gamma-lut v0.1.0

Map an input through a small lookup table with linear interpolation.

Application scenarios

• LED dimming: map a perceptual level to a PWM duty.
• Sensor linearization: correct a nonlinear thermistor or photodiode.
• Any curve stored in flash as a few points, without floating point.

Contract

uint16_t gamma_lut_apply(const uint16_t *table, uint8_t entries,
                         uint16_t input, uint16_t input_max);

• table has entries points covering 0..input_max.
• Inputs at or above input_max return the last point.
• Between points the result is linearly interpolated and rounded.
• entries must be at least 2 for interpolation; 0 or 1 return the first point.
• No allocation; O(1).

Parameters

• More entries: smoother curve, more flash.
• input_max should match the input range; 255, 1023, or 4095 are common.

Memory and timing

• Table is caller-owned; 2 bytes per entry.
• One multiply, one divide, and two 64-bit products per call.

Integration

static const uint16_t gamma[5] = {0, 20, 120, 600, 4095};
uint16_t duty = gamma_lut_apply(gamma, 5, level, 255);

When not to use

• A mathematically simple curve that fits in one multiply; skip the table.
• A curve that changes at runtime; rebuild the table or use a different model.

Verification

make test covers endpoints, interpolation, clamping above the range, and tables with fewer than two points.