projects / thirdparty / gcc.git / blame
| Commit | Line | Data |
|---|---|---|
| c63539ff ML |
1 | .. |
| 2 | Copyright 1988-2022 Free Software Foundation, Inc. | |
| 3 | This is part of the GCC manual. | |
| 4 | For copying conditions, see the copyright.rst file. | |
| 5 | ||
| 6 | .. index:: cross compilation and floating point, floating point and cross compilation | |
| 7 | ||
| 8 | .. _floating-point: | |
| 9 | ||
| 10 | Cross Compilation and Floating Point | |
| 11 | ************************************ | |
| 12 | ||
| 13 | While all modern machines use twos-complement representation for integers, | |
| 14 | there are a variety of representations for floating point numbers. This | |
| 15 | means that in a cross-compiler the representation of floating point numbers | |
| 16 | in the compiled program may be different from that used in the machine | |
| 17 | doing the compilation. | |
| 18 | ||
| 19 | Because different representation systems may offer different amounts of | |
| 20 | range and precision, all floating point constants must be represented in | |
| 21 | the target machine's format. Therefore, the cross compiler cannot | |
| 22 | safely use the host machine's floating point arithmetic; it must emulate | |
| 23 | the target's arithmetic. To ensure consistency, GCC always uses | |
| 24 | emulation to work with floating point values, even when the host and | |
| 25 | target floating point formats are identical. | |
| 26 | ||
| 27 | The following macros are provided by :samp:`real.h` for the compiler to | |
| 28 | use. All parts of the compiler which generate or optimize | |
| 29 | floating-point calculations must use these macros. They may evaluate | |
| 30 | their operands more than once, so operands must not have side effects. | |
| 31 | ||
| 32 | .. c:macro:: REAL_VALUE_TYPE | |
| 33 | ||
| 34 | The C data type to be used to hold a floating point value in the target | |
| 35 | machine's format. Typically this is a ``struct`` containing an | |
| 36 | array of ``HOST_WIDE_INT``, but all code should treat it as an opaque | |
| 37 | quantity. | |
| 38 | ||
| 39 | .. function:: HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE x) | |
| 40 | ||
| 41 | Truncates :samp:`{x}` to a signed integer, rounding toward zero. | |
| 42 | ||
| 43 | .. function:: unsigned HOST_WIDE_INT REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE x) | |
| 44 | ||
| 45 | Truncates :samp:`{x}` to an unsigned integer, rounding toward zero. If | |
| 46 | :samp:`{x}` is negative, returns zero. | |
| 47 | ||
| 48 | .. function:: REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *string, machine_mode mode) | |
| 49 | ||
| 50 | Converts :samp:`{string}` into a floating point number in the target machine's | |
| 51 | representation for mode :samp:`{mode}`. This routine can handle both | |
| 52 | decimal and hexadecimal floating point constants, using the syntax | |
| 53 | defined by the C language for both. | |
| 54 | ||
| 55 | .. function:: int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE x) | |
| 56 | ||
| 57 | Returns 1 if :samp:`{x}` is negative (including negative zero), 0 otherwise. | |
| 58 | ||
| 59 | .. function:: int REAL_VALUE_ISINF (REAL_VALUE_TYPE x) | |
| 60 | ||
| 61 | Determines whether :samp:`{x}` represents infinity (positive or negative). | |
| 62 | ||
| 63 | .. function:: int REAL_VALUE_ISNAN (REAL_VALUE_TYPE x) | |
| 64 | ||
| 65 | Determines whether :samp:`{x}` represents a 'NaN' (not-a-number). | |
| 66 | ||
| 67 | .. function:: REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE x) | |
| 68 | ||
| 69 | Returns the negative of the floating point value :samp:`{x}`. | |
| 70 | ||
| 71 | .. function:: REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE x) | |
| 72 | ||
| 3ed1b4ce | 73 | Returns the absolute value of :samp:`{x}`. |