very very minor performance improvements

Removed the power operator from the uncalled cubic interp function and replaced them with assigned variables under multiplication.

Added t5 variable for akima and removed two instances of multiplication and replaced it with one instance of assignment.

Probably more text so far than performance improvements.

frogpilot/controls/lib/frogpilot_acceleration.py

@@ -1,6 +1,7 @@
 #!/usr/bin/env python3
 import numpy as np
 
+
 def cubic_interp(x, xp, fp):
      """Cubic interpolation using NumPy's native operations for speed."""
      # Boundary conditions
@@ -17,7 +18,10 @@ def cubic_interp(x, xp, fp):
      t = (x - xp[i]) / float(xp[i+1] - xp[i])
 
      # Hermite cubic formula
-     return fp[i]*(1 - 3*t**2 + 2*t**3) + fp[i+1]*(3*t**2 - 2*t**3)
+     t2 = t*t
+     t3 = t2*t
+
+     return fp[i]*(1 - 3*t2 + 2*t3) + fp[i+1]*(3*t2 - 2*t3)
 
 def akima_interp(x, xp, fp):
      """Akima-inspired interpolation with reduced overshoot characteristics."""
@@ -33,10 +37,12 @@ def akima_interp(x, xp, fp):
 
      # Quintic polynomial to reduce overshoot
      t2 = t*t
-     t4 = t2*t2
      t3 = t2*t
-     return (fp[i]*(1 - 10*t3 + 15*t4 - 6*t3*t2)
-             + fp[i+1]*(10*t3 - 15*t4 + 6*t3*t2))
+     t4 = t2*t2
+     t5 = t3*t2
+
+     return (fp[i]*(1 - 10*t3 + 15*t4 - 6*t5) + fp[i+1]*(10*t3 - 15*t4 + 6*t5))
+
 
 from openpilot.selfdrive.controls.lib.longitudinal_planner import A_CRUISE_MIN, get_max_accel