Merge commit '280bc4219151c3f79fe8ca076a2d10df4ff88b34' into develop

This commit is contained in:
assistant-librarian[bot]
2025-12-02 16:14:43 +00:00
parent aef67fef38
commit ebfafa78fe
30 changed files with 1783 additions and 1135 deletions

View File

@@ -66,9 +66,9 @@ constexpr TransferABC FwdTransfer_4x64x1{
{
.thread_cluster_dims =
{.m_block = 1, .m_wave_per_xdl = 32, .n_block = 1, .n_wave_per_xdl = 8},
.epilogue = {.m_per_wave_per_shuffle = 1,
.n_per_wave_per_shuffle = 1,
.scalar_per_vector = 8},
.epilogue = {.m_xdl_per_wave_per_shuffle = 1,
.n_per_wave_per_shuffle = 1,
.scalar_per_vector = 8},
},
};
@@ -99,9 +99,9 @@ constexpr TransferABC FwdTransfer_4x64x1_fp8{
{
.thread_cluster_dims =
{.m_block = 1, .m_wave_per_xdl = 32, .n_block = 1, .n_wave_per_xdl = 8},
.epilogue = {.m_per_wave_per_shuffle = 1,
.n_per_wave_per_shuffle = 1,
.scalar_per_vector = 8},
.epilogue = {.m_xdl_per_wave_per_shuffle = 1,
.n_per_wave_per_shuffle = 1,
.scalar_per_vector = 8},
},
};
@@ -132,9 +132,9 @@ constexpr TransferABC FwdTransfer_4x16x1{
{
.thread_cluster_dims =
{.m_block = 1, .m_wave_per_xdl = 16, .n_block = 1, .n_wave_per_xdl = 4},
.epilogue = {.m_per_wave_per_shuffle = 1,
.n_per_wave_per_shuffle = 1,
.scalar_per_vector = 8},
.epilogue = {.m_xdl_per_wave_per_shuffle = 1,
.n_per_wave_per_shuffle = 1,
.scalar_per_vector = 8},
},
};
@@ -166,9 +166,9 @@ constexpr TransferABC FwdTransfer_4x32x1{
{
.thread_cluster_dims =
{.m_block = 1, .m_wave_per_xdl = 32, .n_block = 1, .n_wave_per_xdl = 4},
.epilogue = {.m_per_wave_per_shuffle = 1,
.n_per_wave_per_shuffle = 1,
.scalar_per_vector = 8},
.epilogue = {.m_xdl_per_wave_per_shuffle = 1,
.n_per_wave_per_shuffle = 1,
.scalar_per_vector = 8},
},
};