enabled dwordx4 for fp16

21cac0e5 · Jing Zhang · 3c71fe87 · 21cac0e5
Commit 21cac0e5 authored Jun 09, 2021 by Jing Zhang
Hide whitespace changes
Inline Side-by-side

Showing with 1 addition and 1 deletion

composable_kernel/include/utility/amd_buffer_addressing_v2.hpp ...sable_kernel/include/utility/amd_buffer_addressing_v2.hpp +1 -1

No files found.
--- a/composable_kernel/include/utility/amd_buffer_addressing_v2.hpp
+++ b/composable_kernel/include/utility/amd_buffer_addressing_v2.hpp
@@ -269,7 +269,7 @@ amd_buffer_load_impl_v2(int32x4_t src_wave_buffer_resource,
        else if constexpr(N == 8)
        {

-#if 1
+#if 0
            vector_type<half_t, 8> tmp;

            tmp.AsType<half4_t>()(Number<0>{}) = __llvm_amdgcn_raw_buffer_load_fp16x4(