s390x/Z14: Change register blocking for SGEMM to 16x4

Change register blocking for SGEMM (and STRMM) on z14 from 8x4 to 16x4
by adjusting SGEMM_DEFAULT_UNROLL_M and choosing the appropriate copy
implementations. Actually make KERNEL.Z14 more flexible, so that the
change in param.h suffices. As a result, performance for SGEMM improves
by around 30% on z15.

On z14, FP SIMD instructions can operate on float-sized scalars in
vector registers, while z13 could do that for double-sized scalars only.
Thus, we can double the amount of elements of C that are held in
registers in an SGEMM kernel.

Signed-off-by: Marius Hillenbrand <mhillen@linux.ibm.com>

kernel/zarch/gemm_vec.c

@@ -220,6 +220,15 @@ typedef FLOAT vector_float __attribute__ ((vector_size (16)));
 	}
 
 
+#if UNROLL_M == 16
+VECTOR_BLOCK(16, 4)
+VECTOR_BLOCK(16, 2)
+VECTOR_BLOCK(16, 1)
+#endif
+#if UNROLL_N == 8
+VECTOR_BLOCK(8, 8)
+VECTOR_BLOCK(4, 8)
+#endif
 VECTOR_BLOCK(8, 4)
 VECTOR_BLOCK(8, 2)
 VECTOR_BLOCK(8, 1)
@@ -284,6 +293,12 @@ static inline void GEBP_block(BLASLONG m, BLASLONG n,
 		return;                                         \
 	}
 
+#if UNROLL_M == 16
+	BLOCK(16, 4); BLOCK(16, 2); BLOCK(16, 1);
+#endif
+#if UNROLL_N == 8
+	BLOCK(8, 8); BLOCK(4, 8);
+#endif
 	BLOCK(8, 4); BLOCK(8, 2); BLOCK(8, 1);
 	BLOCK(4, 4); BLOCK(4, 2); BLOCK(4, 1);