skylakex: Make the sgemm/dgemm beta code robust for a N=0 or M=0 case

in the threading code there are cases where N or M can become 0, and the optimized beta code did not handle this well, leading to a crash during the audit for the crash a few edge conditions on the if statements were found and fixed as well
OpenMathLib · Nov 1, 2018 · dcc5d62 · dcc5d62
1 parent f5595d0
commit dcc5d62
Show file tree

Hide file tree

Showing 2 changed files with 8 additions and 4 deletions.
diff --git a/kernel/x86_64/dgemm_beta_skylakex.c b/kernel/x86_64/dgemm_beta_skylakex.c
@@ -55,6 +55,8 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT beta,
 	return 0;
   }
 
+  if (m == 0 || n == 0)
+	return 0;
 
   c_offset = c;
 
@@ -69,15 +71,15 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT beta,
 
       i = m;
 
-      while (i > 32) {
+      while (i >= 32) {
 	  _mm512_storeu_pd(c_offset1, z_zero);
 	  _mm512_storeu_pd(c_offset1 + 8, z_zero);
 	  _mm512_storeu_pd(c_offset1 + 16, z_zero);
 	  _mm512_storeu_pd(c_offset1 + 24 , z_zero);
 	  c_offset1 += 32;
 	  i -= 32;
       }
-      while (i > 8) {
+      while (i >= 8) {
 	  _mm512_storeu_pd(c_offset1, z_zero);
 	  c_offset1 += 8;
 	  i -= 8;

diff --git a/kernel/x86_64/sgemm_beta_skylakex.c b/kernel/x86_64/sgemm_beta_skylakex.c
@@ -55,6 +55,8 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT beta,
 	return 0;
   }
 
+  if (n == 0 || m == 0)
+	return;
 
   c_offset = c;
 
@@ -71,13 +73,13 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT beta,
 
       i = m;
 
-      while (i > 32) {
+      while (i >= 32) {
 	  _mm512_storeu_ps(c_offset1, z_zero);
 	  _mm512_storeu_ps(c_offset1 + 16, z_zero);
 	  c_offset1 += 32;
 	  i -= 32;
       }
-      while (i > 8) {
+      while (i >= 8) {
 	  _mm256_storeu_ps(c_offset1, y_zero);
 	  c_offset1 += 8;
 	  i -= 8;