shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
llvm-project/llvm/test/CodeGen/AArch64/sve-fixed-length-extract-subvector.ll
Gaëtan Bossu 9828745661
[AArch64][ISel] Select constructive EXT_ZZI pseudo instruction (#152554) 
The patch adds patterns to select the EXT_ZZI_CONSTRUCTIVE pseudo
instead of the EXT_ZZI destructive instruction for vector_splice. This
only works when the two inputs to vector_splice are identical.

Given that registers aren't tied anymore, this gives the register
allocator more freedom and a lot of MOVs get replaced with MOVPRFX.

In some cases however, we could have just chosen the same input and
output register, but regalloc preferred not to. This means we end up
with some test cases now having more instructions: there is now a
MOVPRFX while no MOV was previously needed.
2025-08-15 14:30:24 +01:00

983 lines
37 KiB
LLVM
Raw Permalink Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -aarch64-sve-vector-bits-min=256 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_256
; RUN: llc -aarch64-sve-vector-bits-min=512 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512
; RUN: llc -aarch64-sve-vector-bits-min=2048 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512
target triple = "aarch64-unknown-linux-gnu"
; Note that both the vector.extract intrinsics and SK_ExtractSubvector
; shufflevector instructions get detected as a extract_subvector ISD node in
; SelectionDAG. We'll test both cases for the sake of completeness, even though
; vector.extract intrinsics should get lowered into shufflevector by the time we
; reach the backend.
; i8
; Don't use SVE for 64-bit vectors.
define <4 x i8> @extract_subvector_v8i8(<8 x i8> %op) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v8i8:
; CHECK: // %bb.0:
; CHECK-NEXT: zip2 v0.8b, v0.8b, v0.8b
; CHECK-NEXT: ret
%ret = call <4 x i8> @llvm.vector.extract.v4i8.v8i8(<8 x i8> %op, i64 4)
ret <4 x i8> %ret
}
; Don't use SVE for 128-bit vectors.
define <8 x i8> @extract_subvector_v16i8(<16 x i8> %op) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v16i8:
; CHECK: // %bb.0:
; CHECK-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-NEXT: ret
%ret = call <8 x i8> @llvm.vector.extract.v8i8.v16i8(<16 x i8> %op, i64 8)
ret <8 x i8> %ret
}
define void @extract_subvector_v32i8(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v32i8:
; CHECK: // %bb.0:
; CHECK-NEXT: ldr q0, [x0, #16]
; CHECK-NEXT: str q0, [x1]
; CHECK-NEXT: ret
%op = load <32 x i8>, ptr %a
%ret = call <16 x i8> @llvm.vector.extract.v16i8.v32i8(<32 x i8> %op, i64 16)
store <16 x i8> %ret, ptr %b
ret void
}
define void @extract_v32i8_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(2,2) {
; CHECK-LABEL: extract_v32i8_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #16
; CHECK-NEXT: str q1, [x1]
; CHECK-NEXT: str q0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <32 x i8>, ptr %in
%hi = shufflevector <32 x i8> %b, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
store <16 x i8> %hi, ptr %out
%lo = shufflevector <32 x i8> %b, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
store <16 x i8> %lo, ptr %out2
ret void
}
define void @extract_v32i8_half_unaligned(ptr %in, ptr %out) #0 vscale_range(2,2) {
; CHECK-LABEL: extract_v32i8_half_unaligned:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #16
; CHECK-NEXT: ext v0.16b, v0.16b, v1.16b, #4
; CHECK-NEXT: str q0, [x1]
; CHECK-NEXT: ret
entry:
%b = load <32 x i8>, ptr %in
%d = shufflevector <32 x i8> %b, <32 x i8> poison, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
store <16 x i8> %d, ptr %out
ret void
}
define void @extract_v32i8_quarters(ptr %in, ptr %out, ptr %out2, ptr %out3, ptr %out4) #0 vscale_range(2,2) {
; CHECK-LABEL: extract_v32i8_quarters:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #16
; CHECK-NEXT: movprfx z2, z0
; CHECK-NEXT: ext z2.b, z2.b, z0.b, #24
; CHECK-NEXT: movprfx z3, z0
; CHECK-NEXT: ext z3.b, z3.b, z0.b, #8
; CHECK-NEXT: str d1, [x1]
; CHECK-NEXT: str d2, [x2]
; CHECK-NEXT: str d0, [x3]
; CHECK-NEXT: str d3, [x4]
; CHECK-NEXT: ret
entry:
%b = load <32 x i8>, ptr %in
%hilo = shufflevector <32 x i8> %b, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
store <8 x i8> %hilo, ptr %out
%hihi = shufflevector <32 x i8> %b, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
store <8 x i8> %hihi, ptr %out2
%lolo = shufflevector <32 x i8> %b, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
store <8 x i8> %lolo, ptr %out3
%lohi = shufflevector <32 x i8> %b, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
store <8 x i8> %lohi, ptr %out4
ret void
}
define void @extract_subvector_v64i8(ptr %a, ptr %b) #0 {
; CHECK-LABEL: extract_subvector_v64i8:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.b, vl32
; CHECK-NEXT: mov w8, #32 // =0x20
; CHECK-NEXT: ld1b { z0.b }, p0/z, [x0, x8]
; CHECK-NEXT: st1b { z0.b }, p0, [x1]
; CHECK-NEXT: ret
%op = load <64 x i8>, ptr %a
%ret = call <32 x i8> @llvm.vector.extract.v32i8.v64i8(<64 x i8> %op, i64 32)
store <32 x i8> %ret, ptr %b
ret void
}
define void @extract_v64i8_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(4,4) {
; CHECK-LABEL: extract_v64i8_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: ptrue p0.b, vl32
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #32
; CHECK-NEXT: st1b { z1.b }, p0, [x1]
; CHECK-NEXT: st1b { z0.b }, p0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <64 x i8>, ptr %in
%hi = shufflevector <64 x i8> %b, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
store <32 x i8> %hi, ptr %out
%lo = shufflevector <64 x i8> %b, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
store <32 x i8> %lo, ptr %out2
ret void
}
define void @extract_subvector_v128i8(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: extract_subvector_v128i8:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.b, vl64
; CHECK-NEXT: mov w8, #64 // =0x40
; CHECK-NEXT: ld1b { z0.b }, p0/z, [x0, x8]
; CHECK-NEXT: st1b { z0.b }, p0, [x1]
; CHECK-NEXT: ret
%op = load <128 x i8>, ptr %a
%ret = call <64 x i8> @llvm.vector.extract.v64i8.v128i8(<128 x i8> %op, i64 64)
store <64 x i8> %ret, ptr %b
ret void
}
define void @extract_subvector_v256i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: extract_subvector_v256i8:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.b, vl128
; CHECK-NEXT: mov w8, #128 // =0x80
; CHECK-NEXT: ld1b { z0.b }, p0/z, [x0, x8]
; CHECK-NEXT: st1b { z0.b }, p0, [x1]
; CHECK-NEXT: ret
%op = load <256 x i8>, ptr %a
%ret = call <128 x i8> @llvm.vector.extract.v128i8.v256i8(<256 x i8> %op, i64 128)
store <128 x i8> %ret, ptr %b
ret void
}
; i16
; Don't use SVE for 64-bit vectors.
define <2 x i16> @extract_subvector_v4i16(<4 x i16> %op) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v4i16:
; CHECK: // %bb.0:
; CHECK-NEXT: zip2 v0.4h, v0.4h, v0.4h
; CHECK-NEXT: ret
%ret = call <2 x i16> @llvm.vector.extract.v2i16.v4i16(<4 x i16> %op, i64 2)
ret <2 x i16> %ret
}
; Don't use SVE for 128-bit vectors.
define <4 x i16> @extract_subvector_v8i16(<8 x i16> %op) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v8i16:
; CHECK: // %bb.0:
; CHECK-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-NEXT: ret
%ret = call <4 x i16> @llvm.vector.extract.v4i16.v8i16(<8 x i16> %op, i64 4)
ret <4 x i16> %ret
}
define void @extract_subvector_v16i16(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v16i16:
; CHECK: // %bb.0:
; CHECK-NEXT: ldr q0, [x0, #16]
; CHECK-NEXT: str q0, [x1]
; CHECK-NEXT: ret
%op = load <16 x i16>, ptr %a
%ret = call <8 x i16> @llvm.vector.extract.v8i16.v16i16(<16 x i16> %op, i64 8)
store <8 x i16> %ret, ptr %b
ret void
}
define void @extract_v16i16_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(2,2) {
; CHECK-LABEL: extract_v16i16_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #16
; CHECK-NEXT: str q1, [x1]
; CHECK-NEXT: str q0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <16 x i16>, ptr %in
%hi = shufflevector <16 x i16> %b, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
store <8 x i16> %hi, ptr %out
%lo = shufflevector <16 x i16> %b, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
store <8 x i16> %lo, ptr %out2
ret void
}
define void @extract_subvector_v32i16(ptr %a, ptr %b) #0 {
; CHECK-LABEL: extract_subvector_v32i16:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.h, vl16
; CHECK-NEXT: mov x8, #16 // =0x10
; CHECK-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1]
; CHECK-NEXT: st1h { z0.h }, p0, [x1]
; CHECK-NEXT: ret
%op = load <32 x i16>, ptr %a
%ret = call <16 x i16> @llvm.vector.extract.v16i16.v32i16(<32 x i16> %op, i64 16)
store <16 x i16> %ret, ptr %b
ret void
}
define void @extract_v32i16_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(4,4) {
; CHECK-LABEL: extract_v32i16_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: ptrue p0.h, vl16
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #32
; CHECK-NEXT: st1h { z1.h }, p0, [x1]
; CHECK-NEXT: st1h { z0.h }, p0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <32 x i16>, ptr %in
%hi = shufflevector <32 x i16> %b, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
store <16 x i16> %hi, ptr %out
%lo = shufflevector <32 x i16> %b, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
store <16 x i16> %lo, ptr %out2
ret void
}
define void @extract_subvector_v64i16(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: extract_subvector_v64i16:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.h, vl32
; CHECK-NEXT: mov x8, #32 // =0x20
; CHECK-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1]
; CHECK-NEXT: st1h { z0.h }, p0, [x1]
; CHECK-NEXT: ret
%op = load <64 x i16>, ptr %a
%ret = call <32 x i16> @llvm.vector.extract.v32i16.v64i16(<64 x i16> %op, i64 32)
store <32 x i16> %ret, ptr %b
ret void
}
define void @extract_subvector_v128i16(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: extract_subvector_v128i16:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.h, vl64
; CHECK-NEXT: mov x8, #64 // =0x40
; CHECK-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1]
; CHECK-NEXT: st1h { z0.h }, p0, [x1]
; CHECK-NEXT: ret
%op = load <128 x i16>, ptr %a
%ret = call <64 x i16> @llvm.vector.extract.v64i16.v128i16(<128 x i16> %op, i64 64)
store <64 x i16> %ret, ptr %b
ret void
}
; i32
; Don't use SVE for 64-bit vectors.
define <1 x i32> @extract_subvector_v2i32(<2 x i32> %op) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v2i32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: dup v0.2s, v0.s[1]
; CHECK-NEXT: ret
%ret = call <1 x i32> @llvm.vector.extract.v1i32.v2i32(<2 x i32> %op, i64 1)
ret <1 x i32> %ret
}
; Don't use SVE for 128-bit vectors.
define <2 x i32> @extract_subvector_v4i32(<4 x i32> %op) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v4i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-NEXT: ret
%ret = call <2 x i32> @llvm.vector.extract.v2i32.v4i32(<4 x i32> %op, i64 2)
ret <2 x i32> %ret
}
define void @extract_subvector_v8i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v8i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ldr q0, [x0, #16]
; CHECK-NEXT: str q0, [x1]
; CHECK-NEXT: ret
%op = load <8 x i32>, ptr %a
%ret = call <4 x i32> @llvm.vector.extract.v4i32.v8i32(<8 x i32> %op, i64 4)
store <4 x i32> %ret, ptr %b
ret void
}
define void @extract_v8i32_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(2,2) {
; CHECK-LABEL: extract_v8i32_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #16
; CHECK-NEXT: str q1, [x1]
; CHECK-NEXT: str q0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <8 x i32>, ptr %in
%hi = shufflevector <8 x i32> %b, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
store <4 x i32> %hi, ptr %out
%lo = shufflevector <8 x i32> %b, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
store <4 x i32> %lo, ptr %out2
ret void
}
define void @extract_subvector_v16i32(ptr %a, ptr %b) #0 {
; CHECK-LABEL: extract_subvector_v16i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.s, vl8
; CHECK-NEXT: mov x8, #8 // =0x8
; CHECK-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
; CHECK-NEXT: ret
%op = load <16 x i32>, ptr %a
%ret = call <8 x i32> @llvm.vector.extract.v8i32.v16i32(<16 x i32> %op, i64 8)
store <8 x i32> %ret, ptr %b
ret void
}
define void @extract_v16i32_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(4,4) {
; CHECK-LABEL: extract_v16i32_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: ptrue p0.s, vl8
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #32
; CHECK-NEXT: st1w { z1.s }, p0, [x1]
; CHECK-NEXT: st1w { z0.s }, p0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <16 x i32>, ptr %in
%hi = shufflevector <16 x i32> %b, <16 x i32> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
store <8 x i32> %hi, ptr %out
%lo = shufflevector <16 x i32> %b, <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
store <8 x i32> %lo, ptr %out2
ret void
}
define void @extract_subvector_v32i32(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: extract_subvector_v32i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.s, vl16
; CHECK-NEXT: mov x8, #16 // =0x10
; CHECK-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
; CHECK-NEXT: ret
%op = load <32 x i32>, ptr %a
%ret = call <16 x i32> @llvm.vector.extract.v16i32.v32i32(<32 x i32> %op, i64 16)
store <16 x i32> %ret, ptr %b
ret void
}
define void @extract_subvector_v64i32(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: extract_subvector_v64i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.s, vl32
; CHECK-NEXT: mov x8, #32 // =0x20
; CHECK-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
; CHECK-NEXT: ret
%op = load <64 x i32>, ptr %a
%ret = call <32 x i32> @llvm.vector.extract.v32i32.v64i32(<64 x i32> %op, i64 32)
store <32 x i32> %ret, ptr %b
ret void
}
; i64
; Don't use SVE for 128-bit vectors.
define <1 x i64> @extract_subvector_v2i64(<2 x i64> %op) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v2i64:
; CHECK: // %bb.0:
; CHECK-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-NEXT: ret
%ret = call <1 x i64> @llvm.vector.extract.v1i64.v2i64(<2 x i64> %op, i64 1)
ret <1 x i64> %ret
}
define void @extract_subvector_v4i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v4i64:
; CHECK: // %bb.0:
; CHECK-NEXT: ldr q0, [x0, #16]
; CHECK-NEXT: str q0, [x1]
; CHECK-NEXT: ret
%op = load <4 x i64>, ptr %a
%ret = call <2 x i64> @llvm.vector.extract.v2i64.v4i64(<4 x i64> %op, i64 2)
store <2 x i64> %ret, ptr %b
ret void
}
define void @extract_v4i64_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(2,2) {
; CHECK-LABEL: extract_v4i64_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #16
; CHECK-NEXT: str q1, [x1]
; CHECK-NEXT: str q0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <4 x i64>, ptr %in
%hi = shufflevector <4 x i64> %b, <4 x i64> poison, <2 x i32> <i32 2, i32 3>
store <2 x i64> %hi, ptr %out
%lo = shufflevector <4 x i64> %b, <4 x i64> poison, <2 x i32> <i32 0, i32 1>
store <2 x i64> %lo, ptr %out2
ret void
}
define void @extract_subvector_v8i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v8i64:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.d, vl4
; CHECK-NEXT: mov x8, #4 // =0x4
; CHECK-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
; CHECK-NEXT: st1d { z0.d }, p0, [x1]
; CHECK-NEXT: ret
%op = load <8 x i64>, ptr %a
%ret = call <4 x i64> @llvm.vector.extract.v4i64.v8i64(<8 x i64> %op, i64 4)
store <4 x i64> %ret, ptr %b
ret void
}
define void @extract_v8i64_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(4,4) {
; CHECK-LABEL: extract_v8i64_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: ptrue p0.d, vl4
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #32
; CHECK-NEXT: st1d { z1.d }, p0, [x1]
; CHECK-NEXT: st1d { z0.d }, p0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <8 x i64>, ptr %in
%hi = shufflevector <8 x i64> %b, <8 x i64> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
store <4 x i64> %hi, ptr %out
%lo = shufflevector <8 x i64> %b, <8 x i64> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
store <4 x i64> %lo, ptr %out2
ret void
}
define void @extract_subvector_v16i64(ptr %a, ptr %b) #0 {
; VBITS_GE_256-LABEL: extract_subvector_v16i64:
; VBITS_GE_256: // %bb.0:
; VBITS_GE_256-NEXT: ptrue p0.d, vl4
; VBITS_GE_256-NEXT: mov x8, #12 // =0xc
; VBITS_GE_256-NEXT: mov x9, #8 // =0x8
; VBITS_GE_256-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
; VBITS_GE_256-NEXT: ld1d { z1.d }, p0/z, [x0, x9, lsl #3]
; VBITS_GE_256-NEXT: mov x8, #4 // =0x4
; VBITS_GE_256-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3]
; VBITS_GE_256-NEXT: st1d { z1.d }, p0, [x1]
; VBITS_GE_256-NEXT: ret
;
; VBITS_GE_512-LABEL: extract_subvector_v16i64:
; VBITS_GE_512: // %bb.0:
; VBITS_GE_512-NEXT: ptrue p0.d, vl8
; VBITS_GE_512-NEXT: mov x8, #8 // =0x8
; VBITS_GE_512-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
; VBITS_GE_512-NEXT: st1d { z0.d }, p0, [x1]
; VBITS_GE_512-NEXT: ret
%op = load <16 x i64>, ptr %a
%ret = call <8 x i64> @llvm.vector.extract.v8i64.v16i64(<16 x i64> %op, i64 8)
store <8 x i64> %ret, ptr %b
ret void
}
define void @extract_subvector_v32i64(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: extract_subvector_v32i64:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.d, vl16
; CHECK-NEXT: mov x8, #16 // =0x10
; CHECK-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
; CHECK-NEXT: st1d { z0.d }, p0, [x1]
; CHECK-NEXT: ret
%op = load <32 x i64>, ptr %a
%ret = call <16 x i64> @llvm.vector.extract.v16i64.v32i64(<32 x i64> %op, i64 16)
store <16 x i64> %ret, ptr %b
ret void
}
; f16
; Don't use SVE for 64-bit vectors.
define <2 x half> @extract_subvector_v4f16(<4 x half> %op) vscale_range(16,0) #0 {
; CHECK-LABEL: extract_subvector_v4f16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: dup v0.2s, v0.s[1]
; CHECK-NEXT: ret
%ret = call <2 x half> @llvm.vector.extract.v2f16.v4f16(<4 x half> %op, i64 2)
ret <2 x half> %ret
}
; Don't use SVE for 128-bit vectors.
define <4 x half> @extract_subvector_v8f16(<8 x half> %op) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v8f16:
; CHECK: // %bb.0:
; CHECK-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-NEXT: ret
%ret = call <4 x half> @llvm.vector.extract.v4f16.v8f16(<8 x half> %op, i64 4)
ret <4 x half> %ret
}
define void @extract_subvector_v16f16(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v16f16:
; CHECK: // %bb.0:
; CHECK-NEXT: ldr q0, [x0, #16]
; CHECK-NEXT: str q0, [x1]
; CHECK-NEXT: ret
%op = load <16 x half>, ptr %a
%ret = call <8 x half> @llvm.vector.extract.v8f16.v16f16(<16 x half> %op, i64 8)
store <8 x half> %ret, ptr %b
ret void
}
define void @extract_v16half_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(2,2) {
; CHECK-LABEL: extract_v16half_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #16
; CHECK-NEXT: str q1, [x1]
; CHECK-NEXT: str q0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <16 x half>, ptr %in
%hi = shufflevector <16 x half> %b, <16 x half> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
store <8 x half> %hi, ptr %out
%lo = shufflevector <16 x half> %b, <16 x half> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
store <8 x half> %lo, ptr %out2
ret void
}
define void @extract_subvector_v32f16(ptr %a, ptr %b) #0 {
; CHECK-LABEL: extract_subvector_v32f16:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.h, vl16
; CHECK-NEXT: mov x8, #16 // =0x10
; CHECK-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1]
; CHECK-NEXT: st1h { z0.h }, p0, [x1]
; CHECK-NEXT: ret
%op = load <32 x half>, ptr %a
%ret = call <16 x half> @llvm.vector.extract.v16f16.v32f16(<32 x half> %op, i64 16)
store <16 x half> %ret, ptr %b
ret void
}
define void @extract_v32half_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(4,4) {
; CHECK-LABEL: extract_v32half_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: ptrue p0.h, vl16
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #32
; CHECK-NEXT: st1h { z1.h }, p0, [x1]
; CHECK-NEXT: st1h { z0.h }, p0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <32 x half>, ptr %in
%hi = shufflevector <32 x half> %b, <32 x half> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
store <16 x half> %hi, ptr %out
%lo = shufflevector <32 x half> %b, <32 x half> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
store <16 x half> %lo, ptr %out2
ret void
}
define void @extract_subvector_v64f16(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: extract_subvector_v64f16:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.h, vl32
; CHECK-NEXT: mov x8, #32 // =0x20
; CHECK-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1]
; CHECK-NEXT: st1h { z0.h }, p0, [x1]
; CHECK-NEXT: ret
%op = load <64 x half>, ptr %a
%ret = call <32 x half> @llvm.vector.extract.v32f16.v64f16(<64 x half> %op, i64 32)
store <32 x half> %ret, ptr %b
ret void
}
define void @extract_subvector_v128f16(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: extract_subvector_v128f16:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.h, vl64
; CHECK-NEXT: mov x8, #64 // =0x40
; CHECK-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1]
; CHECK-NEXT: st1h { z0.h }, p0, [x1]
; CHECK-NEXT: ret
%op = load <128 x half>, ptr %a
%ret = call <64 x half> @llvm.vector.extract.v64f16.v128f16(<128 x half> %op, i64 64)
store <64 x half> %ret, ptr %b
ret void
}
; f32
; Don't use SVE for 64-bit vectors.
define <1 x float> @extract_subvector_v2f32(<2 x float> %op) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v2f32:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: dup v0.2s, v0.s[1]
; CHECK-NEXT: ret
%ret = call <1 x float> @llvm.vector.extract.v1f32.v2f32(<2 x float> %op, i64 1)
ret <1 x float> %ret
}
; Don't use SVE for 128-bit vectors.
define <2 x float> @extract_subvector_v4f32(<4 x float> %op) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v4f32:
; CHECK: // %bb.0:
; CHECK-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-NEXT: ret
%ret = call <2 x float> @llvm.vector.extract.v2f32.v4f32(<4 x float> %op, i64 2)
ret <2 x float> %ret
}
define void @extract_subvector_v8f32(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v8f32:
; CHECK: // %bb.0:
; CHECK-NEXT: ldr q0, [x0, #16]
; CHECK-NEXT: str q0, [x1]
; CHECK-NEXT: ret
%op = load <8 x float>, ptr %a
%ret = call <4 x float> @llvm.vector.extract.v4f32.v8f32(<8 x float> %op, i64 4)
store <4 x float> %ret, ptr %b
ret void
}
define void @extract_v8float_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(2,2) {
; CHECK-LABEL: extract_v8float_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #16
; CHECK-NEXT: str q1, [x1]
; CHECK-NEXT: str q0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <8 x float>, ptr %in
%hi = shufflevector <8 x float> %b, <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
store <4 x float> %hi, ptr %out
%lo = shufflevector <8 x float> %b, <8 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
store <4 x float> %lo, ptr %out2
ret void
}
define void @extract_subvector_v16f32(ptr %a, ptr %b) #0 {
; CHECK-LABEL: extract_subvector_v16f32:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.s, vl8
; CHECK-NEXT: mov x8, #8 // =0x8
; CHECK-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
; CHECK-NEXT: ret
%op = load <16 x float>, ptr %a
%ret = call <8 x float> @llvm.vector.extract.v8f32.v16f32(<16 x float> %op, i64 8)
store <8 x float> %ret, ptr %b
ret void
}
define void @extract_v16float_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(4,4) {
; CHECK-LABEL: extract_v16float_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: ptrue p0.s, vl8
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #32
; CHECK-NEXT: st1w { z1.s }, p0, [x1]
; CHECK-NEXT: st1w { z0.s }, p0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <16 x float>, ptr %in
%hi = shufflevector <16 x float> %b, <16 x float> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
store <8 x float> %hi, ptr %out
%lo = shufflevector <16 x float> %b, <16 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
store <8 x float> %lo, ptr %out2
ret void
}
define void @extract_subvector_v32f32(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: extract_subvector_v32f32:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.s, vl16
; CHECK-NEXT: mov x8, #16 // =0x10
; CHECK-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
; CHECK-NEXT: ret
%op = load <32 x float>, ptr %a
%ret = call <16 x float> @llvm.vector.extract.v16f32.v32f32(<32 x float> %op, i64 16)
store <16 x float> %ret, ptr %b
ret void
}
define void @extract_subvector_v64f32(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: extract_subvector_v64f32:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.s, vl32
; CHECK-NEXT: mov x8, #32 // =0x20
; CHECK-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
; CHECK-NEXT: st1w { z0.s }, p0, [x1]
; CHECK-NEXT: ret
%op = load <64 x float>, ptr %a
%ret = call <32 x float> @llvm.vector.extract.v32f32.v64f32(<64 x float> %op, i64 32)
store <32 x float> %ret, ptr %b
ret void
}
; f64
; Don't use SVE for 128-bit vectors.
define <1 x double> @extract_subvector_v2f64(<2 x double> %op) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v2f64:
; CHECK: // %bb.0:
; CHECK-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-NEXT: ret
%ret = call <1 x double> @llvm.vector.extract.v1f64.v2f64(<2 x double> %op, i64 1)
ret <1 x double> %ret
}
define void @extract_subvector_v4f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: extract_subvector_v4f64:
; CHECK: // %bb.0:
; CHECK-NEXT: ldr q0, [x0, #16]
; CHECK-NEXT: str q0, [x1]
; CHECK-NEXT: ret
%op = load <4 x double>, ptr %a
%ret = call <2 x double> @llvm.vector.extract.v2f64.v4f64(<4 x double> %op, i64 2)
store <2 x double> %ret, ptr %b
ret void
}
define void @extract_v4double_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(2,2) {
; CHECK-LABEL: extract_v4double_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #16
; CHECK-NEXT: str q1, [x1]
; CHECK-NEXT: str q0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <4 x double>, ptr %in
%hi = shufflevector <4 x double> %b, <4 x double> poison, <2 x i32> <i32 2, i32 3>
store <2 x double> %hi, ptr %out
%lo = shufflevector <4 x double> %b, <4 x double> poison, <2 x i32> <i32 0, i32 1>
store <2 x double> %lo, ptr %out2
ret void
}
define void @extract_subvector_v8f64(ptr %a, ptr %b) #0 {
; CHECK-LABEL: extract_subvector_v8f64:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.d, vl4
; CHECK-NEXT: mov x8, #4 // =0x4
; CHECK-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
; CHECK-NEXT: st1d { z0.d }, p0, [x1]
; CHECK-NEXT: ret
%op = load <8 x double>, ptr %a
%ret = call <4 x double> @llvm.vector.extract.v4f64.v8f64(<8 x double> %op, i64 4)
store <4 x double> %ret, ptr %b
ret void
}
define void @extract_v8double_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(4,4) {
; CHECK-LABEL: extract_v8double_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr z0, [x0]
; CHECK-NEXT: ptrue p0.d, vl4
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #32
; CHECK-NEXT: st1d { z1.d }, p0, [x1]
; CHECK-NEXT: st1d { z0.d }, p0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <8 x double>, ptr %in
%hi = shufflevector <8 x double> %b, <8 x double> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
store <4 x double> %hi, ptr %out
%lo = shufflevector <8 x double> %b, <8 x double> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
store <4 x double> %lo, ptr %out2
ret void
}
define void @extract_subvector_v16f64(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: extract_subvector_v16f64:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.d, vl8
; CHECK-NEXT: mov x8, #8 // =0x8
; CHECK-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
; CHECK-NEXT: st1d { z0.d }, p0, [x1]
; CHECK-NEXT: ret
%op = load <16 x double>, ptr %a
%ret = call <8 x double> @llvm.vector.extract.v8f64.v16f64(<16 x double> %op, i64 8)
store <8 x double> %ret, ptr %b
ret void
}
define void @extract_subvector_v32f64(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: extract_subvector_v32f64:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.d, vl16
; CHECK-NEXT: mov x8, #16 // =0x10
; CHECK-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
; CHECK-NEXT: st1d { z0.d }, p0, [x1]
; CHECK-NEXT: ret
%op = load <32 x double>, ptr %a
%ret = call <16 x double> @llvm.vector.extract.v16f64.v32f64(<32 x double> %op, i64 16)
store <16 x double> %ret, ptr %b
ret void
}
; bf16
define void @extract_v8bfloat_halves(ptr %in, ptr %out, ptr %out2) #0 {
; CHECK-LABEL: extract_v8bfloat_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr q0, [x0]
; CHECK-NEXT: ext v1.16b, v0.16b, v0.16b, #8
; CHECK-NEXT: str d1, [x1]
; CHECK-NEXT: str d0, [x2]
; CHECK-NEXT: ret
entry:
%b = load <8 x bfloat>, ptr %in
%hi = shufflevector <8 x bfloat> %b, <8 x bfloat> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
store <4 x bfloat> %hi, ptr %out
%lo = shufflevector <8 x bfloat> %b, <8 x bfloat> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
store <4 x bfloat> %lo, ptr %out2
ret void
}
define void @extract_v16bfloat_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(2,2) {
; CHECK-LABEL: extract_v16bfloat_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldp q1, q0, [x0]
; CHECK-NEXT: str q0, [x1]
; CHECK-NEXT: str q1, [x2]
; CHECK-NEXT: ret
entry:
%b = load <16 x bfloat>, ptr %in
%hi = shufflevector <16 x bfloat> %b, <16 x bfloat> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
store <8 x bfloat> %hi, ptr %out
%lo = shufflevector <16 x bfloat> %b, <16 x bfloat> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
store <8 x bfloat> %lo, ptr %out2
ret void
}
define void @extract_v32bfloat_halves(ptr %in, ptr %out, ptr %out2) #0 vscale_range(4,4) {
; CHECK-LABEL: extract_v32bfloat_halves:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldp q0, q1, [x0, #32]
; CHECK-NEXT: ldp q3, q2, [x0]
; CHECK-NEXT: stp q0, q1, [x1]
; CHECK-NEXT: stp q3, q2, [x2]
; CHECK-NEXT: ret
entry:
%b = load <32 x bfloat>, ptr %in
%hi = shufflevector <32 x bfloat> %b, <32 x bfloat> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
store <16 x bfloat> %hi, ptr %out
%lo = shufflevector <32 x bfloat> %b, <32 x bfloat> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
store <16 x bfloat> %lo, ptr %out2
ret void
}
; Test for infinite loop due to fold:
; extract_subvector(insert_subvector(x,y,c1),c2)--> extract_subvector(y,c2-c1)
define void @extract_subvector_legalization_v8i32() vscale_range(2,2) #0 {
; CHECK-LABEL: extract_subvector_legalization_v8i32:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: adrp x8, .LCPI59_0
; CHECK-NEXT: add x8, x8, :lo12:.LCPI59_0
; CHECK-NEXT: ptrue p1.d
; CHECK-NEXT: ldr z0, [x8]
; CHECK-NEXT: movprfx z1, z0
; CHECK-NEXT: ext z1.b, z1.b, z0.b, #16
; CHECK-NEXT: cmeq v0.4s, v0.4s, #0
; CHECK-NEXT: cmeq v1.4s, v1.4s, #0
; CHECK-NEXT: sunpklo z0.d, z0.s
; CHECK-NEXT: sunpklo z1.d, z1.s
; CHECK-NEXT: cmpne p0.d, p1/z, z1.d, #0
; CHECK-NEXT: cmpne p1.d, p1/z, z0.d, #0
; CHECK-NEXT: .LBB59_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: st1d { z0.d }, p1, [x8]
; CHECK-NEXT: st1d { z0.d }, p0, [x8]
; CHECK-NEXT: b .LBB59_1
entry:
%splat = shufflevector <8 x i32> poison, <8 x i32> poison, <8 x i32> zeroinitializer
br label %body
body:
%0 = icmp eq <8 x i32> zeroinitializer, %splat
tail call void @llvm.masked.store.v8f64.p0(<8 x double> poison, ptr poison, i32 8, <8 x i1> %0)
br label %body
}
declare void @llvm.masked.store.v8f64.p0(<8 x double>, ptr nocapture, i32 immarg, <8 x i1>)
declare <4 x i8> @llvm.vector.extract.v4i8.v8i8(<8 x i8>, i64)
declare <8 x i8> @llvm.vector.extract.v8i8.v16i8(<16 x i8>, i64)
declare <16 x i8> @llvm.vector.extract.v16i8.v32i8(<32 x i8>, i64)
declare <32 x i8> @llvm.vector.extract.v32i8.v64i8(<64 x i8>, i64)
declare <64 x i8> @llvm.vector.extract.v64i8.v128i8(<128 x i8>, i64)
declare <128 x i8> @llvm.vector.extract.v128i8.v256i8(<256 x i8>, i64)
declare <2 x i16> @llvm.vector.extract.v2i16.v4i16(<4 x i16>, i64)
declare <4 x i16> @llvm.vector.extract.v4i16.v8i16(<8 x i16>, i64)
declare <8 x i16> @llvm.vector.extract.v8i16.v16i16(<16 x i16>, i64)
declare <16 x i16> @llvm.vector.extract.v16i16.v32i16(<32 x i16>, i64)
declare <32 x i16> @llvm.vector.extract.v32i16.v64i16(<64 x i16>, i64)
declare <64 x i16> @llvm.vector.extract.v64i16.v128i16(<128 x i16>, i64)
declare <1 x i32> @llvm.vector.extract.v1i32.v2i32(<2 x i32>, i64)
declare <2 x i32> @llvm.vector.extract.v2i32.v4i32(<4 x i32>, i64)
declare <4 x i32> @llvm.vector.extract.v4i32.v8i32(<8 x i32>, i64)
declare <8 x i32> @llvm.vector.extract.v8i32.v16i32(<16 x i32>, i64)
declare <16 x i32> @llvm.vector.extract.v16i32.v32i32(<32 x i32>, i64)
declare <32 x i32> @llvm.vector.extract.v32i32.v64i32(<64 x i32>, i64)
declare <1 x i64> @llvm.vector.extract.v1i64.v2i64(<2 x i64>, i64)
declare <2 x i64> @llvm.vector.extract.v2i64.v4i64(<4 x i64>, i64)
declare <4 x i64> @llvm.vector.extract.v4i64.v8i64(<8 x i64>, i64)
declare <8 x i64> @llvm.vector.extract.v8i64.v16i64(<16 x i64>, i64)
declare <16 x i64> @llvm.vector.extract.v16i64.v32i64(<32 x i64>, i64)
declare <2 x half> @llvm.vector.extract.v2f16.v4f16(<4 x half>, i64)
declare <4 x half> @llvm.vector.extract.v4f16.v8f16(<8 x half>, i64)
declare <8 x half> @llvm.vector.extract.v8f16.v16f16(<16 x half>, i64)
declare <16 x half> @llvm.vector.extract.v16f16.v32f16(<32 x half>, i64)
declare <32 x half> @llvm.vector.extract.v32f16.v64f16(<64 x half>, i64)
declare <64 x half> @llvm.vector.extract.v64f16.v128f16(<128 x half>, i64)
declare <1 x float> @llvm.vector.extract.v1f32.v2f32(<2 x float>, i64)
declare <2 x float> @llvm.vector.extract.v2f32.v4f32(<4 x float>, i64)
declare <4 x float> @llvm.vector.extract.v4f32.v8f32(<8 x float>, i64)
declare <8 x float> @llvm.vector.extract.v8f32.v16f32(<16 x float>, i64)
declare <16 x float> @llvm.vector.extract.v16f32.v32f32(<32 x float>, i64)
declare <32 x float> @llvm.vector.extract.v32f32.v64f32(<64 x float>, i64)
declare <1 x double> @llvm.vector.extract.v1f64.v2f64(<2 x double>, i64)
declare <2 x double> @llvm.vector.extract.v2f64.v4f64(<4 x double>, i64)
declare <4 x double> @llvm.vector.extract.v4f64.v8f64(<8 x double>, i64)
declare <8 x double> @llvm.vector.extract.v8f64.v16f64(<16 x double>, i64)
declare <16 x double> @llvm.vector.extract.v16f64.v32f64(<32 x double>, i64)
attributes #0 = { "target-features"="+sve" }
Reference in New Issue View Git Blame Copy Permalink