d:double; inline;
+operator +(const r1:Real48) r:Real48;inline;
+operator +(const r1:Real48;const r2:Real48) r:Real48;inline;
+operator -(const r1:Real48) r:Real48;inline;
+operator -(const r1:Real48;const r2:Real48) r:Real48;inline;
+operator *(const r1:Real48;const r2:Real48) r:Real48;inline;
+operator /(const r1:Real48;const r2:Real48) r:Real48;inline;
+operator =(const r1:Real48;const r2:Real48) b:boolean;inline;
+operator <(const r1:Real48;const r2:Real48) b:boolean;inline;
+operator >(const r1:Real48;const r2:Real48) b:boolean;inline;
+operator >=(const r1:Real48;const r2:Real48) b:boolean;inline;
+operator <=(const r1:Real48;const r2:Real48) b:boolean;inline;
+
+implementation
+
+function Double2Real(d : double) : real48;
+var
+   res : array[0..5] of byte;
+   rrec:TReal48Rec absolute res;
+   drec:TDoubleRec absolute d;
+begin
+  { copy mantissa }
+  rrec.F1 := drec.F1;
+  rrec.F2 := drec.F2 shr 13;
+
+  { copy exponent }
+  { correct exponent: }
+  rrec.E := drec.E - 1023 + 129;
+
+  { set sign }
+  rrec.S := drec.S;
+  double2real:=res;
+end;
+
+operator explicit (r:Real48) d:double;inline;
+begin
+ d := Real2Double(r);
+end;
+
+operator explicit (d:double) r:Real48;inline;
+begin
+ r := Double2Real(d);
+end;
+
+operator := (d:double) r:real48; inline;
+begin
+ r := Double2Real(d);
+end;
+
+operator := (r:real48) d:double; inline;
+begin
+ d := Real2Double(r);
+end;
+
+operator +(const r1:Real48;const r2:Real48) r:Real48;inline;
+begin
+ r := double(r1)+double(r2);
+end;
+
+operator -(const r1:Real48) r:Real48;inline;
+begin
+ r := -double(r1);
+end;
+
+operator +(const r1:Real48) r:Real48;inline;
+begin
+ r := double(r1);
+end;
+
+operator -(const r1:Real48;const r2:Real48) r:Real48;inline;
+begin
+ r := double(r1)-double(r2);
+end;
+
+operator *(const r1:Real48;const r2:Real48) r:Real48;inline;
+begin
+ r := double(r1)*double(r2);
+end;
+
+operator /(const r1:Real48;const r2:Real48) r:Real48;inline;
+begin
+ r := double(r1)/double(r2);
+end;
+
+operator =(const r1:Real48;const r2:Real48) b:boolean;inline;
+begin
+ b := double(r1)=double(r2);
+end;
+
+operator <(const r1:Real48;const r2:Real48) b:boolean;inline;
+begin
+ b := double(r1)<double(r2);
+end;
+
+operator >(const r1:Real48;const r2:Real48) b:boolean;inline;
+begin
+ b := double(r1)>double(r2);
+end;
+
+operator >=(const r1:Real48;const r2:Real48) b:boolean;inline;
+begin
+ b := double(r1)>=double(r2);
+end;
+
+operator <=(const r1:Real48;const r2:Real48) b:boolean;inline;
+begin
+ b := double(r1)<=double(r2);
+end;
+
+end.
+