HP 3000 Manuals HP 3000 Manuals
Major Considerations MPE V and MPE XL have two areas of incompatibility that may make it difficult for you to convert SPL programs to HP C/XL: * The representation of floating-point numbers * Data storage alignment Floating-Point Numbers MPE XL floating-point numbers are represented in the industry-standard IEEE format. This format is different from the MPE V format in bit layout, range, and precision. (Range is governed by the size of the exponent; precision is governed by the size of the fraction.) MPE V 32-bit floating-point numbers: Bit layout: 1-bit sign, 9-bit exponent, 22-bit fraction Nonzero range: Ý8.63617x10-78 to Ý1.157921x1077 MPE XL 32-bit floating-point numbers: Bit layout: 1-bit sign, 8-bit exponent, 23-bit fraction Nonzero range: Ý1.754944x10-38 to Ý3.4028235x1038 MPE V 64-bit floating-point numbers: Bit layout: 1-bit sign, 9-bit exponent, 54-bit fraction Nonzero range: Ý8.63618555094445x10-78 to Ý1.157920892373162x1077 MPE XL 64-bit floating-point numbers: Bit layout: 1-bit sign, 11-bit exponent, 52-bit fraction Nonzero range: Ý2.2250738585072014x10-308 to Ý1.7976931348623157x10308 MPE XL 32-bit floating point has greater precision but a smaller range than MPE V. Thus, it is possible to have a valid MPE V floating-point number that is not representable in MPE XL floating point. On the other hand, MPE XL 64-bit floating-point numbers can handle a much higher range than MPE V 32-bit or 64-bit floating point, but they have less precision than MPE V 64-bit floating point. The data storage formats are quite different, corresponding to the bit representations noted above. Floating-point data stored on disk must be converted or replaced if the programs are converted to HP C/XL. The MPE XL intrinsic HPFPCONVERT may be used to convert floating point data to and from the various representations. See the MPE XL Intrinsics Reference Manual for details. Data Storage Alignment On MPE V, a data item whose size is two bytes or greater is aligned on a two-byte boundary. On MPE XL, a data item is aligned on a boundary not less than the size of the data item itself, that is, a multiple of 1, 2, 4, or 8 bytes. Thus, a character followed by a 64-bit floating-point number would require 10 bytes in MPE V and 16 bytes in MPE XL. In MPE V, the character would start at byte 0, there would be one unused byte, and the floating-point number would start at byte 3. In MPE XL, the character would start at byte 0, there would be seven unused bytes, and the floating point number would start at byte 8. This incompatibility of data storage affects program access to data both in memory and on disk.