150 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 A, B, C have the float type. */5; /* The type size of elements of the matrix multiplication operands is used6; to determine the parameters of the code produced by the optimization7; of the matrix multiplication (e.g. bounds of the loops of the loop8; nest, the innermost loop body). This test checks the form of9; the generated loop nest. See getMicroKernelParams and10; getMacroKernelParams from lib/Transform/ScheduleOptimizer.cpp11; for details. */12; for (i = 0; i < _PB_NI; i++)13; for (j = 0; j < _PB_NJ; j++)14; for (k = 0; k < _PB_NK; ++k)15; C[i][j] += A[i][k] * B[k][j];16;17; CHECK: // 1st level tiling - Tiles18; CHECK-NEXT: for (int c1 = 0; c1 <= 2; c1 += 1) {19; CHECK-NEXT: for (int c3 = 0; c3 <= 1023; c3 += 1)20; CHECK-NEXT: for (int c4 = 384 * c1; c4 <= min(1023, 384 * c1 + 383); c4 += 1)21; CHECK-NEXT: CopyStmt_0(0, c3, c4);22; CHECK-NEXT: for (int c2 = 0; c2 <= 7; c2 += 1) {23; CHECK-NEXT: for (int c6 = 128 * c2; c6 <= 128 * c2 + 127; c6 += 1)24; CHECK-NEXT: for (int c7 = 384 * c1; c7 <= min(1023, 384 * c1 + 383); c7 += 1)25; CHECK-NEXT: CopyStmt_1(0, c1, c2, c6, c7);26; CHECK-NEXT: // 1st level tiling - Points27; CHECK-NEXT: // Register tiling - Tiles28; CHECK-NEXT: for (int c3 = 0; c3 <= 127; c3 += 1)29; CHECK-NEXT: for (int c4 = 0; c4 <= 15; c4 += 1)30; CHECK-NEXT: for (int c5 = 0; c5 <= min(383, -384 * c1 + 1023); c5 += 1) {31; CHECK-NEXT: // Loop Vectorizer Disabled32; CHECK-NEXT: // Register tiling - Points33; CHECK-NEXT: {34; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4, 8 * c3, 384 * c1 + c5);35; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4, 8 * c3 + 1, 384 * c1 + c5);36; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4, 8 * c3 + 2, 384 * c1 + c5);37; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4, 8 * c3 + 3, 384 * c1 + c5);38; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4, 8 * c3 + 4, 384 * c1 + c5);39; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4, 8 * c3 + 5, 384 * c1 + c5);40; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4, 8 * c3 + 6, 384 * c1 + c5);41; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4, 8 * c3 + 7, 384 * c1 + c5);42; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 1, 8 * c3, 384 * c1 + c5);43; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 1, 8 * c3 + 1, 384 * c1 + c5);44; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 1, 8 * c3 + 2, 384 * c1 + c5);45; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 1, 8 * c3 + 3, 384 * c1 + c5);46; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 1, 8 * c3 + 4, 384 * c1 + c5);47; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 1, 8 * c3 + 5, 384 * c1 + c5);48; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 1, 8 * c3 + 6, 384 * c1 + c5);49; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 1, 8 * c3 + 7, 384 * c1 + c5);50; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 2, 8 * c3, 384 * c1 + c5);51; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 2, 8 * c3 + 1, 384 * c1 + c5);52; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 2, 8 * c3 + 2, 384 * c1 + c5);53; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 2, 8 * c3 + 3, 384 * c1 + c5);54; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 2, 8 * c3 + 4, 384 * c1 + c5);55; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 2, 8 * c3 + 5, 384 * c1 + c5);56; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 2, 8 * c3 + 6, 384 * c1 + c5);57; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 2, 8 * c3 + 7, 384 * c1 + c5);58; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 3, 8 * c3, 384 * c1 + c5);59; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 3, 8 * c3 + 1, 384 * c1 + c5);60; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 3, 8 * c3 + 2, 384 * c1 + c5);61; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 3, 8 * c3 + 3, 384 * c1 + c5);62; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 3, 8 * c3 + 4, 384 * c1 + c5);63; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 3, 8 * c3 + 5, 384 * c1 + c5);64; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 3, 8 * c3 + 6, 384 * c1 + c5);65; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 3, 8 * c3 + 7, 384 * c1 + c5);66; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 4, 8 * c3, 384 * c1 + c5);67; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 4, 8 * c3 + 1, 384 * c1 + c5);68; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 4, 8 * c3 + 2, 384 * c1 + c5);69; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 4, 8 * c3 + 3, 384 * c1 + c5);70; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 4, 8 * c3 + 4, 384 * c1 + c5);71; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 4, 8 * c3 + 5, 384 * c1 + c5);72; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 4, 8 * c3 + 6, 384 * c1 + c5);73; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 4, 8 * c3 + 7, 384 * c1 + c5);74; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 5, 8 * c3, 384 * c1 + c5);75; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 5, 8 * c3 + 1, 384 * c1 + c5);76; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 5, 8 * c3 + 2, 384 * c1 + c5);77; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 5, 8 * c3 + 3, 384 * c1 + c5);78; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 5, 8 * c3 + 4, 384 * c1 + c5);79; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 5, 8 * c3 + 5, 384 * c1 + c5);80; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 5, 8 * c3 + 6, 384 * c1 + c5);81; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 5, 8 * c3 + 7, 384 * c1 + c5);82; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 6, 8 * c3, 384 * c1 + c5);83; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 6, 8 * c3 + 1, 384 * c1 + c5);84; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 6, 8 * c3 + 2, 384 * c1 + c5);85; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 6, 8 * c3 + 3, 384 * c1 + c5);86; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 6, 8 * c3 + 4, 384 * c1 + c5);87; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 6, 8 * c3 + 5, 384 * c1 + c5);88; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 6, 8 * c3 + 6, 384 * c1 + c5);89; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 6, 8 * c3 + 7, 384 * c1 + c5);90; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 7, 8 * c3, 384 * c1 + c5);91; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 7, 8 * c3 + 1, 384 * c1 + c5);92; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 7, 8 * c3 + 2, 384 * c1 + c5);93; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 7, 8 * c3 + 3, 384 * c1 + c5);94; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 7, 8 * c3 + 4, 384 * c1 + c5);95; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 7, 8 * c3 + 5, 384 * c1 + c5);96; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 7, 8 * c3 + 6, 384 * c1 + c5);97; CHECK-NEXT: Stmt_for_body6(128 * c2 + 8 * c4 + 7, 8 * c3 + 7, 384 * c1 + c5);98; CHECK-NEXT: }99; CHECK-NEXT: }100; CHECK-NEXT: }101; CHECK-NEXT: }102;103target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"104target triple = "x86_64-unknown-unknown"105 106; Function Attrs: noinline nounwind uwtable107define internal void @kernel_gemm(i32 %ni, i32 %nj, i32 %nk, float %alpha, float %beta, ptr %C, ptr %A, ptr %B) #0 {108entry:109 br label %entry.split110 111entry.split: ; preds = %entry112 br label %for.cond1.preheader113 114for.cond1.preheader: ; preds = %for.inc20, %entry.split115 %indvars.iv41 = phi i64 [ 0, %entry.split ], [ %indvars.iv.next42, %for.inc20 ]116 br label %for.cond4.preheader117 118for.cond4.preheader: ; preds = %for.inc17, %for.cond1.preheader119 %indvars.iv38 = phi i64 [ 0, %for.cond1.preheader ], [ %indvars.iv.next39, %for.inc17 ]120 br label %for.body6121 122for.body6: ; preds = %for.body6, %for.cond4.preheader123 %indvars.iv = phi i64 [ 0, %for.cond4.preheader ], [ %indvars.iv.next, %for.body6 ]124 %arrayidx8 = getelementptr inbounds [1024 x float], ptr %A, i64 %indvars.iv41, i64 %indvars.iv125 %tmp = load float, ptr %arrayidx8, align 4126 %arrayidx12 = getelementptr inbounds [1024 x float], ptr %B, i64 %indvars.iv, i64 %indvars.iv38127 %tmp1 = load float, ptr %arrayidx12, align 4128 %mul = fmul float %tmp, %tmp1129 %arrayidx16 = getelementptr inbounds [1024 x float], ptr %C, i64 %indvars.iv41, i64 %indvars.iv38130 %tmp2 = load float, ptr %arrayidx16, align 4131 %add = fadd float %tmp2, %mul132 store float %add, ptr %arrayidx16, align 4133 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1134 %exitcond = icmp ne i64 %indvars.iv.next, 1024135 br i1 %exitcond, label %for.body6, label %for.inc17136 137for.inc17: ; preds = %for.body6138 %indvars.iv.next39 = add nuw nsw i64 %indvars.iv38, 1139 %exitcond40 = icmp ne i64 %indvars.iv.next39, 1024140 br i1 %exitcond40, label %for.cond4.preheader, label %for.inc20141 142for.inc20: ; preds = %for.inc17143 %indvars.iv.next42 = add nuw nsw i64 %indvars.iv41, 1144 %exitcond43 = icmp ne i64 %indvars.iv.next42, 1024145 br i1 %exitcond43, label %for.cond1.preheader, label %for.end22146 147for.end22: ; preds = %for.inc20148 ret void149}150