Skip to content

GitLab

Commit 21fbf2ce authored by Adam Osewski's avatar Adam Osewski
Browse files

Calculate tile start/end in device kernel.

parent 1de7e9bc
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 176 additions and 154 deletions
include/ck/tensor_operation/gpu/grid/block_to_ctile_map.hpp
View file @ 21fbf2ce
...@@ -269,7 +269,8 @@ struct BlockToCTileMap_KSplit_M00_N0_M01Adapt ...@@ -269,7 +269,8 @@ struct BlockToCTileMap_KSplit_M00_N0_M01Adapt
{ {
} }
__host__ constexpr index_t CalculateGridSize(const CGridDesc_M_N& c_grid_desc_m_n) const __host__ __device__ constexpr index_t
CalculateGridSize(const CGridDesc_M_N& c_grid_desc_m_n) const
{ {
const auto M0 = math::integer_divide_ceil(c_grid_desc_m_n.GetLength(I0), MPerBlock); const auto M0 = math::integer_divide_ceil(c_grid_desc_m_n.GetLength(I0), MPerBlock);
const auto N0 = math::integer_divide_ceil(c_grid_desc_m_n.GetLength(I1), NPerBlock); const auto N0 = math::integer_divide_ceil(c_grid_desc_m_n.GetLength(I1), NPerBlock);
......
include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r4r2.hpp
View file @ 21fbf2ce
...@@ -610,9 +610,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2 ...@@ -610,9 +610,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2
p_c_grid, c_grid_desc_mblock_mperblock_nblock_nperblock.GetElementSpaceSize()); p_c_grid, c_grid_desc_mblock_mperblock_nblock_nperblock.GetElementSpaceSize());
// divide block work by [KBatch, M, N] // divide block work by [KBatch, M, N]
const auto block_work_idx = const auto block_work_idx = block_2_ctile_map.CalculateBottomIndex();
// block_2_ctile_map.CalculateBottomIndex(make_multi_index(get_block_1d_id()));
block_2_ctile_map.CalculateBottomIndex();
if(!block_2_ctile_map.ValidCTileIndex( if(!block_2_ctile_map.ValidCTileIndex(
block_work_idx, block_work_idx,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment