What's the rounding mode for float point operation in rts64plus.lib?

miscellanea

Prodigy 220 points

It seems not work as IEEE754 default mode.

The following program generates different results from rts64plus.lib and rts6700.lib by using the relevant simulator.

float a = (float)9067800;

float b = (float)1041830805;

float c = 1.0 / a;

float d = b * c;

printf("0x%x\n", *((int*)&d));

0x42e5c971 by using rts64plus.lib

0x42e5c972 by using rts6700.lib which is the same as visual studio's output on x86 cpu.

How can I get a IEEE754 float operation with default rounding mode on c64x+ ?

Thanks.

over 16 years ago

0 George Mock over 16 years ago

TI__Guru**** 249440 points

The floating point format used by C6000 devices comes from IEEE 754. But the RTS library functions used to perform floating point operations are not guaranteed to conform to IEEE 754. The particular difference you point out here occurs because the C6700 performs those floating point operations with hardware instructions, but the C64+ performs those operations with RTS functions.

Thanks and regards,

-George