Add msa support for loongson

1. Using core loongson3r3 and loongson3r4 for loongson
2. Add DYNAMIC_ARCH for loongson

Change-Id: I1c6b54dbeca3a0cc31d1222af36a7e9bd6ab54c1
This commit is contained in:
gxw
2020-11-26 14:59:41 +08:00
parent d67babf345
commit 4b548857d6
28 changed files with 682 additions and 361 deletions

View File

@@ -121,7 +121,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define CGEMM_KERNEL_8X1_MSA(OP0, OP1, OP2, OP3, OP4) \
{ \
LD_SP4_INC(pa0, 4, src_a0, src_a1, src_a2, src_a3); \
src_bi = (v4f32) __msa_cast_to_vector_double(*((double *) pb0)); \
src_bi = (v4f32) COPY_DOUBLE_TO_VECTOR(*((double *) pb0)); \
SPLATI_W2_SP(src_bi, 0, src_br, src_bi); \
\
PCKEVOD_W2_SP(src_a1, src_a0, src_a0r, src_a0i); \
@@ -200,7 +200,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define CGEMM_KERNEL_4X1_MSA(OP0, OP1, OP2, OP3, OP4) \
{ \
LD_SP2_INC(pa0, 4, src_a0, src_a1); \
src_bi = (v4f32) __msa_cast_to_vector_double(*((double *) pb0)); \
src_bi = (v4f32) COPY_DOUBLE_TO_VECTOR(*((double *) pb0)); \
SPLATI_W2_SP(src_bi, 0, src_br, src_bi); \
\
PCKEVOD_W2_SP(src_a1, src_a0, src_a0r, src_a0i); \