1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc < %s -mtriple=i386-apple-darwin10 -mcpu=penryn | FileCheck %s 3 4; rdar://7475489 5 6define i32 @test1(i32 %a, i32 %b) nounwind ssp { 7; CHECK-LABEL: test1: 8; CHECK: ## %bb.0: ## %entry 9; CHECK-NEXT: movzbl {{[0-9]+}}(%esp), %eax 10; CHECK-NEXT: xorb {{[0-9]+}}(%esp), %al 11; CHECK-NEXT: testb $64, %al 12; CHECK-NEXT: jne _bar ## TAILCALL 13; CHECK-NEXT: ## %bb.1: ## %bb 14; CHECK-NEXT: jmp _foo ## TAILCALL 15entry: 16 %0 = and i32 %a, 16384 17 %1 = icmp ne i32 %0, 0 18 %2 = and i32 %b, 16384 19 %3 = icmp ne i32 %2, 0 20 %4 = xor i1 %1, %3 21 br i1 %4, label %bb1, label %bb 22 23bb: ; preds = %entry 24 %5 = tail call i32 (...) @foo() nounwind ; <i32> [#uses=1] 25 ret i32 %5 26 27bb1: ; preds = %entry 28 %6 = tail call i32 (...) @bar() nounwind ; <i32> [#uses=1] 29 ret i32 %6 30} 31 32declare i32 @foo(...) 33 34declare i32 @bar(...) 35 36 37; <rdar://problem/7598384>: 38; 39; jCC L1 40; jmp L2 41; L1: 42; ... 43; L2: 44; ... 45; 46; to: 47; 48; jnCC L2 49; L1: 50; ... 51; L2: 52; ... 53define float @test4(float %x, float %y) nounwind readnone optsize ssp { 54; CHECK-LABEL: test4: 55; CHECK: ## %bb.0: ## %entry 56; CHECK-NEXT: pushl %eax 57; CHECK-NEXT: cvtss2sd {{[0-9]+}}(%esp), %xmm1 58; CHECK-NEXT: cvtss2sd {{[0-9]+}}(%esp), %xmm0 59; CHECK-NEXT: mulsd %xmm1, %xmm0 60; CHECK-NEXT: xorpd %xmm1, %xmm1 61; CHECK-NEXT: ucomisd %xmm1, %xmm0 62; CHECK-NEXT: jne LBB1_1 63; CHECK-NEXT: jnp LBB1_2 64; CHECK-NEXT: LBB1_1: ## %bb1 65; CHECK-NEXT: addsd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0 66; CHECK-NEXT: LBB1_2: ## %bb2 67; CHECK-NEXT: cvtsd2ss %xmm0, %xmm0 68; CHECK-NEXT: movss %xmm0, (%esp) 69; CHECK-NEXT: flds (%esp) 70; CHECK-NEXT: popl %eax 71; CHECK-NEXT: retl 72entry: 73 %0 = fpext float %x to double ; <double> [#uses=1] 74 %1 = fpext float %y to double ; <double> [#uses=1] 75 %2 = fmul double %0, %1 ; <double> [#uses=3] 76 %3 = fcmp oeq double %2, 0.000000e+00 ; <i1> [#uses=1] 77 br i1 %3, label %bb2, label %bb1 78 79 80bb1: ; preds = %entry 81 %4 = fadd double %2, -1.000000e+00 ; <double> [#uses=1] 82 br label %bb2 83 84bb2: ; preds = %entry, %bb1 85 %.0.in = phi double [ %4, %bb1 ], [ %2, %entry ] ; <double> [#uses=1] 86 %.0 = fptrunc double %.0.in to float ; <float> [#uses=1] 87 ret float %.0 88} 89 90declare i32 @llvm.x86.sse41.ptestz(<4 x float> %p1, <4 x float> %p2) nounwind 91declare i32 @llvm.x86.sse41.ptestc(<4 x float> %p1, <4 x float> %p2) nounwind 92 93define <4 x float> @test5(<4 x float> %a, <4 x float> %b) nounwind { 94; CHECK-LABEL: test5: 95; CHECK: ## %bb.0: ## %entry 96; CHECK-NEXT: ptest %xmm0, %xmm0 97; CHECK-NEXT: jne LBB2_2 98; CHECK-NEXT: ## %bb.1: ## %bb1 99; CHECK-NEXT: addps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 100; CHECK-NEXT: movaps %xmm1, %xmm0 101; CHECK-NEXT: retl 102; CHECK-NEXT: LBB2_2: ## %bb2 103; CHECK-NEXT: divps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 104; CHECK-NEXT: movaps %xmm1, %xmm0 105; CHECK-NEXT: retl 106entry: 107 108 %res = call i32 @llvm.x86.sse41.ptestz(<4 x float> %a, <4 x float> %a) nounwind 109 %one = icmp ne i32 %res, 0 110 br i1 %one, label %bb1, label %bb2 111 112bb1: 113 %c = fadd <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 114 br label %return 115 116bb2: 117 %d = fdiv <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 118 br label %return 119 120return: 121 %e = phi <4 x float> [%c, %bb1], [%d, %bb2] 122 ret <4 x float> %e 123} 124 125define <4 x float> @test7(<4 x float> %a, <4 x float> %b) nounwind { 126; CHECK-LABEL: test7: 127; CHECK: ## %bb.0: ## %entry 128; CHECK-NEXT: ptest %xmm0, %xmm0 129; CHECK-NEXT: jne LBB3_2 130; CHECK-NEXT: ## %bb.1: ## %bb1 131; CHECK-NEXT: addps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 132; CHECK-NEXT: movaps %xmm1, %xmm0 133; CHECK-NEXT: retl 134; CHECK-NEXT: LBB3_2: ## %bb2 135; CHECK-NEXT: divps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 136; CHECK-NEXT: movaps %xmm1, %xmm0 137; CHECK-NEXT: retl 138entry: 139 140 %res = call i32 @llvm.x86.sse41.ptestz(<4 x float> %a, <4 x float> %a) nounwind 141 %one = trunc i32 %res to i1 142 br i1 %one, label %bb1, label %bb2 143 144bb1: 145 %c = fadd <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 146 br label %return 147 148bb2: 149 %d = fdiv <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 150 br label %return 151 152return: 153 %e = phi <4 x float> [%c, %bb1], [%d, %bb2] 154 ret <4 x float> %e 155} 156 157define <4 x float> @test8(<4 x float> %a, <4 x float> %b) nounwind { 158; CHECK-LABEL: test8: 159; CHECK: ## %bb.0: ## %entry 160; CHECK-NEXT: ptest %xmm0, %xmm0 161; CHECK-NEXT: jae LBB4_2 162; CHECK-NEXT: ## %bb.1: ## %bb1 163; CHECK-NEXT: addps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 164; CHECK-NEXT: movaps %xmm1, %xmm0 165; CHECK-NEXT: retl 166; CHECK-NEXT: LBB4_2: ## %bb2 167; CHECK-NEXT: divps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 168; CHECK-NEXT: movaps %xmm1, %xmm0 169; CHECK-NEXT: retl 170entry: 171 172 %res = call i32 @llvm.x86.sse41.ptestc(<4 x float> %a, <4 x float> %a) nounwind 173 %one = icmp ne i32 %res, 0 174 br i1 %one, label %bb1, label %bb2 175 176bb1: 177 %c = fadd <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 178 br label %return 179 180bb2: 181 %d = fdiv <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 182 br label %return 183 184return: 185 %e = phi <4 x float> [%c, %bb1], [%d, %bb2] 186 ret <4 x float> %e 187} 188 189define <4 x float> @test10(<4 x float> %a, <4 x float> %b) nounwind { 190; CHECK-LABEL: test10: 191; CHECK: ## %bb.0: ## %entry 192; CHECK-NEXT: ptest %xmm0, %xmm0 193; CHECK-NEXT: jae LBB5_2 194; CHECK-NEXT: ## %bb.1: ## %bb1 195; CHECK-NEXT: addps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 196; CHECK-NEXT: movaps %xmm1, %xmm0 197; CHECK-NEXT: retl 198; CHECK-NEXT: LBB5_2: ## %bb2 199; CHECK-NEXT: divps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 200; CHECK-NEXT: movaps %xmm1, %xmm0 201; CHECK-NEXT: retl 202entry: 203 204 %res = call i32 @llvm.x86.sse41.ptestc(<4 x float> %a, <4 x float> %a) nounwind 205 %one = trunc i32 %res to i1 206 br i1 %one, label %bb1, label %bb2 207 208bb1: 209 %c = fadd <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 210 br label %return 211 212bb2: 213 %d = fdiv <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 214 br label %return 215 216return: 217 %e = phi <4 x float> [%c, %bb1], [%d, %bb2] 218 ret <4 x float> %e 219} 220 221define <4 x float> @test11(<4 x float> %a, <4 x float> %b) nounwind { 222; CHECK-LABEL: test11: 223; CHECK: ## %bb.0: ## %entry 224; CHECK-NEXT: ptest %xmm0, %xmm0 225; CHECK-NEXT: jne LBB6_2 226; CHECK-NEXT: ## %bb.1: ## %bb1 227; CHECK-NEXT: addps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 228; CHECK-NEXT: movaps %xmm1, %xmm0 229; CHECK-NEXT: retl 230; CHECK-NEXT: LBB6_2: ## %bb2 231; CHECK-NEXT: divps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 232; CHECK-NEXT: movaps %xmm1, %xmm0 233; CHECK-NEXT: retl 234entry: 235 236 %res = call i32 @llvm.x86.sse41.ptestz(<4 x float> %a, <4 x float> %a) nounwind 237 %one = icmp eq i32 %res, 1 238 br i1 %one, label %bb1, label %bb2 239 240bb1: 241 %c = fadd <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 242 br label %return 243 244bb2: 245 %d = fdiv <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 246 br label %return 247 248return: 249 %e = phi <4 x float> [%c, %bb1], [%d, %bb2] 250 ret <4 x float> %e 251} 252 253define <4 x float> @test12(<4 x float> %a, <4 x float> %b) nounwind { 254; CHECK-LABEL: test12: 255; CHECK: ## %bb.0: ## %entry 256; CHECK-NEXT: ptest %xmm0, %xmm0 257; CHECK-NEXT: je LBB7_2 258; CHECK-NEXT: ## %bb.1: ## %bb1 259; CHECK-NEXT: addps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 260; CHECK-NEXT: movaps %xmm1, %xmm0 261; CHECK-NEXT: retl 262; CHECK-NEXT: LBB7_2: ## %bb2 263; CHECK-NEXT: divps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1 264; CHECK-NEXT: movaps %xmm1, %xmm0 265; CHECK-NEXT: retl 266entry: 267 268 %res = call i32 @llvm.x86.sse41.ptestz(<4 x float> %a, <4 x float> %a) nounwind 269 %one = icmp ne i32 %res, 1 270 br i1 %one, label %bb1, label %bb2 271 272bb1: 273 %c = fadd <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 274 br label %return 275 276bb2: 277 %d = fdiv <4 x float> %b, < float 1.000000e+002, float 2.000000e+002, float 3.000000e+002, float 4.000000e+002 > 278 br label %return 279 280return: 281 %e = phi <4 x float> [%c, %bb1], [%d, %bb2] 282 ret <4 x float> %e 283} 284 285