3e47d8deba
When a node is picked, we should reset its next cluster candidate to null before releasing its successors/predecessors.
2526 lines
89 KiB
LLVM
2526 lines
89 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mattr=+sve -force-streaming-compatible < %s | FileCheck %s
|
|
; RUN: llc -mattr=+sme -force-streaming < %s | FileCheck %s
|
|
; RUN: llc -force-streaming-compatible < %s | FileCheck %s --check-prefix=NONEON-NOSVE
|
|
|
|
target triple = "aarch64-unknown-linux-gnu"
|
|
|
|
;
|
|
; UADDV
|
|
;
|
|
|
|
define i8 @uaddv_v8i8(<8 x i8> %a) {
|
|
; CHECK-LABEL: uaddv_v8i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.b, vl8
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: uaddv d0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uaddv_v8i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldrb w12, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: ldrb w13, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrb w14, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: add w8, w9, w8
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: add w12, w13, w12
|
|
; NONEON-NOSVE-NEXT: add w10, w11, w10
|
|
; NONEON-NOSVE-NEXT: add w10, w12, w10
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w14
|
|
; NONEON-NOSVE-NEXT: add w8, w10, w8
|
|
; NONEON-NOSVE-NEXT: add w0, w8, w9
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i8 @llvm.vector.reduce.add.v8i8(<8 x i8> %a)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i8 @uaddv_v16i8(<16 x i8> %a) {
|
|
; CHECK-LABEL: uaddv_v16i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uaddv d0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uaddv_v16i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #5]
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #1]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp]
|
|
; NONEON-NOSVE-NEXT: ldrb w12, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: ldrb w13, [sp, #3]
|
|
; NONEON-NOSVE-NEXT: ldrb w14, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: add w10, w11, w10
|
|
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldrb w15, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add w11, w14, w13
|
|
; NONEON-NOSVE-NEXT: add w9, w12, w9
|
|
; NONEON-NOSVE-NEXT: ldrb w16, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: ldrb w12, [sp, #7]
|
|
; NONEON-NOSVE-NEXT: add w10, w10, w11
|
|
; NONEON-NOSVE-NEXT: ldrb w13, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w16
|
|
; NONEON-NOSVE-NEXT: ldrb w14, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: add w12, w12, w15
|
|
; NONEON-NOSVE-NEXT: ldrb w15, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: add w8, w13, w8
|
|
; NONEON-NOSVE-NEXT: ldrb w16, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldrb w13, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: add w12, w12, w14
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w11
|
|
; NONEON-NOSVE-NEXT: add w9, w10, w9
|
|
; NONEON-NOSVE-NEXT: add w10, w12, w16
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w15
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w10
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w13
|
|
; NONEON-NOSVE-NEXT: add w0, w9, w8
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %a)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i8 @uaddv_v32i8(ptr %a) {
|
|
; CHECK-LABEL: uaddv_v32i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: add z0.b, z1.b, z0.b
|
|
; CHECK-NEXT: uaddv d0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uaddv_v32i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #17]
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #1]
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp]
|
|
; NONEON-NOSVE-NEXT: ldrb w12, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldrb w13, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: add w8, w9, w8
|
|
; NONEON-NOSVE-NEXT: ldrb w14, [sp, #19]
|
|
; NONEON-NOSVE-NEXT: ldrb w15, [sp, #3]
|
|
; NONEON-NOSVE-NEXT: add w9, w11, w10
|
|
; NONEON-NOSVE-NEXT: add w10, w13, w12
|
|
; NONEON-NOSVE-NEXT: ldrb w12, [sp, #21]
|
|
; NONEON-NOSVE-NEXT: ldrb w13, [sp, #5]
|
|
; NONEON-NOSVE-NEXT: add w11, w15, w14
|
|
; NONEON-NOSVE-NEXT: add w8, w9, w8
|
|
; NONEON-NOSVE-NEXT: ldrb w16, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldrb w17, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: add w9, w10, w11
|
|
; NONEON-NOSVE-NEXT: add w10, w13, w12
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldrb w12, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: add w14, w17, w16
|
|
; NONEON-NOSVE-NEXT: ldrb w13, [sp, #23]
|
|
; NONEON-NOSVE-NEXT: ldrb w15, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add w10, w14, w10
|
|
; NONEON-NOSVE-NEXT: ldrb w14, [sp, #7]
|
|
; NONEON-NOSVE-NEXT: add w11, w12, w11
|
|
; NONEON-NOSVE-NEXT: ldrb w12, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w9
|
|
; NONEON-NOSVE-NEXT: add w9, w10, w11
|
|
; NONEON-NOSVE-NEXT: add w10, w14, w13
|
|
; NONEON-NOSVE-NEXT: ldrb w13, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w9
|
|
; NONEON-NOSVE-NEXT: add w11, w15, w12
|
|
; NONEON-NOSVE-NEXT: ldrb w12, [sp, #25]
|
|
; NONEON-NOSVE-NEXT: ldrb w14, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: add w9, w10, w11
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #27]
|
|
; NONEON-NOSVE-NEXT: ldrb w15, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: add w10, w13, w12
|
|
; NONEON-NOSVE-NEXT: ldrb w12, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: ldrb w13, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w10
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldrb w16, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: add w11, w12, w11
|
|
; NONEON-NOSVE-NEXT: ldrb w12, [sp, #29]
|
|
; NONEON-NOSVE-NEXT: ldrb w17, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: add w10, w13, w10
|
|
; NONEON-NOSVE-NEXT: ldrb w13, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: add w14, w15, w14
|
|
; NONEON-NOSVE-NEXT: add w10, w11, w10
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: add w9, w9, w14
|
|
; NONEON-NOSVE-NEXT: add w12, w13, w12
|
|
; NONEON-NOSVE-NEXT: ldrb w13, [sp, #31]
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w9
|
|
; NONEON-NOSVE-NEXT: add w10, w10, w12
|
|
; NONEON-NOSVE-NEXT: add w11, w16, w11
|
|
; NONEON-NOSVE-NEXT: add w10, w10, w11
|
|
; NONEON-NOSVE-NEXT: add w11, w17, w13
|
|
; NONEON-NOSVE-NEXT: add w9, w10, w11
|
|
; NONEON-NOSVE-NEXT: add w0, w8, w9
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <32 x i8>, ptr %a
|
|
%res = call i8 @llvm.vector.reduce.add.v32i8(<32 x i8> %op)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i16 @uaddv_v4i16(<4 x i16> %a) {
|
|
; CHECK-LABEL: uaddv_v4i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: uaddv d0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uaddv_v4i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add w8, w9, w8
|
|
; NONEON-NOSVE-NEXT: add w10, w11, w10
|
|
; NONEON-NOSVE-NEXT: add w0, w10, w8
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> %a)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i16 @uaddv_v8i16(<8 x i16> %a) {
|
|
; CHECK-LABEL: uaddv_v8i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uaddv d0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uaddv_v8i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: ldrh w12, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: ldrh w13, [sp]
|
|
; NONEON-NOSVE-NEXT: add w8, w9, w8
|
|
; NONEON-NOSVE-NEXT: ldrh w14, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: add w12, w13, w12
|
|
; NONEON-NOSVE-NEXT: add w10, w11, w10
|
|
; NONEON-NOSVE-NEXT: add w10, w12, w10
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w14
|
|
; NONEON-NOSVE-NEXT: add w8, w10, w8
|
|
; NONEON-NOSVE-NEXT: add w0, w8, w9
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> %a)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i16 @uaddv_v16i16(ptr %a) {
|
|
; CHECK-LABEL: uaddv_v16i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: add z0.h, z1.h, z0.h
|
|
; CHECK-NEXT: uaddv d0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uaddv_v16i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp]
|
|
; NONEON-NOSVE-NEXT: ldrh w12, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldrh w13, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: add w8, w9, w8
|
|
; NONEON-NOSVE-NEXT: ldrh w14, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldrh w15, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: add w9, w11, w10
|
|
; NONEON-NOSVE-NEXT: add w12, w13, w12
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: add w13, w15, w14
|
|
; NONEON-NOSVE-NEXT: add w8, w9, w8
|
|
; NONEON-NOSVE-NEXT: ldrh w16, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldrh w17, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add w9, w12, w13
|
|
; NONEON-NOSVE-NEXT: add w10, w11, w10
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldrh w15, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: add w14, w17, w16
|
|
; NONEON-NOSVE-NEXT: ldrh w12, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldrh w13, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: add w10, w14, w10
|
|
; NONEON-NOSVE-NEXT: add w11, w15, w11
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w9
|
|
; NONEON-NOSVE-NEXT: add w9, w10, w11
|
|
; NONEON-NOSVE-NEXT: add w8, w8, w9
|
|
; NONEON-NOSVE-NEXT: add w9, w13, w12
|
|
; NONEON-NOSVE-NEXT: add w0, w8, w9
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x i16>, ptr %a
|
|
%res = call i16 @llvm.vector.reduce.add.v16i16(<16 x i16> %op)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i32 @uaddv_v2i32(<2 x i32> %a) {
|
|
; CHECK-LABEL: uaddv_v2i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: uaddv d0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uaddv_v2i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldp w9, w8, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add w0, w9, w8
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> %a)
|
|
ret i32 %res
|
|
}
|
|
|
|
define i32 @uaddv_v4i32(<4 x i32> %a) {
|
|
; CHECK-LABEL: uaddv_v4i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uaddv d0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uaddv_v4i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldp w9, w8, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldp w11, w10, [sp], #16
|
|
; NONEON-NOSVE-NEXT: add w10, w11, w10
|
|
; NONEON-NOSVE-NEXT: add w8, w9, w8
|
|
; NONEON-NOSVE-NEXT: add w0, w10, w8
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %a)
|
|
ret i32 %res
|
|
}
|
|
|
|
define i32 @uaddv_v8i32(ptr %a) {
|
|
; CHECK-LABEL: uaddv_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: add z0.s, z1.s, z0.s
|
|
; CHECK-NEXT: uaddv d0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uaddv_v8i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldp w9, w8, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldp w11, w10, [sp]
|
|
; NONEON-NOSVE-NEXT: ldp w12, w13, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldp w14, w15, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add w8, w10, w8
|
|
; NONEON-NOSVE-NEXT: add w9, w11, w9
|
|
; NONEON-NOSVE-NEXT: add w8, w9, w8
|
|
; NONEON-NOSVE-NEXT: add w10, w14, w12
|
|
; NONEON-NOSVE-NEXT: add w11, w15, w13
|
|
; NONEON-NOSVE-NEXT: add w9, w10, w11
|
|
; NONEON-NOSVE-NEXT: add w0, w8, w9
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x i32>, ptr %a
|
|
%res = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> %op)
|
|
ret i32 %res
|
|
}
|
|
|
|
define i64 @uaddv_v2i64(<2 x i64> %a) {
|
|
; CHECK-LABEL: uaddv_v2i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uaddv d0, p0, z0.d
|
|
; CHECK-NEXT: fmov x0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uaddv_v2i64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldp x9, x8, [sp], #16
|
|
; NONEON-NOSVE-NEXT: add x0, x9, x8
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %a)
|
|
ret i64 %res
|
|
}
|
|
|
|
define i64 @uaddv_v4i64(ptr %a) {
|
|
; CHECK-LABEL: uaddv_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: add z0.d, z1.d, z0.d
|
|
; CHECK-NEXT: uaddv d0, p0, z0.d
|
|
; CHECK-NEXT: fmov x0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uaddv_v4i64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldp x9, x8, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldp x11, x10, [sp], #32
|
|
; NONEON-NOSVE-NEXT: add x8, x10, x8
|
|
; NONEON-NOSVE-NEXT: add x9, x11, x9
|
|
; NONEON-NOSVE-NEXT: add x0, x9, x8
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x i64>, ptr %a
|
|
%res = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %op)
|
|
ret i64 %res
|
|
}
|
|
|
|
;
|
|
; SMAXV
|
|
;
|
|
|
|
define i8 @smaxv_v8i8(<8 x i8> %a) {
|
|
; CHECK-LABEL: smaxv_v8i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.b, vl8
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: smaxv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: smaxv_v8i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i8 @llvm.vector.reduce.smax.v8i8(<8 x i8> %a)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i8 @smaxv_v16i8(<16 x i8> %a) {
|
|
; CHECK-LABEL: smaxv_v16i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: smaxv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: smaxv_v16i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #1]
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp]
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #3]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #5]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #7]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i8 @llvm.vector.reduce.smax.v16i8(<16 x i8> %a)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i8 @smaxv_v32i8(ptr %a) {
|
|
; CHECK-LABEL: smaxv_v32i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: smax z0.b, p0/m, z0.b, z1.b
|
|
; CHECK-NEXT: smaxv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: smaxv_v32i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #17]
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #1]
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #19]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #3]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #21]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #5]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #23]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #7]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #25]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #27]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #29]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #31]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <32 x i8>, ptr %a
|
|
%res = call i8 @llvm.vector.reduce.smax.v32i8(<32 x i8> %op)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i16 @smaxv_v4i16(<4 x i16> %a) {
|
|
; CHECK-LABEL: smaxv_v4i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: smaxv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: smaxv_v4i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i16 @llvm.vector.reduce.smax.v4i16(<4 x i16> %a)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i16 @smaxv_v8i16(<8 x i16> %a) {
|
|
; CHECK-LABEL: smaxv_v8i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: smaxv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: smaxv_v8i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp]
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i16 @llvm.vector.reduce.smax.v8i16(<8 x i16> %a)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i16 @smaxv_v16i16(ptr %a) {
|
|
; CHECK-LABEL: smaxv_v16i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: smax z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: smaxv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: smaxv_v16i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x i16>, ptr %a
|
|
%res = call i16 @llvm.vector.reduce.smax.v16i16(<16 x i16> %op)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i32 @smaxv_v2i32(<2 x i32> %a) {
|
|
; CHECK-LABEL: smaxv_v2i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: smaxv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: smaxv_v2i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldp w9, w8, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w0, w9, w8, gt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i32 @llvm.vector.reduce.smax.v2i32(<2 x i32> %a)
|
|
ret i32 %res
|
|
}
|
|
|
|
define i32 @smaxv_v4i32(<4 x i32> %a) {
|
|
; CHECK-LABEL: smaxv_v4i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: smaxv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: smaxv_v4i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldp w9, w8, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, gt
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> %a)
|
|
ret i32 %res
|
|
}
|
|
|
|
define i32 @smaxv_v8i32(ptr %a) {
|
|
; CHECK-LABEL: smaxv_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: smax z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: smaxv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: smaxv_v8i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldp w11, w8, [sp]
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: ldp w11, w12, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, gt
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w10, w11, w10, gt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, gt
|
|
; NONEON-NOSVE-NEXT: cmp w12, w9
|
|
; NONEON-NOSVE-NEXT: csel w9, w12, w9, gt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, gt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x i32>, ptr %a
|
|
%res = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> %op)
|
|
ret i32 %res
|
|
}
|
|
|
|
; No NEON 64-bit vector SMAXV support. Use SVE.
|
|
define i64 @smaxv_v2i64(<2 x i64> %a) {
|
|
; CHECK-LABEL: smaxv_v2i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: smaxv d0, p0, z0.d
|
|
; CHECK-NEXT: fmov x0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: smaxv_v2i64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldp x9, x8, [sp], #16
|
|
; NONEON-NOSVE-NEXT: cmp x9, x8
|
|
; NONEON-NOSVE-NEXT: csel x0, x9, x8, gt
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i64 @llvm.vector.reduce.smax.v2i64(<2 x i64> %a)
|
|
ret i64 %res
|
|
}
|
|
|
|
define i64 @smaxv_v4i64(ptr %a) {
|
|
; CHECK-LABEL: smaxv_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: smax z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: smaxv d0, p0, z0.d
|
|
; CHECK-NEXT: fmov x0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: smaxv_v4i64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldp x8, x10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr x9, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr x11, [sp], #32
|
|
; NONEON-NOSVE-NEXT: cmp x8, x9
|
|
; NONEON-NOSVE-NEXT: csel x8, x8, x9, gt
|
|
; NONEON-NOSVE-NEXT: cmp x11, x10
|
|
; NONEON-NOSVE-NEXT: csel x9, x11, x10, gt
|
|
; NONEON-NOSVE-NEXT: cmp x9, x8
|
|
; NONEON-NOSVE-NEXT: csel x0, x9, x8, gt
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x i64>, ptr %a
|
|
%res = call i64 @llvm.vector.reduce.smax.v4i64(<4 x i64> %op)
|
|
ret i64 %res
|
|
}
|
|
|
|
;
|
|
; SMINV
|
|
;
|
|
|
|
define i8 @sminv_v8i8(<8 x i8> %a) {
|
|
; CHECK-LABEL: sminv_v8i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.b, vl8
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: sminv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: sminv_v8i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i8 @llvm.vector.reduce.smin.v8i8(<8 x i8> %a)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i8 @sminv_v16i8(<16 x i8> %a) {
|
|
; CHECK-LABEL: sminv_v16i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: sminv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: sminv_v16i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #1]
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp]
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #3]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #5]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #7]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i8 @llvm.vector.reduce.smin.v16i8(<16 x i8> %a)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i8 @sminv_v32i8(ptr %a) {
|
|
; CHECK-LABEL: sminv_v32i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: smin z0.b, p0/m, z0.b, z1.b
|
|
; CHECK-NEXT: sminv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: sminv_v32i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #17]
|
|
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #1]
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #19]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #3]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #21]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #5]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #23]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #7]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #25]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #27]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #29]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #31]
|
|
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <32 x i8>, ptr %a
|
|
%res = call i8 @llvm.vector.reduce.smin.v32i8(<32 x i8> %op)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i16 @sminv_v4i16(<4 x i16> %a) {
|
|
; CHECK-LABEL: sminv_v4i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: sminv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: sminv_v4i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i16 @llvm.vector.reduce.smin.v4i16(<4 x i16> %a)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i16 @sminv_v8i16(<8 x i16> %a) {
|
|
; CHECK-LABEL: sminv_v8i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: sminv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: sminv_v8i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp]
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i16 @llvm.vector.reduce.smin.v8i16(<8 x i16> %a)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i16 @sminv_v16i16(ptr %a) {
|
|
; CHECK-LABEL: sminv_v16i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: smin z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: sminv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: sminv_v16i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldrsh w11, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x i16>, ptr %a
|
|
%res = call i16 @llvm.vector.reduce.smin.v16i16(<16 x i16> %op)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i32 @sminv_v2i32(<2 x i32> %a) {
|
|
; CHECK-LABEL: sminv_v2i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: sminv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: sminv_v2i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldp w9, w8, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w0, w9, w8, lt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i32 @llvm.vector.reduce.smin.v2i32(<2 x i32> %a)
|
|
ret i32 %res
|
|
}
|
|
|
|
define i32 @sminv_v4i32(<4 x i32> %a) {
|
|
; CHECK-LABEL: sminv_v4i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: sminv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: sminv_v4i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldp w9, w8, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lt
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> %a)
|
|
ret i32 %res
|
|
}
|
|
|
|
define i32 @sminv_v8i32(ptr %a) {
|
|
; CHECK-LABEL: sminv_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: smin z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: sminv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: sminv_v8i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldp w11, w8, [sp]
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: ldp w11, w12, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lt
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w10, w11, w10, lt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lt
|
|
; NONEON-NOSVE-NEXT: cmp w12, w9
|
|
; NONEON-NOSVE-NEXT: csel w9, w12, w9, lt
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lt
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x i32>, ptr %a
|
|
%res = call i32 @llvm.vector.reduce.smin.v8i32(<8 x i32> %op)
|
|
ret i32 %res
|
|
}
|
|
|
|
; No NEON 64-bit vector SMINV support. Use SVE.
|
|
define i64 @sminv_v2i64(<2 x i64> %a) {
|
|
; CHECK-LABEL: sminv_v2i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: sminv d0, p0, z0.d
|
|
; CHECK-NEXT: fmov x0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: sminv_v2i64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldp x9, x8, [sp], #16
|
|
; NONEON-NOSVE-NEXT: cmp x9, x8
|
|
; NONEON-NOSVE-NEXT: csel x0, x9, x8, lt
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i64 @llvm.vector.reduce.smin.v2i64(<2 x i64> %a)
|
|
ret i64 %res
|
|
}
|
|
|
|
define i64 @sminv_v4i64(ptr %a) {
|
|
; CHECK-LABEL: sminv_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: smin z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: sminv d0, p0, z0.d
|
|
; CHECK-NEXT: fmov x0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: sminv_v4i64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldp x8, x10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr x9, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr x11, [sp], #32
|
|
; NONEON-NOSVE-NEXT: cmp x8, x9
|
|
; NONEON-NOSVE-NEXT: csel x8, x8, x9, lt
|
|
; NONEON-NOSVE-NEXT: cmp x11, x10
|
|
; NONEON-NOSVE-NEXT: csel x9, x11, x10, lt
|
|
; NONEON-NOSVE-NEXT: cmp x9, x8
|
|
; NONEON-NOSVE-NEXT: csel x0, x9, x8, lt
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x i64>, ptr %a
|
|
%res = call i64 @llvm.vector.reduce.smin.v4i64(<4 x i64> %op)
|
|
ret i64 %res
|
|
}
|
|
|
|
;
|
|
; UMAXV
|
|
;
|
|
|
|
define i8 @umaxv_v8i8(<8 x i8> %a) {
|
|
; CHECK-LABEL: umaxv_v8i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.b, vl8
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: umaxv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: umaxv_v8i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i8 @llvm.vector.reduce.umax.v8i8(<8 x i8> %a)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i8 @umaxv_v16i8(<16 x i8> %a) {
|
|
; CHECK-LABEL: umaxv_v16i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: umaxv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: umaxv_v16i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #1]
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp]
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #3]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #5]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #7]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i8 @llvm.vector.reduce.umax.v16i8(<16 x i8> %a)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i8 @umaxv_v32i8(ptr %a) {
|
|
; CHECK-LABEL: umaxv_v32i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: umax z0.b, p0/m, z0.b, z1.b
|
|
; CHECK-NEXT: umaxv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: umaxv_v32i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #17]
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #1]
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #19]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #3]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #21]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #5]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #23]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #7]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #25]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #27]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #29]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #31]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <32 x i8>, ptr %a
|
|
%res = call i8 @llvm.vector.reduce.umax.v32i8(<32 x i8> %op)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i16 @umaxv_v4i16(<4 x i16> %a) {
|
|
; CHECK-LABEL: umaxv_v4i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: umaxv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: umaxv_v4i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i16 @llvm.vector.reduce.umax.v4i16(<4 x i16> %a)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i16 @umaxv_v8i16(<8 x i16> %a) {
|
|
; CHECK-LABEL: umaxv_v8i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: umaxv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: umaxv_v8i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp]
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i16 @llvm.vector.reduce.umax.v8i16(<8 x i16> %a)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i16 @umaxv_v16i16(ptr %a) {
|
|
; CHECK-LABEL: umaxv_v16i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: umax z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: umaxv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: umaxv_v16i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x i16>, ptr %a
|
|
%res = call i16 @llvm.vector.reduce.umax.v16i16(<16 x i16> %op)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i32 @umaxv_v2i32(<2 x i32> %a) {
|
|
; CHECK-LABEL: umaxv_v2i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: umaxv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: umaxv_v2i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldp w9, w8, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w0, w9, w8, hi
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i32 @llvm.vector.reduce.umax.v2i32(<2 x i32> %a)
|
|
ret i32 %res
|
|
}
|
|
|
|
define i32 @umaxv_v4i32(<4 x i32> %a) {
|
|
; CHECK-LABEL: umaxv_v4i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: umaxv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: umaxv_v4i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldp w9, w8, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, hi
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i32 @llvm.vector.reduce.umax.v4i32(<4 x i32> %a)
|
|
ret i32 %res
|
|
}
|
|
|
|
define i32 @umaxv_v8i32(ptr %a) {
|
|
; CHECK-LABEL: umaxv_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: umax z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: umaxv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: umaxv_v8i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldp w11, w8, [sp]
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: ldp w11, w12, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, hi
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w10, w11, w10, hi
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, hi
|
|
; NONEON-NOSVE-NEXT: cmp w12, w9
|
|
; NONEON-NOSVE-NEXT: csel w9, w12, w9, hi
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, hi
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x i32>, ptr %a
|
|
%res = call i32 @llvm.vector.reduce.umax.v8i32(<8 x i32> %op)
|
|
ret i32 %res
|
|
}
|
|
|
|
; No NEON 64-bit vector UMAXV support. Use SVE.
|
|
define i64 @umaxv_v2i64(<2 x i64> %a) {
|
|
; CHECK-LABEL: umaxv_v2i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: umaxv d0, p0, z0.d
|
|
; CHECK-NEXT: fmov x0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: umaxv_v2i64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldp x9, x8, [sp], #16
|
|
; NONEON-NOSVE-NEXT: cmp x9, x8
|
|
; NONEON-NOSVE-NEXT: csel x0, x9, x8, hi
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i64 @llvm.vector.reduce.umax.v2i64(<2 x i64> %a)
|
|
ret i64 %res
|
|
}
|
|
|
|
define i64 @umaxv_v4i64(ptr %a) {
|
|
; CHECK-LABEL: umaxv_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: umax z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: umaxv d0, p0, z0.d
|
|
; CHECK-NEXT: fmov x0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: umaxv_v4i64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldp x8, x10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr x9, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr x11, [sp], #32
|
|
; NONEON-NOSVE-NEXT: cmp x8, x9
|
|
; NONEON-NOSVE-NEXT: csel x8, x8, x9, hi
|
|
; NONEON-NOSVE-NEXT: cmp x11, x10
|
|
; NONEON-NOSVE-NEXT: csel x9, x11, x10, hi
|
|
; NONEON-NOSVE-NEXT: cmp x9, x8
|
|
; NONEON-NOSVE-NEXT: csel x0, x9, x8, hi
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x i64>, ptr %a
|
|
%res = call i64 @llvm.vector.reduce.umax.v4i64(<4 x i64> %op)
|
|
ret i64 %res
|
|
}
|
|
|
|
;
|
|
; UMINV
|
|
;
|
|
|
|
define i8 @uminv_v8i8(<8 x i8> %a) {
|
|
; CHECK-LABEL: uminv_v8i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.b, vl8
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: uminv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uminv_v8i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i8 @llvm.vector.reduce.umin.v8i8(<8 x i8> %a)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i8 @uminv_v16i8(<16 x i8> %a) {
|
|
; CHECK-LABEL: uminv_v16i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uminv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uminv_v16i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #1]
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp]
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #3]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #5]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #7]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i8 @llvm.vector.reduce.umin.v16i8(<16 x i8> %a)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i8 @uminv_v32i8(ptr %a) {
|
|
; CHECK-LABEL: uminv_v32i8:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.b, vl16
|
|
; CHECK-NEXT: umin z0.b, p0/m, z0.b, z1.b
|
|
; CHECK-NEXT: uminv b0, p0, z0.b
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uminv_v32i8:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #17]
|
|
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #1]
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #19]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #3]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #21]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #5]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #23]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #7]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #25]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #9]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #27]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #11]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #29]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #13]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #31]
|
|
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #15]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <32 x i8>, ptr %a
|
|
%res = call i8 @llvm.vector.reduce.umin.v32i8(<32 x i8> %op)
|
|
ret i8 %res
|
|
}
|
|
|
|
define i16 @uminv_v4i16(<4 x i16> %a) {
|
|
; CHECK-LABEL: uminv_v4i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: uminv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uminv_v4i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i16 @llvm.vector.reduce.umin.v4i16(<4 x i16> %a)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i16 @uminv_v8i16(<8 x i16> %a) {
|
|
; CHECK-LABEL: uminv_v8i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uminv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uminv_v8i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp]
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i16 @llvm.vector.reduce.umin.v8i16(<8 x i16> %a)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i16 @uminv_v16i16(ptr %a) {
|
|
; CHECK-LABEL: uminv_v16i16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: umin z0.h, p0/m, z0.h, z1.h
|
|
; CHECK-NEXT: uminv h0, p0, z0.h
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uminv_v16i16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldrh w11, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x i16>, ptr %a
|
|
%res = call i16 @llvm.vector.reduce.umin.v16i16(<16 x i16> %op)
|
|
ret i16 %res
|
|
}
|
|
|
|
define i32 @uminv_v2i32(<2 x i32> %a) {
|
|
; CHECK-LABEL: uminv_v2i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
|
|
; CHECK-NEXT: uminv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uminv_v2i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldp w9, w8, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w0, w9, w8, lo
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i32 @llvm.vector.reduce.umin.v2i32(<2 x i32> %a)
|
|
ret i32 %res
|
|
}
|
|
|
|
define i32 @uminv_v4i32(<4 x i32> %a) {
|
|
; CHECK-LABEL: uminv_v4i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uminv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uminv_v4i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldp w9, w8, [sp]
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lo
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i32 @llvm.vector.reduce.umin.v4i32(<4 x i32> %a)
|
|
ret i32 %res
|
|
}
|
|
|
|
define i32 @uminv_v8i32(ptr %a) {
|
|
; CHECK-LABEL: uminv_v8i32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: umin z0.s, p0/m, z0.s, z1.s
|
|
; CHECK-NEXT: uminv s0, p0, z0.s
|
|
; CHECK-NEXT: fmov w0, s0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uminv_v8i32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldp w11, w8, [sp]
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w9, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: cmp w9, w8
|
|
; NONEON-NOSVE-NEXT: ldp w11, w12, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: csel w8, w9, w8, lo
|
|
; NONEON-NOSVE-NEXT: ldp w10, w9, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: cmp w11, w10
|
|
; NONEON-NOSVE-NEXT: csel w10, w11, w10, lo
|
|
; NONEON-NOSVE-NEXT: cmp w8, w10
|
|
; NONEON-NOSVE-NEXT: csel w8, w8, w10, lo
|
|
; NONEON-NOSVE-NEXT: cmp w12, w9
|
|
; NONEON-NOSVE-NEXT: csel w9, w12, w9, lo
|
|
; NONEON-NOSVE-NEXT: cmp w8, w9
|
|
; NONEON-NOSVE-NEXT: csel w0, w8, w9, lo
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x i32>, ptr %a
|
|
%res = call i32 @llvm.vector.reduce.umin.v8i32(<8 x i32> %op)
|
|
ret i32 %res
|
|
}
|
|
|
|
; No NEON 64-bit vector UMINV support. Use SVE.
|
|
define i64 @uminv_v2i64(<2 x i64> %a) {
|
|
; CHECK-LABEL: uminv_v2i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; CHECK-NEXT: uminv d0, p0, z0.d
|
|
; CHECK-NEXT: fmov x0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uminv_v2i64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldp x9, x8, [sp], #16
|
|
; NONEON-NOSVE-NEXT: cmp x9, x8
|
|
; NONEON-NOSVE-NEXT: csel x0, x9, x8, lo
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call i64 @llvm.vector.reduce.umin.v2i64(<2 x i64> %a)
|
|
ret i64 %res
|
|
}
|
|
|
|
define i64 @uminv_v4i64(ptr %a) {
|
|
; CHECK-LABEL: uminv_v4i64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q1, q0, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: umin z0.d, p0/m, z0.d, z1.d
|
|
; CHECK-NEXT: uminv d0, p0, z0.d
|
|
; CHECK-NEXT: fmov x0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: uminv_v4i64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldp x8, x10, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr x9, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr x11, [sp], #32
|
|
; NONEON-NOSVE-NEXT: cmp x8, x9
|
|
; NONEON-NOSVE-NEXT: csel x8, x8, x9, lo
|
|
; NONEON-NOSVE-NEXT: cmp x11, x10
|
|
; NONEON-NOSVE-NEXT: csel x9, x11, x10, lo
|
|
; NONEON-NOSVE-NEXT: cmp x9, x8
|
|
; NONEON-NOSVE-NEXT: csel x0, x9, x8, lo
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x i64>, ptr %a
|
|
%res = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> %op)
|
|
ret i64 %res
|
|
}
|
|
|
|
declare i8 @llvm.vector.reduce.add.v8i8(<8 x i8>)
|
|
declare i8 @llvm.vector.reduce.add.v16i8(<16 x i8>)
|
|
declare i8 @llvm.vector.reduce.add.v32i8(<32 x i8>)
|
|
|
|
declare i16 @llvm.vector.reduce.add.v4i16(<4 x i16>)
|
|
declare i16 @llvm.vector.reduce.add.v8i16(<8 x i16>)
|
|
declare i16 @llvm.vector.reduce.add.v16i16(<16 x i16>)
|
|
|
|
declare i32 @llvm.vector.reduce.add.v2i32(<2 x i32>)
|
|
declare i32 @llvm.vector.reduce.add.v4i32(<4 x i32>)
|
|
declare i32 @llvm.vector.reduce.add.v8i32(<8 x i32>)
|
|
|
|
declare i64 @llvm.vector.reduce.add.v2i64(<2 x i64>)
|
|
declare i64 @llvm.vector.reduce.add.v4i64(<4 x i64>)
|
|
|
|
declare i8 @llvm.vector.reduce.smax.v8i8(<8 x i8>)
|
|
declare i8 @llvm.vector.reduce.smax.v16i8(<16 x i8>)
|
|
declare i8 @llvm.vector.reduce.smax.v32i8(<32 x i8>)
|
|
|
|
declare i16 @llvm.vector.reduce.smax.v4i16(<4 x i16>)
|
|
declare i16 @llvm.vector.reduce.smax.v8i16(<8 x i16>)
|
|
declare i16 @llvm.vector.reduce.smax.v16i16(<16 x i16>)
|
|
|
|
declare i32 @llvm.vector.reduce.smax.v2i32(<2 x i32>)
|
|
declare i32 @llvm.vector.reduce.smax.v4i32(<4 x i32>)
|
|
declare i32 @llvm.vector.reduce.smax.v8i32(<8 x i32>)
|
|
|
|
declare i64 @llvm.vector.reduce.smax.v2i64(<2 x i64>)
|
|
declare i64 @llvm.vector.reduce.smax.v4i64(<4 x i64>)
|
|
|
|
declare i8 @llvm.vector.reduce.smin.v8i8(<8 x i8>)
|
|
declare i8 @llvm.vector.reduce.smin.v16i8(<16 x i8>)
|
|
declare i8 @llvm.vector.reduce.smin.v32i8(<32 x i8>)
|
|
|
|
declare i16 @llvm.vector.reduce.smin.v4i16(<4 x i16>)
|
|
declare i16 @llvm.vector.reduce.smin.v8i16(<8 x i16>)
|
|
declare i16 @llvm.vector.reduce.smin.v16i16(<16 x i16>)
|
|
|
|
declare i32 @llvm.vector.reduce.smin.v2i32(<2 x i32>)
|
|
declare i32 @llvm.vector.reduce.smin.v4i32(<4 x i32>)
|
|
declare i32 @llvm.vector.reduce.smin.v8i32(<8 x i32>)
|
|
|
|
declare i64 @llvm.vector.reduce.smin.v2i64(<2 x i64>)
|
|
declare i64 @llvm.vector.reduce.smin.v4i64(<4 x i64>)
|
|
|
|
declare i8 @llvm.vector.reduce.umax.v8i8(<8 x i8>)
|
|
declare i8 @llvm.vector.reduce.umax.v16i8(<16 x i8>)
|
|
declare i8 @llvm.vector.reduce.umax.v32i8(<32 x i8>)
|
|
|
|
declare i16 @llvm.vector.reduce.umax.v4i16(<4 x i16>)
|
|
declare i16 @llvm.vector.reduce.umax.v8i16(<8 x i16>)
|
|
declare i16 @llvm.vector.reduce.umax.v16i16(<16 x i16>)
|
|
|
|
declare i32 @llvm.vector.reduce.umax.v2i32(<2 x i32>)
|
|
declare i32 @llvm.vector.reduce.umax.v4i32(<4 x i32>)
|
|
declare i32 @llvm.vector.reduce.umax.v8i32(<8 x i32>)
|
|
|
|
declare i64 @llvm.vector.reduce.umax.v2i64(<2 x i64>)
|
|
declare i64 @llvm.vector.reduce.umax.v4i64(<4 x i64>)
|
|
|
|
declare i8 @llvm.vector.reduce.umin.v8i8(<8 x i8>)
|
|
declare i8 @llvm.vector.reduce.umin.v16i8(<16 x i8>)
|
|
declare i8 @llvm.vector.reduce.umin.v32i8(<32 x i8>)
|
|
|
|
declare i16 @llvm.vector.reduce.umin.v4i16(<4 x i16>)
|
|
declare i16 @llvm.vector.reduce.umin.v8i16(<8 x i16>)
|
|
declare i16 @llvm.vector.reduce.umin.v16i16(<16 x i16>)
|
|
|
|
declare i32 @llvm.vector.reduce.umin.v2i32(<2 x i32>)
|
|
declare i32 @llvm.vector.reduce.umin.v4i32(<4 x i32>)
|
|
declare i32 @llvm.vector.reduce.umin.v8i32(<8 x i32>)
|
|
|
|
declare i64 @llvm.vector.reduce.umin.v2i64(<2 x i64>)
|
|
declare i64 @llvm.vector.reduce.umin.v4i64(<4 x i64>)
|