[Hotfix] correct occupancy calculation (#1451)

ROCm · Dec 16, 2024 · 41767d9 · 41767d9
1 parent 0ed9795
commit 41767d9
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/tensilelite/Tensile/KernelWriterAssembly.py b/tensilelite/Tensile/KernelWriterAssembly.py
@@ -97,7 +97,7 @@ def getOccupancy(self, numThreads, vgprs, sgprs, ldsSize, accvgprs=0, doubleVgpr
       vgprLimitedOccupancy    = self.getVgprOccupancy(numThreads, vgprs,          doubleVgpr)
       accvgprLimitedOccupancy = self.getVgprOccupancy(numThreads, accvgprs,       doubleVgpr)
     else:
-      vgprLimitedOccupancy    = self.getVgprOccupancy(numThreads, ceil(vgprs//8)*8+accvgprs, doubleVgpr)
+      vgprLimitedOccupancy    = self.getVgprOccupancy(numThreads, ceil(vgprs/8)*8+accvgprs, doubleVgpr)
       accvgprLimitedOccupancy = vgprLimitedOccupancy
     sgprLimitedOccupancy = self.getSgprOccupancy(sgprs)