Author: Rito Takeuchi Date: 2026-10-19T14:52:32+02:00 New Revision: 4635269e3782522f0d00759f544227c93df0bcc9
URL: https://github.com/llvm/llvm-project/commit/4635269e3782522f0d00759f544227c93df0bcc9 DIFF: https://github.com/llvm/llvm-project/commit/4635269e3782522f0d00759f544227c93df0bcc9.diff LOG: [X86] Fix constant vector i64 signed div/rem on 32-bit targets (#229244) Only mark v4i64 and v8i64 [SU]MUL_LOHI as Custom in 64-bit mode. Their custom lowering unrolls to scalar i64 operations, whose results cannot be expanded by the type legalizer on 32-bit targets. Add crash regression coverage for signed division and remainder by positive and negative constants with AVX2, AVX512F, and AVX512DQ+VL. Fixes #228656 (cherry picked from commit e3e894a951586e653dd7910cad39804cb30ce9fd) Added: llvm/test/CodeGen/X86/vector-i64-div-rem-32bit.ll Modified: llvm/lib/Target/X86/X86ISelLowering.cpp Removed: ################################################################################ diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp index 6e367a94cd9bc..e4d4da09ce9b7 100644 --- a/llvm/lib/Target/X86/X86ISelLowering.cpp +++ b/llvm/lib/Target/X86/X86ISelLowering.cpp @@ -1754,9 +1754,12 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM, if (HasInt256) { setOperationAction(ISD::MULHU, MVT::v4i64, Custom); // Custom so the combiner keeps full products as [SU]MUL_LOHI, not - // MULH[SU]. - setOperationAction(ISD::UMUL_LOHI, MVT::v4i64, Custom); - setOperationAction(ISD::SMUL_LOHI, MVT::v4i64, Custom); + // MULH[SU]. The custom lowering unrolls to scalar i64 [SU]MUL_LOHI, + // which is only legalizable when i64 is a legal type. + if (Subtarget.is64Bit()) { + setOperationAction(ISD::UMUL_LOHI, MVT::v4i64, Custom); + setOperationAction(ISD::SMUL_LOHI, MVT::v4i64, Custom); + } setOperationAction(ISD::VSELECT, MVT::v32i8, Legal); // Custom legalize 2x32 to get a little better code. @@ -2029,8 +2032,10 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM, setOperationAction(ISD::MUL, MVT::v64i8, Custom); setOperationAction(ISD::MULHU, MVT::v8i64, Custom); - setOperationAction(ISD::UMUL_LOHI, MVT::v8i64, Custom); - setOperationAction(ISD::SMUL_LOHI, MVT::v8i64, Custom); + if (Subtarget.is64Bit()) { + setOperationAction(ISD::UMUL_LOHI, MVT::v8i64, Custom); + setOperationAction(ISD::SMUL_LOHI, MVT::v8i64, Custom); + } setOperationAction(ISD::MULHU, MVT::v16i32, Custom); setOperationAction(ISD::MULHS, MVT::v16i32, Custom); setOperationAction(ISD::MULHS, MVT::v32i16, HasBWI ? Legal : Custom); diff --git a/llvm/test/CodeGen/X86/vector-i64-div-rem-32bit.ll b/llvm/test/CodeGen/X86/vector-i64-div-rem-32bit.ll new file mode 100644 index 0000000000000..b54d5809f7aba --- /dev/null +++ b/llvm/test/CodeGen/X86/vector-i64-div-rem-32bit.ll @@ -0,0 +1,1593 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc < %s -mtriple=i686-unknown-linux-gnu -mattr=+avx2 | FileCheck %s --check-prefix=AVX2 +; RUN: llc < %s -mtriple=i686-unknown-linux-gnu -mattr=+avx512f,-avx512dq | FileCheck %s --check-prefix=AVX512F +; RUN: llc < %s -mtriple=i686-unknown-linux-gnu -mattr=+avx512f,+avx512dq,+avx512vl | FileCheck %s --check-prefix=AVX512DQ + +; Constant signed division/remainder must not unroll vector MUL_LOHI into +; scalar i64 MUL_LOHI on 32-bit targets, where i64 is an illegal type. + +define <4 x i64> @sdiv_v4i64(<4 x i64> %a) nounwind { +; AVX2-LABEL: sdiv_v4i64: +; AVX2: # %bb.0: +; AVX2-NEXT: pushl %ebp +; AVX2-NEXT: pushl %ebx +; AVX2-NEXT: pushl %edi +; AVX2-NEXT: pushl %esi +; AVX2-NEXT: subl $76, %esp +; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, %esi +; AVX2-NEXT: movl %edx, %edi +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, %ebx +; AVX2-NEXT: movl %edx, %ebp +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovd %eax, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vmovd %esi, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX2-NEXT: addl $76, %esp +; AVX2-NEXT: popl %esi +; AVX2-NEXT: popl %edi +; AVX2-NEXT: popl %ebx +; AVX2-NEXT: popl %ebp +; AVX2-NEXT: retl +; +; AVX512F-LABEL: sdiv_v4i64: +; AVX512F: # %bb.0: +; AVX512F-NEXT: pushl %ebp +; AVX512F-NEXT: pushl %ebx +; AVX512F-NEXT: pushl %edi +; AVX512F-NEXT: pushl %esi +; AVX512F-NEXT: subl $76, %esp +; AVX512F-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, %esi +; AVX512F-NEXT: movl %edx, %edi +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX512F-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, %ebx +; AVX512F-NEXT: movl %edx, %ebp +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovd %eax, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vmovd %esi, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX512F-NEXT: addl $76, %esp +; AVX512F-NEXT: popl %esi +; AVX512F-NEXT: popl %edi +; AVX512F-NEXT: popl %ebx +; AVX512F-NEXT: popl %ebp +; AVX512F-NEXT: retl +; +; AVX512DQ-LABEL: sdiv_v4i64: +; AVX512DQ: # %bb.0: +; AVX512DQ-NEXT: vpsraq $32, %ymm0, %ymm1 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [277094665,0,277094665,0,277094665,0,277094665,0] +; AVX512DQ-NEXT: vpmuldq %ymm2, %ymm1, %ymm3 +; AVX512DQ-NEXT: vpmuludq %ymm2, %ymm0, %ymm2 +; AVX512DQ-NEXT: vpsrlq $32, %ymm2, %ymm2 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm3, %ymm2 +; AVX512DQ-NEXT: vpsraq $32, %ymm2, %ymm3 +; AVX512DQ-NEXT: vpxor %xmm4, %xmm4, %xmm4 +; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[3],ymm2[4],ymm4[5],ymm2[6],ymm4[7] +; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm4 = ymm0[0],ymm4[1],ymm0[2],ymm4[3],ymm0[4],ymm4[5],ymm0[6],ymm4[7] +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm5 = [2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295] +; AVX512DQ-NEXT: vpmullq %ymm5, %ymm4, %ymm4 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm4, %ymm2 +; AVX512DQ-NEXT: vpsraq $32, %ymm2, %ymm2 +; AVX512DQ-NEXT: vpmuldq %ymm5, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpaddq %ymm3, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpaddq %ymm0, %ymm1, %ymm0 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm0, %ymm0 +; AVX512DQ-NEXT: vpsrlq $63, %ymm0, %ymm1 +; AVX512DQ-NEXT: vpsraq $5, %ymm0, %ymm0 +; AVX512DQ-NEXT: vpaddq %ymm1, %ymm0, %ymm0 +; AVX512DQ-NEXT: retl + %r = sdiv <4 x i64> %a, splat (i64 62) + ret <4 x i64> %r +} + +define <4 x i64> @srem_v4i64(<4 x i64> %a) nounwind { +; AVX2-LABEL: srem_v4i64: +; AVX2: # %bb.0: +; AVX2-NEXT: pushl %ebp +; AVX2-NEXT: pushl %ebx +; AVX2-NEXT: pushl %edi +; AVX2-NEXT: pushl %esi +; AVX2-NEXT: subl $76, %esp +; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, %esi +; AVX2-NEXT: movl %edx, %edi +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, %ebx +; AVX2-NEXT: movl %edx, %ebp +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovd %eax, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vmovd %esi, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX2-NEXT: addl $76, %esp +; AVX2-NEXT: popl %esi +; AVX2-NEXT: popl %edi +; AVX2-NEXT: popl %ebx +; AVX2-NEXT: popl %ebp +; AVX2-NEXT: retl +; +; AVX512F-LABEL: srem_v4i64: +; AVX512F: # %bb.0: +; AVX512F-NEXT: pushl %ebp +; AVX512F-NEXT: pushl %ebx +; AVX512F-NEXT: pushl %edi +; AVX512F-NEXT: pushl %esi +; AVX512F-NEXT: subl $76, %esp +; AVX512F-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, %esi +; AVX512F-NEXT: movl %edx, %edi +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX512F-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, %ebx +; AVX512F-NEXT: movl %edx, %ebp +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovd %eax, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vmovd %esi, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX512F-NEXT: addl $76, %esp +; AVX512F-NEXT: popl %esi +; AVX512F-NEXT: popl %edi +; AVX512F-NEXT: popl %ebx +; AVX512F-NEXT: popl %ebp +; AVX512F-NEXT: retl +; +; AVX512DQ-LABEL: srem_v4i64: +; AVX512DQ: # %bb.0: +; AVX512DQ-NEXT: vpsraq $32, %ymm0, %ymm1 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [277094665,0,277094665,0,277094665,0,277094665,0] +; AVX512DQ-NEXT: vpmuldq %ymm2, %ymm1, %ymm3 +; AVX512DQ-NEXT: vpmuludq %ymm2, %ymm0, %ymm2 +; AVX512DQ-NEXT: vpsrlq $32, %ymm2, %ymm2 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm3, %ymm2 +; AVX512DQ-NEXT: vpsraq $32, %ymm2, %ymm3 +; AVX512DQ-NEXT: vpxor %xmm4, %xmm4, %xmm4 +; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[3],ymm2[4],ymm4[5],ymm2[6],ymm4[7] +; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm4 = ymm0[0],ymm4[1],ymm0[2],ymm4[3],ymm0[4],ymm4[5],ymm0[6],ymm4[7] +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm5 = [2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295] +; AVX512DQ-NEXT: vpmullq %ymm5, %ymm4, %ymm4 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm4, %ymm2 +; AVX512DQ-NEXT: vpsraq $32, %ymm2, %ymm2 +; AVX512DQ-NEXT: vpmuldq %ymm5, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpaddq %ymm3, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpaddq %ymm0, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpsrlq $63, %ymm1, %ymm2 +; AVX512DQ-NEXT: vpsraq $5, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpmullq {{\.?LCPI[0-9]+_[0-9]+}}{1to4}, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpsubq %ymm1, %ymm0, %ymm0 +; AVX512DQ-NEXT: retl + %r = srem <4 x i64> %a, splat (i64 62) + ret <4 x i64> %r +} + +define <8 x i64> @sdiv_v8i64(<8 x i64> %a) nounwind { +; AVX2-LABEL: sdiv_v8i64: +; AVX2: # %bb.0: +; AVX2-NEXT: pushl %ebp +; AVX2-NEXT: pushl %ebx +; AVX2-NEXT: pushl %edi +; AVX2-NEXT: pushl %esi +; AVX2-NEXT: subl $156, %esp +; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vextractps $1, %xmm1, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm1, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, %esi +; AVX2-NEXT: movl %edx, %edi +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, %ebx +; AVX2-NEXT: movl %edx, %ebp +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: vmovd %eax, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX2-NEXT: vmovd %esi, %xmm1 +; AVX2-NEXT: vpinsrd $1, %edi, %xmm1, %xmm1 +; AVX2-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX2-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: # xmm2 = mem[0],zero,zero,zero +; AVX2-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm2 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm2, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm2, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 4-byte Folded Reload +; AVX2-NEXT: # xmm0 = mem[0],zero,zero,zero +; AVX2-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: addl $156, %esp +; AVX2-NEXT: popl %esi +; AVX2-NEXT: popl %edi +; AVX2-NEXT: popl %ebx +; AVX2-NEXT: popl %ebp +; AVX2-NEXT: retl +; +; AVX512F-LABEL: sdiv_v8i64: +; AVX512F: # %bb.0: +; AVX512F-NEXT: pushl %ebp +; AVX512F-NEXT: pushl %ebx +; AVX512F-NEXT: pushl %edi +; AVX512F-NEXT: pushl %esi +; AVX512F-NEXT: subl $188, %esp +; AVX512F-NEXT: vmovups %zmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 64-byte Spill +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf32x4 $2, %zmm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, %esi +; AVX512F-NEXT: movl %edx, %edi +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf32x4 $3, %zmm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, %ebx +; AVX512F-NEXT: movl %edx, %ebp +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: vmovd %eax, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX512F-NEXT: vmovd %esi, %xmm1 +; AVX512F-NEXT: vpinsrd $1, %edi, %xmm1, %xmm1 +; AVX512F-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX512F-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: # xmm2 = mem[0],zero,zero,zero +; AVX512F-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm2 # 64-byte Reload +; AVX512F-NEXT: vextractps $3, %xmm2, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm2, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX512F-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX512F-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 4-byte Folded Reload +; AVX512F-NEXT: # xmm0 = mem[0],zero,zero,zero +; AVX512F-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX512F-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%e{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload +; AVX512F-NEXT: addl $188, %esp +; AVX512F-NEXT: popl %esi +; AVX512F-NEXT: popl %edi +; AVX512F-NEXT: popl %ebx +; AVX512F-NEXT: popl %ebp +; AVX512F-NEXT: retl +; +; AVX512DQ-LABEL: sdiv_v8i64: +; AVX512DQ: # %bb.0: +; AVX512DQ-NEXT: vpsraq $32, %zmm0, %zmm1 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} zmm2 = [277094665,0,277094665,0,277094665,0,277094665,0,277094665,0,277094665,0,277094665,0,277094665,0] +; AVX512DQ-NEXT: vpmuldq %zmm2, %zmm1, %zmm3 +; AVX512DQ-NEXT: vpmuludq %zmm2, %zmm0, %zmm2 +; AVX512DQ-NEXT: vpsrlq $32, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm3, %zmm2 +; AVX512DQ-NEXT: vpsraq $32, %zmm2, %zmm3 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} zmm4 = [4294967295,0,4294967295,0,4294967295,0,4294967295,0,4294967295,0,4294967295,0,4294967295,0,4294967295,0] +; AVX512DQ-NEXT: vpandq %zmm4, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpandq %zmm4, %zmm0, %zmm4 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} zmm5 = [2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295] +; AVX512DQ-NEXT: vpmullq %zmm5, %zmm4, %zmm4 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm4, %zmm2 +; AVX512DQ-NEXT: vpsraq $32, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpmuldq %zmm5, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpaddq %zmm3, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpaddq %zmm0, %zmm1, %zmm0 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm0, %zmm0 +; AVX512DQ-NEXT: vpsrlq $63, %zmm0, %zmm1 +; AVX512DQ-NEXT: vpsraq $5, %zmm0, %zmm0 +; AVX512DQ-NEXT: vpaddq %zmm1, %zmm0, %zmm0 +; AVX512DQ-NEXT: retl + %r = sdiv <8 x i64> %a, splat (i64 62) + ret <8 x i64> %r +} + +define <8 x i64> @srem_v8i64(<8 x i64> %a) nounwind { +; AVX2-LABEL: srem_v8i64: +; AVX2: # %bb.0: +; AVX2-NEXT: pushl %ebp +; AVX2-NEXT: pushl %ebx +; AVX2-NEXT: pushl %edi +; AVX2-NEXT: pushl %esi +; AVX2-NEXT: subl $156, %esp +; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vextractps $1, %xmm1, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm1, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, %esi +; AVX2-NEXT: movl %edx, %edi +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, %ebx +; AVX2-NEXT: movl %edx, %ebp +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: vmovd %eax, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX2-NEXT: vmovd %esi, %xmm1 +; AVX2-NEXT: vpinsrd $1, %edi, %xmm1, %xmm1 +; AVX2-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX2-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: # xmm2 = mem[0],zero,zero,zero +; AVX2-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm2 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm2, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm2, (%esp) +; AVX2-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 4-byte Folded Reload +; AVX2-NEXT: # xmm0 = mem[0],zero,zero,zero +; AVX2-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: addl $156, %esp +; AVX2-NEXT: popl %esi +; AVX2-NEXT: popl %edi +; AVX2-NEXT: popl %ebx +; AVX2-NEXT: popl %ebp +; AVX2-NEXT: retl +; +; AVX512F-LABEL: srem_v8i64: +; AVX512F: # %bb.0: +; AVX512F-NEXT: pushl %ebp +; AVX512F-NEXT: pushl %ebx +; AVX512F-NEXT: pushl %edi +; AVX512F-NEXT: pushl %esi +; AVX512F-NEXT: subl $188, %esp +; AVX512F-NEXT: vmovups %zmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 64-byte Spill +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf32x4 $2, %zmm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, %esi +; AVX512F-NEXT: movl %edx, %edi +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf32x4 $3, %zmm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, %ebx +; AVX512F-NEXT: movl %edx, %ebp +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: vmovd %eax, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX512F-NEXT: vmovd %esi, %xmm1 +; AVX512F-NEXT: vpinsrd $1, %edi, %xmm1, %xmm1 +; AVX512F-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX512F-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: # xmm2 = mem[0],zero,zero,zero +; AVX512F-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm2 # 64-byte Reload +; AVX512F-NEXT: vextractps $3, %xmm2, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm2, (%esp) +; AVX512F-NEXT: movl $0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX512F-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX512F-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 4-byte Folded Reload +; AVX512F-NEXT: # xmm0 = mem[0],zero,zero,zero +; AVX512F-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX512F-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%e{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload +; AVX512F-NEXT: addl $188, %esp +; AVX512F-NEXT: popl %esi +; AVX512F-NEXT: popl %edi +; AVX512F-NEXT: popl %ebx +; AVX512F-NEXT: popl %ebp +; AVX512F-NEXT: retl +; +; AVX512DQ-LABEL: srem_v8i64: +; AVX512DQ: # %bb.0: +; AVX512DQ-NEXT: vpsraq $32, %zmm0, %zmm1 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} zmm2 = [277094665,0,277094665,0,277094665,0,277094665,0,277094665,0,277094665,0,277094665,0,277094665,0] +; AVX512DQ-NEXT: vpmuldq %zmm2, %zmm1, %zmm3 +; AVX512DQ-NEXT: vpmuludq %zmm2, %zmm0, %zmm2 +; AVX512DQ-NEXT: vpsrlq $32, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm3, %zmm2 +; AVX512DQ-NEXT: vpsraq $32, %zmm2, %zmm3 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} zmm4 = [4294967295,0,4294967295,0,4294967295,0,4294967295,0,4294967295,0,4294967295,0,4294967295,0,4294967295,0] +; AVX512DQ-NEXT: vpandq %zmm4, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpandq %zmm4, %zmm0, %zmm4 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} zmm5 = [2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295,2216757314,4294967295] +; AVX512DQ-NEXT: vpmullq %zmm5, %zmm4, %zmm4 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm4, %zmm2 +; AVX512DQ-NEXT: vpsraq $32, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpmuldq %zmm5, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpaddq %zmm3, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpaddq %zmm0, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpsrlq $63, %zmm1, %zmm2 +; AVX512DQ-NEXT: vpsraq $5, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpmullq {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpsubq %zmm1, %zmm0, %zmm0 +; AVX512DQ-NEXT: retl + %r = srem <8 x i64> %a, splat (i64 62) + ret <8 x i64> %r +} + +define <4 x i64> @sdiv_v4i64_negative(<4 x i64> %a) nounwind { +; AVX2-LABEL: sdiv_v4i64_negative: +; AVX2: # %bb.0: +; AVX2-NEXT: pushl %ebp +; AVX2-NEXT: pushl %ebx +; AVX2-NEXT: pushl %edi +; AVX2-NEXT: pushl %esi +; AVX2-NEXT: subl $76, %esp +; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, %esi +; AVX2-NEXT: movl %edx, %edi +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, %ebx +; AVX2-NEXT: movl %edx, %ebp +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovd %eax, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vmovd %esi, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX2-NEXT: addl $76, %esp +; AVX2-NEXT: popl %esi +; AVX2-NEXT: popl %edi +; AVX2-NEXT: popl %ebx +; AVX2-NEXT: popl %ebp +; AVX2-NEXT: retl +; +; AVX512F-LABEL: sdiv_v4i64_negative: +; AVX512F: # %bb.0: +; AVX512F-NEXT: pushl %ebp +; AVX512F-NEXT: pushl %ebx +; AVX512F-NEXT: pushl %edi +; AVX512F-NEXT: pushl %esi +; AVX512F-NEXT: subl $76, %esp +; AVX512F-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, %esi +; AVX512F-NEXT: movl %edx, %edi +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX512F-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, %ebx +; AVX512F-NEXT: movl %edx, %ebp +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovd %eax, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vmovd %esi, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX512F-NEXT: addl $76, %esp +; AVX512F-NEXT: popl %esi +; AVX512F-NEXT: popl %edi +; AVX512F-NEXT: popl %ebx +; AVX512F-NEXT: popl %ebp +; AVX512F-NEXT: retl +; +; AVX512DQ-LABEL: sdiv_v4i64_negative: +; AVX512DQ: # %bb.0: +; AVX512DQ-NEXT: vpsraq $32, %ymm0, %ymm1 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [4017872631,0,4017872631,0,4017872631,0,4017872631,0] +; AVX512DQ-NEXT: vpmullq %ymm2, %ymm1, %ymm3 +; AVX512DQ-NEXT: vpmuludq %ymm2, %ymm0, %ymm2 +; AVX512DQ-NEXT: vpsrlq $32, %ymm2, %ymm2 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm3, %ymm2 +; AVX512DQ-NEXT: vpsraq $32, %ymm2, %ymm3 +; AVX512DQ-NEXT: vpxor %xmm4, %xmm4, %xmm4 +; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[3],ymm2[4],ymm4[5],ymm2[6],ymm4[7] +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm4 = [2078209981,0,2078209981,0,2078209981,0,2078209981,0] +; AVX512DQ-NEXT: vpmuludq %ymm4, %ymm0, %ymm5 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm5, %ymm2 +; AVX512DQ-NEXT: vpsraq $32, %ymm2, %ymm2 +; AVX512DQ-NEXT: vpmuldq %ymm4, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpaddq %ymm3, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpsubq %ymm0, %ymm1, %ymm0 +; AVX512DQ-NEXT: vpsrlq $63, %ymm0, %ymm1 +; AVX512DQ-NEXT: vpsraq $5, %ymm0, %ymm0 +; AVX512DQ-NEXT: vpaddq %ymm1, %ymm0, %ymm0 +; AVX512DQ-NEXT: retl + %r = sdiv <4 x i64> %a, splat (i64 -62) + ret <4 x i64> %r +} + +define <4 x i64> @srem_v4i64_negative(<4 x i64> %a) nounwind { +; AVX2-LABEL: srem_v4i64_negative: +; AVX2: # %bb.0: +; AVX2-NEXT: pushl %ebp +; AVX2-NEXT: pushl %ebx +; AVX2-NEXT: pushl %edi +; AVX2-NEXT: pushl %esi +; AVX2-NEXT: subl $76, %esp +; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, %esi +; AVX2-NEXT: movl %edx, %edi +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, %ebx +; AVX2-NEXT: movl %edx, %ebp +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovd %eax, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vmovd %esi, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX2-NEXT: addl $76, %esp +; AVX2-NEXT: popl %esi +; AVX2-NEXT: popl %edi +; AVX2-NEXT: popl %ebx +; AVX2-NEXT: popl %ebp +; AVX2-NEXT: retl +; +; AVX512F-LABEL: srem_v4i64_negative: +; AVX512F: # %bb.0: +; AVX512F-NEXT: pushl %ebp +; AVX512F-NEXT: pushl %ebx +; AVX512F-NEXT: pushl %edi +; AVX512F-NEXT: pushl %esi +; AVX512F-NEXT: subl $76, %esp +; AVX512F-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, %esi +; AVX512F-NEXT: movl %edx, %edi +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX512F-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, %ebx +; AVX512F-NEXT: movl %edx, %ebp +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovd %eax, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vmovd %esi, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX512F-NEXT: addl $76, %esp +; AVX512F-NEXT: popl %esi +; AVX512F-NEXT: popl %edi +; AVX512F-NEXT: popl %ebx +; AVX512F-NEXT: popl %ebp +; AVX512F-NEXT: retl +; +; AVX512DQ-LABEL: srem_v4i64_negative: +; AVX512DQ: # %bb.0: +; AVX512DQ-NEXT: vpsraq $32, %ymm0, %ymm1 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [4017872631,0,4017872631,0,4017872631,0,4017872631,0] +; AVX512DQ-NEXT: vpmullq %ymm2, %ymm1, %ymm3 +; AVX512DQ-NEXT: vpmuludq %ymm2, %ymm0, %ymm2 +; AVX512DQ-NEXT: vpsrlq $32, %ymm2, %ymm2 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm3, %ymm2 +; AVX512DQ-NEXT: vpsraq $32, %ymm2, %ymm3 +; AVX512DQ-NEXT: vpxor %xmm4, %xmm4, %xmm4 +; AVX512DQ-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[3],ymm2[4],ymm4[5],ymm2[6],ymm4[7] +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm4 = [2078209981,0,2078209981,0,2078209981,0,2078209981,0] +; AVX512DQ-NEXT: vpmuludq %ymm4, %ymm0, %ymm5 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm5, %ymm2 +; AVX512DQ-NEXT: vpsraq $32, %ymm2, %ymm2 +; AVX512DQ-NEXT: vpmuldq %ymm4, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpaddq %ymm3, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpsubq %ymm0, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpsrlq $63, %ymm1, %ymm2 +; AVX512DQ-NEXT: vpsraq $5, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpaddq %ymm2, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpmullq {{\.?LCPI[0-9]+_[0-9]+}}{1to4}, %ymm1, %ymm1 +; AVX512DQ-NEXT: vpsubq %ymm1, %ymm0, %ymm0 +; AVX512DQ-NEXT: retl + %r = srem <4 x i64> %a, splat (i64 -62) + ret <4 x i64> %r +} + +define <8 x i64> @sdiv_v8i64_negative(<8 x i64> %a) nounwind { +; AVX2-LABEL: sdiv_v8i64_negative: +; AVX2: # %bb.0: +; AVX2-NEXT: pushl %ebp +; AVX2-NEXT: pushl %ebx +; AVX2-NEXT: pushl %edi +; AVX2-NEXT: pushl %esi +; AVX2-NEXT: subl $156, %esp +; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vextractps $1, %xmm1, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm1, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, %esi +; AVX2-NEXT: movl %edx, %edi +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: movl %eax, %ebx +; AVX2-NEXT: movl %edx, %ebp +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: vmovd %eax, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX2-NEXT: vmovd %esi, %xmm1 +; AVX2-NEXT: vpinsrd $1, %edi, %xmm1, %xmm1 +; AVX2-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX2-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: # xmm2 = mem[0],zero,zero,zero +; AVX2-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm2 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm2, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm2, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 4-byte Folded Reload +; AVX2-NEXT: # xmm0 = mem[0],zero,zero,zero +; AVX2-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __divdi3 +; AVX2-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: addl $156, %esp +; AVX2-NEXT: popl %esi +; AVX2-NEXT: popl %edi +; AVX2-NEXT: popl %ebx +; AVX2-NEXT: popl %ebp +; AVX2-NEXT: retl +; +; AVX512F-LABEL: sdiv_v8i64_negative: +; AVX512F: # %bb.0: +; AVX512F-NEXT: pushl %ebp +; AVX512F-NEXT: pushl %ebx +; AVX512F-NEXT: pushl %edi +; AVX512F-NEXT: pushl %esi +; AVX512F-NEXT: subl $188, %esp +; AVX512F-NEXT: vmovups %zmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 64-byte Spill +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf32x4 $2, %zmm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, %esi +; AVX512F-NEXT: movl %edx, %edi +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf32x4 $3, %zmm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: movl %eax, %ebx +; AVX512F-NEXT: movl %edx, %ebp +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: vmovd %eax, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX512F-NEXT: vmovd %esi, %xmm1 +; AVX512F-NEXT: vpinsrd $1, %edi, %xmm1, %xmm1 +; AVX512F-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX512F-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: # xmm2 = mem[0],zero,zero,zero +; AVX512F-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm2 # 64-byte Reload +; AVX512F-NEXT: vextractps $3, %xmm2, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm2, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX512F-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX512F-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 4-byte Folded Reload +; AVX512F-NEXT: # xmm0 = mem[0],zero,zero,zero +; AVX512F-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __divdi3 +; AVX512F-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX512F-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%e{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload +; AVX512F-NEXT: addl $188, %esp +; AVX512F-NEXT: popl %esi +; AVX512F-NEXT: popl %edi +; AVX512F-NEXT: popl %ebx +; AVX512F-NEXT: popl %ebp +; AVX512F-NEXT: retl +; +; AVX512DQ-LABEL: sdiv_v8i64_negative: +; AVX512DQ: # %bb.0: +; AVX512DQ-NEXT: vpsraq $32, %zmm0, %zmm1 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} zmm2 = [4017872631,0,4017872631,0,4017872631,0,4017872631,0,4017872631,0,4017872631,0,4017872631,0,4017872631,0] +; AVX512DQ-NEXT: vpmullq %zmm2, %zmm1, %zmm3 +; AVX512DQ-NEXT: vpmuludq %zmm2, %zmm0, %zmm2 +; AVX512DQ-NEXT: vpsrlq $32, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm3, %zmm2 +; AVX512DQ-NEXT: vpsraq $32, %zmm2, %zmm3 +; AVX512DQ-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} zmm4 = [2078209981,0,2078209981,0,2078209981,0,2078209981,0,2078209981,0,2078209981,0,2078209981,0,2078209981,0] +; AVX512DQ-NEXT: vpmuludq %zmm4, %zmm0, %zmm5 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm5, %zmm2 +; AVX512DQ-NEXT: vpsraq $32, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpmuldq %zmm4, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpaddq %zmm3, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpsubq %zmm0, %zmm1, %zmm0 +; AVX512DQ-NEXT: vpsrlq $63, %zmm0, %zmm1 +; AVX512DQ-NEXT: vpsraq $5, %zmm0, %zmm0 +; AVX512DQ-NEXT: vpaddq %zmm1, %zmm0, %zmm0 +; AVX512DQ-NEXT: retl + %r = sdiv <8 x i64> %a, splat (i64 -62) + ret <8 x i64> %r +} + +define <8 x i64> @srem_v8i64_negative(<8 x i64> %a) nounwind { +; AVX2-LABEL: srem_v8i64_negative: +; AVX2: # %bb.0: +; AVX2-NEXT: pushl %ebp +; AVX2-NEXT: pushl %ebx +; AVX2-NEXT: pushl %edi +; AVX2-NEXT: pushl %esi +; AVX2-NEXT: subl $156, %esp +; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vextractps $1, %xmm1, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm1, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, %esi +; AVX2-NEXT: movl %edx, %edi +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: movl %eax, %ebx +; AVX2-NEXT: movl %edx, %ebp +; AVX2-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX2-NEXT: vmovss %xmm0, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: vmovd %eax, %xmm0 +; AVX2-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX2-NEXT: vmovd %esi, %xmm1 +; AVX2-NEXT: vpinsrd $1, %edi, %xmm1, %xmm1 +; AVX2-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX2-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: # xmm2 = mem[0],zero,zero,zero +; AVX2-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX2-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm2 # 32-byte Reload +; AVX2-NEXT: vextractps $3, %xmm2, {{[0-9]+}}(%esp) +; AVX2-NEXT: vextractps $2, %xmm2, (%esp) +; AVX2-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX2-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX2-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 4-byte Folded Reload +; AVX2-NEXT: # xmm0 = mem[0],zero,zero,zero +; AVX2-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload +; AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: calll __moddi3 +; AVX2-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX2-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX2-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload +; AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload +; AVX2-NEXT: addl $156, %esp +; AVX2-NEXT: popl %esi +; AVX2-NEXT: popl %edi +; AVX2-NEXT: popl %ebx +; AVX2-NEXT: popl %ebp +; AVX2-NEXT: retl +; +; AVX512F-LABEL: srem_v8i64_negative: +; AVX512F: # %bb.0: +; AVX512F-NEXT: pushl %ebp +; AVX512F-NEXT: pushl %ebx +; AVX512F-NEXT: pushl %edi +; AVX512F-NEXT: pushl %esi +; AVX512F-NEXT: subl $188, %esp +; AVX512F-NEXT: vmovups %zmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 64-byte Spill +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf32x4 $2, %zmm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, %esi +; AVX512F-NEXT: movl %edx, %edi +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm0 # 64-byte Reload +; AVX512F-NEXT: vextractf32x4 $3, %zmm0, %xmm0 +; AVX512F-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vextractps $3, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: movl %eax, %ebx +; AVX512F-NEXT: movl %edx, %ebp +; AVX512F-NEXT: vmovaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vmovss %xmm0, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: vmovd %eax, %xmm0 +; AVX512F-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %ebp, %xmm0, %xmm0 +; AVX512F-NEXT: vmovd %esi, %xmm1 +; AVX512F-NEXT: vpinsrd $1, %edi, %xmm1, %xmm1 +; AVX512F-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 4-byte Folded Reload +; AVX512F-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: # xmm2 = mem[0],zero,zero,zero +; AVX512F-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $2, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vpinsrd $3, {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 4-byte Folded Reload +; AVX512F-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %zmm2 # 64-byte Reload +; AVX512F-NEXT: vextractps $3, %xmm2, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vextractps $2, %xmm2, (%esp) +; AVX512F-NEXT: movl $-1, {{[0-9]+}}(%esp) +; AVX512F-NEXT: movl $-62, {{[0-9]+}}(%esp) +; AVX512F-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 +; AVX512F-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill +; AVX512F-NEXT: vmovd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 4-byte Folded Reload +; AVX512F-NEXT: # xmm0 = mem[0],zero,zero,zero +; AVX512F-NEXT: vpinsrd $1, {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload +; AVX512F-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill +; AVX512F-NEXT: vzeroupper +; AVX512F-NEXT: calll __moddi3 +; AVX512F-NEXT: vmovdqa {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload +; AVX512F-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 +; AVX512F-NEXT: vpinsrd $3, %edx, %xmm0, %xmm0 +; AVX512F-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; AVX512F-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%e{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload +; AVX512F-NEXT: addl $188, %esp +; AVX512F-NEXT: popl %esi +; AVX512F-NEXT: popl %edi +; AVX512F-NEXT: popl %ebx +; AVX512F-NEXT: popl %ebp +; AVX512F-NEXT: retl +; +; AVX512DQ-LABEL: srem_v8i64_negative: +; AVX512DQ: # %bb.0: +; AVX512DQ-NEXT: vpsraq $32, %zmm0, %zmm1 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} zmm2 = [4017872631,0,4017872631,0,4017872631,0,4017872631,0,4017872631,0,4017872631,0,4017872631,0,4017872631,0] +; AVX512DQ-NEXT: vpmullq %zmm2, %zmm1, %zmm3 +; AVX512DQ-NEXT: vpmuludq %zmm2, %zmm0, %zmm2 +; AVX512DQ-NEXT: vpsrlq $32, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm3, %zmm2 +; AVX512DQ-NEXT: vpsraq $32, %zmm2, %zmm3 +; AVX512DQ-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} zmm4 = [2078209981,0,2078209981,0,2078209981,0,2078209981,0,2078209981,0,2078209981,0,2078209981,0,2078209981,0] +; AVX512DQ-NEXT: vpmuludq %zmm4, %zmm0, %zmm5 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm5, %zmm2 +; AVX512DQ-NEXT: vpsraq $32, %zmm2, %zmm2 +; AVX512DQ-NEXT: vpmuldq %zmm4, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpaddq %zmm3, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpsubq %zmm0, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpsrlq $63, %zmm1, %zmm2 +; AVX512DQ-NEXT: vpsraq $5, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpaddq %zmm2, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpmullq {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %zmm1, %zmm1 +; AVX512DQ-NEXT: vpsubq %zmm1, %zmm0, %zmm0 +; AVX512DQ-NEXT: retl + %r = srem <8 x i64> %a, splat (i64 -62) + ret <8 x i64> %r +} _______________________________________________ llvm-branch-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
