135 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc < %s -mtriple=x86_64-linux -mcpu=corei7 | FileCheck %s --check-prefixes=SSE-LINUX3; RUN: llc < %s -mtriple=x86_64-win32 -mcpu=corei7 | FileCheck %s --check-prefixes=SSE-MSVC4; RUN: llc < %s -mtriple=x86_64-linux -mcpu=corei7-avx | FileCheck %s --check-prefixes=AVX-LINUX5; RUN: llc < %s -mtriple=x86_64-win32 -mcpu=corei7-avx | FileCheck %s --check-prefixes=AVX-MSVC6 7; llc should share constant pool entries between this integer vector8; and this floating-point vector since they have the same encoding.9; FIXME: AVX is duplicating broadcasts10 11define void @share_v4i32_v4f32(ptr %p, ptr %q, i1 %t) nounwind {12; SSE-LINUX-LABEL: share_v4i32_v4f32:13; SSE-LINUX: # %bb.0: # %entry14; SSE-LINUX-NEXT: movaps {{.*#+}} xmm0 = [1073741824,1073741824,1073741824,1073741824]15; SSE-LINUX-NEXT: .p2align 416; SSE-LINUX-NEXT: .LBB0_1: # %loop17; SSE-LINUX-NEXT: # =>This Inner Loop Header: Depth=118; SSE-LINUX-NEXT: movaps %xmm0, (%rdi)19; SSE-LINUX-NEXT: movaps %xmm0, (%rsi)20; SSE-LINUX-NEXT: testb $1, %dl21; SSE-LINUX-NEXT: jne .LBB0_122; SSE-LINUX-NEXT: # %bb.2: # %ret23; SSE-LINUX-NEXT: retq24;25; SSE-MSVC-LABEL: share_v4i32_v4f32:26; SSE-MSVC: # %bb.0: # %entry27; SSE-MSVC-NEXT: movaps {{.*#+}} xmm0 = [1073741824,1073741824,1073741824,1073741824]28; SSE-MSVC-NEXT: .p2align 429; SSE-MSVC-NEXT: .LBB0_1: # %loop30; SSE-MSVC-NEXT: # =>This Inner Loop Header: Depth=131; SSE-MSVC-NEXT: movaps %xmm0, (%rcx)32; SSE-MSVC-NEXT: movaps %xmm0, (%rdx)33; SSE-MSVC-NEXT: testb $1, %r8b34; SSE-MSVC-NEXT: jne .LBB0_135; SSE-MSVC-NEXT: # %bb.2: # %ret36; SSE-MSVC-NEXT: retq37;38; AVX-LINUX-LABEL: share_v4i32_v4f32:39; AVX-LINUX: # %bb.0: # %entry40; AVX-LINUX-NEXT: vbroadcastss {{.*#+}} xmm0 = [1073741824,1073741824,1073741824,1073741824]41; AVX-LINUX-NEXT: vbroadcastss {{.*#+}} xmm1 = [1073741824,1073741824,1073741824,1073741824]42; AVX-LINUX-NEXT: .p2align 443; AVX-LINUX-NEXT: .LBB0_1: # %loop44; AVX-LINUX-NEXT: # =>This Inner Loop Header: Depth=145; AVX-LINUX-NEXT: vmovaps %xmm0, (%rdi)46; AVX-LINUX-NEXT: vmovaps %xmm1, (%rsi)47; AVX-LINUX-NEXT: testb $1, %dl48; AVX-LINUX-NEXT: jne .LBB0_149; AVX-LINUX-NEXT: # %bb.2: # %ret50; AVX-LINUX-NEXT: retq51;52; AVX-MSVC-LABEL: share_v4i32_v4f32:53; AVX-MSVC: # %bb.0: # %entry54; AVX-MSVC-NEXT: vbroadcastss {{.*#+}} xmm0 = [1073741824,1073741824,1073741824,1073741824]55; AVX-MSVC-NEXT: vbroadcastss {{.*#+}} xmm1 = [1073741824,1073741824,1073741824,1073741824]56; AVX-MSVC-NEXT: .p2align 457; AVX-MSVC-NEXT: .LBB0_1: # %loop58; AVX-MSVC-NEXT: # =>This Inner Loop Header: Depth=159; AVX-MSVC-NEXT: vmovaps %xmm0, (%rcx)60; AVX-MSVC-NEXT: vmovaps %xmm1, (%rdx)61; AVX-MSVC-NEXT: testb $1, %r8b62; AVX-MSVC-NEXT: jne .LBB0_163; AVX-MSVC-NEXT: # %bb.2: # %ret64; AVX-MSVC-NEXT: retq65entry:66 br label %loop67loop:68 store <4 x i32><i32 1073741824, i32 1073741824, i32 1073741824, i32 1073741824>, ptr %p69 store <4 x float><float 2.0, float 2.0, float 2.0, float 2.0>, ptr %q70 br i1 %t, label %loop, label %ret71ret:72 ret void73}74 75define void @store_repeated_constants(ptr %lo, ptr %hi) {76; SSE-LINUX-LABEL: store_repeated_constants:77; SSE-LINUX: # %bb.0:78; SSE-LINUX-NEXT: xorps %xmm0, %xmm079; SSE-LINUX-NEXT: movaps %xmm0, 48(%rdi)80; SSE-LINUX-NEXT: movsd {{.*#+}} xmm1 = [18446744073709551615,0]81; SSE-LINUX-NEXT: movaps %xmm1, 32(%rdi)82; SSE-LINUX-NEXT: movaps %xmm1, 16(%rdi)83; SSE-LINUX-NEXT: movaps %xmm1, (%rdi)84; SSE-LINUX-NEXT: movaps %xmm0, 32(%rsi)85; SSE-LINUX-NEXT: movaps %xmm0, 48(%rsi)86; SSE-LINUX-NEXT: movaps %xmm1, (%rsi)87; SSE-LINUX-NEXT: movaps {{.*#+}} xmm0 = [0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255]88; SSE-LINUX-NEXT: movaps %xmm0, 16(%rsi)89; SSE-LINUX-NEXT: retq90;91; SSE-MSVC-LABEL: store_repeated_constants:92; SSE-MSVC: # %bb.0:93; SSE-MSVC-NEXT: xorps %xmm0, %xmm094; SSE-MSVC-NEXT: movaps %xmm0, 48(%rcx)95; SSE-MSVC-NEXT: movsd {{.*#+}} xmm1 = [18446744073709551615,0]96; SSE-MSVC-NEXT: movaps %xmm1, 32(%rcx)97; SSE-MSVC-NEXT: movaps %xmm1, 16(%rcx)98; SSE-MSVC-NEXT: movaps %xmm1, (%rcx)99; SSE-MSVC-NEXT: movaps %xmm0, 32(%rdx)100; SSE-MSVC-NEXT: movaps %xmm0, 48(%rdx)101; SSE-MSVC-NEXT: movaps %xmm1, (%rdx)102; SSE-MSVC-NEXT: movaps {{.*#+}} xmm0 = [0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255]103; SSE-MSVC-NEXT: movaps %xmm0, 16(%rdx)104; SSE-MSVC-NEXT: retq105;106; AVX-LINUX-LABEL: store_repeated_constants:107; AVX-LINUX: # %bb.0:108; AVX-LINUX-NEXT: vmovaps {{.*#+}} ymm0 = [18446744073709551615,0,18446744073709551615,0]109; AVX-LINUX-NEXT: vmovaps %ymm0, (%rdi)110; AVX-LINUX-NEXT: vmovaps {{.*#+}} ymm0 = [18446744073709551615,0,0,18446744073709551615]111; AVX-LINUX-NEXT: vmovaps %xmm0, %xmm1112; AVX-LINUX-NEXT: vmovaps %ymm1, 32(%rdi)113; AVX-LINUX-NEXT: vxorps %xmm1, %xmm1, %xmm1114; AVX-LINUX-NEXT: vmovaps %ymm1, 32(%rsi)115; AVX-LINUX-NEXT: vmovaps %ymm0, (%rsi)116; AVX-LINUX-NEXT: vzeroupper117; AVX-LINUX-NEXT: retq118;119; AVX-MSVC-LABEL: store_repeated_constants:120; AVX-MSVC: # %bb.0:121; AVX-MSVC-NEXT: vmovaps {{.*#+}} ymm0 = [18446744073709551615,0,18446744073709551615,0]122; AVX-MSVC-NEXT: vmovaps %ymm0, (%rcx)123; AVX-MSVC-NEXT: vmovaps {{.*#+}} ymm0 = [18446744073709551615,0,0,18446744073709551615]124; AVX-MSVC-NEXT: vmovaps %xmm0, %xmm1125; AVX-MSVC-NEXT: vmovaps %ymm1, 32(%rcx)126; AVX-MSVC-NEXT: vxorps %xmm1, %xmm1, %xmm1127; AVX-MSVC-NEXT: vmovaps %ymm1, 32(%rdx)128; AVX-MSVC-NEXT: vmovaps %ymm0, (%rdx)129; AVX-MSVC-NEXT: vzeroupper130; AVX-MSVC-NEXT: retq131 store volatile <8 x i64> <i64 -1, i64 0, i64 -1, i64 0, i64 -1, i64 0, i64 0, i64 0>, ptr %lo, align 64132 store volatile <8 x i64> <i64 -1, i64 0, i64 0, i64 -1, i64 0, i64 0, i64 0, i64 0>, ptr %hi, align 64133 ret void134}135