Merge commit 'b683d01b' into amd-main

* commit 'b683d01b':
  Use extra #undef when including ba/ex API headers.
  Minor preprocessor/header cleanup.
  Fixed typo in cpp guard in bli_util_ft.h.
  Defined eqsc, eqv, eqm to test object equality.
  Defined setijv, getijv to set/get vector elements.
  Minor API breakage in bli_pack API.
  Add err_t* "return" parameter to malloc functions.
  Always stay initialized after BLAS compat calls.
  Renamed membrk files/vars/functions to pba.
  Switch allocator mutexes to static initialization.

AMD-Internal: [CPUPL-2698]
Change-Id: Ied2ca8619f144d4b8a7123ac45a1be0dda3875df

sandbox/gemmlike/bls_l3_packm_a.c

@@ -79,7 +79,7 @@ void PASTECH2(bls_,ch,opname) \
 			   the current function before the other threads have a chance to
 			   copy from it. (A barrier would fix that race condition, but then
 			   again, I prefer to keep barriers to a minimum.) */ \
-			bli_membrk_acquire_m \
+			bli_pba_acquire_m \
 			( \
 			  rntm, \
 			  size_needed, \
@@ -123,12 +123,12 @@ void PASTECH2(bls_,ch,opname) \
 				   above for why the acquisition needs to be directly to
 				   the chief thread's passed-in mem_t and not a local
 				   (temporary) mem_t. */ \
-				bli_membrk_release \
+				bli_pba_release \
 				( \
 				  rntm, \
 				  mem \
 				); \
-				bli_membrk_acquire_m \
+				bli_pba_acquire_m \
 				( \
 				  rntm, \
 				  size_needed, \
@@ -182,7 +182,7 @@ void PASTECH2(bls_,ch,opname) \
 		   is allocated, which it should be. */ \
 		if ( bli_mem_is_alloc( mem ) ) \
 		{ \
-			bli_membrk_release \
+			bli_pba_release \
 			( \
 			  rntm, \
 			  mem \

sandbox/gemmlike/bls_l3_packm_b.c

@@ -79,7 +79,7 @@ void PASTECH2(bls_,ch,opname) \
 			   the current function before the other threads have a chance to
 			   copy from it. (A barrier would fix that race condition, but then
 			   again, I prefer to keep barriers to a minimum.) */ \
-			bli_membrk_acquire_m \
+			bli_pba_acquire_m \
 			( \
 			  rntm, \
 			  size_needed, \
@@ -123,12 +123,12 @@ void PASTECH2(bls_,ch,opname) \
 				   above for why the acquisition needs to be directly to
 				   the chief thread's passed-in mem_t and not a local
 				   (temporary) mem_t. */ \
-				bli_membrk_release \
+				bli_pba_release \
 				( \
 				  rntm, \
 				  mem \
 				); \
-				bli_membrk_acquire_m \
+				bli_pba_acquire_m \
 				( \
 				  rntm, \
 				  size_needed, \
@@ -182,7 +182,7 @@ void PASTECH2(bls_,ch,opname) \
 		   is allocated, which it should be. */ \
 		if ( bli_mem_is_alloc( mem ) ) \
 		{ \
-			bli_membrk_release \
+			bli_pba_release \
 			( \
 			  rntm, \
 			  mem \

sandbox/gemmlike/thread/bls_l3_decor_openmp.c

@@ -75,7 +75,7 @@ void bls_l3_thread_decorator
 	// Set the packing block allocator field of the rntm. This will be
 	// inherited by all of the child threads when they make local copies of
 	// the rntm below.
-	bli_membrk_rntm_set_membrk( rntm );
+	bli_pba_rntm_set_pba( rntm );
 
 	// Allcoate a global communicator for the root thrinfo_t structures.
 	thrcomm_t* restrict gl_comm = bli_thrcomm_create( rntm, n_threads );

sandbox/gemmlike/thread/bls_l3_decor_pthreads.c

@@ -140,7 +140,7 @@ void bls_l3_thread_decorator
 	// Set the packing block allocator field of the rntm. This will be
 	// inherited by all of the child threads when they make local copies of
 	// the rntm below.
-	bli_membrk_rntm_set_membrk( rntm );
+	bli_pba_rntm_set_pba( rntm );
 
 	// Allocate a global communicator for the root thrinfo_t structures.
 	thrcomm_t* restrict gl_comm = bli_thrcomm_create( rntm, n_threads );

sandbox/gemmlike/thread/bls_l3_decor_single.c

@@ -68,7 +68,7 @@ void bls_l3_thread_decorator
 	bli_sba_rntm_set_pool( 0, array, rntm );
 
 	// Set the packing block allocator field of the rntm.
-	bli_membrk_rntm_set_membrk( rntm );
+	bli_pba_rntm_set_pba( rntm );
 
 #ifndef SKIP_THRINFO_TREE
 	// Allcoate a global communicator for the root thrinfo_t structures.

sandbox/ref99/old/packm/blx_l3_packm.c

@@ -45,7 +45,7 @@ void blx_l3_packm
        thrinfo_t* thread
      )
 {
-	membrk_t* membrk;
+	pba_t* pba;
 	packbuf_t pack_buf_type;
 	mem_t*    cntl_mem_p;
 	siz_t     size_needed;
@@ -71,7 +71,7 @@ void blx_l3_packm
 	if ( size_needed == 0 ) return;
 
 	// Query the memory broker from the context.
-	membrk = bli_cntx_get_membrk( cntx );
+	pba = bli_cntx_get_pba( cntx );
 
 	// Query the pack buffer type from the control tree node.
 	pack_buf_type = bli_cntl_packm_params_pack_buf_type( cntl );
@@ -91,9 +91,9 @@ void blx_l3_packm
 		{
 			// The chief thread acquires a block from the memory broker
 			// and saves the associated mem_t entry to local_mem_s.
-			bli_membrk_acquire_m
+			bli_pba_acquire_m
 			(
-			  membrk,
+			  pba,
 			  size_needed,
 			  pack_buf_type,
 			  &local_mem_s
@@ -130,10 +130,10 @@ void blx_l3_packm
 				// The chief thread releases the existing block associated with
 				// the mem_t entry in the control tree, and then re-acquires a
 				// new block, saving the associated mem_t entry to local_mem_s.
-				bli_membrk_release( cntl_mem_p );
-				bli_membrk_acquire_m
+				bli_pba_release( cntl_mem_p );
+				bli_pba_acquire_m
 				(
-				  membrk,
+				  pba,
 				  size_needed,
 				  pack_buf_type,
 				  &local_mem_s

sandbox/ref99/old/thread/blx_gemm_thread.c

@@ -147,7 +147,7 @@ void blx_gemm_thread
 	// Set the packing block allocator field of the rntm. This will be
 	// inherited by all of the child threads when they make local copies of
 	// the rntm below.
-	bli_membrk_rntm_set_membrk( rntm );
+	bli_pba_rntm_set_pba( rntm );
 
 	// Allocate a global communicator for the root thrinfo_t structures.
 	thrcomm_t* restrict gl_comm = bli_thrcomm_create( rntm, n_threads );