warp-lang 1.9.0__py3-none-win_amd64.whl → 1.10.0rc2__py3-none-win_amd64.whl

warp/native/vec.h

@@ -149,6 +149,17 @@ using vec2d = vec_t<2,double>;
 using vec3d = vec_t<3,double>;
 using vec4d = vec_t<4,double>;
 
+// Type trait to detect if a type is a vec_t
+template<typename T>
+struct is_vector {
+    static constexpr bool value = false;
+};
+
+template<unsigned Length, typename Type>
+struct is_vector<vec_t<Length, Type>> {
+    static constexpr bool value = true;
+};
+
 template<unsigned Length, typename Type>
 inline CUDA_CALLABLE vec_t<Length, Type> operator - (const vec_t<Length, Type>& x)
 {
@@ -343,17 +354,6 @@ inline CUDA_CALLABLE vec_t<Length, Type> add(vec_t<Length, Type> a, vec_t<Length
     return ret;
 }
 
-template<unsigned Length, typename Type>
-inline CUDA_CALLABLE vec_t<Length, Type> add(Type a, vec_t<Length, Type> b)
-{
-    vec_t<Length, Type> ret;
-    for( unsigned i=0; i < Length; ++i )
-    {
-        ret[i] = a + b[i];
-    }
-    return ret;
-}
-
 template<typename Type>
 inline CUDA_CALLABLE vec_t<2, Type> add(vec_t<2, Type> a, vec_t<2, Type> b)
 {
@@ -378,18 +378,6 @@ inline CUDA_CALLABLE vec_t<Length, Type> sub(vec_t<Length, Type> a, vec_t<Length
     return ret;
 }
 
-template<unsigned Length, typename Type>
-inline CUDA_CALLABLE vec_t<Length, Type> sub(Type a, vec_t<Length, Type> b)
-{
-    vec_t<Length, Type> ret;
-    for (unsigned i=0; i < Length; ++i)
-    {
-        ret[i] = Type(a - b[i]);
-    }
-
-    return ret;
-}
-
 template<typename Type>
 inline CUDA_CALLABLE vec_t<2, Type> sub(vec_t<2, Type> a, vec_t<2, Type> b)
 {
@@ -427,6 +415,139 @@ inline CUDA_CALLABLE vec_t<3, Type> mod(vec_t<3, Type> a, vec_t<3, Type> b)
     return vec_t<3, Type>(mod(a.c[0], b.c[0]), mod(a.c[1], b.c[1]), mod(a.c[2], b.c[2]));
 }
 
+// bitwise AND
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE vec_t<Length, Type> bit_and(vec_t<Length, Type> a, vec_t<Length, Type> b)
+{
+    vec_t<Length, Type> ret;
+    for( unsigned i=0; i < Length; ++i )
+    {
+        ret[i] = Type(a[i] & b[i]);
+    }
+    return ret;
+}
+
+template<typename Type>
+inline CUDA_CALLABLE vec_t<2, Type> bit_and(vec_t<2, Type> a, vec_t<2, Type> b)
+{
+    return vec_t<2, Type>( a.c[0] & b.c[0], a.c[1] & b.c[1]);
+}
+
+template<typename Type>
+inline CUDA_CALLABLE vec_t<3, Type> bit_and(vec_t<3, Type> a, vec_t<3, Type> b)
+{
+    return vec_t<3, Type>( a.c[0] & b.c[0], a.c[1] & b.c[1], a.c[2] & b.c[2]);
+}
+
+// bitwise OR
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE vec_t<Length, Type> bit_or(vec_t<Length, Type> a, vec_t<Length, Type> b)
+{
+    vec_t<Length, Type> ret;
+    for( unsigned i=0; i < Length; ++i )
+    {
+        ret[i] = Type(a[i] | b[i]);
+    }
+    return ret;
+}
+
+template<typename Type>
+inline CUDA_CALLABLE vec_t<2, Type> bit_or(vec_t<2, Type> a, vec_t<2, Type> b)
+{
+    return vec_t<2, Type>( a.c[0] | b.c[0], a.c[1] | b.c[1]);
+}
+
+template<typename Type>
+inline CUDA_CALLABLE vec_t<3, Type> bit_or(vec_t<3, Type> a, vec_t<3, Type> b)
+{
+    return vec_t<3, Type>( a.c[0] | b.c[0], a.c[1] | b.c[1], a.c[2] | b.c[2]);
+}
+
+// bitwise XOR
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE vec_t<Length, Type> bit_xor(vec_t<Length, Type> a, vec_t<Length, Type> b)
+{
+    vec_t<Length, Type> ret;
+    for( unsigned i=0; i < Length; ++i )
+    {
+        ret[i] = Type(a[i] ^ b[i]);
+    }
+    return ret;
+}
+
+template<typename Type>
+inline CUDA_CALLABLE vec_t<2, Type> bit_xor(vec_t<2, Type> a, vec_t<2, Type> b)
+{
+    return vec_t<2, Type>( a.c[0] ^ b.c[0], a.c[1] ^ b.c[1]);
+}
+
+template<typename Type>
+inline CUDA_CALLABLE vec_t<3, Type> bit_xor(vec_t<3, Type> a, vec_t<3, Type> b)
+{
+    return vec_t<3, Type>( a.c[0] ^ b.c[0], a.c[1] ^ b.c[1], a.c[2] ^ b.c[2]);
+}
+
+// left shift
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE vec_t<Length, Type> lshift(vec_t<Length, Type> a, vec_t<Length, Type> b)
+{
+    vec_t<Length, Type> ret;
+    for( unsigned i=0; i < Length; ++i )
+    {
+        ret[i] = Type(a[i] << b[i]);
+    }
+    return ret;
+}
+
+template<typename Type>
+inline CUDA_CALLABLE vec_t<2, Type> lshift(vec_t<2, Type> a, vec_t<2, Type> b)
+{
+    return vec_t<2, Type>( a.c[0] << b.c[0], a.c[1] << b.c[1]);
+}
+
+template<typename Type>
+inline CUDA_CALLABLE vec_t<3, Type> lshift(vec_t<3, Type> a, vec_t<3, Type> b)
+{
+    return vec_t<3, Type>( a.c[0] << b.c[0], a.c[1] << b.c[1], a.c[2] << b.c[2]);
+}
+
+// right shift
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE vec_t<Length, Type> rshift(vec_t<Length, Type> a, vec_t<Length, Type> b)
+{
+    vec_t<Length, Type> ret;
+    for( unsigned i=0; i < Length; ++i )
+    {
+        ret[i] = Type(a[i] >> b[i]);
+    }
+    return ret;
+}
+
+template<typename Type>
+inline CUDA_CALLABLE vec_t<2, Type> rshift(vec_t<2, Type> a, vec_t<2, Type> b)
+{
+    return vec_t<2, Type>( a.c[0] >> b.c[0], a.c[1] >> b.c[1]);
+}
+
+template<typename Type>
+inline CUDA_CALLABLE vec_t<3, Type> rshift(vec_t<3, Type> a, vec_t<3, Type> b)
+{
+    return vec_t<3, Type>( a.c[0] >> b.c[0], a.c[1] >> b.c[1], a.c[2] >> b.c[2]);
+}
+
+// invert
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE vec_t<Length,Type> invert(vec_t<Length,Type> v)
+{
+    vec_t<Length,Type> ret;
+    for (unsigned i=0; i < Length; ++i)
+    {
+        ret[i] = ~v[i];
+    }
+
+    return ret;
+}
+
 // dot product:
 template<unsigned Length, typename Type>
 inline CUDA_CALLABLE Type dot(vec_t<Length, Type> a, vec_t<Length, Type> b)
@@ -747,6 +868,183 @@ inline CUDA_CALLABLE void adj_sub_inplace(
 }
 
 
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void bit_and_inplace(vec_t<Length, Type>& v, int idx, Type value)
+{
+#ifndef NDEBUG
+    if (idx < -(int)Length || idx >= (int)Length)
+    {
+        printf("vec index %d out of bounds at %s %d\n", idx, __FILE__, __LINE__);
+        assert(0);
+    }
+#endif
+
+    if (idx < 0)
+    {
+        idx += Length;
+    }
+
+    v[idx] &= value;
+}
+
+
+template<unsigned SliceLength, unsigned Length, typename Type>
+inline CUDA_CALLABLE void bit_and_inplace(vec_t<Length, Type>& v, slice_t slice, const vec_t<SliceLength, Type> &a)
+{
+    assert(slice.start >= 0 && slice.start <= (int)Length);
+    assert(slice.stop >= -1 && slice.stop <= (int)Length);
+    assert(slice.step != 0 && slice.step < 0 ? slice.start >= slice.stop : slice.start <= slice.stop);
+    assert(slice_get_length(slice) == SliceLength);
+
+    bool is_reversed = slice.step < 0;
+
+    int ii = 0;
+    for (
+        int i = slice.start;
+        is_reversed ? (i > slice.stop) : (i < slice.stop);
+        i += slice.step
+    )
+    {
+        v[i] &= a[ii];
+        ++ii;
+    }
+
+    assert(ii == SliceLength);
+}
+
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_bit_and_inplace(
+    vec_t<Length, Type>& v, int idx, Type value,
+    vec_t<Length, Type>& adj_v, int adj_idx, Type& adj_value
+) {}
+
+
+template<unsigned SliceLength, unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_bit_and_inplace(
+    const vec_t<Length, Type>& v, slice_t slice, const vec_t<SliceLength, Type> &a,
+    vec_t<Length, Type>& adj_v, slice_t& adj_slice, vec_t<SliceLength, Type>& adj_a
+) {}
+
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void bit_or_inplace(vec_t<Length, Type>& v, int idx, Type value)
+{
+#ifndef NDEBUG
+    if (idx < -(int)Length || idx >= (int)Length)
+    {
+        printf("vec index %d out of bounds at %s %d\n", idx, __FILE__, __LINE__);
+        assert(0);
+    }
+#endif
+
+    if (idx < 0)
+    {
+        idx += Length;
+    }
+
+    v[idx] |= value;
+}
+
+
+template<unsigned SliceLength, unsigned Length, typename Type>
+inline CUDA_CALLABLE void bit_or_inplace(vec_t<Length, Type>& v, slice_t slice, const vec_t<SliceLength, Type> &a)
+{
+    assert(slice.start >= 0 && slice.start <= (int)Length);
+    assert(slice.stop >= -1 && slice.stop <= (int)Length);
+    assert(slice.step != 0 && slice.step < 0 ? slice.start >= slice.stop : slice.start <= slice.stop);
+    assert(slice_get_length(slice) == SliceLength);
+
+    bool is_reversed = slice.step < 0;
+
+    int ii = 0;
+    for (
+        int i = slice.start;
+        is_reversed ? (i > slice.stop) : (i < slice.stop);
+        i += slice.step
+    )
+    {
+        v[i] |= a[ii];
+        ++ii;
+    }
+
+    assert(ii == SliceLength);
+}
+
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_bit_or_inplace(
+    vec_t<Length, Type>& v, int idx, Type value,
+    vec_t<Length, Type>& adj_v, int adj_idx, Type& adj_value
+) {}
+
+
+template<unsigned SliceLength, unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_bit_or_inplace(
+    const vec_t<Length, Type>& v, slice_t slice, const vec_t<SliceLength, Type> &a,
+    vec_t<Length, Type>& adj_v, slice_t& adj_slice, vec_t<SliceLength, Type>& adj_a
+) {}
+
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void bit_xor_inplace(vec_t<Length, Type>& v, int idx, Type value)
+{
+#ifndef NDEBUG
+    if (idx < -(int)Length || idx >= (int)Length)
+    {
+        printf("vec index %d out of bounds at %s %d\n", idx, __FILE__, __LINE__);
+        assert(0);
+    }
+#endif
+
+    if (idx < 0)
+    {
+        idx += Length;
+    }
+
+    v[idx] ^= value;
+}
+
+
+template<unsigned SliceLength, unsigned Length, typename Type>
+inline CUDA_CALLABLE void bit_xor_inplace(vec_t<Length, Type>& v, slice_t slice, const vec_t<SliceLength, Type> &a)
+{
+    assert(slice.start >= 0 && slice.start <= (int)Length);
+    assert(slice.stop >= -1 && slice.stop <= (int)Length);
+    assert(slice.step != 0 && slice.step < 0 ? slice.start >= slice.stop : slice.start <= slice.stop);
+    assert(slice_get_length(slice) == SliceLength);
+
+    bool is_reversed = slice.step < 0;
+
+    int ii = 0;
+    for (
+        int i = slice.start;
+        is_reversed ? (i > slice.stop) : (i < slice.stop);
+        i += slice.step
+    )
+    {
+        v[i] ^= a[ii];
+        ++ii;
+    }
+
+    assert(ii == SliceLength);
+}
+
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_bit_xor_inplace(
+    vec_t<Length, Type>& v, int idx, Type value,
+    vec_t<Length, Type>& adj_v, int adj_idx, Type& adj_value
+) {}
+
+
+template<unsigned SliceLength, unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_bit_xor_inplace(
+    const vec_t<Length, Type>& v, slice_t slice, const vec_t<SliceLength, Type> &a,
+    vec_t<Length, Type>& adj_v, slice_t& adj_slice, vec_t<SliceLength, Type>& adj_a
+) {}
+
+
 template<unsigned Length, typename Type>
 inline CUDA_CALLABLE void assign_inplace(vec_t<Length, Type>& v, int idx, Type value)
 {
@@ -1303,21 +1601,6 @@ inline CUDA_CALLABLE void adj_add(vec_t<Length, Type> a, vec_t<Length, Type> b,
     adj_b += adj_ret;
 }
 
-template<unsigned Length, typename Type>
-inline CUDA_CALLABLE void adj_add(
-    Type a, vec_t<Length, Type> b,
-    Type& adj_a, vec_t<Length, Type>& adj_b,
-    const vec_t<Length, Type>& adj_ret
-)
-{
-    for (unsigned i = 0; i < Length; ++i)
-    {
-        adj_a += adj_ret.c[i];
-    }
-
-    adj_b += adj_ret;
-}
-
 template<typename Type>
 inline CUDA_CALLABLE void adj_add(vec_t<2, Type> a, vec_t<2, Type> b, vec_t<2, Type>& adj_a, vec_t<2, Type>& adj_b, const vec_t<2, Type>& adj_ret)
 {
@@ -1345,21 +1628,6 @@ inline CUDA_CALLABLE void adj_sub(vec_t<Length, Type> a, vec_t<Length, Type> b,
     adj_b -= adj_ret;
 }
 
-template<unsigned Length, typename Type>
-inline CUDA_CALLABLE void adj_sub(
-    Type a, vec_t<Length, Type> b,
-    Type& adj_a, vec_t<Length, Type>& adj_b,
-    const vec_t<Length, Type>& adj_ret
-)
-{
-    for (unsigned i = 0; i < Length; ++i)
-    {
-        adj_a += adj_ret.c[i];
-    }
-
-    adj_b -= adj_ret;
-}
-
 template<typename Type>
 inline CUDA_CALLABLE void adj_sub(vec_t<2, Type> a, vec_t<2, Type> b, vec_t<2, Type>& adj_a, vec_t<2, Type>& adj_b, const vec_t<2, Type>& adj_ret)
 {
@@ -1395,6 +1663,90 @@ inline CUDA_CALLABLE void adj_mod(vec_t<3, Type> a, vec_t<3, Type> b, vec_t<3, T
 {
 }
 
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_bit_and(vec_t<Length, Type> a, vec_t<Length, Type> b, vec_t<Length, Type>& adj_a, vec_t<Length, Type>& adj_b, const vec_t<Length, Type>& adj_ret)
+{
+}
+
+template<typename Type>
+inline CUDA_CALLABLE void adj_bit_and(vec_t<2, Type> a, vec_t<2, Type> b, vec_t<2, Type>& adj_a, vec_t<2, Type>& adj_b, const vec_t<2, Type>& adj_ret)
+{
+}
+
+template<typename Type>
+inline CUDA_CALLABLE void adj_bit_and(vec_t<3, Type> a, vec_t<3, Type> b, vec_t<3, Type>& adj_a, vec_t<3, Type>& adj_b, const vec_t<3, Type>& adj_ret)
+{
+}
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_bit_or(vec_t<Length, Type> a, vec_t<Length, Type> b, vec_t<Length, Type>& adj_a, vec_t<Length, Type>& adj_b, const vec_t<Length, Type>& adj_ret)
+{
+}
+
+template<typename Type>
+inline CUDA_CALLABLE void adj_bit_or(vec_t<2, Type> a, vec_t<2, Type> b, vec_t<2, Type>& adj_a, vec_t<2, Type>& adj_b, const vec_t<2, Type>& adj_ret)
+{
+}
+
+template<typename Type>
+inline CUDA_CALLABLE void adj_bit_or(vec_t<3, Type> a, vec_t<3, Type> b, vec_t<3, Type>& adj_a, vec_t<3, Type>& adj_b, const vec_t<3, Type>& adj_ret)
+{
+}
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_bit_xor(vec_t<Length, Type> a, vec_t<Length, Type> b, vec_t<Length, Type>& adj_a, vec_t<Length, Type>& adj_b, const vec_t<Length, Type>& adj_ret)
+{
+}
+
+template<typename Type>
+inline CUDA_CALLABLE void adj_bit_xor(vec_t<2, Type> a, vec_t<2, Type> b, vec_t<2, Type>& adj_a, vec_t<2, Type>& adj_b, const vec_t<2, Type>& adj_ret)
+{
+}
+
+template<typename Type>
+inline CUDA_CALLABLE void adj_bit_xor(vec_t<3, Type> a, vec_t<3, Type> b, vec_t<3, Type>& adj_a, vec_t<3, Type>& adj_b, const vec_t<3, Type>& adj_ret)
+{
+}
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_lshift(vec_t<Length, Type> a, vec_t<Length, Type> b, vec_t<Length, Type>& adj_a, vec_t<Length, Type>& adj_b, const vec_t<Length, Type>& adj_ret)
+{
+}
+
+template<typename Type>
+inline CUDA_CALLABLE void adj_lshift(vec_t<2, Type> a, vec_t<2, Type> b, vec_t<2, Type>& adj_a, vec_t<2, Type>& adj_b, const vec_t<2, Type>& adj_ret)
+{
+}
+
+template<typename Type>
+inline CUDA_CALLABLE void adj_lshift(vec_t<3, Type> a, vec_t<3, Type> b, vec_t<3, Type>& adj_a, vec_t<3, Type>& adj_b, const vec_t<3, Type>& adj_ret)
+{
+}
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_rshift(vec_t<Length, Type> a, vec_t<Length, Type> b, vec_t<Length, Type>& adj_a, vec_t<Length, Type>& adj_b, const vec_t<Length, Type>& adj_ret)
+{
+}
+
+template<typename Type>
+inline CUDA_CALLABLE void adj_rshift(vec_t<2, Type> a, vec_t<2, Type> b, vec_t<2, Type>& adj_a, vec_t<2, Type>& adj_b, const vec_t<2, Type>& adj_ret)
+{
+}
+
+template<typename Type>
+inline CUDA_CALLABLE void adj_rshift(vec_t<3, Type> a, vec_t<3, Type> b, vec_t<3, Type>& adj_a, vec_t<3, Type>& adj_b, const vec_t<3, Type>& adj_ret)
+{
+}
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE void adj_invert(
+    const vec_t<Length,Type>& v,
+    vec_t<Length,Type>& adj_v,
+    const vec_t<Length,Type>& adj_ret
+)
+{
+}
+
 template<unsigned Length, typename Type>
 inline CUDA_CALLABLE void adj_dot(vec_t<Length, Type> a, vec_t<Length, Type> b, vec_t<Length, Type>& adj_a, vec_t<Length, Type>& adj_b, const Type adj_ret)
 {
@@ -1679,6 +2031,42 @@ inline CUDA_CALLABLE vec_t<Length, Type> atomic_max(vec_t<Length, Type> * addr,
     return ret;
 }
 
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE vec_t<Length, Type> atomic_and(vec_t<Length, Type> * addr, vec_t<Length, Type> value)
+{
+    vec_t<Length, Type> ret;
+    for( unsigned i=0; i < Length; ++i )
+    {
+        ret[i] = atomic_and(&(addr -> c[i]), value[i]);
+    }
+
+    return ret;
+}
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE vec_t<Length, Type> atomic_or(vec_t<Length, Type> * addr, vec_t<Length, Type> value)
+{
+    vec_t<Length, Type> ret;
+    for( unsigned i=0; i < Length; ++i )
+    {
+        ret[i] = atomic_or(&(addr -> c[i]), value[i]);
+    }
+
+    return ret;
+}
+
+template<unsigned Length, typename Type>
+inline CUDA_CALLABLE vec_t<Length, Type> atomic_xor(vec_t<Length, Type> * addr, vec_t<Length, Type> value)
+{
+    vec_t<Length, Type> ret;
+    for( unsigned i=0; i < Length; ++i )
+    {
+        ret[i] = atomic_xor(&(addr -> c[i]), value[i]);
+    }
+
+    return ret;
+}
+
 template<unsigned Length, typename Type>
 inline CUDA_CALLABLE void adj_atomic_minmax(
     vec_t<Length,Type> *addr,
@@ -1736,6 +2124,34 @@ template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_add(vec_t<Length,
 template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_add(vec_t<Length, int64>* buf, const vec_t<Length, int64> &value) { }
 template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_add(vec_t<Length, uint64>* buf, const vec_t<Length, uint64> &value) { }
 
+// for bitwise operations we do not accumulate gradients
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_and(vec_t<Length, int8>* buf, const vec_t<Length, int8> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_and(vec_t<Length, uint8>* buf, const vec_t<Length, uint8> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_and(vec_t<Length, int16>* buf, const vec_t<Length, int16> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_and(vec_t<Length, uint16>* buf, const vec_t<Length, uint16> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_and(vec_t<Length, int32>* buf, const vec_t<Length, int32> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_and(vec_t<Length, uint32>* buf, const vec_t<Length, uint32> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_and(vec_t<Length, int64>* buf, const vec_t<Length, int64> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_and(vec_t<Length, uint64>* buf, const vec_t<Length, uint64> &value) { }
+
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_or(vec_t<Length, int8>* buf, const vec_t<Length, int8> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_or(vec_t<Length, uint8>* buf, const vec_t<Length, uint8> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_or(vec_t<Length, int16>* buf, const vec_t<Length, int16> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_or(vec_t<Length, uint16>* buf, const vec_t<Length, uint16> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_or(vec_t<Length, int32>* buf, const vec_t<Length, int32> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_or(vec_t<Length, uint32>* buf, const vec_t<Length, uint32> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_or(vec_t<Length, int64>* buf, const vec_t<Length, int64> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_or(vec_t<Length, uint64>* buf, const vec_t<Length, uint64> &value) { }
+
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_xor(vec_t<Length, int8>* buf, const vec_t<Length, int8> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_xor(vec_t<Length, uint8>* buf, const vec_t<Length, uint8> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_xor(vec_t<Length, int16>* buf, const vec_t<Length, int16> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_xor(vec_t<Length, uint16>* buf, const vec_t<Length, uint16> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_xor(vec_t<Length, int32>* buf, const vec_t<Length, int32> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_xor(vec_t<Length, uint32>* buf, const vec_t<Length, uint32> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_xor(vec_t<Length, int64>* buf, const vec_t<Length, int64> &value) { }
+template<unsigned Length> CUDA_CALLABLE inline void adj_atomic_xor(vec_t<Length, uint64>* buf, const vec_t<Length, uint64> &value) { }
+
 
 // adjoints for some of the constructors, used in intersect.h
 inline CUDA_CALLABLE void adj_vec2(float x, float y, float& adj_x, float& adj_y, const vec2& adj_ret)

warp/native/version.h

@@ -0,0 +1,23 @@
+/*
+ * SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-License-Identifier: Apache-2.0
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef WP_VERSION_H
+#define WP_VERSION_H
+
+#define WP_VERSION_STRING "1.10.0rc2"
+
+#endif // WP_VERSION_H

warp/native/volume.cpp

@@ -205,7 +205,7 @@ void wp_volume_get_buffer_info(uint64_t id, void** buf, uint64_t* size)
 
 void wp_volume_get_voxel_size(uint64_t id, float* dx, float* dy, float* dz)
 {
-    *dx = *dx = *dz = 0.0f;
+    *dx = *dy = *dz = 0.0f;
 
     const VolumeDesc* volume;
     if (volume_get_descriptor(id, volume))

warp/native/volume.cu

@@ -19,6 +19,7 @@
 #include "volume_impl.h"
 #include "warp.h"
 
+extern CUcontext get_current_context();
 
 __global__ void volume_get_leaf_coords(const uint32_t leaf_count, pnanovdb_coord_t *leaf_coords,
                                        const pnanovdb_buf_t buf)

warp/native/volume.h

@@ -161,7 +161,7 @@ CUDA_CALLABLE inline void pnano_read(T &result, pnanovdb_buf_t buf, PNANOVDB_INO
     pnano_read<T>(result, buf, address);
 }
 
-/// regular grid accessor (values stored in leafs)
+/// regular grid accessor (values stored in leaves)
 
 struct value_accessor_base
 {

warp/native/volume_builder.cu

@@ -16,6 +16,8 @@
  */
 
 #include "volume_builder.h"
+#include "cuda_util.h"
+#include "warp.h"
 
 #include <nanovdb/tools/cuda/PointsToGrid.cuh>