diff --git a/src/lib.rs b/src/lib.rs
index c09e522..4b59a9a 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -16,6 +16,7 @@ mod tests;
 
 use core::cmp::PartialOrd;
 use core::ffi::{c_char, CStr};
+use core::fmt::Display;
 use core::ops::{Add, Div, Mul, Neg, Rem, Sub};
 
 // DType {{{
@@ -72,6 +73,7 @@ pub trait DType:
     + Rem<Output = Self>
     + Sub<Output = Self>
     + PartialOrd
+    + Display
     + FromF32
 {
     fn powi(self, rhs: i32) -> Self;
@@ -395,7 +397,7 @@ fn pq_oetf_common<T: DType>(f: T, m2: T) -> T {
     let numerator: T = T::ff32(PQEOTF_C2).fma(y_pow_m1, PQEOTF_C1.to_dt());
     let denominator: T = T::ff32(PQEOTF_C3).fma(y_pow_m1, 1.0.to_dt());
 
-    (numerator / denominator).powf(m2)
+    (numerator / denominator).spowf(m2)
 }
 
 /// Dolby Perceptual Quantizer Electro-Optical Transfer Function primarily used for ICtCP
@@ -544,7 +546,7 @@ impl TryFrom<&str> for Space {
     }
 }
 
-impl core::fmt::Display for Space {
+impl Display for Space {
     fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
         core::fmt::write(
             f,
diff --git a/src/tests.rs b/src/tests.rs
index ebf09fd..70ad7aa 100644
--- a/src/tests.rs
+++ b/src/tests.rs
@@ -240,7 +240,7 @@ fn individual() {
         ("CIELAB->XYZ", CIELAB, XYZ, cielab_to_xyz),
         ("XYZ->OKLAB", XYZ, OKLAB, xyz_to_oklab),
         ("OKLAB->XYZ", OKLAB, XYZ, oklab_to_xyz),
-        //("XYZ->JZAZBZ", XYZ, JZAZBZ, xyz_to_jzazbz)
+        //("XYZ->JZAZBZ", XYZ, JZAZBZ, xyz_to_jzazbz),
         ("JZAZBZ->XYZ", JZAZBZ, XYZ, jzazbz_to_xyz),
         ("CIELAB->CIELCH", CIELAB, CIELCH, lab_to_lch),
         ("CIELCH->CIELAB", CIELCH, CIELAB, lch_to_lab),
@@ -426,26 +426,30 @@ fn nan_checks() {
     let it = [1e+3, -1e+3, 1e-3, -1e-3];
     // do these at f32 to faster approach bounds
     let fns: &[(&'static str, fn(&mut [f32; 3]))] = &[
-        ("hsv_forwards", srgb_to_hsv),
-        ("hsv_backwards", hsv_to_srgb),
-        ("lrgb_forwards", srgb_to_lrgb),
-        ("lrgb_backwards", lrgb_to_srgb),
-        ("xyz_forwards", lrgb_to_xyz),
-        ("xyz_backwards", xyz_to_lrgb),
-        ("lab_forwards", xyz_to_cielab),
-        ("lab_backwards", cielab_to_xyz),
-        ("lch_forwards", lab_to_lch),
-        ("lch_backwards", lch_to_lab),
-        ("oklab_forwards", xyz_to_oklab),
-        ("oklab_backwards", oklab_to_xyz),
-        //("jzazbz_forwards", xyz_to_jzazbz), // ugh
-        ("jzazbz_backwards", jzazbz_to_xyz),
+        ("srgb_to_hsv", srgb_to_hsv),
+        ("hsv_to_srgb", hsv_to_srgb),
+        ("srgb_to_lrgb", srgb_to_lrgb),
+        ("lrgb_to_srgb", lrgb_to_srgb),
+        ("lrgb_to_xyz", lrgb_to_xyz),
+        ("xyz_to_lrgb", xyz_to_lrgb),
+        ("xyz_to_cielab", xyz_to_cielab),
+        ("cielab_to_xyz", cielab_to_xyz),
+        ("lab_to_lch", lab_to_lch),
+        ("lch_to_lab", lch_to_lab),
+        ("xyz_to_oklab", xyz_to_oklab),
+        ("oklab_to_xyz", oklab_to_xyz),
+        // fails hard in the PQ function with (N/D)^P
+        // Succeeds in F64 but graphics is almost always run in F32
+        //("xyz_to_jzazbz", xyz_to_jzazbz),
+        ("jzazbz_to_xyz", jzazbz_to_xyz),
+        ("_lrgb_to_ictcp", _lrgb_to_ictcp),
+        ("_ictcp_to_lrgb", _ictcp_to_lrgb),
     ];
     for (label, func) in fns {
         for a in it.iter() {
             for b in it.iter() {
                 for c in it.iter() {
-                    let from: [f32; 3] = [*a, *b, *c];
+                    let from = [*a, *b, *c];
                     let mut to = from;
                     func(&mut to);
                     if to.iter().any(|c| !c.is_finite()) {