Hip Refactor for element_binary_kernels, unary kernels, and embedding kernels

[Bob-Chen222]

Description of changes:
Hip Refactor for element_binary_kernels, unary kernels, and embedding kernels

Related Issues:

Linked Issues:

• Issue Hip kernel refactor #1296

Issues closed by this PR:

• Closes #

---

This change is 

[reyna-abhyankar]

comments

[Bob-Chen222]

Reviewable status: 0 of 3 files reviewed, 6 unresolved discussions (waiting on @reyna-abhyankar)

---

lib/kernels/src/hip/element_binary_kernels.cpp line 108 at r1 (raw file):

Previously, reyna-abhyankar (Reyna Abhyankar) wrote…

I think the previous one is correct? I only see miopenOpTensorAdd in the docs

Done:)

---

lib/kernels/src/hip/element_binary_kernels.cpp line 262 at r1 (raw file):

Previously, reyna-abhyankar (Reyna Abhyankar) wrote…

Why? &alpha2 is referenced in the else block below?

Yeah alpha2 should never be used. Now deleted

---

lib/kernels/src/hip/element_binary_kernels.cpp line 415 at r1 (raw file):

Previously, reyna-abhyankar (Reyna Abhyankar) wrote…

Are you sure we need the dim3?

Yeah, I checked, and dim3 is not necessary. Removed!

---

lib/kernels/src/hip/element_unary_kernels.cpp line 71 at r1 (raw file):

Previously, reyna-abhyankar (Reyna Abhyankar) wrote…

Why change to CUDNN?

Done.

---

lib/kernels/src/hip/element_unary_kernels.cpp line 238 at r1 (raw file):

Previously, reyna-abhyankar (Reyna Abhyankar) wrote…

dot syntax instead of m->

Done.

---

lib/kernels/src/hip/embedding_kernels.cpp line 424 at r1 (raw file):

Previously, reyna-abhyankar (Reyna Abhyankar) wrote…

if dim3 is necessary, why not use it here too?

Done.

[reyna-abhyankar]

Reviewable status: 0 of 3 files reviewed, all discussions resolved

---

lib/kernels/src/hip/element_binary_kernels.cpp line 262 at r1 (raw file):

Previously, Bob-Chen222 (Bob Chen) wrote…

Yeah alpha2 should never be used. Now deleted

Line 283 still has alpha2?

[Bob-Chen222]

Reviewable status: 0 of 3 files reviewed, all discussions resolved (waiting on @reyna-abhyankar)

---

lib/kernels/src/hip/element_binary_kernels.cpp line 262 at r1 (raw file):

Previously, reyna-abhyankar (Reyna Abhyankar) wrote…

Line 283 still has alpha2?

Do you mean line 314? I think line 283 has already been alpha. Line 314 is also changed.

[reyna-abhyankar]

Reviewable status: 0 of 3 files reviewed, all discussions resolved (waiting on @Bob-Chen222)

---

lib/kernels/src/hip/element_binary_kernels.cpp line 262 at r1 (raw file):

Previously, Bob-Chen222 (Bob Chen) wrote…

Do you mean line 314? I think line 283 has already been alpha. Line 314 is also changed.

I guess I'm asking why was alpha2 removed at all?

[Bob-Chen222]

Reviewable status: 0 of 3 files reviewed, all discussions resolved (waiting on @Bob-Chen222)

---

lib/kernels/src/hip/element_binary_kernels.cpp line 262 at r1 (raw file):

Previously, reyna-abhyankar (Reyna Abhyankar) wrote…

I guess I'm asking why was alpha2 removed at all?

Oh it is because in cuda version there is no alpha2.