Added info on simdgroup availability (#586)

* Add info on simdgroup availability

* "osx" not "os x"

* Update metal_matmul.py

* Update metal_matmul.py
This commit is contained in:
calledit
2023-02-23 22:59:02 +01:00
committed by GitHub
parent d22e19536b
commit 81f7c6800a

View File

@@ -25,7 +25,7 @@ FLOPS = N*N*N*2
prog = CLProgram("test", f"""
#include <metal_stdlib>
#include <metal_simdgroup_matrix>
#include <metal_simdgroup_matrix> // Available from Metal version 2.3 released with OS X 11.0+
using namespace metal;
kernel void test(device float *a, device const float *data1, device const float *data2, uint3 gid [[thread_position_in_grid]], uint3 xid [[threadgroup_position_in_grid]], uint3 lid [[thread_position_in_threadgroup]], uint sidx [[simdgroup_index_in_threadgroup]]) {{
// 1-2 simd groups