samvv/bolt
1
0
Fork 0
Code Issues 24 Pull requests Projects Releases Packages Wiki Activity Actions
bolt/deps/llvm-18.1.8/llvm/test/CodeGen/AArch64/arm64-neon-simd-shift.ll
Sam Vervaeck 174b6f4d4e
Embed LLVM 18.1.8
2025-02-14 19:21:04 +01:00

821 lines
29 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc < %s -verify-machineinstrs -mtriple=arm64-none-linux-gnu -mattr=+neon | FileCheck %s
define <8 x i8> @test_vshr_n_s8(<8 x i8> %a) {
; CHECK-LABEL: test_vshr_n_s8:
; CHECK: // %bb.0:
; CHECK-NEXT: sshr v0.8b, v0.8b, #3
; CHECK-NEXT: ret
%vshr_n = ashr <8 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
ret <8 x i8> %vshr_n
}
define <4 x i16> @test_vshr_n_s16(<4 x i16> %a) {
; CHECK-LABEL: test_vshr_n_s16:
; CHECK: // %bb.0:
; CHECK-NEXT: sshr v0.4h, v0.4h, #3
; CHECK-NEXT: ret
%vshr_n = ashr <4 x i16> %a, <i16 3, i16 3, i16 3, i16 3>
ret <4 x i16> %vshr_n
}
define <2 x i32> @test_vshr_n_s32(<2 x i32> %a) {
; CHECK-LABEL: test_vshr_n_s32:
; CHECK: // %bb.0:
; CHECK-NEXT: sshr v0.2s, v0.2s, #3
; CHECK-NEXT: ret
%vshr_n = ashr <2 x i32> %a, <i32 3, i32 3>
ret <2 x i32> %vshr_n
}
define <16 x i8> @test_vshrq_n_s8(<16 x i8> %a) {
; CHECK-LABEL: test_vshrq_n_s8:
; CHECK: // %bb.0:
; CHECK-NEXT: sshr v0.16b, v0.16b, #3
; CHECK-NEXT: ret
%vshr_n = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
ret <16 x i8> %vshr_n
}
define <8 x i16> @test_vshrq_n_s16(<8 x i16> %a) {
; CHECK-LABEL: test_vshrq_n_s16:
; CHECK: // %bb.0:
; CHECK-NEXT: sshr v0.8h, v0.8h, #3
; CHECK-NEXT: ret
%vshr_n = ashr <8 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
ret <8 x i16> %vshr_n
}
define <4 x i32> @test_vshrq_n_s32(<4 x i32> %a) {
; CHECK-LABEL: test_vshrq_n_s32:
; CHECK: // %bb.0:
; CHECK-NEXT: sshr v0.4s, v0.4s, #3
; CHECK-NEXT: ret
%vshr_n = ashr <4 x i32> %a, <i32 3, i32 3, i32 3, i32 3>
ret <4 x i32> %vshr_n
}
define <2 x i64> @test_vshrq_n_s64(<2 x i64> %a) {
; CHECK-LABEL: test_vshrq_n_s64:
; CHECK: // %bb.0:
; CHECK-NEXT: sshr v0.2d, v0.2d, #3
; CHECK-NEXT: ret
%vshr_n = ashr <2 x i64> %a, <i64 3, i64 3>
ret <2 x i64> %vshr_n
}
define <8 x i8> @test_vshr_n_u8(<8 x i8> %a) {
; CHECK-LABEL: test_vshr_n_u8:
; CHECK: // %bb.0:
; CHECK-NEXT: ushr v0.8b, v0.8b, #3
; CHECK-NEXT: ret
%vshr_n = lshr <8 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
ret <8 x i8> %vshr_n
}
define <4 x i16> @test_vshr_n_u16(<4 x i16> %a) {
; CHECK-LABEL: test_vshr_n_u16:
; CHECK: // %bb.0:
; CHECK-NEXT: ushr v0.4h, v0.4h, #3
; CHECK-NEXT: ret
%vshr_n = lshr <4 x i16> %a, <i16 3, i16 3, i16 3, i16 3>
ret <4 x i16> %vshr_n
}
define <2 x i32> @test_vshr_n_u32(<2 x i32> %a) {
; CHECK-LABEL: test_vshr_n_u32:
; CHECK: // %bb.0:
; CHECK-NEXT: ushr v0.2s, v0.2s, #3
; CHECK-NEXT: ret
%vshr_n = lshr <2 x i32> %a, <i32 3, i32 3>
ret <2 x i32> %vshr_n
}
define <16 x i8> @test_vshrq_n_u8(<16 x i8> %a) {
; CHECK-LABEL: test_vshrq_n_u8:
; CHECK: // %bb.0:
; CHECK-NEXT: ushr v0.16b, v0.16b, #3
; CHECK-NEXT: ret
%vshr_n = lshr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
ret <16 x i8> %vshr_n
}
define <8 x i16> @test_vshrq_n_u16(<8 x i16> %a) {
; CHECK-LABEL: test_vshrq_n_u16:
; CHECK: // %bb.0:
; CHECK-NEXT: ushr v0.8h, v0.8h, #3
; CHECK-NEXT: ret
%vshr_n = lshr <8 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
ret <8 x i16> %vshr_n
}
define <4 x i32> @test_vshrq_n_u32(<4 x i32> %a) {
; CHECK-LABEL: test_vshrq_n_u32:
; CHECK: // %bb.0:
; CHECK-NEXT: ushr v0.4s, v0.4s, #3
; CHECK-NEXT: ret
%vshr_n = lshr <4 x i32> %a, <i32 3, i32 3, i32 3, i32 3>
ret <4 x i32> %vshr_n
}
define <2 x i64> @test_vshrq_n_u64(<2 x i64> %a) {
; CHECK-LABEL: test_vshrq_n_u64:
; CHECK: // %bb.0:
; CHECK-NEXT: ushr v0.2d, v0.2d, #3
; CHECK-NEXT: ret
%vshr_n = lshr <2 x i64> %a, <i64 3, i64 3>
ret <2 x i64> %vshr_n
}
define <8 x i8> @test_vsra_n_s8(<8 x i8> %a, <8 x i8> %b) {
; CHECK-LABEL: test_vsra_n_s8:
; CHECK: // %bb.0:
; CHECK-NEXT: ssra v0.8b, v1.8b, #3
; CHECK-NEXT: ret
%vsra_n = ashr <8 x i8> %b, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
%1 = add <8 x i8> %vsra_n, %a
ret <8 x i8> %1
}
define <4 x i16> @test_vsra_n_s16(<4 x i16> %a, <4 x i16> %b) {
; CHECK-LABEL: test_vsra_n_s16:
; CHECK: // %bb.0:
; CHECK-NEXT: ssra v0.4h, v1.4h, #3
; CHECK-NEXT: ret
%vsra_n = ashr <4 x i16> %b, <i16 3, i16 3, i16 3, i16 3>
%1 = add <4 x i16> %vsra_n, %a
ret <4 x i16> %1
}
define <2 x i32> @test_vsra_n_s32(<2 x i32> %a, <2 x i32> %b) {
; CHECK-LABEL: test_vsra_n_s32:
; CHECK: // %bb.0:
; CHECK-NEXT: ssra v0.2s, v1.2s, #3
; CHECK-NEXT: ret
%vsra_n = ashr <2 x i32> %b, <i32 3, i32 3>
%1 = add <2 x i32> %vsra_n, %a
ret <2 x i32> %1
}
define <16 x i8> @test_vsraq_n_s8(<16 x i8> %a, <16 x i8> %b) {
; CHECK-LABEL: test_vsraq_n_s8:
; CHECK: // %bb.0:
; CHECK-NEXT: ssra v0.16b, v1.16b, #3
; CHECK-NEXT: ret
%vsra_n = ashr <16 x i8> %b, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
%1 = add <16 x i8> %vsra_n, %a
ret <16 x i8> %1
}
define <8 x i16> @test_vsraq_n_s16(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_vsraq_n_s16:
; CHECK: // %bb.0:
; CHECK-NEXT: ssra v0.8h, v1.8h, #3
; CHECK-NEXT: ret
%vsra_n = ashr <8 x i16> %b, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
%1 = add <8 x i16> %vsra_n, %a
ret <8 x i16> %1
}
define <4 x i32> @test_vsraq_n_s32(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_vsraq_n_s32:
; CHECK: // %bb.0:
; CHECK-NEXT: ssra v0.4s, v1.4s, #3
; CHECK-NEXT: ret
%vsra_n = ashr <4 x i32> %b, <i32 3, i32 3, i32 3, i32 3>
%1 = add <4 x i32> %vsra_n, %a
ret <4 x i32> %1
}
define <2 x i64> @test_vsraq_n_s64(<2 x i64> %a, <2 x i64> %b) {
; CHECK-LABEL: test_vsraq_n_s64:
; CHECK: // %bb.0:
; CHECK-NEXT: ssra v0.2d, v1.2d, #3
; CHECK-NEXT: ret
%vsra_n = ashr <2 x i64> %b, <i64 3, i64 3>
%1 = add <2 x i64> %vsra_n, %a
ret <2 x i64> %1
}
define <8 x i8> @test_vsra_n_u8(<8 x i8> %a, <8 x i8> %b) {
; CHECK-LABEL: test_vsra_n_u8:
; CHECK: // %bb.0:
; CHECK-NEXT: usra v0.8b, v1.8b, #3
; CHECK-NEXT: ret
%vsra_n = lshr <8 x i8> %b, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
%1 = add <8 x i8> %vsra_n, %a
ret <8 x i8> %1
}
define <4 x i16> @test_vsra_n_u16(<4 x i16> %a, <4 x i16> %b) {
; CHECK-LABEL: test_vsra_n_u16:
; CHECK: // %bb.0:
; CHECK-NEXT: usra v0.4h, v1.4h, #3
; CHECK-NEXT: ret
%vsra_n = lshr <4 x i16> %b, <i16 3, i16 3, i16 3, i16 3>
%1 = add <4 x i16> %vsra_n, %a
ret <4 x i16> %1
}
define <2 x i32> @test_vsra_n_u32(<2 x i32> %a, <2 x i32> %b) {
; CHECK-LABEL: test_vsra_n_u32:
; CHECK: // %bb.0:
; CHECK-NEXT: usra v0.2s, v1.2s, #3
; CHECK-NEXT: ret
%vsra_n = lshr <2 x i32> %b, <i32 3, i32 3>
%1 = add <2 x i32> %vsra_n, %a
ret <2 x i32> %1
}
define <16 x i8> @test_vsraq_n_u8(<16 x i8> %a, <16 x i8> %b) {
; CHECK-LABEL: test_vsraq_n_u8:
; CHECK: // %bb.0:
; CHECK-NEXT: usra v0.16b, v1.16b, #3
; CHECK-NEXT: ret
%vsra_n = lshr <16 x i8> %b, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
%1 = add <16 x i8> %vsra_n, %a
ret <16 x i8> %1
}
define <8 x i16> @test_vsraq_n_u16(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_vsraq_n_u16:
; CHECK: // %bb.0:
; CHECK-NEXT: usra v0.8h, v1.8h, #3
; CHECK-NEXT: ret
%vsra_n = lshr <8 x i16> %b, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
%1 = add <8 x i16> %vsra_n, %a
ret <8 x i16> %1
}
define <4 x i32> @test_vsraq_n_u32(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_vsraq_n_u32:
; CHECK: // %bb.0:
; CHECK-NEXT: usra v0.4s, v1.4s, #3
; CHECK-NEXT: ret
%vsra_n = lshr <4 x i32> %b, <i32 3, i32 3, i32 3, i32 3>
%1 = add <4 x i32> %vsra_n, %a
ret <4 x i32> %1
}
define <2 x i64> @test_vsraq_n_u64(<2 x i64> %a, <2 x i64> %b) {
; CHECK-LABEL: test_vsraq_n_u64:
; CHECK: // %bb.0:
; CHECK-NEXT: usra v0.2d, v1.2d, #3
; CHECK-NEXT: ret
%vsra_n = lshr <2 x i64> %b, <i64 3, i64 3>
%1 = add <2 x i64> %vsra_n, %a
ret <2 x i64> %1
}
define <8 x i8> @test_vshrn_n_s16(<8 x i16> %a) {
; CHECK-LABEL: test_vshrn_n_s16:
; CHECK: // %bb.0:
; CHECK-NEXT: shrn v0.8b, v0.8h, #3
; CHECK-NEXT: ret
%1 = ashr <8 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
%vshrn_n = trunc <8 x i16> %1 to <8 x i8>
ret <8 x i8> %vshrn_n
}
define <4 x i16> @test_vshrn_n_s32(<4 x i32> %a) {
; CHECK-LABEL: test_vshrn_n_s32:
; CHECK: // %bb.0:
; CHECK-NEXT: shrn v0.4h, v0.4s, #9
; CHECK-NEXT: ret
%1 = ashr <4 x i32> %a, <i32 9, i32 9, i32 9, i32 9>
%vshrn_n = trunc <4 x i32> %1 to <4 x i16>
ret <4 x i16> %vshrn_n
}
define <2 x i32> @test_vshrn_n_s64(<2 x i64> %a) {
; CHECK-LABEL: test_vshrn_n_s64:
; CHECK: // %bb.0:
; CHECK-NEXT: shrn v0.2s, v0.2d, #19
; CHECK-NEXT: ret
%1 = ashr <2 x i64> %a, <i64 19, i64 19>
%vshrn_n = trunc <2 x i64> %1 to <2 x i32>
ret <2 x i32> %vshrn_n
}
define <8 x i8> @test_vshrn_n_u16(<8 x i16> %a) {
; CHECK-LABEL: test_vshrn_n_u16:
; CHECK: // %bb.0:
; CHECK-NEXT: shrn v0.8b, v0.8h, #3
; CHECK-NEXT: ret
%1 = lshr <8 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
%vshrn_n = trunc <8 x i16> %1 to <8 x i8>
ret <8 x i8> %vshrn_n
}
define <4 x i16> @test_vshrn_n_u32(<4 x i32> %a) {
; CHECK-LABEL: test_vshrn_n_u32:
; CHECK: // %bb.0:
; CHECK-NEXT: shrn v0.4h, v0.4s, #9
; CHECK-NEXT: ret
%1 = lshr <4 x i32> %a, <i32 9, i32 9, i32 9, i32 9>
%vshrn_n = trunc <4 x i32> %1 to <4 x i16>
ret <4 x i16> %vshrn_n
}
define <2 x i32> @test_vshrn_n_u64(<2 x i64> %a) {
; CHECK-LABEL: test_vshrn_n_u64:
; CHECK: // %bb.0:
; CHECK-NEXT: shrn v0.2s, v0.2d, #19
; CHECK-NEXT: ret
%1 = lshr <2 x i64> %a, <i64 19, i64 19>
%vshrn_n = trunc <2 x i64> %1 to <2 x i32>
ret <2 x i32> %vshrn_n
}
define <16 x i8> @test_vshrn_high_n_s16(<8 x i8> %a, <8 x i16> %b) {
; CHECK-LABEL: test_vshrn_high_n_s16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: shrn2 v0.16b, v1.8h, #3
; CHECK-NEXT: ret
%1 = ashr <8 x i16> %b, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
%vshrn_n = trunc <8 x i16> %1 to <8 x i8>
%2 = bitcast <8 x i8> %a to <1 x i64>
%3 = bitcast <8 x i8> %vshrn_n to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %2, <1 x i64> %3, <2 x i32> <i32 0, i32 1>
%4 = bitcast <2 x i64> %shuffle.i to <16 x i8>
ret <16 x i8> %4
}
define <8 x i16> @test_vshrn_high_n_s32(<4 x i16> %a, <4 x i32> %b) {
; CHECK-LABEL: test_vshrn_high_n_s32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: shrn2 v0.8h, v1.4s, #9
; CHECK-NEXT: ret
%1 = ashr <4 x i32> %b, <i32 9, i32 9, i32 9, i32 9>
%vshrn_n = trunc <4 x i32> %1 to <4 x i16>
%2 = bitcast <4 x i16> %a to <1 x i64>
%3 = bitcast <4 x i16> %vshrn_n to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %2, <1 x i64> %3, <2 x i32> <i32 0, i32 1>
%4 = bitcast <2 x i64> %shuffle.i to <8 x i16>
ret <8 x i16> %4
}
define <4 x i32> @test_vshrn_high_n_s64(<2 x i32> %a, <2 x i64> %b) {
; CHECK-LABEL: test_vshrn_high_n_s64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: shrn2 v0.4s, v1.2d, #19
; CHECK-NEXT: ret
%1 = bitcast <2 x i32> %a to <1 x i64>
%2 = ashr <2 x i64> %b, <i64 19, i64 19>
%vshrn_n = trunc <2 x i64> %2 to <2 x i32>
%3 = bitcast <2 x i32> %vshrn_n to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %3, <2 x i32> <i32 0, i32 1>
%4 = bitcast <2 x i64> %shuffle.i to <4 x i32>
ret <4 x i32> %4
}
define <16 x i8> @test_vshrn_high_n_u16(<8 x i8> %a, <8 x i16> %b) {
; CHECK-LABEL: test_vshrn_high_n_u16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: shrn2 v0.16b, v1.8h, #3
; CHECK-NEXT: ret
%1 = lshr <8 x i16> %b, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
%vshrn_n = trunc <8 x i16> %1 to <8 x i8>
%2 = bitcast <8 x i8> %a to <1 x i64>
%3 = bitcast <8 x i8> %vshrn_n to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %2, <1 x i64> %3, <2 x i32> <i32 0, i32 1>
%4 = bitcast <2 x i64> %shuffle.i to <16 x i8>
ret <16 x i8> %4
}
define <8 x i16> @test_vshrn_high_n_u32(<4 x i16> %a, <4 x i32> %b) {
; CHECK-LABEL: test_vshrn_high_n_u32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: shrn2 v0.8h, v1.4s, #9
; CHECK-NEXT: ret
%1 = lshr <4 x i32> %b, <i32 9, i32 9, i32 9, i32 9>
%vshrn_n = trunc <4 x i32> %1 to <4 x i16>
%2 = bitcast <4 x i16> %a to <1 x i64>
%3 = bitcast <4 x i16> %vshrn_n to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %2, <1 x i64> %3, <2 x i32> <i32 0, i32 1>
%4 = bitcast <2 x i64> %shuffle.i to <8 x i16>
ret <8 x i16> %4
}
define <4 x i32> @test_vshrn_high_n_u64(<2 x i32> %a, <2 x i64> %b) {
; CHECK-LABEL: test_vshrn_high_n_u64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: shrn2 v0.4s, v1.2d, #19
; CHECK-NEXT: ret
%1 = bitcast <2 x i32> %a to <1 x i64>
%2 = lshr <2 x i64> %b, <i64 19, i64 19>
%vshrn_n = trunc <2 x i64> %2 to <2 x i32>
%3 = bitcast <2 x i32> %vshrn_n to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %3, <2 x i32> <i32 0, i32 1>
%4 = bitcast <2 x i64> %shuffle.i to <4 x i32>
ret <4 x i32> %4
}
define <16 x i8> @test_vqshrun_high_n_s16(<8 x i8> %a, <8 x i16> %b) {
; CHECK-LABEL: test_vqshrun_high_n_s16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqshrun2 v0.16b, v1.8h, #3
; CHECK-NEXT: ret
%vqshrun = tail call <8 x i8> @llvm.aarch64.neon.sqshrun.v8i8(<8 x i16> %b, i32 3)
%1 = bitcast <8 x i8> %a to <1 x i64>
%2 = bitcast <8 x i8> %vqshrun to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <16 x i8>
ret <16 x i8> %3
}
define <8 x i16> @test_vqshrun_high_n_s32(<4 x i16> %a, <4 x i32> %b) {
; CHECK-LABEL: test_vqshrun_high_n_s32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqshrun2 v0.8h, v1.4s, #9
; CHECK-NEXT: ret
%vqshrun = tail call <4 x i16> @llvm.aarch64.neon.sqshrun.v4i16(<4 x i32> %b, i32 9)
%1 = bitcast <4 x i16> %a to <1 x i64>
%2 = bitcast <4 x i16> %vqshrun to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <8 x i16>
ret <8 x i16> %3
}
define <4 x i32> @test_vqshrun_high_n_s64(<2 x i32> %a, <2 x i64> %b) {
; CHECK-LABEL: test_vqshrun_high_n_s64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqshrun2 v0.4s, v1.2d, #19
; CHECK-NEXT: ret
%1 = bitcast <2 x i32> %a to <1 x i64>
%vqshrun = tail call <2 x i32> @llvm.aarch64.neon.sqshrun.v2i32(<2 x i64> %b, i32 19)
%2 = bitcast <2 x i32> %vqshrun to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <4 x i32>
ret <4 x i32> %3
}
define <16 x i8> @test_vrshrn_high_n_s16(<8 x i8> %a, <8 x i16> %b) {
; CHECK-LABEL: test_vrshrn_high_n_s16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: rshrn2 v0.16b, v1.8h, #3
; CHECK-NEXT: ret
%vrshrn = tail call <8 x i8> @llvm.aarch64.neon.rshrn.v8i8(<8 x i16> %b, i32 3)
%1 = bitcast <8 x i8> %a to <1 x i64>
%2 = bitcast <8 x i8> %vrshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <16 x i8>
ret <16 x i8> %3
}
define <8 x i16> @test_vrshrn_high_n_s32(<4 x i16> %a, <4 x i32> %b) {
; CHECK-LABEL: test_vrshrn_high_n_s32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: rshrn2 v0.8h, v1.4s, #9
; CHECK-NEXT: ret
%vrshrn = tail call <4 x i16> @llvm.aarch64.neon.rshrn.v4i16(<4 x i32> %b, i32 9)
%1 = bitcast <4 x i16> %a to <1 x i64>
%2 = bitcast <4 x i16> %vrshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <8 x i16>
ret <8 x i16> %3
}
define <4 x i32> @test_vrshrn_high_n_s64(<2 x i32> %a, <2 x i64> %b) {
; CHECK-LABEL: test_vrshrn_high_n_s64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: rshrn2 v0.4s, v1.2d, #19
; CHECK-NEXT: ret
%1 = bitcast <2 x i32> %a to <1 x i64>
%vrshrn = tail call <2 x i32> @llvm.aarch64.neon.rshrn.v2i32(<2 x i64> %b, i32 19)
%2 = bitcast <2 x i32> %vrshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <4 x i32>
ret <4 x i32> %3
}
define <16 x i8> @test_vqrshrun_high_n_s16(<8 x i8> %a, <8 x i16> %b) {
; CHECK-LABEL: test_vqrshrun_high_n_s16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqrshrun2 v0.16b, v1.8h, #3
; CHECK-NEXT: ret
%vqrshrun = tail call <8 x i8> @llvm.aarch64.neon.sqrshrun.v8i8(<8 x i16> %b, i32 3)
%1 = bitcast <8 x i8> %a to <1 x i64>
%2 = bitcast <8 x i8> %vqrshrun to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <16 x i8>
ret <16 x i8> %3
}
define <8 x i16> @test_vqrshrun_high_n_s32(<4 x i16> %a, <4 x i32> %b) {
; CHECK-LABEL: test_vqrshrun_high_n_s32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqrshrun2 v0.8h, v1.4s, #9
; CHECK-NEXT: ret
%vqrshrun = tail call <4 x i16> @llvm.aarch64.neon.sqrshrun.v4i16(<4 x i32> %b, i32 9)
%1 = bitcast <4 x i16> %a to <1 x i64>
%2 = bitcast <4 x i16> %vqrshrun to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <8 x i16>
ret <8 x i16> %3
}
define <4 x i32> @test_vqrshrun_high_n_s64(<2 x i32> %a, <2 x i64> %b) {
; CHECK-LABEL: test_vqrshrun_high_n_s64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqrshrun2 v0.4s, v1.2d, #19
; CHECK-NEXT: ret
%1 = bitcast <2 x i32> %a to <1 x i64>
%vqrshrun = tail call <2 x i32> @llvm.aarch64.neon.sqrshrun.v2i32(<2 x i64> %b, i32 19)
%2 = bitcast <2 x i32> %vqrshrun to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <4 x i32>
ret <4 x i32> %3
}
define <16 x i8> @test_vqshrn_high_n_s16(<8 x i8> %a, <8 x i16> %b) {
; CHECK-LABEL: test_vqshrn_high_n_s16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqshrn2 v0.16b, v1.8h, #3
; CHECK-NEXT: ret
%vqshrn = tail call <8 x i8> @llvm.aarch64.neon.sqshrn.v8i8(<8 x i16> %b, i32 3)
%1 = bitcast <8 x i8> %a to <1 x i64>
%2 = bitcast <8 x i8> %vqshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <16 x i8>
ret <16 x i8> %3
}
define <8 x i16> @test_vqshrn_high_n_s32(<4 x i16> %a, <4 x i32> %b) {
; CHECK-LABEL: test_vqshrn_high_n_s32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqshrn2 v0.8h, v1.4s, #9
; CHECK-NEXT: ret
%vqshrn = tail call <4 x i16> @llvm.aarch64.neon.sqshrn.v4i16(<4 x i32> %b, i32 9)
%1 = bitcast <4 x i16> %a to <1 x i64>
%2 = bitcast <4 x i16> %vqshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <8 x i16>
ret <8 x i16> %3
}
define <4 x i32> @test_vqshrn_high_n_s64(<2 x i32> %a, <2 x i64> %b) {
; CHECK-LABEL: test_vqshrn_high_n_s64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqshrn2 v0.4s, v1.2d, #19
; CHECK-NEXT: ret
%1 = bitcast <2 x i32> %a to <1 x i64>
%vqshrn = tail call <2 x i32> @llvm.aarch64.neon.sqshrn.v2i32(<2 x i64> %b, i32 19)
%2 = bitcast <2 x i32> %vqshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <4 x i32>
ret <4 x i32> %3
}
define <16 x i8> @test_vqshrn_high_n_u16(<8 x i8> %a, <8 x i16> %b) {
; CHECK-LABEL: test_vqshrn_high_n_u16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: uqshrn2 v0.16b, v1.8h, #3
; CHECK-NEXT: ret
%vqshrn = tail call <8 x i8> @llvm.aarch64.neon.uqshrn.v8i8(<8 x i16> %b, i32 3)
%1 = bitcast <8 x i8> %a to <1 x i64>
%2 = bitcast <8 x i8> %vqshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <16 x i8>
ret <16 x i8> %3
}
define <8 x i16> @test_vqshrn_high_n_u32(<4 x i16> %a, <4 x i32> %b) {
; CHECK-LABEL: test_vqshrn_high_n_u32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: uqshrn2 v0.8h, v1.4s, #9
; CHECK-NEXT: ret
%vqshrn = tail call <4 x i16> @llvm.aarch64.neon.uqshrn.v4i16(<4 x i32> %b, i32 9)
%1 = bitcast <4 x i16> %a to <1 x i64>
%2 = bitcast <4 x i16> %vqshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <8 x i16>
ret <8 x i16> %3
}
define <4 x i32> @test_vqshrn_high_n_u64(<2 x i32> %a, <2 x i64> %b) {
; CHECK-LABEL: test_vqshrn_high_n_u64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: uqshrn2 v0.4s, v1.2d, #19
; CHECK-NEXT: ret
%1 = bitcast <2 x i32> %a to <1 x i64>
%vqshrn = tail call <2 x i32> @llvm.aarch64.neon.uqshrn.v2i32(<2 x i64> %b, i32 19)
%2 = bitcast <2 x i32> %vqshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <4 x i32>
ret <4 x i32> %3
}
define <16 x i8> @test_vqrshrn_high_n_s16(<8 x i8> %a, <8 x i16> %b) {
; CHECK-LABEL: test_vqrshrn_high_n_s16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqrshrn2 v0.16b, v1.8h, #3
; CHECK-NEXT: ret
%vqrshrn = tail call <8 x i8> @llvm.aarch64.neon.sqrshrn.v8i8(<8 x i16> %b, i32 3)
%1 = bitcast <8 x i8> %a to <1 x i64>
%2 = bitcast <8 x i8> %vqrshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <16 x i8>
ret <16 x i8> %3
}
define <8 x i16> @test_vqrshrn_high_n_s32(<4 x i16> %a, <4 x i32> %b) {
; CHECK-LABEL: test_vqrshrn_high_n_s32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqrshrn2 v0.8h, v1.4s, #9
; CHECK-NEXT: ret
%vqrshrn = tail call <4 x i16> @llvm.aarch64.neon.sqrshrn.v4i16(<4 x i32> %b, i32 9)
%1 = bitcast <4 x i16> %a to <1 x i64>
%2 = bitcast <4 x i16> %vqrshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <8 x i16>
ret <8 x i16> %3
}
define <4 x i32> @test_vqrshrn_high_n_s64(<2 x i32> %a, <2 x i64> %b) {
; CHECK-LABEL: test_vqrshrn_high_n_s64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: sqrshrn2 v0.4s, v1.2d, #19
; CHECK-NEXT: ret
%1 = bitcast <2 x i32> %a to <1 x i64>
%vqrshrn = tail call <2 x i32> @llvm.aarch64.neon.sqrshrn.v2i32(<2 x i64> %b, i32 19)
%2 = bitcast <2 x i32> %vqrshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <4 x i32>
ret <4 x i32> %3
}
define <16 x i8> @test_vqrshrn_high_n_u16(<8 x i8> %a, <8 x i16> %b) {
; CHECK-LABEL: test_vqrshrn_high_n_u16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: uqrshrn2 v0.16b, v1.8h, #3
; CHECK-NEXT: ret
%vqrshrn = tail call <8 x i8> @llvm.aarch64.neon.uqrshrn.v8i8(<8 x i16> %b, i32 3)
%1 = bitcast <8 x i8> %a to <1 x i64>
%2 = bitcast <8 x i8> %vqrshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <16 x i8>
ret <16 x i8> %3
}
define <8 x i16> @test_vqrshrn_high_n_u32(<4 x i16> %a, <4 x i32> %b) {
; CHECK-LABEL: test_vqrshrn_high_n_u32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: uqrshrn2 v0.8h, v1.4s, #9
; CHECK-NEXT: ret
%vqrshrn = tail call <4 x i16> @llvm.aarch64.neon.uqrshrn.v4i16(<4 x i32> %b, i32 9)
%1 = bitcast <4 x i16> %a to <1 x i64>
%2 = bitcast <4 x i16> %vqrshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <8 x i16>
ret <8 x i16> %3
}
define <4 x i32> @test_vqrshrn_high_n_u64(<2 x i32> %a, <2 x i64> %b) {
; CHECK-LABEL: test_vqrshrn_high_n_u64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: uqrshrn2 v0.4s, v1.2d, #19
; CHECK-NEXT: ret
%1 = bitcast <2 x i32> %a to <1 x i64>
%vqrshrn = tail call <2 x i32> @llvm.aarch64.neon.uqrshrn.v2i32(<2 x i64> %b, i32 19)
%2 = bitcast <2 x i32> %vqrshrn to <1 x i64>
%shuffle.i = shufflevector <1 x i64> %1, <1 x i64> %2, <2 x i32> <i32 0, i32 1>
%3 = bitcast <2 x i64> %shuffle.i to <4 x i32>
ret <4 x i32> %3
}
declare <8 x i8> @llvm.aarch64.neon.sqshrun.v8i8(<8 x i16>, i32)
declare <4 x i16> @llvm.aarch64.neon.sqshrun.v4i16(<4 x i32>, i32)
declare <2 x i32> @llvm.aarch64.neon.sqshrun.v2i32(<2 x i64>, i32)
declare <8 x i8> @llvm.aarch64.neon.rshrn.v8i8(<8 x i16>, i32)
declare <4 x i16> @llvm.aarch64.neon.rshrn.v4i16(<4 x i32>, i32)
declare <2 x i32> @llvm.aarch64.neon.rshrn.v2i32(<2 x i64>, i32)
declare <8 x i8> @llvm.aarch64.neon.sqrshrun.v8i8(<8 x i16>, i32)
declare <4 x i16> @llvm.aarch64.neon.sqrshrun.v4i16(<4 x i32>, i32)
declare <2 x i32> @llvm.aarch64.neon.sqrshrun.v2i32(<2 x i64>, i32)
declare <8 x i8> @llvm.aarch64.neon.sqshrn.v8i8(<8 x i16>, i32)
declare <4 x i16> @llvm.aarch64.neon.sqshrn.v4i16(<4 x i32>, i32)
declare <2 x i32> @llvm.aarch64.neon.sqshrn.v2i32(<2 x i64>, i32)
declare <8 x i8> @llvm.aarch64.neon.uqshrn.v8i8(<8 x i16>, i32)
declare <4 x i16> @llvm.aarch64.neon.uqshrn.v4i16(<4 x i32>, i32)
declare <2 x i32> @llvm.aarch64.neon.uqshrn.v2i32(<2 x i64>, i32)
declare <8 x i8> @llvm.aarch64.neon.sqrshrn.v8i8(<8 x i16>, i32)
declare <4 x i16> @llvm.aarch64.neon.sqrshrn.v4i16(<4 x i32>, i32)
declare <2 x i32> @llvm.aarch64.neon.sqrshrn.v2i32(<2 x i64>, i32)
declare <8 x i8> @llvm.aarch64.neon.uqrshrn.v8i8(<8 x i16>, i32)
declare <4 x i16> @llvm.aarch64.neon.uqrshrn.v4i16(<4 x i32>, i32)
declare <2 x i32> @llvm.aarch64.neon.uqrshrn.v2i32(<2 x i64>, i32)
declare <2 x float> @llvm.aarch64.neon.vcvtfxs2fp.v2f32.v2i32(<2 x i32>, i32)
declare <4 x float> @llvm.aarch64.neon.vcvtfxs2fp.v4f32.v4i32(<4 x i32>, i32)
declare <2 x double> @llvm.aarch64.neon.vcvtfxs2fp.v2f64.v2i64(<2 x i64>, i32)
declare <2 x float> @llvm.aarch64.neon.vcvtfxu2fp.v2f32.v2i32(<2 x i32>, i32)
declare <4 x float> @llvm.aarch64.neon.vcvtfxu2fp.v4f32.v4i32(<4 x i32>, i32)
declare <2 x double> @llvm.aarch64.neon.vcvtfxu2fp.v2f64.v2i64(<2 x i64>, i32)
declare <2 x i32> @llvm.aarch64.neon.vcvtfp2fxs.v2i32.v2f32(<2 x float>, i32)
declare <4 x i32> @llvm.aarch64.neon.vcvtfp2fxs.v4i32.v4f32(<4 x float>, i32)
declare <2 x i64> @llvm.aarch64.neon.vcvtfp2fxs.v2i64.v2f64(<2 x double>, i32)
declare <2 x i32> @llvm.aarch64.neon.vcvtfp2fxu.v2i32.v2f32(<2 x float>, i32)
declare <4 x i32> @llvm.aarch64.neon.vcvtfp2fxu.v4i32.v4f32(<4 x float>, i32)
declare <2 x i64> @llvm.aarch64.neon.vcvtfp2fxu.v2i64.v2f64(<2 x double>, i32)
define <1 x i64> @test_vcvt_n_s64_f64(<1 x double> %a) {
; CHECK-LABEL: test_vcvt_n_s64_f64:
; CHECK: // %bb.0:
; CHECK-NEXT: fcvtzs d0, d0, #64
; CHECK-NEXT: ret
%1 = tail call <1 x i64> @llvm.aarch64.neon.vcvtfp2fxs.v1i64.v1f64(<1 x double> %a, i32 64)
ret <1 x i64> %1
}
define <1 x i64> @test_vcvt_n_u64_f64(<1 x double> %a) {
; CHECK-LABEL: test_vcvt_n_u64_f64:
; CHECK: // %bb.0:
; CHECK-NEXT: fcvtzu d0, d0, #64
; CHECK-NEXT: ret
%1 = tail call <1 x i64> @llvm.aarch64.neon.vcvtfp2fxu.v1i64.v1f64(<1 x double> %a, i32 64)
ret <1 x i64> %1
}
define <1 x double> @test_vcvt_n_f64_s64(<1 x i64> %a) {
; CHECK-LABEL: test_vcvt_n_f64_s64:
; CHECK: // %bb.0:
; CHECK-NEXT: scvtf d0, d0, #64
; CHECK-NEXT: ret
%1 = tail call <1 x double> @llvm.aarch64.neon.vcvtfxs2fp.v1f64.v1i64(<1 x i64> %a, i32 64)
ret <1 x double> %1
}
define <1 x double> @test_vcvt_n_f64_u64(<1 x i64> %a) {
; CHECK-LABEL: test_vcvt_n_f64_u64:
; CHECK: // %bb.0:
; CHECK-NEXT: ucvtf d0, d0, #64
; CHECK-NEXT: ret
%1 = tail call <1 x double> @llvm.aarch64.neon.vcvtfxu2fp.v1f64.v1i64(<1 x i64> %a, i32 64)
ret <1 x double> %1
}
declare <1 x i64> @llvm.aarch64.neon.vcvtfp2fxs.v1i64.v1f64(<1 x double>, i32)
declare <1 x i64> @llvm.aarch64.neon.vcvtfp2fxu.v1i64.v1f64(<1 x double>, i32)
declare <1 x double> @llvm.aarch64.neon.vcvtfxs2fp.v1f64.v1i64(<1 x i64>, i32)
declare <1 x double> @llvm.aarch64.neon.vcvtfxu2fp.v1f64.v1i64(<1 x i64>, i32)
Reference in a new issue View git blame Copy permalink