Merge pull request #5175 from shubhamsvc/dgemv_thread_throttling

Add thread throttling profile for DGEMV on NEOVERSEV1

interface/gemv.c

@@ -70,11 +70,22 @@ static int (*gemv_thread[])(BLASLONG, BLASLONG, FLOAT, FLOAT *, BLASLONG,  FLOAT
 
 #if defined(DYNAMIC_ARCH) || defined(NEOVERSEV1)
 static inline int get_gemv_optimal_nthreads_neoversev1(BLASLONG MN, int ncpu) {
-  return
-      MN < 25600L    ? 1
-    : MN < 63001L    ? MIN(ncpu, 4)
-    : MN < 459684L   ? MIN(ncpu, 16)
-    : ncpu;
+  #ifdef DOUBLE
+      return (MN < 8100L)      ? 1
+          : (MN < 12100L)     ? MIN(ncpu, 2)
+          : (MN < 36100L)     ? MIN(ncpu, 4)
+          : (MN < 84100L)     ? MIN(ncpu, 8)
+          : (MN < 348100L)    ? MIN(ncpu, 16)
+          : (MN < 435600L)    ? MIN(ncpu, 24)
+          : (MN < 810000L)    ? MIN(ncpu, 32)
+          : (MN < 1050625L)   ? MIN(ncpu, 40)
+          : ncpu;
+  #else
+      return (MN < 25600L)     ? 1
+          : (MN < 63001L)     ? MIN(ncpu, 4)
+          : (MN < 459684L)    ? MIN(ncpu, 16)
+          : ncpu;
+  #endif
 }
 #endif
 
@@ -96,11 +107,11 @@ static inline int get_gemv_optimal_nthreads(BLASLONG MN) {
     return num_cpu_avail(4);
   return 1;
 #endif
-#if defined(NEOVERSEV1) && !defined(COMPLEX) && !defined(DOUBLE) && !defined(BFLOAT16)
+#if defined(NEOVERSEV1) && !defined(COMPLEX) && !defined(BFLOAT16)
   return get_gemv_optimal_nthreads_neoversev1(MN, ncpu);
 #elif defined(NEOVERSEV2) && !defined(COMPLEX) && !defined(DOUBLE) && !defined(BFLOAT16)
   return get_gemv_optimal_nthreads_neoversev2(MN, ncpu);
-#elif defined(DYNAMIC_ARCH) && !defined(COMPLEX) && !defined(DOUBLE) && !defined(BFLOAT16)
+#elif defined(DYNAMIC_ARCH) && !defined(COMPLEX) && !defined(BFLOAT16)
   if (strcmp(gotoblas_corename(), "neoversev1") == 0) {
     return get_gemv_optimal_nthreads_neoversev1(MN, ncpu);
   }