amd/blis
1
0
Fork 0
mirror of https://github.com/amd/blis.git synced 2026-04-30 20:41:13 +00:00
Code Issues Packages Projects Releases Wiki Activity

Disabled default packing of matrices in batch_gemm of FP32 (#55)

Browse Source 
AMD-Internal: SWLCSG-3527
This commit is contained in:
Vankadari, Meghana
2025-06-17 10:53:05 +05:30
committed by GitHub
parent 8649cdc14b
commit 26e5c63781
1 changed files with 10 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
addon/aocl_gemm/aocl_batch_gemm_f32f32f32of32.c
View File

@@ -161,6 +161,11 @@ AOCL_BGEMM_MATMUL(float,float,float,float,f32f32f32of32)
mtag_a[bs_i] = PACK;
}
if( bli_is_trans(blis_transa ) )
{
mtag_b[bs_i] = PACK;
}
// swap m & n in case of col-major matrices
m_local[bs_i] = n[bs_i];
n_local[bs_i] = m[bs_i];
@@ -206,6 +211,11 @@ AOCL_BGEMM_MATMUL(float,float,float,float,f32f32f32of32)
mtag_a[bs_i] = PACK;
}
if( bli_is_trans(blis_transb ) && ( mtag_b[bs_i] == UNPACKED ) )
{
mtag_b[bs_i] = PACK;
}
// copy the values of m & n
m_local[bs_i] = m[bs_i];
n_local[bs_i] = n[bs_i];
@@ -218,14 +228,6 @@ AOCL_BGEMM_MATMUL(float,float,float,float,f32f32f32of32)
rs_c[bs_i] = ldc[bs_i];
cs_c[bs_i] = 1;
// By default enable packing for B matrix. Before the 5 loop, based on
// the input dimensions, the smart threading logic will adjust it
// (disable/enable) accordingly.
if ( mtag_b[bs_i] == UNPACKED )
{
mtag_b[bs_i] = PACK;
}
err_t err = lpgemm_translate_to_post_ops_list
(
post_op_unparsed[bs_i], post_op_list[bs_i],
@@ -259,10 +261,6 @@ AOCL_BGEMM_MATMUL(float,float,float,float,f32f32f32of32)
#else
// Setting pack A and B by default for non open mp case.
bli_rntm_set_pack_a( 1, &rntm_g );
bli_rntm_set_pack_b( 1, &rntm_g );
batch_lpgemm_f32f32f32of32_thread_decorator
(
batch_size, m_local, n_local, k,