enable the SGEMM/SKX C based kernel

In QA the final bug was found so now the sklyakex sgemm C based kernel can
be activated....

Author: fenrus75

kernel/x86_64/KERNEL.SKYLAKEX

@@ -1,6 +1,11 @@
 include $(KERNELDIR)/KERNEL.HASWELL
 
-SGEMMKERNEL    =  sgemm_kernel_16x4_skylakex.S
+SGEMMKERNEL    =  sgemm_kernel_16x4_skylakex.c
+
+SGEMMINCOPY    =  ../generic/gemm_ncopy_16.c
+SGEMMITCOPY    =  sgemm_tcopy_16_skylakex.c
+SGEMMONCOPY    =  sgemm_ncopy_4_skylakex.c
+SGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
 
 DGEMMKERNEL    =  dgemm_kernel_4x8_skylakex.c
 
@@ -9,5 +14,5 @@ DGEMMITCOPY    =  dgemm_tcopy_8_skylakex.c
 DGEMMONCOPY    =  dgemm_ncopy_8_skylakex.c
 DGEMMOTCOPY    =  dgemm_tcopy_8_skylakex.c
 
-SGEMM_BETA = ../generic/gemm_beta.c
+SGEMM_BETA = sgemm_beta_skylakex.c
 DGEMM_BETA = dgemm_beta_skylakex.c

kernel/x86_64/sgemm_beta_skylakex.c

@@ -60,8 +60,10 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT beta,
 
   if (beta == ZERO){
     __m512 z_zero;
+    __m256 y_zero;
 
     z_zero = _mm512_setzero_ps();
+    y_zero = _mm256_setzero_ps();
     j = n;
     do {
       c_offset1 = c_offset;
@@ -71,14 +73,12 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT beta,
 
       while (i > 32) {
 	  _mm512_storeu_ps(c_offset1, z_zero);
-	  _mm512_storeu_ps(c_offset1 + 8, z_zero);
 	  _mm512_storeu_ps(c_offset1 + 16, z_zero);
-	  _mm512_storeu_ps(c_offset1 + 24 , z_zero);
 	  c_offset1 += 32;
 	  i -= 32;
       }
       while (i > 8) {
-	  _mm512_storeu_ps(c_offset1, z_zero);
+	  _mm256_storeu_ps(c_offset1, y_zero);
 	  c_offset1 += 8;
 	  i -= 8;
       }