94 lines · plain
1; RUN: opt -S %loadNPMPolly -polly-pattern-matching-based-opts=false -polly-vectorizer=stripmine '-passes=polly-custom<opt-isl;ast>' -polly-print-ast -disable-output < %s | FileCheck %s2; CHECK: // 1st level tiling - Tiles3; CHECK-NEXT: #pragma known-parallel4; CHECK-NEXT: for (int c0 = 0; c0 <= floord(ni - 1, 32); c0 += 1)5; CHECK-NEXT: for (int c1 = 0; c1 <= floord(nj - 1, 32); c1 += 1)6; CHECK-NEXT: #pragma minimal dependence distance: 17; CHECK-NEXT: for (int c2 = 0; c2 <= floord(nk - 1, 32); c2 += 1) {8; CHECK-NEXT: // 1st level tiling - Points9; CHECK-NEXT: for (int c3 = 0; c3 <= min(31, ni - 32 * c0 - 1); c3 += 1) {10; CHECK-NEXT: for (int c4 = 0; c4 <= min(7, -8 * c1 + nj / 4 - 1); c4 += 1)11; CHECK-NEXT: #pragma minimal dependence distance: 112; CHECK-NEXT: for (int c5 = 0; c5 <= min(31, nk - 32 * c2 - 1); c5 += 1) {13; CHECK-NEXT: // SIMD14; CHECK-NEXT: #pragma simd15; CHECK-NEXT: for (int c6 = 0; c6 <= 3; c6 += 1)16; CHECK-NEXT: Stmt_for_body_6(32 * c0 + c3, 32 * c1 + 4 * c4 + c6, 32 * c2 + c5);17; CHECK-NEXT: }18; CHECK-NEXT: if (32 * c1 + 31 >= nj)19; CHECK-NEXT: #pragma minimal dependence distance: 120; CHECK-NEXT: for (int c5 = 0; c5 <= min(31, nk - 32 * c2 - 1); c5 += 1) {21; CHECK-NEXT: // SIMD22; CHECK-NEXT: #pragma simd23; CHECK-NEXT: for (int c6 = 0; c6 < nj % 4; c6 += 1)24; CHECK-NEXT: Stmt_for_body_6(32 * c0 + c3, -(nj % 4) + nj + c6, 32 * c2 + c5);25; CHECK-NEXT: }26; CHECK-NEXT: }27; CHECK-NEXT: }28 29; Function Attrs: nounwind uwtable30define void @kernel_gemm(i32 %ni, i32 %nj, i32 %nk, double %alpha, double %beta, ptr %C, ptr %A, ptr %B) #0 {31entry:32 %cmp.27 = icmp sgt i32 %ni, 033 br i1 %cmp.27, label %for.cond.1.preheader.lr.ph, label %for.end.2234 35for.cond.1.preheader.lr.ph: ; preds = %entry36 br label %for.cond.1.preheader37 38for.cond.1.preheader: ; preds = %for.cond.1.preheader.lr.ph, %for.inc.2039 %indvars.iv33 = phi i64 [ 0, %for.cond.1.preheader.lr.ph ], [ %indvars.iv.next34, %for.inc.20 ]40 %cmp2.25 = icmp sgt i32 %nj, 041 br i1 %cmp2.25, label %for.cond.4.preheader.lr.ph, label %for.inc.2042 43for.cond.4.preheader.lr.ph: ; preds = %for.cond.1.preheader44 br label %for.cond.4.preheader45 46for.cond.4.preheader: ; preds = %for.cond.4.preheader.lr.ph, %for.inc.1747 %indvars.iv29 = phi i64 [ 0, %for.cond.4.preheader.lr.ph ], [ %indvars.iv.next30, %for.inc.17 ]48 %cmp5.23 = icmp sgt i32 %nk, 049 br i1 %cmp5.23, label %for.body.6.lr.ph, label %for.inc.1750 51for.body.6.lr.ph: ; preds = %for.cond.4.preheader52 br label %for.body.653 54for.body.6: ; preds = %for.body.6.lr.ph, %for.body.655 %indvars.iv = phi i64 [ 0, %for.body.6.lr.ph ], [ %indvars.iv.next, %for.body.6 ]56 %arrayidx8 = getelementptr inbounds [1024 x double], ptr %A, i64 %indvars.iv33, i64 %indvars.iv57 %0 = load double, ptr %arrayidx8, align 858 %arrayidx12 = getelementptr inbounds [1024 x double], ptr %B, i64 %indvars.iv, i64 %indvars.iv2959 %1 = load double, ptr %arrayidx12, align 860 %mul = fmul double %0, %161 %arrayidx16 = getelementptr inbounds [1024 x double], ptr %C, i64 %indvars.iv33, i64 %indvars.iv2962 %2 = load double, ptr %arrayidx16, align 863 %add = fadd double %2, %mul64 store double %add, ptr %arrayidx16, align 865 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 166 %lftr.wideiv = trunc i64 %indvars.iv.next to i3267 %exitcond = icmp ne i32 %lftr.wideiv, %nk68 br i1 %exitcond, label %for.body.6, label %for.cond.4.for.inc.17_crit_edge69 70for.cond.4.for.inc.17_crit_edge: ; preds = %for.body.671 br label %for.inc.1772 73for.inc.17: ; preds = %for.cond.4.for.inc.17_crit_edge, %for.cond.4.preheader74 %indvars.iv.next30 = add nuw nsw i64 %indvars.iv29, 175 %lftr.wideiv31 = trunc i64 %indvars.iv.next30 to i3276 %exitcond32 = icmp ne i32 %lftr.wideiv31, %nj77 br i1 %exitcond32, label %for.cond.4.preheader, label %for.cond.1.for.inc.20_crit_edge78 79for.cond.1.for.inc.20_crit_edge: ; preds = %for.inc.1780 br label %for.inc.2081 82for.inc.20: ; preds = %for.cond.1.for.inc.20_crit_edge, %for.cond.1.preheader83 %indvars.iv.next34 = add nuw nsw i64 %indvars.iv33, 184 %lftr.wideiv35 = trunc i64 %indvars.iv.next34 to i3285 %exitcond36 = icmp ne i32 %lftr.wideiv35, %ni86 br i1 %exitcond36, label %for.cond.1.preheader, label %for.cond.for.end.22_crit_edge87 88for.cond.for.end.22_crit_edge: ; preds = %for.inc.2089 br label %for.end.2290 91for.end.22: ; preds = %for.cond.for.end.22_crit_edge, %entry92 ret void93}94