; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve,+neon < %s | FileCheck %s ; Inserting a truncated (i64 to i32) element from the bottom 128-bits of any vector type into a NEON vector should use INS (element) of the ; truncated size to avoid pointless GPR trips. define <2 x i32> @test_s_trunc_d_lane0(<2 x i32> %a, <1 x i64> %b) { ; CHECK-LABEL: test_s_trunc_d_lane0: ; CHECK: // %bb.0: ; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0 ; CHECK-NEXT: // kill: def $d1 killed $d1 def $q1 ; CHECK-NEXT: mov v0.s[0], v1.s[0] ; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0 ; CHECK-NEXT: ret %c = extractelement <1 x i64> %b, i32 0 %d = trunc i64 %c to i32 %e = insertelement <2 x i32> %a, i32 %d, i64 0 ret <2 x i32> %e } define <2 x i32> @test_s_trunc_d_qlane1(<2 x i32> %a, <2 x i64> %b) { ; CHECK-LABEL: test_s_trunc_d_qlane1: ; CHECK: // %bb.0: ; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0 ; CHECK-NEXT: mov v0.s[0], v1.s[2] ; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0 ; CHECK-NEXT: ret %c = extractelement <2 x i64> %b, i32 1 %d = trunc i64 %c to i32 %e = insertelement <2 x i32> %a, i32 %d, i64 0 ret <2 x i32> %e } define <4 x i32> @test_qs_trunc_d_lane0(<4 x i32> %a, <1 x i64> %b) { ; CHECK-LABEL: test_qs_trunc_d_lane0: ; CHECK: // %bb.0: ; CHECK-NEXT: // kill: def $d1 killed $d1 def $q1 ; CHECK-NEXT: mov v0.s[0], v1.s[0] ; CHECK-NEXT: ret %c = extractelement <1 x i64> %b, i32 0 %d = trunc i64 %c to i32 %e = insertelement <4 x i32> %a, i32 %d, i64 0 ret <4 x i32> %e } define <4 x i32> @test_qs_trunc_d_qlane1(<4 x i32> %a, <2 x i64> %b) { ; CHECK-LABEL: test_qs_trunc_d_qlane1: ; CHECK: // %bb.0: ; CHECK-NEXT: mov v0.s[3], v1.s[2] ; CHECK-NEXT: ret %c = extractelement <2 x i64> %b, i32 1 %d = trunc i64 %c to i32 %e = insertelement <4 x i32> %a, i32 %d, i64 3 ret <4 x i32> %e } ; ---- From the bottom 128b of an SVE vector define <2 x i32> @test_s_trunc_dsve_lane0(<2 x i32> %a, %b) { ; CHECK-LABEL: test_s_trunc_dsve_lane0: ; CHECK: // %bb.0: ; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0 ; CHECK-NEXT: mov v0.s[0], v1.s[0] ; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0 ; CHECK-NEXT: ret %c = extractelement %b, i32 0 %d = trunc i64 %c to i32 %e = insertelement <2 x i32> %a, i32 %d, i64 0 ret <2 x i32> %e } define <2 x i32> @test_s_trunc_dsve_lane1(<2 x i32> %a, %b) { ; CHECK-LABEL: test_s_trunc_dsve_lane1: ; CHECK: // %bb.0: ; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0 ; CHECK-NEXT: mov v0.s[1], v1.s[2] ; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0 ; CHECK-NEXT: ret %c = extractelement %b, i32 1 %d = trunc i64 %c to i32 %e = insertelement <2 x i32> %a, i32 %d, i64 1 ret <2 x i32> %e } ; (negative test) Extracted element is not within V-register. define <2 x i32> @test_s_trunc_dsve_lane2(<2 x i32> %a, %b) { ; CHECK-LABEL: test_s_trunc_dsve_lane2: ; CHECK: // %bb.0: ; CHECK-NEXT: mov z1.s, z1.s[4] ; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0 ; CHECK-NEXT: fmov w8, s1 ; CHECK-NEXT: mov v0.s[1], w8 ; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0 ; CHECK-NEXT: ret %c = extractelement %b, i32 2 %d = trunc i64 %c to i32 %e = insertelement <2 x i32> %a, i32 %d, i64 1 ret <2 x i32> %e } define <4 x i32> @test_qs_trunc_dsve_lane0(<4 x i32> %a, %b) { ; CHECK-LABEL: test_qs_trunc_dsve_lane0: ; CHECK: // %bb.0: ; CHECK-NEXT: mov v0.s[0], v1.s[0] ; CHECK-NEXT: ret %c = extractelement %b, i32 0 %d = trunc i64 %c to i32 %e = insertelement <4 x i32> %a, i32 %d, i64 0 ret <4 x i32> %e } define <4 x i32> @test_qs_trunc_dsve_lane1(<4 x i32> %a, %b) { ; CHECK-LABEL: test_qs_trunc_dsve_lane1: ; CHECK: // %bb.0: ; CHECK-NEXT: mov v0.s[3], v1.s[2] ; CHECK-NEXT: ret %c = extractelement %b, i32 1 %d = trunc i64 %c to i32 %e = insertelement <4 x i32> %a, i32 %d, i64 3 ret <4 x i32> %e } ; (negative test) Extracted element is not within V-register. define <4 x i32> @test_qs_trunc_dsve_lane2(<4 x i32> %a, %b) { ; CHECK-LABEL: test_qs_trunc_dsve_lane2: ; CHECK: // %bb.0: ; CHECK-NEXT: mov z1.s, z1.s[4] ; CHECK-NEXT: fmov w8, s1 ; CHECK-NEXT: mov v0.s[3], w8 ; CHECK-NEXT: ret %c = extractelement %b, i32 2 %d = trunc i64 %c to i32 %e = insertelement <4 x i32> %a, i32 %d, i64 3 ret <4 x i32> %e }