v0.5.1: fix small bug

FindDefinition · Dec 26, 2023 · 20ac6f9 · 20ac6f9
1 parent b380181
commit 20ac6f9
Show file tree

Hide file tree

Showing 4 changed files with 6 additions and 3 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,4 +1,7 @@
 # Changelog
+## [0.5.1] - 2023-12-26
+### Fixed
+- fix a small bug in `mp_helper.h`
 
 ## [0.5.0] - 2023-11-15
 ### Added

diff --git a/cumm/tensorview/__init__.py b/cumm/tensorview/__init__.py
@@ -485,7 +485,7 @@ def numpy_view(self):
 
 ALL_TV_TENSOR_DTYPES = set([
     bool_, float16, float32, float64, int8, int16, int32, int64, uint8, uint16,
-    uint32, uint64, tf32, custom16, custom32, custom48, custom64, custom80,
+    uint32, uint64, tf32, bfloat16, custom16, custom32, custom48, custom64, custom80,
     custom96, custom128
 ])
 

diff --git a/include/tensorview/core/mp_helper.h b/include/tensorview/core/mp_helper.h
@@ -512,7 +512,7 @@ template <class L, class F> constexpr F mp_for_each(F &&f) {
 #endif
 
 template <class L, class F> TV_HOST_DEVICE_INLINE constexpr F mp_for_each_cuda(F &&f) {
-  return detail::mp_for_each_impl(mp_rename<L, mp_list>(), std::forward<F>(f));
+  return detail::mp_for_each_impl_cuda(mp_rename<L, mp_list>(), std::forward<F>(f));
 }
 
 template <unsigned N, unsigned... Ns> struct mp_prod_int {

diff --git a/version.txt b/version.txt
@@ -1 +1 @@
-0.5.0
+0.5.1