summaryrefslogtreecommitdiff
path: root/ggml/src
diff options
context:
space:
mode:
Diffstat (limited to 'ggml/src')
-rw-r--r--ggml/src/iqk/iqk_mul_mat.cpp12
1 files changed, 12 insertions, 0 deletions
diff --git a/ggml/src/iqk/iqk_mul_mat.cpp b/ggml/src/iqk/iqk_mul_mat.cpp
index 29ae9d53..45d804a4 100644
--- a/ggml/src/iqk/iqk_mul_mat.cpp
+++ b/ggml/src/iqk/iqk_mul_mat.cpp
@@ -16142,7 +16142,13 @@ struct FlashQKV {
std::memcpy(S, fms.S, nq1*sizeof(float));
auto R = qkv_cache;
for (int j = 0; j < nq1; ++j) {
+#ifdef __aarch64__
+ for (int i = 0; i < D/F16::block_size; ++i) {
+ F16::store(qkv + F16::block_size*i, F16::load(R + F16::block_size*i));
+ }
+#else
std::memcpy(qkv, R, D*sizeof(float));
+#endif
qkv += stride_qkv;
R += D;
}
@@ -16162,7 +16168,13 @@ struct FlashQKV {
std::memcpy(S, fms.S, q_step*sizeof(float));
auto R = qkv_cache;
for (int j = 0; j < q_step; ++j) {
+#ifdef __aarch64__
+ for (int i = 0; i < D/F16::block_size; ++i) {
+ F16::store(qkv + F16::block_size*i, F16::load(R + F16::block_size*i));
+ }
+#else
std::memcpy(qkv, R, D*sizeof(float));
+#endif
qkv += stride_qkv;
R += D;
}