[HARDWARE][POWER] optimize math functions of VSX power (#47321)
Signed-off-by: Akash Kaothalkar <akashkaothalkar@akashs-mbp.bl1-in.ibm.com> Signed-off-by: Akash Kaothalkar <akash.kaothalkar@ibm.com> Signed-off-by: Akash kaothalkar <akash.kaothalkar@ibm.com> Signed-off-by: Rukhaiya <bibirukhaiya123@gmail.com> Co-authored-by: Akash Kaothalkar <akashkaothalkar@akashs-mbp.bl1-in.ibm.com> Co-authored-by: Akash Kaothalkar <akash.kaothalkar@ibm.com>
This commit is contained in:
co-authored by
Akash Kaothalkar
Akash Kaothalkar
parent
c5b66233b2
commit
c85d72076a
@@ -323,8 +323,6 @@ class AttentionImpl<ISA::VSX, scalar_t, head_dim> {
|
||||
const int64_t num_blocks_stride, const int64_t cache_head_num_stride,
|
||||
const int64_t block_size, const int64_t block_size_stride,
|
||||
const float k_inv = 0.0f, const float v_inv = 0.0f) {
|
||||
// k_inv and v_inv are unused on VSX: FP8 KV cache is not supported on
|
||||
// PowerPC. The parameters are present to match the common interface.
|
||||
#pragma omp parallel for collapse(2)
|
||||
for (int64_t token_idx = 0; token_idx < token_num; ++token_idx) {
|
||||
for (int64_t head_idx = 0; head_idx < head_num; ++head_idx) {
|
||||
|
||||
Reference in New Issue
Block a user