ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-19 22:39:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_BUILDER] Convolution forward transfer concepts. (#3535)

Browse Source 
* Rename member variable to better reflect its actuall meaning.

* Add transfer checks for conv fwd xdl.

* Validate tensor layouts & vector size conv fwd v3.

* Add combined transfer concepts.

* Add transfer concepts for conv fwd factories.

* Fix clang format

* Add helper instruction to get max mem vector instruction width.

* Apply review comments.

* Rename thread cluster access(->arrange) order concept

* FIx merge artifacts.

* Add generic access order limits into block transfer concept.
This commit is contained in:
Adam Osewski
2026-01-19 10:54:10 +01:00
committed by GitHub
parent fe40a5d139
commit 1a6d1b59ef
13 changed files with 570 additions and 157 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
experimental/builder/test/test_conv_description.cpp
View File

@@ -128,26 +128,26 @@ struct DefaultAlgorithm
ckb::test::Transfer<> transfer{
.a =
{
.block_transfer = {.k0 = 1, .m_n = 128, .k1 = 2},
.lds_transfer = {.src_vector_dim = 2,
.src_scalar_per_vector = 2,
.lds_dst_scalar_per_vector = 2,
.is_direct_load = false,
.lds_padding = false},
.block_transfer_access_order = {.order = {0, 1, 2}},
.src_access_order = {.order = {0, 1, 2}},
.block_transfer = {.k0 = 1, .m_n = 128, .k1 = 2},
.lds_transfer = {.src_vector_dim = 2,
.src_scalar_per_vector = 2,
.lds_dst_scalar_per_vector = 2,
.is_direct_load = false,
.lds_padding = false},
.thread_cluster_arrange_order = {.order = {0, 1, 2}},
.src_access_order = {.order = {0, 1, 2}},
},
.b =
{
.block_transfer = {.k0 = 1, .m_n = 128, .k1 = 2},
.lds_transfer = {.src_vector_dim = 2,
.src_scalar_per_vector = 2,
.lds_dst_scalar_per_vector = 2,
.is_direct_load = false,
.lds_padding = false},
.block_transfer_access_order = {.order = {0, 1, 2}},
.src_access_order = {.order = {0, 1, 2}},
.block_transfer = {.k0 = 1, .m_n = 128, .k1 = 2},
.lds_transfer = {.src_vector_dim = 2,
.src_scalar_per_vector = 2,
.lds_dst_scalar_per_vector = 2,
.is_direct_load = false,
.lds_padding = false},
.thread_cluster_arrange_order = {.order = {0, 1, 2}},
.src_access_order = {.order = {0, 1, 2}},
},
.c =
{