| ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py | |
| ; RUN: opt < %s -passes=slp-vectorizer,dce -S -mtriple=i386-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s | |
| target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32-S128" | |
| target triple = "i386-apple-macosx10.8.0" | |
| ; int foo(ptr A, int n, int m) { | |
| ; double sum = 0, v1 = 2, v0 = 3; | |
| ; for (int i=0; i < n; ++i) | |
| ; sum += 7*A[i*2] + 7*A[i*2+1]; | |
| ; return sum; | |
| ; } | |
| define i32 @reduce(ptr nocapture %A, i32 %n, i32 %m) { | |
| ; CHECK-LABEL: @reduce( | |
| ; CHECK-NEXT: entry: | |
| ; CHECK-NEXT: [[CMP13:%.*]] = icmp sgt i32 [[N:%.*]], 0 | |
| ; CHECK-NEXT: br i1 [[CMP13]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]] | |
| ; CHECK: for.body: | |
| ; CHECK-NEXT: [[I_015:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ] | |
| ; CHECK-NEXT: [[SUM_014:%.*]] = phi double [ [[ADD6:%.*]], [[FOR_BODY]] ], [ 0.000000e+00, [[ENTRY]] ] | |
| ; CHECK-NEXT: [[MUL:%.*]] = shl nsw i32 [[I_015]], 1 | |
| ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[A:%.*]], i32 [[MUL]] | |
| ; CHECK-NEXT: [[TMP0:%.*]] = load <2 x double>, ptr [[ARRAYIDX]], align 4 | |
| ; CHECK-NEXT: [[TMP1:%.*]] = fmul <2 x double> [[TMP0]], splat (double 7.000000e+00) | |
| ; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x double> [[TMP1]], i64 0 | |
| ; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x double> [[TMP1]], i64 1 | |
| ; CHECK-NEXT: [[ADD5:%.*]] = fadd double [[TMP2]], [[TMP3]] | |
| ; CHECK-NEXT: [[ADD6]] = fadd double [[SUM_014]], [[ADD5]] | |
| ; CHECK-NEXT: [[INC]] = add nsw i32 [[I_015]], 1 | |
| ; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], [[N]] | |
| ; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY]] | |
| ; CHECK: for.cond.for.end_crit_edge: | |
| ; CHECK-NEXT: [[PHITMP:%.*]] = fptosi double [[ADD6]] to i32 | |
| ; CHECK-NEXT: br label [[FOR_END]] | |
| ; CHECK: for.end: | |
| ; CHECK-NEXT: [[SUM_0_LCSSA:%.*]] = phi i32 [ [[PHITMP]], [[FOR_COND_FOR_END_CRIT_EDGE]] ], [ 0, [[ENTRY]] ] | |
| ; CHECK-NEXT: ret i32 [[SUM_0_LCSSA]] | |
| ; | |
| entry: | |
| %cmp13 = icmp sgt i32 %n, 0 | |
| br i1 %cmp13, label %for.body, label %for.end | |
| for.body: ; preds = %entry, %for.body | |
| %i.015 = phi i32 [ %inc, %for.body ], [ 0, %entry ] | |
| %sum.014 = phi double [ %add6, %for.body ], [ 0.000000e+00, %entry ] | |
| %mul = shl nsw i32 %i.015, 1 | |
| %arrayidx = getelementptr inbounds double, ptr %A, i32 %mul | |
| %0 = load double, ptr %arrayidx, align 4 | |
| %mul1 = fmul double %0, 7.000000e+00 | |
| %add12 = or disjoint i32 %mul, 1 | |
| %arrayidx3 = getelementptr inbounds double, ptr %A, i32 %add12 | |
| %1 = load double, ptr %arrayidx3, align 4 | |
| %mul4 = fmul double %1, 7.000000e+00 | |
| %add5 = fadd double %mul1, %mul4 | |
| %add6 = fadd double %sum.014, %add5 | |
| %inc = add nsw i32 %i.015, 1 | |
| %exitcond = icmp eq i32 %inc, %n | |
| br i1 %exitcond, label %for.cond.for.end_crit_edge, label %for.body | |
| for.cond.for.end_crit_edge: ; preds = %for.body | |
| %phitmp = fptosi double %add6 to i32 | |
| br label %for.end | |
| for.end: ; preds = %for.cond.for.end_crit_edge, %entry | |
| %sum.0.lcssa = phi i32 [ %phitmp, %for.cond.for.end_crit_edge ], [ 0, %entry ] | |
| ret i32 %sum.0.lcssa | |
| } | |
| ; PR43948 - https://bugs.llvm.org/show_bug.cgi?id=43948 | |
| ; The extra use of a non-vectorized element of a reduction must not be killed. | |
| define i32 @horiz_max_multiple_uses(ptr %x, ptr %p) { | |
| ; CHECK-LABEL: @horiz_max_multiple_uses( | |
| ; CHECK-NEXT: [[X4:%.*]] = getelementptr [32 x i32], ptr [[X:%.*]], i64 0, i64 4 | |
| ; CHECK-NEXT: [[X5:%.*]] = getelementptr [32 x i32], ptr [[X]], i64 0, i64 5 | |
| ; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, ptr [[X]], align 4 | |
| ; CHECK-NEXT: [[T6:%.*]] = load i32, ptr [[X4]], align 4 | |
| ; CHECK-NEXT: [[T7:%.*]] = load i32, ptr [[X5]], align 4 | |
| ; CHECK-NEXT: [[RDX4:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP1]]) | |
| ; CHECK-NEXT: [[MAX_ROOT_CMP1:%.*]] = icmp sgt i32 [[RDX4]], [[T6]] | |
| ; CHECK-NEXT: [[MAX_ROOT_SEL:%.*]] = select i1 [[MAX_ROOT_CMP1]], i32 [[RDX4]], i32 [[T6]] | |
| ; CHECK-NEXT: [[T18:%.*]] = call i32 @llvm.smax.i32(i32 [[MAX_ROOT_SEL]], i32 [[T7]]) | |
| ; CHECK-NEXT: [[THREE_OR_FOUR:%.*]] = select i1 [[MAX_ROOT_CMP1]], i32 3, i32 4 | |
| ; CHECK-NEXT: store i32 [[THREE_OR_FOUR]], ptr [[P:%.*]], align 8 | |
| ; CHECK-NEXT: ret i32 [[T18]] | |
| ; | |
| %x1 = getelementptr [32 x i32], ptr %x, i64 0, i64 1 | |
| %x2 = getelementptr [32 x i32], ptr %x, i64 0, i64 2 | |
| %x3 = getelementptr [32 x i32], ptr %x, i64 0, i64 3 | |
| %x4 = getelementptr [32 x i32], ptr %x, i64 0, i64 4 | |
| %x5 = getelementptr [32 x i32], ptr %x, i64 0, i64 5 | |
| %t0 = load i32, ptr %x | |
| %t1 = load i32, ptr %x1 | |
| %t2 = load i32, ptr %x2 | |
| %t3 = load i32, ptr %x3 | |
| %t4 = load i32, ptr %x4 | |
| %t5 = load i32, ptr %x5 | |
| %s5 = call i32 @llvm.smax(i32 %t0, i32 %t1) | |
| %t8 = call i32 @llvm.smax(i32 %s5, i32 %t2) | |
| %rdx4 = call i32 @llvm.smax(i32 %t8, i32 %t3) | |
| %MAX_ROOT_CMP = icmp sgt i32 %rdx4, %t4 | |
| %MAX_ROOT_SEL = select i1 %MAX_ROOT_CMP, i32 %rdx4, i32 %t4 | |
| %t17 = call i32 @llvm.smax(i32 %MAX_ROOT_SEL, i32 %t5) | |
| %three_or_four = select i1 %MAX_ROOT_CMP, i32 3, i32 4 | |
| store i32 %three_or_four, ptr %p, align 8 | |
| ret i32 %t17 | |
| } | |
| ; This is a miscompile (see the undef operand) and/or test for invalid IR. | |
| define i1 @bad_insertpoint_rdx(ptr %p) #0 { | |
| ; CHECK-LABEL: @bad_insertpoint_rdx( | |
| ; CHECK-NEXT: [[T0:%.*]] = load i32, ptr [[P:%.*]], align 16 | |
| ; CHECK-NEXT: [[CMP23:%.*]] = icmp sgt i32 [[T0]], 0 | |
| ; CHECK-NEXT: [[SPEC_SELECT:%.*]] = select i1 [[CMP23]], i32 [[T0]], i32 0 | |
| ; CHECK-NEXT: [[ARRAYIDX22_1:%.*]] = getelementptr inbounds [8 x i32], ptr [[P]], i64 0, i64 1 | |
| ; CHECK-NEXT: [[T1:%.*]] = load i32, ptr [[ARRAYIDX22_1]], align 4 | |
| ; CHECK-NEXT: [[CMP23_1:%.*]] = icmp sgt i32 [[T1]], [[SPEC_SELECT]] | |
| ; CHECK-NEXT: [[SPEC_STORE_SELECT87:%.*]] = zext i1 [[CMP23_1]] to i32 | |
| ; CHECK-NEXT: [[SPEC_SELECT88:%.*]] = select i1 [[CMP23_1]], i32 [[T1]], i32 [[SPEC_SELECT]] | |
| ; CHECK-NEXT: [[CMP23_2:%.*]] = icmp sgt i32 [[SPEC_STORE_SELECT87]], [[SPEC_SELECT88]] | |
| ; CHECK-NEXT: ret i1 [[CMP23_2]] | |
| ; | |
| %t0 = load i32, ptr %p, align 16 | |
| %cmp23 = icmp sgt i32 %t0, 0 | |
| %spec.select = select i1 %cmp23, i32 %t0, i32 0 | |
| %arrayidx22.1 = getelementptr inbounds [8 x i32], ptr %p, i64 0, i64 1 | |
| %t1 = load i32, ptr %arrayidx22.1, align 4 | |
| %cmp23.1 = icmp sgt i32 %t1, %spec.select | |
| %spec.store.select87 = zext i1 %cmp23.1 to i32 | |
| %spec.select88 = select i1 %cmp23.1, i32 %t1, i32 %spec.select | |
| %cmp23.2 = icmp sgt i32 %spec.store.select87, %spec.select88 | |
| ret i1 %cmp23.2 | |
| } |