[OpenMP] Avoid remainder operations for loop index values on a collapsed loop nest.

[gtbercea]

Summary

Change the strategy for computing loop index variables after collapsing a loop nest via the collapse clause by replacing the expensive remainder operation with multiplications and additions.

[ABataev]

lib/Sema/SemaOpenMP.cpp

5594

No need for .get() here

5595

Use preincrement

5604

1. Preincrement.
2. Format.

[gtbercea]

Invert accumulation direction.

[gtbercea]

Fix update.

[ABataev]

lib/Sema/SemaOpenMP.cpp

5599–5613

Do you need to initialize Iter here?

[gtbercea]

Remove redundant initalization.

[ABataev]

LG

[gtbercea]

Fix.

lib/Sema/SemaOpenMP.cpp

checkOpenMPLoop(OpenMPDirectiveKind DKind, Expr *CollapseLoopCountExpr,
 
   // Build updates and final values of the loop counters.
   bool HasErrors = false;
   Built.Counters.resize(NestedLoopCount);
   Built.Inits.resize(NestedLoopCount);
   Built.Updates.resize(NestedLoopCount);
   Built.Finals.resize(NestedLoopCount);
   {
-    ExprResult Div;
-    // Go from inner nested loop to outer.
-    for (int Cnt = NestedLoopCount - 1; Cnt >= 0; --Cnt) {
+    // We implement the following algorithm for obtaining the
+    // original loop iteration variable values based on the
+    // value of the collapsed loop iteration variable IV.
+    //
+    // Let n+1 be the number of collapsed loops in the nest.
+    // Iteration variables (I0, I1, .... In)
+    // Iteration counts (N0, N1, ... Nn)
+    //
+    // Acc = IV;
+    //
+    // To compute Ik for loop k, 0 <= k <= n, generate:
+    //    Prod = N(k+1) * N(k+2) * ... * Nn;
+    //    Ik = Acc / Prod;
+    //    Acc -= Ik * Prod;
+    //
+    ExprResult Acc = IV;
+    for (unsigned int Cnt = 0; Cnt < NestedLoopCount; ++Cnt) {
       LoopIterationSpace &IS = IterSpaces[Cnt];
       SourceLocation UpdLoc = IS.IncSrcRange.getBegin();
-      // Build: Iter = (IV / Div) % IS.NumIters
-      // where Div is product of previous iterations' IS.NumIters.
       ExprResult Iter;
-      if (Div.isUsable()) {
-        Iter =
-            SemaRef.BuildBinOp(CurScope, UpdLoc, BO_Div, IV.get(), Div.get());
-      } else {
-        Iter = IV;
-        assert((Cnt == (int)NestedLoopCount - 1) &&
-               "unusable div expected on first iteration only");
-      }
 
-      if (Cnt != 0 && Iter.isUsable())
-        Iter = SemaRef.BuildBinOp(CurScope, UpdLoc, BO_Rem, Iter.get(),
-                                  IS.NumIterations);
+      // Compute prod
+      ExprResult Prod =
+          SemaRef.ActOnIntegerConstant(SourceLocation(), 1).get();
+      for (unsigned int K = Cnt+1; K < NestedLoopCount; ++K)
+        Prod = SemaRef.BuildBinOp(CurScope, UpdLoc, BO_Mul, Prod.get(),
+                                  IterSpaces[K].NumIterations);
+
+      // Iter = Acc / Prod
+      // If there is at least one more inner loop to avoid
+      // multiplication by 1.
+      if (Cnt + 1 < NestedLoopCount)
+        Iter = SemaRef.BuildBinOp(CurScope, UpdLoc, BO_Div,
+                                  Acc.get(), Prod.get());
+      else
+        Iter = Acc;
       if (!Iter.isUsable()) {
         HasErrors = true;
         break;
       }
 
+      // Update Acc:
+      // Acc -= Iter * Prod
+      // Check if there is at least one more inner loop to avoid
+      // multiplication by 1.
+      if (Cnt + 1 < NestedLoopCount)
+        Prod = SemaRef.BuildBinOp(CurScope, UpdLoc, BO_Mul,
+                                  Iter.get(), Prod.get());
+      else
+        Prod = Iter;
+      Acc = SemaRef.BuildBinOp(CurScope, UpdLoc, BO_Sub,
+                               Acc.get(), Prod.get());
+
       // Build update: IS.CounterVar(Private) = IS.Start + Iter * IS.Step
       auto *VD = cast<VarDecl>(cast<DeclRefExpr>(IS.CounterVar)->getDecl());
       DeclRefExpr *CounterVar = buildDeclRefExpr(
           SemaRef, VD, IS.CounterVar->getType(), IS.CounterVar->getExprLoc(),
           /*RefersToCapture=*/true);
       ExprResult Init = buildCounterInit(SemaRef, CurScope, UpdLoc, CounterVar,
                                          IS.CounterInit, Captures);
       if (!Init.isUsable()) {
for (unsigned int Cnt = 0; Cnt < NestedLoopCount; ++Cnt) {
       ExprResult Final = buildCounterUpdate(
           SemaRef, CurScope, UpdLoc, CounterVar, IS.CounterInit,
           IS.NumIterations, IS.CounterStep, IS.Subtract, &Captures);
       if (!Final.isUsable()) {
         HasErrors = true;
         break;
       }
 
-      // Build Div for the next iteration: Div <- Div * IS.NumIters
-      if (Cnt != 0) {
-        if (Div.isUnset())
-          Div = IS.NumIterations;
-        else
-          Div = SemaRef.BuildBinOp(CurScope, UpdLoc, BO_Mul, Div.get(),
-                                   IS.NumIterations);
-
-        // Add parentheses (for debugging purposes only).
-        if (Div.isUsable())
-          Div = tryBuildCapture(SemaRef, Div.get(), Captures);
-        if (!Div.isUsable()) {
-          HasErrors = true;
-          break;
-        }
-      }
       if (!Update.isUsable() || !Final.isUsable()) {
         HasErrors = true;
         break;
       }
       // Save results
       Built.Counters[Cnt] = IS.CounterVar;
       Built.PrivateCounters[Cnt] = IS.PrivateCounterVar;
       Built.Inits[Cnt] = Init.get();

test/OpenMP/for_codegen.cpp

void loop_with_counter_collapse() {
   // LIFETIME: call void @llvm.lifetime.end
   // LIFETIME: call void @llvm.lifetime.end
   // LIFETIME: call void @llvm.lifetime.end
   // LIFETIME: call void @llvm.lifetime.end
   // LIFETIME: call void @llvm.lifetime.end
   // LIFETIME: call void @llvm.lifetime.end
   // LIFETIME: call void @llvm.lifetime.end
   // LIFETIME: call void @llvm.lifetime.end
-  // LIFETIME: call void @llvm.lifetime.end
   #pragma omp for collapse(2)
   for (int i = 0; i < 4; i++) {
     for (int j = i; j < 4; j++) {
     }
   }
 }
 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
 void without_schedule_clause(float *a, float *b, float *c, float *d) {

test/OpenMP/for_simd_codegen.cpp

 // Loop counters i and j updates:
 // CHECK: [[IV1:%.+]] = load i64, i64* [[T1_OMP_IV]]
 // CHECK-NEXT: [[I_1:%.+]] = sdiv i64 [[IV1]], 4
 // CHECK-NEXT: [[I_1_MUL1:%.+]] = mul nsw i64 [[I_1]], 1
 // CHECK-NEXT: [[I_1_ADD0:%.+]] = add nsw i64 0, [[I_1_MUL1]]
 // CHECK-NEXT: [[I_2:%.+]] = trunc i64 [[I_1_ADD0]] to i32
 // CHECK-NEXT: store i32 [[I_2]], i32*
 // CHECK: [[IV2:%.+]] = load i64, i64* [[T1_OMP_IV]]
-// CHECK-NEXT: [[J_1:%.+]] = srem i64 [[IV2]], 4
+// CHECK-NEXT: [[IV2_1:%.+]] = load i64, i64* [[T1_OMP_IV]]
+// CHECK-NEXT: [[DIV_2:%.+]] = sdiv i64 [[IV2_1]], 4
+// CHECK-NEXT: [[MUL_2:%.+]] = mul nsw i64 [[DIV_2]], 4
+// CHECK-NEXT: [[J_1:%.+]] = sub nsw i64 [[IV2]], [[MUL_2]]
 // CHECK-NEXT: [[J_2:%.+]] = mul nsw i64 [[J_1]], 2
 // CHECK-NEXT: [[J_2_ADD0:%.+]] = add nsw i64 0, [[J_2]]
 // CHECK-NEXT: store i64 [[J_2_ADD0]], i64*
 // simd.for.inc:
 // CHECK: [[IV3:%.+]] = load i64, i64* [[T1_OMP_IV]]
 // CHECK-NEXT: [[INC:%.+]] = add nsw i64 [[IV3]], 1
 // CHECK-NEXT: store i64 [[INC]], i64* [[T1_OMP_IV]]
 // CHECK-NEXT: br label {{%.+}}

 // CHECK: [[COLL1_BODY]]:
 // Start of body: calculate i from index:
 // CHECK: [[IV1:%.+]] = load i32, i32* [[OMP_IV]]
 // Calculation of the loop counters values.
 // CHECK: [[CALC_I_1:%.+]] = udiv i32 [[IV1]], 60
 // CHECK-NEXT: [[CALC_I_1_MUL1:%.+]] = mul i32 [[CALC_I_1]], 1
 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 1, [[CALC_I_1_MUL1]]
 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
+
 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]
-// CHECK-NEXT: [[CALC_J_1:%.+]] = udiv i32 [[IV1_2]], 20
-// CHECK-NEXT: [[CALC_J_2:%.+]] = urem i32 [[CALC_J_1]], 3
+// CHECK: [[IV1_2_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[CALC_J_1:%.+]] = udiv i32 [[IV1_2_1]], 60
+// CHECK-NEXT: [[MUL_1:%.+]] = mul i32 [[CALC_J_1]], 60
+// CHECK-NEXT: [[SUB_3:%.+]] = sub i32 [[IV1_2]], [[MUL_1]]
+// CHECK-NEXT: [[CALC_J_2:%.+]] = udiv i32 [[SUB_3]], 20
 // CHECK-NEXT: [[CALC_J_2_MUL1:%.+]] = mul i32 [[CALC_J_2]], 1
 // CHECK-NEXT: [[CALC_J_3:%.+]] = add i32 2, [[CALC_J_2_MUL1]]
 // CHECK-NEXT: store i32 [[CALC_J_3]], i32* [[LC_J:.+]]
+
 // CHECK: [[IV1_3:%.+]] = load i32, i32* [[OMP_IV]]
-// CHECK-NEXT: [[CALC_K_1:%.+]] = udiv i32 [[IV1_3]], 5
-// CHECK-NEXT: [[CALC_K_2:%.+]] = urem i32 [[CALC_K_1]], 4
-// CHECK-NEXT: [[CALC_K_2_MUL1:%.+]] = mul i32 [[CALC_K_2]], 1
-// CHECK-NEXT: [[CALC_K_3:%.+]] = add i32 3, [[CALC_K_2_MUL1]]
-// CHECK-NEXT: store i32 [[CALC_K_3]], i32* [[LC_K:.+]]
+// CHECK: [[IV1_3_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_1:%.+]] = udiv i32 [[IV1_3_1]], 60
+// CHECK-NEXT: [[MUL_2:%.+]] = mul i32 [[DIV_1]], 60
+// CHECK-NEXT: [[ADD_3:%.+]] = sub i32 [[IV1_3]], [[MUL_2]]
+
 // CHECK: [[IV1_4:%.+]] = load i32, i32* [[OMP_IV]]
-// CHECK-NEXT: [[CALC_L_1:%.+]] = urem i32 [[IV1_4]], 5
-// CHECK-NEXT: [[CALC_L_1_MUL1:%.+]] = mul i32 [[CALC_L_1]], 1
-// CHECK-NEXT: [[CALC_L_2:%.+]] = add i32 4, [[CALC_L_1_MUL1]]
+// CHECK: [[IV1_4_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_2:%.+]] = udiv i32 [[IV1_4_1]], 60
+// CHECK-NEXT: [[MUL_3:%.+]] = mul i32 [[DIV_2]], 60
+// CHECK-NEXT: [[SUB_6:%.+]] = sub i32 [[IV1_4]], [[MUL_3]]
+// CHECK-NEXT: [[DIV_3:%.+]] = udiv i32 [[SUB_6]], 20
+// CHECK-NEXT: [[MUL_4:%.+]] = mul i32 [[DIV_3]], 20
+// CHECK-NEXT: [[SUB_7:%.+]] = sub i32 [[ADD_3]], [[MUL_4]]
+// CHECK-NEXT: [[DIV_4:%.+]] = udiv i32 [[SUB_7]], 5
+// CHECK-NEXT: [[MUL_5:%.+]] = mul i32 [[DIV_4]], 1
+// CHECK-NEXT: [[ADD_6:%.+]] = add i32 3, [[MUL_5]]
+// CHECK-NEXT: store i32 [[ADD_6]], i32* [[LC_K:.+]]
+
+// CHECK: [[IV1_5:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_5_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_5:%.+]] = udiv i32 [[IV1_5_1]], 60
+// CHECK-NEXT: [[MUL_6:%.+]] = mul i32 [[DIV_5]], 60
+// CHECK-NEXT: [[ADD_7:%.+]] = sub i32 [[IV1_5]], [[MUL_6]]
+
+// CHECK: [[IV1_6:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_6_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_6:%.+]] = udiv i32 [[IV1_6_1]], 60
+// CHECK-NEXT: [[MUL_7:%.+]] = mul i32 [[DIV_6]], 60
+// CHECK-NEXT: [[SUB_10:%.+]] = sub i32 [[IV1_6]], [[MUL_7]]
+// CHECK-NEXT: [[DIV_7:%.+]] = udiv i32 [[SUB_10]], 20
+// CHECK-NEXT: [[MUL_8:%.+]] = mul i32 [[DIV_7]], 20
+// CHECK-NEXT: [[ADD_9:%.+]] = sub i32 [[ADD_7]], [[MUL_8]]
+
+// CHECK: [[IV1_7:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_7_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_8:%.+]] = udiv i32 [[IV1_7_1]], 60
+// CHECK-NEXT: [[MUL_9:%.+]] = mul i32 [[DIV_8]], 60
+// CHECK-NEXT: [[ADD_10:%.+]] = sub i32 [[IV1_7]], [[MUL_9]]
+
+// CHECK: [[IV1_8:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_8_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_3:%.+]] = udiv i32 [[IV1_8_1]], 60
+// CHECK-NEXT: [[MUL_4:%.+]] = mul i32 [[DIV_3]], 60
+// CHECK-NEXT: [[SUB_7:%.+]] = sub i32 [[IV1_8]], [[MUL_4]]
+// CHECK-NEXT: [[DIV_4:%.+]] = udiv i32 [[SUB_7]], 20
+// CHECK-NEXT: [[MUL_5:%.+]] = mul i32 [[DIV_4]], 20
+// CHECK-NEXT: [[SUB_8:%.+]] = sub i32 [[ADD_10]], [[MUL_5]]
+// CHECK-NEXT: [[DIV_5:%.+]] = udiv i32 [[SUB_8]], 5
+// CHECK-NEXT: [[MUL_6:%.+]] = mul i32 [[DIV_5]], 5
+// CHECK-NEXT: [[SUB_9:%.+]] = sub i32 [[ADD_9]], [[MUL_6]]
+// CHECK-NEXT: [[MUL_6:%.+]] = mul i32 [[SUB_9]], 1
+// CHECK-NEXT: [[CALC_L_2:%.+]] = add i32 4, [[MUL_6]]
 // CHECK-NEXT: [[CALC_L_3:%.+]] = trunc i32 [[CALC_L_2]] to i16
 // CHECK-NEXT: store i16 [[CALC_L_3]], i16* [[LC_L:.+]]
 // ... loop body ...
 // End of body: store into a[i]:
 // CHECK: store float [[RESULT:%.+]], float* [[RESULT_ADDR:%.+]]
     float res = b[j] * c[k];
     a[i] = res * d[l];
 // CHECK: [[IV2:%.+]] = load i32, i32* [[OMP_IV]]

test/OpenMP/parallel_for_simd_codegen.cpp

 // CHECK: [[COLL1_BODY]]:
 // Start of body: calculate i from index:
 // CHECK: [[IV1:%.+]] = load i32, i32* [[OMP_IV]]
 // Calculation of the loop counters values.
 // CHECK: [[CALC_I_1:%.+]] = udiv i32 [[IV1]], 60
 // CHECK-NEXT: [[CALC_I_1_MUL1:%.+]] = mul i32 [[CALC_I_1]], 1
 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 1, [[CALC_I_1_MUL1]]
 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
+
 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]
-// CHECK-NEXT: [[CALC_J_1:%.+]] = udiv i32 [[IV1_2]], 20
-// CHECK-NEXT: [[CALC_J_2:%.+]] = urem i32 [[CALC_J_1]], 3
+// CHECK: [[IV1_2_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[CALC_J_1:%.+]] = udiv i32 [[IV1_2_1]], 60
+// CHECK-NEXT: [[MUL_1:%.+]] = mul i32 [[CALC_J_1]], 60
+// CHECK-NEXT: [[SUB_3:%.+]] = sub i32 [[IV1_2]], [[MUL_1]]
+// CHECK-NEXT: [[CALC_J_2:%.+]] = udiv i32 [[SUB_3]], 20
 // CHECK-NEXT: [[CALC_J_2_MUL1:%.+]] = mul i32 [[CALC_J_2]], 1
 // CHECK-NEXT: [[CALC_J_3:%.+]] = add i32 2, [[CALC_J_2_MUL1]]
 // CHECK-NEXT: store i32 [[CALC_J_3]], i32* [[LC_J:.+]]
+
 // CHECK: [[IV1_3:%.+]] = load i32, i32* [[OMP_IV]]
-// CHECK-NEXT: [[CALC_K_1:%.+]] = udiv i32 [[IV1_3]], 5
-// CHECK-NEXT: [[CALC_K_2:%.+]] = urem i32 [[CALC_K_1]], 4
-// CHECK-NEXT: [[CALC_K_2_MUL1:%.+]] = mul i32 [[CALC_K_2]], 1
-// CHECK-NEXT: [[CALC_K_3:%.+]] = add i32 3, [[CALC_K_2_MUL1]]
-// CHECK-NEXT: store i32 [[CALC_K_3]], i32* [[LC_K:.+]]
+// CHECK: [[IV1_3_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_1:%.+]] = udiv i32 [[IV1_3_1]], 60
+// CHECK-NEXT: [[MUL_2:%.+]] = mul i32 [[DIV_1]], 60
+// CHECK-NEXT: [[ADD_3:%.+]] = sub i32 [[IV1_3]], [[MUL_2]]
+
 // CHECK: [[IV1_4:%.+]] = load i32, i32* [[OMP_IV]]
-// CHECK-NEXT: [[CALC_L_1:%.+]] = urem i32 [[IV1_4]], 5
-// CHECK-NEXT: [[CALC_L_1_MUL1:%.+]] = mul i32 [[CALC_L_1]], 1
-// CHECK-NEXT: [[CALC_L_2:%.+]] = add i32 4, [[CALC_L_1_MUL1]]
+// CHECK: [[IV1_4_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_2:%.+]] = udiv i32 [[IV1_4_1]], 60
+// CHECK-NEXT: [[MUL_3:%.+]] = mul i32 [[DIV_2]], 60
+// CHECK-NEXT: [[SUB_6:%.+]] = sub i32 [[IV1_4]], [[MUL_3]]
+// CHECK-NEXT: [[DIV_3:%.+]] = udiv i32 [[SUB_6]], 20
+// CHECK-NEXT: [[MUL_4:%.+]] = mul i32 [[DIV_3]], 20
+// CHECK-NEXT: [[SUB_7:%.+]] = sub i32 [[ADD_3]], [[MUL_4]]
+// CHECK-NEXT: [[DIV_4:%.+]] = udiv i32 [[SUB_7]], 5
+// CHECK-NEXT: [[MUL_5:%.+]] = mul i32 [[DIV_4]], 1
+// CHECK-NEXT: [[ADD_6:%.+]] = add i32 3, [[MUL_5]]
+// CHECK-NEXT: store i32 [[ADD_6]], i32* [[LC_K:.+]]
+
+// CHECK: [[IV1_5:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_5_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_5:%.+]] = udiv i32 [[IV1_5_1]], 60
+// CHECK-NEXT: [[MUL_6:%.+]] = mul i32 [[DIV_5]], 60
+// CHECK-NEXT: [[ADD_7:%.+]] = sub i32 [[IV1_5]], [[MUL_6]]
+
+// CHECK: [[IV1_6:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_6_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_6:%.+]] = udiv i32 [[IV1_6_1]], 60
+// CHECK-NEXT: [[MUL_7:%.+]] = mul i32 [[DIV_6]], 60
+// CHECK-NEXT: [[SUB_10:%.+]] = sub i32 [[IV1_6]], [[MUL_7]]
+// CHECK-NEXT: [[DIV_7:%.+]] = udiv i32 [[SUB_10]], 20
+// CHECK-NEXT: [[MUL_8:%.+]] = mul i32 [[DIV_7]], 20
+// CHECK-NEXT: [[ADD_9:%.+]] = sub i32 [[ADD_7]], [[MUL_8]]
+
+// CHECK: [[IV1_7:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_7_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_8:%.+]] = udiv i32 [[IV1_7_1]], 60
+// CHECK-NEXT: [[MUL_9:%.+]] = mul i32 [[DIV_8]], 60
+// CHECK-NEXT: [[ADD_10:%.+]] = sub i32 [[IV1_7]], [[MUL_9]]
+
+// CHECK: [[IV1_8:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_8_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_3:%.+]] = udiv i32 [[IV1_8_1]], 60
+// CHECK-NEXT: [[MUL_4:%.+]] = mul i32 [[DIV_3]], 60
+// CHECK-NEXT: [[SUB_7:%.+]] = sub i32 [[IV1_8]], [[MUL_4]]
+// CHECK-NEXT: [[DIV_4:%.+]] = udiv i32 [[SUB_7]], 20
+// CHECK-NEXT: [[MUL_5:%.+]] = mul i32 [[DIV_4]], 20
+// CHECK-NEXT: [[SUB_8:%.+]] = sub i32 [[ADD_10]], [[MUL_5]]
+// CHECK-NEXT: [[DIV_5:%.+]] = udiv i32 [[SUB_8]], 5
+// CHECK-NEXT: [[MUL_6:%.+]] = mul i32 [[DIV_5]], 5
+// CHECK-NEXT: [[SUB_9:%.+]] = sub i32 [[ADD_9]], [[MUL_6]]
+// CHECK-NEXT: [[MUL_6:%.+]] = mul i32 [[SUB_9]], 1
+// CHECK-NEXT: [[CALC_L_2:%.+]] = add i32 4, [[MUL_6]]
 // CHECK-NEXT: [[CALC_L_3:%.+]] = trunc i32 [[CALC_L_2]] to i16
 // CHECK-NEXT: store i16 [[CALC_L_3]], i16* [[LC_L:.+]]
 // ... loop body ...
 // End of body: store into a[i]:
 // CHECK: store float [[RESULT:%.+]], float* [[RESULT_ADDR:%.+]]
     float res = b[j] * c[k];
     a[i] = res * d[l];
 // CHECK: [[IV2:%.+]] = load i32, i32* [[OMP_IV]]

 // Loop counters i and j updates:
 // CHECK: [[IV1:%.+]] = load i64, i64* [[T1_OMP_IV]]
 // CHECK-NEXT: [[I_1:%.+]] = sdiv i64 [[IV1]], 4
 // CHECK-NEXT: [[I_1_MUL1:%.+]] = mul nsw i64 [[I_1]], 1
 // CHECK-NEXT: [[I_1_ADD0:%.+]] = add nsw i64 0, [[I_1_MUL1]]
 // CHECK-NEXT: [[I_2:%.+]] = trunc i64 [[I_1_ADD0]] to i32
 // CHECK-NEXT: store i32 [[I_2]], i32*
 // CHECK: [[IV2:%.+]] = load i64, i64* [[T1_OMP_IV]]
-// CHECK-NEXT: [[J_1:%.+]] = srem i64 [[IV2]], 4
+// CHECK: [[IV2_1:%.+]] = load i64, i64* [[T1_OMP_IV]]
+// CHECK-NEXT: [[DIV_1:%.+]] = sdiv i64 [[IV2_1]], 4
+// CHECK-NEXT: [[MUL_1:%.+]] = mul nsw i64 [[DIV_1]], 4
+// CHECK-NEXT: [[J_1:%.+]] = sub nsw i64 [[IV2]], [[MUL_1]]
 // CHECK-NEXT: [[J_2:%.+]] = mul nsw i64 [[J_1]], 2
 // CHECK-NEXT: [[J_2_ADD0:%.+]] = add nsw i64 0, [[J_2]]
 // CHECK-NEXT: store i64 [[J_2_ADD0]], i64*
 // simd.for.inc:
 // CHECK: [[IV3:%.+]] = load i64, i64* [[T1_OMP_IV]]
 // CHECK-NEXT: [[INC:%.+]] = add nsw i64 [[IV3]], 1
 // CHECK-NEXT: store i64 [[INC]], i64*
 // CHECK-NEXT: br label {{%.+}}

test/OpenMP/simd_codegen.cpp

 // Loop counters i and j updates:
 // CHECK: [[IV1:%.+]] = load i64, i64* [[T1_OMP_IV]]{{.*}}!llvm.access.group
 // CHECK-NEXT: [[I_1:%.+]] = sdiv i64 [[IV1]], 4
 // CHECK-NEXT: [[I_1_MUL1:%.+]] = mul nsw i64 [[I_1]], 1
 // CHECK-NEXT: [[I_1_ADD0:%.+]] = add nsw i64 0, [[I_1_MUL1]]
 // CHECK-NEXT: [[I_2:%.+]] = trunc i64 [[I_1_ADD0]] to i32
 // CHECK-NEXT: store i32 [[I_2]], i32* {{%.+}}{{.*}}!llvm.access.group
 // CHECK: [[IV2:%.+]] = load i64, i64* [[T1_OMP_IV]]{{.*}}!llvm.access.group
-// CHECK-NEXT: [[J_1:%.+]] = srem i64 [[IV2]], 4
-// CHECK-NEXT: [[J_2:%.+]] = mul nsw i64 [[J_1]], 2
+// CHECK: [[IV2_1:%.+]] = load i64, i64* [[T1_OMP_IV]]{{.*}}!llvm.access.group
+// CHECK-NEXT: [[J_1_DIV1:%.+]] = sdiv i64 [[IV2_1]], 4
+// CHECK-NEXT: [[J_1_MUL1:%.+]] = mul nsw i64 [[J_1_DIV1]], 4
+// CHECK-NEXT: [[J_1_SUB0:%.+]] = sub nsw i64 [[IV2]], [[J_1_MUL1]]
+// CHECK-NEXT: [[J_2:%.+]] = mul nsw i64 [[J_1_SUB0]], 2
 // CHECK-NEXT: [[J_2_ADD0:%.+]] = add nsw i64 0, [[J_2]]
 // CHECK-NEXT: store i64 [[J_2_ADD0]], i64* {{%.+}}{{.*}}!llvm.access.group
 // simd.for.inc:
 // CHECK: [[IV3:%.+]] = load i64, i64* [[T1_OMP_IV]]{{.*}}!llvm.access.group
 // CHECK-NEXT: [[INC:%.+]] = add nsw i64 [[IV3]], 1
 // CHECK-NEXT: store i64 [[INC]], i64* [[T1_OMP_IV]]{{.*}}!llvm.access.group
 // CHECK-NEXT: br label {{%.+}}
 // CHECK: [[T1_END]]:

 // CHECK: [[COLL1_BODY]]:
 // Start of body: calculate i from index:
 // CHECK: [[IV1:%.+]] = load i32, i32* [[OMP_IV]]{{.+}}!llvm.access.group
 // Calculation of the loop counters values.
 // CHECK: [[CALC_I_1:%.+]] = udiv i32 [[IV1]], 60
 // CHECK-NEXT: [[CALC_I_1_MUL1:%.+]] = mul i32 [[CALC_I_1]], 1
 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 1, [[CALC_I_1_MUL1]]
 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
+
 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.+}}!llvm.access.group
-// CHECK-NEXT: [[CALC_J_1:%.+]] = udiv i32 [[IV1_2]], 20
-// CHECK-NEXT: [[CALC_J_2:%.+]] = urem i32 [[CALC_J_1]], 3
+// CHECK: [[IV1_2_1:%.+]] = load i32, i32* [[OMP_IV]]{{.+}}!llvm.access.group
+// CHECK-NEXT: [[CALC_J_1:%.+]] = udiv i32 [[IV1_2_1]], 60
+// CHECK-NEXT: [[MUL_1:%.+]] = mul i32 [[CALC_J_1]], 60
+// CHECK-NEXT: [[SUB_3:%.+]] = sub i32 [[IV1_2]], [[MUL_1]]
+// CHECK-NEXT: [[CALC_J_2:%.+]] = udiv i32 [[SUB_3]], 20
 // CHECK-NEXT: [[CALC_J_2_MUL1:%.+]] = mul i32 [[CALC_J_2]], 1
 // CHECK-NEXT: [[CALC_J_3:%.+]] = add i32 2, [[CALC_J_2_MUL1]]
 // CHECK-NEXT: store i32 [[CALC_J_3]], i32* [[LC_J:.+]]
+
 // CHECK: [[IV1_3:%.+]] = load i32, i32* [[OMP_IV]]{{.+}}!llvm.access.group
-// CHECK-NEXT: [[CALC_K_1:%.+]] = udiv i32 [[IV1_3]], 5
-// CHECK-NEXT: [[CALC_K_2:%.+]] = urem i32 [[CALC_K_1]], 4
-// CHECK-NEXT: [[CALC_K_2_MUL1:%.+]] = mul i32 [[CALC_K_2]], 1
-// CHECK-NEXT: [[CALC_K_3:%.+]] = add i32 3, [[CALC_K_2_MUL1]]
-// CHECK-NEXT: store i32 [[CALC_K_3]], i32* [[LC_K:.+]]
-// CHECK: [[IV1_4:%.+]] = load i32, i32* [[OMP_IV]]{{.+}}!llvm.access.group
-// CHECK-NEXT: [[CALC_L_1:%.+]] = urem i32 [[IV1_4]], 5
-// CHECK-NEXT: [[CALC_L_1_MUL1:%.+]] = mul i32 [[CALC_L_1]], 1
-// CHECK-NEXT: [[CALC_L_2:%.+]] = add i32 4, [[CALC_L_1_MUL1]]
+// CHECK: [[IV1_3_1:%.+]] = load i32, i32* [[OMP_IV]]{{.+}}!llvm.access.group
+// CHECK-NEXT: [[DIV_1:%.+]] = udiv i32 [[IV1_3_1]], 60
+// CHECK-NEXT: [[MUL_2:%.+]] = mul i32 [[DIV_1]], 60
+// CHECK-NEXT: [[ADD_3:%.+]] = sub i32 [[IV1_3]], [[MUL_2]]
+
+// CHECK: [[IV1_4:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_4_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_2:%.+]] = udiv i32 [[IV1_4_1]], 60
+// CHECK-NEXT: [[MUL_3:%.+]] = mul i32 [[DIV_2]], 60
+// CHECK-NEXT: [[SUB_6:%.+]] = sub i32 [[IV1_4]], [[MUL_3]]
+// CHECK-NEXT: [[DIV_3:%.+]] = udiv i32 [[SUB_6]], 20
+// CHECK-NEXT: [[MUL_4:%.+]] = mul i32 [[DIV_3]], 20
+// CHECK-NEXT: [[ADD_5:%.+]] = sub i32 [[ADD_3]], [[MUL_4]]
+// CHECK-NEXT: [[DIV_4:%.+]] = udiv i32 [[ADD_5]], 5
+// CHECK-NEXT: [[MUL_5:%.+]] = mul i32 [[DIV_4]], 1
+// CHECK-NEXT: [[ADD_6:%.+]] = add i32 3, [[MUL_5]]
+// CHECK-NEXT: store i32 [[ADD_6]], i32* [[LC_K:.+]]
+
+// CHECK: [[IV1_5:%.+]] = load i32, i32* [[OMP_IV]]{{.+}}!llvm.access.group
+// CHECK: [[IV1_5_1:%.+]] = load i32, i32* [[OMP_IV]]{{.+}}!llvm.access.group
+// CHECK-NEXT: [[DIV_5:%.+]] = udiv i32 [[IV1_5_1]], 60
+// CHECK-NEXT: [[MUL_6:%.+]] = mul i32 [[DIV_5]], 60
+// CHECK-NEXT: [[SUB_7:%.+]] = sub i32 [[IV1_5]], [[MUL_6]]
+
+// CHECK: [[IV1_6:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_6_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_6:%.+]] = udiv i32 [[IV1_6_1]], 60
+// CHECK-NEXT: [[MUL_7:%.+]] = mul i32 [[DIV_6]], 60
+// CHECK-NEXT: [[SUB_10:%.+]] = sub i32 [[IV1_6]], [[MUL_7]]
+// CHECK-NEXT: [[DIV_7:%.+]] = udiv i32 [[SUB_10]], 20
+// CHECK-NEXT: [[MUL_8:%.+]] = mul i32 [[DIV_7]], 20
+// CHECK-NEXT: [[SUB_11:%.+]] = sub i32 [[SUB_7]], [[MUL_8]]
+
+// CHECK: [[IV1_7:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_7_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_8:%.+]] = udiv i32 [[IV1_7_1]], 60
+// CHECK-NEXT: [[MUL_9:%.+]] = mul i32 [[DIV_8]], 60
+// CHECK-NEXT: [[SUB_12:%.+]] = sub i32 [[IV1_7]], [[MUL_9]]
+
+// CHECK: [[IV1_8:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK: [[IV1_8_1:%.+]] = load i32, i32* [[OMP_IV]]
+// CHECK-NEXT: [[DIV_3:%.+]] = udiv i32 [[IV1_8_1]], 60
+// CHECK-NEXT: [[MUL_4:%.+]] = mul i32 [[DIV_3]], 60
+// CHECK-NEXT: [[SUB_7:%.+]] = sub i32 [[IV1_8]], [[MUL_4]]
+// CHECK-NEXT: [[DIV_4:%.+]] = udiv i32 [[SUB_7]], 20
+// CHECK-NEXT: [[MUL_5:%.+]] = mul i32 [[DIV_4]], 20
+// CHECK-NEXT: [[SUB_8:%.+]] = sub i32 [[SUB_12]], [[MUL_5]]
+// CHECK-NEXT: [[DIV_5:%.+]] = udiv i32 [[SUB_8]], 5
+// CHECK-NEXT: [[MUL_6:%.+]] = mul i32 [[DIV_5]], 5
+// CHECK-NEXT: [[SUB_9:%.+]] = sub i32 [[SUB_11]], [[MUL_6]]
+// CHECK-NEXT: [[MUL_6:%.+]] = mul i32 [[SUB_9]], 1
+// CHECK-NEXT: [[CALC_L_2:%.+]] = add i32 4, [[MUL_6]]
 // CHECK-NEXT: [[CALC_L_3:%.+]] = trunc i32 [[CALC_L_2]] to i16
 // CHECK-NEXT: store i16 [[CALC_L_3]], i16* [[LC_L:.+]]
 // ... loop body ...
 // End of body: store into a[i]:
 // CHECK: store float [[RESULT:%.+]], float* [[RESULT_ADDR:%.+]]{{.+}}!llvm.access.group
     float res = b[j] * c[k];
     a[i] = res * d[l];
 // CHECK: [[IV2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}!llvm.access.group