URL
https://opencores.org/ocsvn/openrisc/openrisc/trunk
Subversion Repositories openrisc
[/] [openrisc/] [trunk/] [gnu-old/] [gcc-4.2.2/] [gcc/] [testsuite/] [gcc.dg/] [dfp/] [nan-2.c] - Rev 154
Go to most recent revision | Compare with Previous | Blame | View Log
/* { dg-options "-std=gnu99" } */ /* N1150 4: Characteristics of decimal floating types (not explicit) C99 5.2.4.2.2: Characteristics of floating types. A few simple checks on arithmetic operations. Based on nan-1.c with the consideration of negative zero. */ extern void abort (void); int main() { _Decimal32 d32; _Decimal64 d64; _Decimal128 d128; /* Verify that division by negative zero produces a negative infinity result. */ d32 = 123.45f; if (d32/-0.0df != -__builtin_infd32()) abort(); if (123.45df/-0.0df != -__builtin_infd32()) abort(); d64 = 123.45f; if (d64/-0.0dd != -__builtin_infd64()) abort(); if (123.45dd/-0.0dd != -__builtin_infd64()) abort(); d128 = 123.45f; if (d128/-0.0dl != -__builtin_infd64()) abort(); if (123.45dl/-0.0dl != -__builtin_infd128()) abort(); d32 = 0.0df; if (!__builtin_isnand32(-(d32/-0.0df))) abort(); if (!__builtin_isnand32(-(0.0df/-0.0df))) abort(); d64 = 0.0dd; if (!__builtin_isnand64(-(d64/-0.0dd))) abort(); if (!__builtin_isnand64(-(0.0dd/-0.0dd))) abort(); d128 = 0.0dl; if (!__builtin_isnand128(-(d128/-0.0dl))) abort(); if (!__builtin_isnand128(-(0.0dl/-0.0dl))) abort(); return 0; }
Go to most recent revision | Compare with Previous | Blame | View Log