amd/blis
1
0
Fork 0
mirror of https://github.com/amd/blis.git synced 2026-05-11 17:50:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

Consolidated code in level-3 _front() functions.

Browse Source 
Details:
- Reduced a code segment that appears in all of the bli_*_front()
  functions except for bli_gemm_front(). Previously, the code looked
  like this (taken from bli_herk_front()):

    if ( bli_cntx_method( cntx ) == BLIS_NAT )
    {
        bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &a_local );
        bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &ah_local );
    }
    else // if ( bli_cntx_method( cntx ) != BLIS_NAT )
    {
        pack_t schema_a = bli_cntx_schema_a_block( cntx );
        pack_t schema_b = bli_cntx_schema_b_panel( cntx );

        bli_obj_set_pack_schema( schema_a, &a_local );
        bli_obj_set_pack_schema( schema_b, &ah_local );
    }

  This code segment is part of a sort-of-hack that allows us to
  communicate the pack schemas into the level-3 thread decorator, which
  needs them so that they can be passed into bli_l3_cntl_create_if(),
  where the control tree is created. However, the first conditional case
  above is unnecessary because the second case is fully generalized.
  That is, even in the native case, the context contains correct,
  queryable schemas. Thus, these code segments were reduced to something
  like:

    pack_t schema_a = bli_cntx_schema_a_block( cntx );
    pack_t schema_b = bli_cntx_schema_b_panel( cntx );

    bli_obj_set_pack_schema( schema_a, &a_local );
    bli_obj_set_pack_schema( schema_b, &ah_local );

  There's always a small chance that the seemingly unnecessary code
  in the first branch case has some special use that is not apparent to
  me, but the testsuite's default input parameters seem to think this
  commit will be fine.
This commit is contained in:
Field G. Van Zee
2020-12-04 16:40:59 -06:00
parent 7038bbaa05
commit 0cef09aa92
11 changed files with 47 additions and 130 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
frame/1m/packm/bli_packm_cxk.c
View File

@@ -74,8 +74,9 @@ void PASTEMAC(ch,opname) \
the outer (panel_dim_max - panel_dim) rows or columns of the
micropanel. (Note that these rows/columns correspond to values
beyond the edge of matrix A.) The kernel intrinsically knows its
own panel_dim_max, since that corresponds to the packm kernel's
leading dimension. However, we *do* need to pass in panel_len_max
own panel_dim_max, since that corresponds to the packm micropanel's
normal width (corresponding to the gemm microkernel's register
blocksize (mr or nr). However, we *do* need to pass in panel_len_max
because the bottom-right edge case of trsm_lu will need all
elements above the extended diagonal and beyond (to the right of)
the bottom-right element to be initialized to zero so the trsm

16
frame/3/gemmt/bli_gemmt_front.c
View File

@@ -110,19 +110,11 @@ void bli_gemmt_front
// bli_l3_cntl_create_if()). This allows us to access the schemas from
// the control tree, which hopefully reduces some confusion, particularly
// in bli_packm_init().
if ( bli_cntx_method( cntx ) == BLIS_NAT )
{
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &a_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &b_local );
}
else // if ( bli_cntx_method( cntx ) != BLIS_NAT )
{
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
}
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
// Invoke the internal back-end via the thread handler.
bli_l3_thread_decorator

16
frame/3/hemm/bli_hemm_front.c
View File

@@ -155,19 +155,11 @@ void bli_hemm_front
// bli_l3_cntl_create_if()). This allows us to access the schemas from
// the control tree, which hopefully reduces some confusion, particularly
// in bli_packm_init().
if ( bli_cntx_method( cntx ) == BLIS_NAT )
{
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &a_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &b_local );
}
else // if ( bli_cntx_method( cntx ) != BLIS_NAT )
{
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
}
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
// Invoke the internal back-end.
bli_l3_thread_decorator

22
frame/3/her2k/bli_her2k_front.c
View File

@@ -124,23 +124,13 @@ void bli_her2k_front
// bli_l3_cntl_create_if()). This allows us to access the schemas from
// the control tree, which hopefully reduces some confusion, particularly
// in bli_packm_init().
if ( bli_cntx_method( cntx ) == BLIS_NAT )
{
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &a_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &bh_local );
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &b_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &ah_local );
}
else // if ( bli_cntx_method( cntx ) != BLIS_NAT )
{
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &bh_local );
bli_obj_set_pack_schema( schema_a, &b_local );
bli_obj_set_pack_schema( schema_b, &ah_local );
}
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &bh_local );
bli_obj_set_pack_schema( schema_a, &b_local );
bli_obj_set_pack_schema( schema_b, &ah_local );
// Invoke herk twice, using beta only the first time.

16
frame/3/herk/bli_herk_front.c
View File

@@ -104,19 +104,11 @@ void bli_herk_front
// bli_l3_cntl_create_if()). This allows us to access the schemas from
// the control tree, which hopefully reduces some confusion, particularly
// in bli_packm_init().
if ( bli_cntx_method( cntx ) == BLIS_NAT )
{
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &a_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &ah_local );
}
else // if ( bli_cntx_method( cntx ) != BLIS_NAT )
{
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &ah_local );
}
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &ah_local );
// Invoke the internal back-end.
bli_l3_thread_decorator

16
frame/3/symm/bli_symm_front.c
View File

@@ -154,19 +154,11 @@ void bli_symm_front
// bli_l3_cntl_create_if()). This allows us to access the schemas from
// the control tree, which hopefully reduces some confusion, particularly
// in bli_packm_init().
if ( bli_cntx_method( cntx ) == BLIS_NAT )
{
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &a_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &b_local );
}
else // if ( bli_cntx_method( cntx ) != BLIS_NAT )
{
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
}
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
// Invoke the internal back-end.
bli_l3_thread_decorator

22
frame/3/syr2k/bli_syr2k_front.c
View File

@@ -105,23 +105,13 @@ void bli_syr2k_front
// bli_l3_cntl_create_if()). This allows us to access the schemas from
// the control tree, which hopefully reduces some confusion, particularly
// in bli_packm_init().
if ( bli_cntx_method( cntx ) == BLIS_NAT )
{
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &a_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &bt_local );
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &b_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &at_local );
}
else // if ( bli_cntx_method( cntx ) != BLIS_NAT )
{
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &bt_local );
bli_obj_set_pack_schema( schema_a, &b_local );
bli_obj_set_pack_schema( schema_b, &at_local );
}
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &bt_local );
bli_obj_set_pack_schema( schema_a, &b_local );
bli_obj_set_pack_schema( schema_b, &at_local );
// Invoke herk twice, using beta only the first time.

16
frame/3/syrk/bli_syrk_front.c
View File

@@ -107,19 +107,11 @@ void bli_syrk_front
// bli_l3_cntl_create_if()). This allows us to access the schemas from
// the control tree, which hopefully reduces some confusion, particularly
// in bli_packm_init().
if ( bli_cntx_method( cntx ) == BLIS_NAT )
{
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &a_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &at_local );
}
else // if ( bli_cntx_method( cntx ) != BLIS_NAT )
{
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &at_local );
}
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &at_local );
// Invoke the internal back-end.
bli_l3_thread_decorator

16
frame/3/trmm/bli_trmm_front.c
View File

@@ -173,19 +173,11 @@ void bli_trmm_front
// bli_l3_cntl_create_if()). This allows us to access the schemas from
// the control tree, which hopefully reduces some confusion, particularly
// in bli_packm_init().
if ( bli_cntx_method( cntx ) == BLIS_NAT )
{
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &a_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &b_local );
}
else // if ( bli_cntx_method( cntx ) != BLIS_NAT )
{
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
}
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
// Invoke the internal back-end.
bli_l3_thread_decorator

16
frame/3/trmm3/bli_trmm3_front.c
View File

@@ -165,19 +165,11 @@ void bli_trmm3_front
// bli_l3_cntl_create_if()). This allows us to access the schemas from
// the control tree, which hopefully reduces some confusion, particularly
// in bli_packm_init().
if ( bli_cntx_method( cntx ) == BLIS_NAT )
{
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &a_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &b_local );
}
else // if ( bli_cntx_method( cntx ) != BLIS_NAT )
{
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
}
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
// Invoke the internal back-end.
bli_l3_thread_decorator

16
frame/3/trsm/bli_trsm_front.c
View File

@@ -147,19 +147,11 @@ void bli_trsm_front
// bli_l3_cntl_create_if()). This allows us to access the schemas from
// the control tree, which hopefully reduces some confusion, particularly
// in bli_packm_init().
if ( bli_cntx_method( cntx ) == BLIS_NAT )
{
bli_obj_set_pack_schema( BLIS_PACKED_ROW_PANELS, &a_local );
bli_obj_set_pack_schema( BLIS_PACKED_COL_PANELS, &b_local );
}
else // if ( bli_cntx_method( cntx ) != BLIS_NAT )
{
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
pack_t schema_a = bli_cntx_schema_a_block( cntx );
pack_t schema_b = bli_cntx_schema_b_panel( cntx );
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
}
bli_obj_set_pack_schema( schema_a, &a_local );
bli_obj_set_pack_schema( schema_b, &b_local );
// Invoke the internal back-end.
bli_l3_thread_decorator