leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

CUTLASS 3.3.0 (#1167)

Browse Source 
* Release 3.3.0

Adds support for mixed precision GEMMs On Hopper and Ampere
Adds support for < 16B aligned GEMMs on Hopper
Enhancements to EVT
Enhancements to Python interface
Enhancements to Sub-byte type handling in CuTe
Several other bug-fixes and performance improvements.

* minor doc update
This commit is contained in:
Pradeep Ramani
2023-11-02 11:09:05 -04:00
committed by GitHub
parent 922fb5108b
commit c008b4aea8
263 changed files with 16214 additions and 5008 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/cute/numeric/math.hpp
+4 -2
View File
@@ -130,6 +130,8 @@ has_single_bit(T x) {
}
// Smallest number of bits needed to represent the given value
// For x == 0, this is 0
// For x != 0, this is 1 + floor(log2(x))
// bit_width( 0b0000 ) = 0
// bit_width( 0b0001 ) = 1
// bit_width( 0b0010 ) = 2
@@ -203,7 +205,7 @@ CUTE_HOST_DEVICE constexpr
T
rotl(T x, int s) {
constexpr int N = numeric_limits<T>::digits;
return s == 0 ? x : s > 0 ? (x << s) | (x >> (N - s)) : rotr(x, -s);
return static_cast<T>(s == 0 ? x : s > 0 ? (x << s) | (x >> (N - s)) : rotr(x, -s));
}
// Computes the result of circular bitwise right-rotation
@@ -212,7 +214,7 @@ CUTE_HOST_DEVICE constexpr
T
rotr(T x, int s) {
constexpr int N = numeric_limits<T>::digits;
return s == 0 ? x : s > 0 ? (x >> s) | (x << (N - s)) : rotl(x, -s);
return static_cast<T>(s == 0 ? x : s > 0 ? (x >> s) | (x << (N - s)) : rotl(x, -s));
}
// Counts the number of consecutive 0 bits, starting from the most significant bit