Packed A matrix stride update to account for fringe cases.

-When A matrix is packed, it is packed in blocks of MRxKC, to form a
whole packed MCxKC block. If the m value is not a multiple of MR, then
the m % MR block is packed in a different manner as opposed to the MR
blocks. Subsequently the strides of the packed MR block and m % MR
blocks are different and the same needs to be updated when calling the
GEMV kernels with packed A matrix.
-Fixes to address compiler warnings.

AMD-Internal: [SWLCSG-3359]
Change-Id: I7f47afbc9cd92536cb375431d74d9b8bca7bab44

addon/aocl_gemm/aocl_gemm_bf16_utils.c

@@ -652,7 +652,7 @@ AOCL_GEMM_REORDER(int8_t, bf16s4f32of32)
 	b.cs = cs_b;
 	b.width = n;
 	b.length = k;
-	b.mtag = input_mat_type;
+	b.mat_type = input_mat_type;
 
 	reorderb_nr64_bf16s4f32of32(&b, &b_reorder, &rntm_g, lcntx_g);
 }

addon/aocl_gemm/frame/bf16bf16f32/lpgemm_reorder_bf16.c

@@ -403,7 +403,7 @@ void reorderb_nr64_bf16s4f32of32
 	dim_t cs_b = b->cs;
 	dim_t n = b->width;
 	dim_t k = b->length;
-	AOCL_MATRIX_TYPE mtag = b->mtag;
+	AOCL_MATRIX_TYPE mat_type = b->mat_type;
 
 	dim_t rs_b_reorder;
 	dim_t cs_b_reorder;
@@ -495,13 +495,16 @@ void reorderb_nr64_bf16s4f32of32
 				// st = ( jc_cur_loop * k )    <traverse blocks 1,2,3,4>
 				//    + ( n_sub_updated * pc ) <traverse block 5>
 				//    + ( NC' * kc0_updated)   <traverse block 6>
-				((pack_s4)lcntx->packb_fun_ptr)(
-					((int8_t *)b_reorder->storage.aligned_buffer) +
-						( (jc_cur_loop * k_updated) + (n_sub_updated * pc) +
-						(jc_cur_loop_rem * kc0_updated) ) / 2,
-					(((int8_t *)b->storage.aligned_buffer) +
-					  ( (rs_b * pc) + (jc * cs_b) ) / 2),
-					rs_b, cs_b, nc0, kc0, &rs_b_reorder, &cs_b_reorder, NULL, mtag);
+				((pack_s4)lcntx->packb_fun_ptr)
+				(
+				  ( ( int8_t* )b_reorder->storage.aligned_buffer ) +
+					( ( jc_cur_loop * k_updated ) + ( n_sub_updated * pc ) +
+					( jc_cur_loop_rem * kc0_updated ) ) / 2,
+				  ( ( ( int8_t* )b->storage.aligned_buffer ) +
+				  ( ( rs_b * pc ) + ( jc * cs_b ) ) / 2 ),
+				  rs_b, cs_b, nc0, kc0, &rs_b_reorder, &cs_b_reorder,
+				  NULL, mat_type
+				);
 			}
 
 			adjust_B_panel_reordered_jc(&jc, jc_cur_loop);
@@ -637,4 +640,4 @@ void reorderb_mxp_nr64_f32obf16
 	b_reorder->rs = rs_b_reorder;
 	b_reorder->cs = cs_b_reorder;
 	b_reorder->mtag = REORDERED;
-}
+}

addon/aocl_gemm/frame/lpgemm_types.h

@@ -136,6 +136,7 @@ typedef struct
 	dim_t cs;
 
 	AOCL_MEMORY_TAG mtag;
+	AOCL_MATRIX_TYPE mat_type;
 
 	lpgemm_mem_t storage;
 } lpgemm_obj_t;

addon/aocl_gemm/frame/s8s8s32/lpgemm_s8s8s32.c

@@ -211,6 +211,8 @@ LPGEMV(int8_t,int8_t,int32_t,s8s8s32o32)
 	}
 	else
 	{
+		dim_t gemm_MR = lcntx->blksz.MR;
+
 		dim_t jc_start, jc_end;
 		thread_jc.n_way = ( thread_jc.n_way == 1 ) ?
 			( thread->n_threads ) : ( thread_jc.n_way );
@@ -246,6 +248,10 @@ LPGEMV(int8_t,int8_t,int32_t,s8s8s32o32)
 			  1, k,
 			  &rs_a_use, &cs_a_use
 			);
+			get_packa_strides_mfringe_u8s8s32os32
+			(
+			  &rs_a_use, &cs_a_use, gemm_MR, 1
+			);
 
 			a_use = pack_a_buffer_s8s8s32os32;
 		}

addon/aocl_gemm/frame/threading/lpgemm_thread_decor_openmp.c

@@ -53,12 +53,18 @@ BLIS_INLINE void calculate_n_threads_per_gemm
 	rntm_t* rntm_g
   )
 {
-	*n_threads = bli_rntm_num_threads( rntm_g ); \
-	*n_gemms_in_parallel = -1; \
-	if( *n_threads == 1 ) *n_gemms_in_parallel = 1; \
-	else if( *n_gemms_in_parallel < 1 ) *n_gemms_in_parallel = bli_min(*n_threads, batch_size); \
-	/* ToDo: All the leftover thrads might go under-utilized. Could be optimized further. */ \
-	*n_threads_per_gemm = ( *n_threads ) / *n_gemms_in_parallel;
+	*n_threads = bli_rntm_num_threads( rntm_g );
+	*n_gemms_in_parallel = -1;
+	if( *n_threads == 1 )
+	{
+		( *n_gemms_in_parallel ) = 1;
+	}
+	else if( *n_gemms_in_parallel < 1 )
+	{
+		( *n_gemms_in_parallel ) = bli_min( ( *n_threads ), batch_size );
+	}
+	/* ToDo: All the leftover thrads might go under-utilized. Could be optimized further. */
+	( *n_threads_per_gemm ) = ( *n_threads ) / ( *n_gemms_in_parallel );
 }
 
 BLIS_INLINE dim_t next_factor

addon/aocl_gemm/frame/u8s8s32/lpgemm_u8s8s32.c

@@ -4,7 +4,7 @@
    An object-based framework for developing high-performance BLAS-like
    libraries.
 
-   Copyright (C) 2022 - 2024, Advanced Micro Devices, Inc. All rights reserved.
+   Copyright (C) 2022 - 2025, Advanced Micro Devices, Inc. All rights reserved.
 
    Redistribution and use in source and binary forms, with or without
    modification, are permitted provided that the following conditions are
@@ -194,6 +194,8 @@ LPGEMV(uint8_t,int8_t,int32_t,u8s8s32os32)
 	}
 	else
 	{
+		dim_t gemm_MR = lcntx->blksz.MR;
+
 		// Compute the JC loop thread range for the current thread.
 		dim_t jc_start, jc_end;
 		thread_jc.n_way = ( thread_jc.n_way == 1 ) ?
@@ -227,6 +229,10 @@ LPGEMV(uint8_t,int8_t,int32_t,u8s8s32os32)
 			  1, k,
 			  &rs_a_use, &cs_a_use
 			);
+			get_packa_strides_mfringe_u8s8s32os32
+			(
+			  &rs_a_use, &cs_a_use, gemm_MR, 1
+			);
 
 			a_use = pack_a_buffer;
 		}

addon/aocl_gemm/kernels/u8s8s32/lpgemm_packa.h

@@ -4,7 +4,7 @@
    An object-based framework for developing high-performance BLAS-like
    libraries.
 
-   Copyright (C) 2022 - 2024, Advanced Micro Devices, Inc. All rights reserved.
+   Copyright (C) 2022 - 2025, Advanced Micro Devices, Inc. All rights reserved.
 
    Redistribution and use in source and binary forms, with or without
    modification, are permitted provided that the following conditions are
@@ -35,6 +35,20 @@
 #ifndef BLIS_GEMM_INT8_PACKA
 #define BLIS_GEMM_INT8_PACKA
 
+// The strides needs to be updated based on the m_fringe value to account
+// for different schemas used to pack A fringe cases.
+BLIS_INLINE void get_packa_strides_mfringe_u8s8s32os32
+     (
+       dim_t* rs,
+       dim_t* cs,
+       dim_t MR,
+       dim_t m_fringe
+     )
+{
+	( *rs ) = 4;
+	( *cs ) = ( ( *cs ) / MR ) * m_fringe;
+}
+
 typedef void (*packa_s32)
      (
        uint8_t*,

bench/bench_aocl_gemm/bench_input.txt

@@ -1,3 +1,4 @@
+r t n n r 1 128 64 1 128 128 *:none
 c n t n n 32 128 2 32 128 32 bf16bf16f32of32:bias=na,swish
 r n n n r 6 1 4 4 16 16 bf16s4f32of32:pre_op_scale=scalar,pre_op_scale_type=bf16,group_size=2
 r n n n r 6 1 4 4 16 16 bf16s4f32of32:pre_op_zp=vector,pre_op_scale=scalar,pre_op_scale_type=bf16,group_size=2