120 lines · plain
1; RUN: opt %loadNPMPolly -polly-pattern-matching-based-opts=true -polly-target-throughput-vector-fma=1 -polly-target-latency-vector-fma=8 -polly-target-1st-cache-level-associativity=8 -polly-target-2nd-cache-level-associativity=8 -polly-target-1st-cache-level-size=32768 -polly-target-vector-register-bitwidth=256 -polly-target-2nd-cache-level-size=262144 '-passes=polly-custom<opt-isl;ast>' -polly-print-ast -disable-output < %s | FileCheck %s2;3; /* C := A * B + C */4; /* Elements of the matrices B, C have the double type. */5; /* Elements of the matrix A have the float type. */6; /* The type size of elements of the matrix multiplication operands is used7; to determine the parameters of the code produced by the optimization8; of the matrix multiplication (e.g. bounds of the loops of the loop9; nest, the innermost loop body). This test checks the form of10; the generated loop nest. See getMicroKernelParams and11; getMacroKernelParams from lib/Transform/ScheduleOptimizer.cpp12; for details. */13; for (i = 0; i < _PB_NI; i++)14; for (j = 0; j < _PB_NJ; j++)15; for (k = 0; k < _PB_NK; ++k)16; C[i][j] += A[i][k] * B[k][j];17;18; CHECK: // 1st level tiling - Tiles19; CHECK-NEXT: for (int c1 = 0; c1 <= 3; c1 += 1) {20; CHECK-NEXT: for (int c3 = 0; c3 <= 1023; c3 += 1)21; CHECK-NEXT: for (int c4 = 256 * c1; c4 <= 256 * c1 + 255; c4 += 1)22; CHECK-NEXT: CopyStmt_0(0, c3, c4);23; CHECK-NEXT: for (int c2 = 0; c2 <= 10; c2 += 1) {24; CHECK-NEXT: for (int c6 = 96 * c2; c6 <= min(1023, 96 * c2 + 95); c6 += 1)25; CHECK-NEXT: for (int c7 = 256 * c1; c7 <= 256 * c1 + 255; c7 += 1)26; CHECK-NEXT: CopyStmt_1(0, c1, c2, c6, c7);27; CHECK-NEXT: // 1st level tiling - Points28; CHECK-NEXT: // Register tiling - Tiles29; CHECK-NEXT: for (int c3 = 0; c3 <= 127; c3 += 1)30; CHECK-NEXT: for (int c4 = 0; c4 <= min(23, -24 * c2 + 255); c4 += 1)31; CHECK-NEXT: for (int c5 = 0; c5 <= 255; c5 += 1) {32; CHECK-NEXT: // Loop Vectorizer Disabled33; CHECK-NEXT: // Register tiling - Points34; CHECK-NEXT: {35; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4, 8 * c3, 256 * c1 + c5);36; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4, 8 * c3 + 1, 256 * c1 + c5);37; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4, 8 * c3 + 2, 256 * c1 + c5);38; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4, 8 * c3 + 3, 256 * c1 + c5);39; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4, 8 * c3 + 4, 256 * c1 + c5);40; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4, 8 * c3 + 5, 256 * c1 + c5);41; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4, 8 * c3 + 6, 256 * c1 + c5);42; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4, 8 * c3 + 7, 256 * c1 + c5);43; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 1, 8 * c3, 256 * c1 + c5);44; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 1, 8 * c3 + 1, 256 * c1 + c5);45; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 1, 8 * c3 + 2, 256 * c1 + c5);46; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 1, 8 * c3 + 3, 256 * c1 + c5);47; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 1, 8 * c3 + 4, 256 * c1 + c5);48; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 1, 8 * c3 + 5, 256 * c1 + c5);49; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 1, 8 * c3 + 6, 256 * c1 + c5);50; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 1, 8 * c3 + 7, 256 * c1 + c5);51; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 2, 8 * c3, 256 * c1 + c5);52; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 2, 8 * c3 + 1, 256 * c1 + c5);53; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 2, 8 * c3 + 2, 256 * c1 + c5);54; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 2, 8 * c3 + 3, 256 * c1 + c5);55; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 2, 8 * c3 + 4, 256 * c1 + c5);56; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 2, 8 * c3 + 5, 256 * c1 + c5);57; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 2, 8 * c3 + 6, 256 * c1 + c5);58; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 2, 8 * c3 + 7, 256 * c1 + c5);59; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 3, 8 * c3, 256 * c1 + c5);60; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 3, 8 * c3 + 1, 256 * c1 + c5);61; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 3, 8 * c3 + 2, 256 * c1 + c5);62; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 3, 8 * c3 + 3, 256 * c1 + c5);63; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 3, 8 * c3 + 4, 256 * c1 + c5);64; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 3, 8 * c3 + 5, 256 * c1 + c5);65; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 3, 8 * c3 + 6, 256 * c1 + c5);66; CHECK-NEXT: Stmt_for_body6(96 * c2 + 4 * c4 + 3, 8 * c3 + 7, 256 * c1 + c5);67; CHECK-NEXT: }68; CHECK-NEXT: }69; CHECK-NEXT: }70; CHECK-NEXT: }71;72target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"73target triple = "x86_64-unknown-unknown"74 75; Function Attrs: noinline nounwind uwtable76define internal void @kernel_gemm(i32 %ni, i32 %nj, i32 %nk, double %alpha, double %beta, ptr %C, ptr %A, ptr %B) #0 {77entry:78 br label %entry.split79 80entry.split: ; preds = %entry81 br label %for.cond1.preheader82 83for.cond1.preheader: ; preds = %for.inc20, %entry.split84 %indvars.iv41 = phi i64 [ 0, %entry.split ], [ %indvars.iv.next42, %for.inc20 ]85 br label %for.cond4.preheader86 87for.cond4.preheader: ; preds = %for.inc17, %for.cond1.preheader88 %indvars.iv38 = phi i64 [ 0, %for.cond1.preheader ], [ %indvars.iv.next39, %for.inc17 ]89 br label %for.body690 91for.body6: ; preds = %for.body6, %for.cond4.preheader92 %indvars.iv = phi i64 [ 0, %for.cond4.preheader ], [ %indvars.iv.next, %for.body6 ]93 %arrayidx8 = getelementptr inbounds [1024 x float], ptr %A, i64 %indvars.iv41, i64 %indvars.iv94 %tmp = load float, ptr %arrayidx8, align 495 %conv = fpext float %tmp to double96 %arrayidx12 = getelementptr inbounds [1024 x double], ptr %B, i64 %indvars.iv, i64 %indvars.iv3897 %tmp1 = load double, ptr %arrayidx12, align 898 %mul = fmul double %conv, %tmp199 %arrayidx16 = getelementptr inbounds [1024 x double], ptr %C, i64 %indvars.iv41, i64 %indvars.iv38100 %tmp2 = load double, ptr %arrayidx16, align 8101 %add = fadd double %tmp2, %mul102 store double %add, ptr %arrayidx16, align 8103 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1104 %exitcond = icmp ne i64 %indvars.iv.next, 1024105 br i1 %exitcond, label %for.body6, label %for.inc17106 107for.inc17: ; preds = %for.body6108 %indvars.iv.next39 = add nuw nsw i64 %indvars.iv38, 1109 %exitcond40 = icmp ne i64 %indvars.iv.next39, 1024110 br i1 %exitcond40, label %for.cond4.preheader, label %for.inc20111 112for.inc20: ; preds = %for.inc17113 %indvars.iv.next42 = add nuw nsw i64 %indvars.iv41, 1114 %exitcond43 = icmp ne i64 %indvars.iv.next42, 1024115 br i1 %exitcond43, label %for.cond1.preheader, label %for.end22116 117for.end22: ; preds = %for.inc20118 ret void119}120