float_compare

Summary. Compare two XFLs under a mode flag and return a boolean result.

Signature.

int64_t float_compare(int64_t float1, int64_t float2, uint32_t mode);

Parameters.

Name	Type	Description
float1	int64_t	Left-hand XFL.
float2	int64_t	Right-hand XFL.
mode	uint32_t	A bitmask of the compare-mode flags below.

Comparison flags:

Constant	Value	Meaning
COMPARE_EQUAL	1	float1 == float2
COMPARE_LESS	2	float1 < float2
COMPARE_GREATER	4	float1 > float2

Flags may be OR-ed to form <= (COMPARE_LESS \| COMPARE_EQUAL = 3), >= (COMPARE_GREATER \| COMPARE_EQUAL = 5), and != (COMPARE_LESS \| COMPARE_GREATER = 6).

Return value. 1 if the relationship selected by mode holds, 0 if it does not. Errors: INVALID_FLOAT (-10024) if either argument is invalid; INVALID_ARGUMENT (-7) if mode is 0, has all three of EQUAL/LESS/GREATER set (7), or sets any bit outside the low three; XFL_OVERFLOW (-30) if the comparison overflows internally.

Common failure patterns.

• mode == 0, mode == 7, or a stray high bit (e.g. 8) → INVALID_ARGUMENT.
• Treating the return as -1/0/1: it is strictly 1 (true) or 0 (false).

Caveats / notes.

• Comparisons respect sign and magnitude across the whole XFL range, including mixed-sign and very large/small values.

Minimal example.

if (float_compare(amount, threshold, COMPARE_LESS) == 1)
    rollback(SBUF("below threshold"), 1);

Practical example.

#define GT   0b100U   // COMPARE_GREATER
#define LTE  0b011U   // COMPARE_LESS | COMPARE_EQUAL

// invalid XFL arguments are rejected
ASSERT(float_compare(-1, -2, COMPARE_EQUAL) == INVALID_FLOAT);
// invalid modes are rejected
ASSERT(float_compare(0, 0, 0)       == INVALID_ARGUMENT);
ASSERT(float_compare(0, 0, 0b111U)  == INVALID_ARGUMENT);
// logic: 0 < 1, and 0 <= 1
ASSERT(float_compare(0, float_one(), LTE) == 1);
ASSERT(float_compare(0, float_one(), GT)  == 0);

Related APIs. float_one, float_sum.