Fix coverity issues (#2530)

In the upstream: triton-lang/triton#4967

---------

Signed-off-by: Anatoly Myachev <[email protected]>

Author: anmyachev

lib/Analysis/AxisInfo.cpp

@@ -895,7 +895,7 @@ class ShLIOpAxisInfoVisitor final : public BinaryOpVisitorImpl<arith::ShLIOp> {
       lhsDivisibility = 1;
     }
     auto numBits = log2Int(lhsDivisibility);
-    return multiplyDivisor(lhsDivisibility, 1 << shift);
+    return multiplyDivisor(lhsDivisibility, 1ll << shift);
   }
 
   int64_t getConstancy(arith::ShLIOp op, const AxisInfo &lhs,

lib/Conversion/TritonGPUToLLVM/ScanOpToLLVM.cpp

@@ -76,7 +76,7 @@ static void warpScan(SmallVector<SmallVector<Value>> &srcValues,
         acc[j] = select(mask, tempAcc[j], acc[j]);
       }
     }
-    srcValues[srcIndex] = acc;
+    srcValues[srcIndex] = std::move(acc);
   }
 }
 
@@ -128,8 +128,8 @@ static void AddPartialReduce(SmallVector<SmallVector<Value>> &srcValues,
                              ConversionPatternRewriter &rewriter,
                              const TargetInfoBase &targetInfo,
                              ScanLoweringHelper &helper,
-                             SmallVector<Value> smemBases,
-                             SmallVector<Type> smemTypes, Value warpId,
+                             ArrayRef<Value> smemBases,
+                             ArrayRef<Type> smemTypes, Value warpId,
                              Value laneIdAxis, Value parallelLaneId) {
   Location loc = helper.getLoc();
   unsigned numParallelLane = helper.getNonAxisNumThreadsPerCTA();
@@ -224,7 +224,7 @@ static void AddPartialReduce(SmallVector<SmallVector<Value>> &srcValues,
                                 srcValues[srcIndex - i * elementStride][j]);
         }
       }
-      srcValues[srcIndex - i * elementStride] = laneValue;
+      srcValues[srcIndex - i * elementStride] = std::move(laneValue);
     }
     // For the next chunk start back from the value containing the
     // accumulated value of all the warps.
@@ -303,7 +303,7 @@ static void AddPartialReduceOneWarp(SmallVector<SmallVector<Value>> &srcValues,
                      srcValues[srcIndex - i * elementStride][j], laneValue[j]);
         }
       }
-      srcValues[srcIndex - i * elementStride] = laneValue;
+      srcValues[srcIndex - i * elementStride] = std::move(laneValue);
     }
     // For the next chunk start back from the value containing the
     // accumulated value of all the warps.

third_party/intel/lib/Analysis/AxisInfo.cpp

@@ -900,7 +900,7 @@ class ShLIOpAxisInfoVisitor final : public BinaryOpVisitorImpl<arith::ShLIOp> {
       lhsDivisibility = 1;
     }
     auto numBits = log2Int(lhsDivisibility);
-    return multiplyDivisor(lhsDivisibility, 1 << shift);
+    return multiplyDivisor(lhsDivisibility, 1ll << shift);
   }
 
   int64_t getConstancy(arith::ShLIOp op, const AxisInfo &lhs,

third_party/intel/lib/TritonIntelGPUToLLVM/ScanOpToLLVM.cpp

@@ -74,7 +74,7 @@ static void warpScan(SmallVector<SmallVector<Value>> &srcValues,
         acc[j] = select(mask, tempAcc[j], acc[j]);
       }
     }
-    srcValues[srcIndex] = acc;
+    srcValues[srcIndex] = std::move(acc);
   }
 }
 
@@ -126,8 +126,8 @@ static void AddPartialReduce(SmallVector<SmallVector<Value>> &srcValues,
                              ConversionPatternRewriter &rewriter,
                              const TargetInfoBase &targetInfo,
                              ScanLoweringHelper &helper,
-                             SmallVector<Value> smemBases,
-                             SmallVector<Type> smemTypes, Value warpId,
+                             ArrayRef<Value> smemBases,
+                             ArrayRef<Type> smemTypes, Value warpId,
                              Value laneIdAxis, Value parallelLaneId) {
   Location loc = helper.getLoc();
   unsigned numParallelLane = helper.getNonAxisNumThreadsPerCTA();
@@ -222,7 +222,7 @@ static void AddPartialReduce(SmallVector<SmallVector<Value>> &srcValues,
                                 srcValues[srcIndex - i * elementStride][j]);
         }
       }
-      srcValues[srcIndex - i * elementStride] = laneValue;
+      srcValues[srcIndex - i * elementStride] = std::move(laneValue);
     }
     // For the next chunk start back from the value containing the
     // accumulated value of all the warps.
@@ -301,7 +301,7 @@ static void AddPartialReduceOneWarp(SmallVector<SmallVector<Value>> &srcValues,
                      srcValues[srcIndex - i * elementStride][j], laneValue[j]);
         }
       }
-      srcValues[srcIndex - i * elementStride] = laneValue;
+      srcValues[srcIndex - i * elementStride] = std::move(laneValue);
     }
     // For the next chunk start back from the value containing the
     // accumulated value of all the warps.

third_party/intel/lib/TritonIntelGPUTransforms/Coalesce.cpp

@@ -40,7 +40,7 @@ struct CoalescePass
       llvm::dbgs() << "\n";
     });
 
-    auto contiguity = axisInfoAnalysis.getAxisInfo(ptr)->getContiguity();
+    const auto &contiguity = axisInfoAnalysis.getAxisInfo(ptr)->getContiguity();
     SmallVector<unsigned> order = argSort(contiguity);
     LDBG("order=[" << triton::join(order, ", ") << "]");