Texas Instruments TMS320C3x - Manual (Page 309)

TMS320C32 Memory Overview
10-6
10.2.3.2 16- or 32-Bit Floating-Point Data Types
The ’C32 supports 16- or 32-bit floating point data. For 16-bit floating-point
reads, the eight MSBs are the signed exponent and the eight LSBs are the
signed mantissa (see Section 5.3.2,
’C32 Short Floating-Point Format for
External 16-Bit Data, on page 5-6). When a 16-bit floating-point value is loaded
into a 40-bit register, the external memory interface zero fills the least signifi-
cant 24 bits of the register. When a 16-bit floating-point value is used as a 32-bit
on-chip input operand, the external memory interface zero fills the 16 LSBs of
the 32-bit input operand. The 32-bit floating-point data access is identical to
that of the ’C30 and ’C31.