derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
3319 lines
114 KiB
LLVM
3319 lines
114 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -enable-subreg-liveness -mattr=+sve -force-streaming-compatible < %s | FileCheck %s
|
|
; RUN: llc -enable-subreg-liveness -mattr=+sme -force-streaming < %s | FileCheck %s
|
|
; RUN: llc -enable-subreg-liveness -force-streaming-compatible < %s | FileCheck %s --check-prefix=NONEON-NOSVE
|
|
|
|
target triple = "aarch64-unknown-linux-gnu"
|
|
|
|
;
|
|
; CEIL -> FRINTP
|
|
;
|
|
|
|
define <2 x half> @frintp_v2f16(<2 x half> %op) {
|
|
; CHECK-LABEL: frintp_v2f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frintp z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintp_v2f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x half> @llvm.ceil.v2f16(<2 x half> %op)
|
|
ret <2 x half> %res
|
|
}
|
|
|
|
define <4 x half> @frintp_v4f16(<4 x half> %op) {
|
|
; CHECK-LABEL: frintp_v4f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frintp z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintp_v4f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x half> @llvm.ceil.v4f16(<4 x half> %op)
|
|
ret <4 x half> %res
|
|
}
|
|
|
|
define <8 x half> @frintp_v8f16(<8 x half> %op) {
|
|
; CHECK-LABEL: frintp_v8f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frintp z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintp_v8f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <8 x half> @llvm.ceil.v8f16(<8 x half> %op)
|
|
ret <8 x half> %res
|
|
}
|
|
|
|
define void @frintp_v16f16(ptr %a) {
|
|
; CHECK-LABEL: frintp_v16f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frintp z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: frintp z1.h, p0/m, z1.h
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintp_v16f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #62]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #60]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #58]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #54]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #52]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #50]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #46]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #44]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #42]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #38]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #36]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #34]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x half>, ptr %a
|
|
%res = call <16 x half> @llvm.ceil.v16f16(<16 x half> %op)
|
|
store <16 x half> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x float> @frintp_v2f32(<2 x float> %op) {
|
|
; CHECK-LABEL: frintp_v2f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: frintp z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintp_v2f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintp s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x float> @llvm.ceil.v2f32(<2 x float> %op)
|
|
ret <2 x float> %res
|
|
}
|
|
|
|
define <4 x float> @frintp_v4f32(<4 x float> %op) {
|
|
; CHECK-LABEL: frintp_v4f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frintp z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintp_v4f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frintp s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintp s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x float> @llvm.ceil.v4f32(<4 x float> %op)
|
|
ret <4 x float> %res
|
|
}
|
|
|
|
define void @frintp_v8f32(ptr %a) {
|
|
; CHECK-LABEL: frintp_v8f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frintp z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: frintp z1.s, p0/m, z1.s
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintp_v8f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: frintp s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: frintp s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frintp s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintp s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintp s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x float>, ptr %a
|
|
%res = call <8 x float> @llvm.ceil.v8f32(<8 x float> %op)
|
|
store <8 x float> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <1 x double> @frintp_v1f64(<1 x double> %op) {
|
|
; CHECK-LABEL: frintp_v1f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: frintp d0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintp_v1f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: frintp d0, d0
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <1 x double> @llvm.ceil.v1f64(<1 x double> %op)
|
|
ret <1 x double> %res
|
|
}
|
|
|
|
define <2 x double> @frintp_v2f64(<2 x double> %op) {
|
|
; CHECK-LABEL: frintp_v2f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frintp z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintp_v2f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintp d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintp d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x double> @llvm.ceil.v2f64(<2 x double> %op)
|
|
ret <2 x double> %res
|
|
}
|
|
|
|
define void @frintp_v4f64(ptr %a) {
|
|
; CHECK-LABEL: frintp_v4f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frintp z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: frintp z1.d, p0/m, z1.d
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintp_v4f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frintp d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frintp d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintp d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintp d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x double>, ptr %a
|
|
%res = call <4 x double> @llvm.ceil.v4f64(<4 x double> %op)
|
|
store <4 x double> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; FLOOR -> FRINTM
|
|
;
|
|
|
|
define <2 x half> @frintm_v2f16(<2 x half> %op) {
|
|
; CHECK-LABEL: frintm_v2f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frintm z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintm_v2f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x half> @llvm.floor.v2f16(<2 x half> %op)
|
|
ret <2 x half> %res
|
|
}
|
|
|
|
define <4 x half> @frintm_v4f16(<4 x half> %op) {
|
|
; CHECK-LABEL: frintm_v4f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frintm z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintm_v4f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x half> @llvm.floor.v4f16(<4 x half> %op)
|
|
ret <4 x half> %res
|
|
}
|
|
|
|
define <8 x half> @frintm_v8f16(<8 x half> %op) {
|
|
; CHECK-LABEL: frintm_v8f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frintm z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintm_v8f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <8 x half> @llvm.floor.v8f16(<8 x half> %op)
|
|
ret <8 x half> %res
|
|
}
|
|
|
|
define void @frintm_v16f16(ptr %a) {
|
|
; CHECK-LABEL: frintm_v16f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frintm z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: frintm z1.h, p0/m, z1.h
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintm_v16f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #62]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #60]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #58]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #54]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #52]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #50]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #46]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #44]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #42]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #38]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #36]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #34]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x half>, ptr %a
|
|
%res = call <16 x half> @llvm.floor.v16f16(<16 x half> %op)
|
|
store <16 x half> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x float> @frintm_v2f32(<2 x float> %op) {
|
|
; CHECK-LABEL: frintm_v2f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: frintm z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintm_v2f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintm s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x float> @llvm.floor.v2f32(<2 x float> %op)
|
|
ret <2 x float> %res
|
|
}
|
|
|
|
define <4 x float> @frintm_v4f32(<4 x float> %op) {
|
|
; CHECK-LABEL: frintm_v4f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frintm z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintm_v4f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frintm s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintm s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x float> @llvm.floor.v4f32(<4 x float> %op)
|
|
ret <4 x float> %res
|
|
}
|
|
|
|
define void @frintm_v8f32(ptr %a) {
|
|
; CHECK-LABEL: frintm_v8f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frintm z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: frintm z1.s, p0/m, z1.s
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintm_v8f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: frintm s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: frintm s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frintm s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintm s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintm s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x float>, ptr %a
|
|
%res = call <8 x float> @llvm.floor.v8f32(<8 x float> %op)
|
|
store <8 x float> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <1 x double> @frintm_v1f64(<1 x double> %op) {
|
|
; CHECK-LABEL: frintm_v1f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: frintm d0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintm_v1f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: frintm d0, d0
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <1 x double> @llvm.floor.v1f64(<1 x double> %op)
|
|
ret <1 x double> %res
|
|
}
|
|
|
|
define <2 x double> @frintm_v2f64(<2 x double> %op) {
|
|
; CHECK-LABEL: frintm_v2f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frintm z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintm_v2f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintm d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintm d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x double> @llvm.floor.v2f64(<2 x double> %op)
|
|
ret <2 x double> %res
|
|
}
|
|
|
|
define void @frintm_v4f64(ptr %a) {
|
|
; CHECK-LABEL: frintm_v4f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frintm z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: frintm z1.d, p0/m, z1.d
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintm_v4f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frintm d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frintm d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintm d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintm d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x double>, ptr %a
|
|
%res = call <4 x double> @llvm.floor.v4f64(<4 x double> %op)
|
|
store <4 x double> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; FNEARBYINT -> FRINTI
|
|
;
|
|
|
|
define <2 x half> @frinti_v2f16(<2 x half> %op) {
|
|
; CHECK-LABEL: frinti_v2f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frinti z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinti_v2f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x half> @llvm.nearbyint.v2f16(<2 x half> %op)
|
|
ret <2 x half> %res
|
|
}
|
|
|
|
define <4 x half> @frinti_v4f16(<4 x half> %op) {
|
|
; CHECK-LABEL: frinti_v4f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frinti z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinti_v4f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x half> @llvm.nearbyint.v4f16(<4 x half> %op)
|
|
ret <4 x half> %res
|
|
}
|
|
|
|
define <8 x half> @frinti_v8f16(<8 x half> %op) {
|
|
; CHECK-LABEL: frinti_v8f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frinti z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinti_v8f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <8 x half> @llvm.nearbyint.v8f16(<8 x half> %op)
|
|
ret <8 x half> %res
|
|
}
|
|
|
|
define void @frinti_v16f16(ptr %a) {
|
|
; CHECK-LABEL: frinti_v16f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frinti z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: frinti z1.h, p0/m, z1.h
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinti_v16f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #62]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #60]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #58]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #54]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #52]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #50]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #46]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #44]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #42]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #38]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #36]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #34]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x half>, ptr %a
|
|
%res = call <16 x half> @llvm.nearbyint.v16f16(<16 x half> %op)
|
|
store <16 x half> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x float> @frinti_v2f32(<2 x float> %op) {
|
|
; CHECK-LABEL: frinti_v2f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: frinti z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinti_v2f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frinti s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x float> @llvm.nearbyint.v2f32(<2 x float> %op)
|
|
ret <2 x float> %res
|
|
}
|
|
|
|
define <4 x float> @frinti_v4f32(<4 x float> %op) {
|
|
; CHECK-LABEL: frinti_v4f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frinti z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinti_v4f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frinti s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frinti s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x float> @llvm.nearbyint.v4f32(<4 x float> %op)
|
|
ret <4 x float> %res
|
|
}
|
|
|
|
define void @frinti_v8f32(ptr %a) {
|
|
; CHECK-LABEL: frinti_v8f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frinti z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: frinti z1.s, p0/m, z1.s
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinti_v8f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: frinti s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: frinti s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frinti s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frinti s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frinti s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x float>, ptr %a
|
|
%res = call <8 x float> @llvm.nearbyint.v8f32(<8 x float> %op)
|
|
store <8 x float> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <1 x double> @frinti_v1f64(<1 x double> %op) {
|
|
; CHECK-LABEL: frinti_v1f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: frinti d0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinti_v1f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: frinti d0, d0
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <1 x double> @llvm.nearbyint.v1f64(<1 x double> %op)
|
|
ret <1 x double> %res
|
|
}
|
|
|
|
define <2 x double> @frinti_v2f64(<2 x double> %op) {
|
|
; CHECK-LABEL: frinti_v2f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frinti z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinti_v2f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frinti d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frinti d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x double> @llvm.nearbyint.v2f64(<2 x double> %op)
|
|
ret <2 x double> %res
|
|
}
|
|
|
|
define void @frinti_v4f64(ptr %a) {
|
|
; CHECK-LABEL: frinti_v4f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frinti z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: frinti z1.d, p0/m, z1.d
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinti_v4f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frinti d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frinti d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frinti d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frinti d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x double>, ptr %a
|
|
%res = call <4 x double> @llvm.nearbyint.v4f64(<4 x double> %op)
|
|
store <4 x double> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; RINT -> FRINTX
|
|
;
|
|
|
|
define <2 x half> @frintx_v2f16(<2 x half> %op) {
|
|
; CHECK-LABEL: frintx_v2f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frintx z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintx_v2f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x half> @llvm.rint.v2f16(<2 x half> %op)
|
|
ret <2 x half> %res
|
|
}
|
|
|
|
define <4 x half> @frintx_v4f16(<4 x half> %op) {
|
|
; CHECK-LABEL: frintx_v4f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frintx z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintx_v4f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x half> @llvm.rint.v4f16(<4 x half> %op)
|
|
ret <4 x half> %res
|
|
}
|
|
|
|
define <8 x half> @frintx_v8f16(<8 x half> %op) {
|
|
; CHECK-LABEL: frintx_v8f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frintx z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintx_v8f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <8 x half> @llvm.rint.v8f16(<8 x half> %op)
|
|
ret <8 x half> %res
|
|
}
|
|
|
|
define void @frintx_v16f16(ptr %a) {
|
|
; CHECK-LABEL: frintx_v16f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frintx z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: frintx z1.h, p0/m, z1.h
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintx_v16f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #62]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #60]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #58]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #54]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #52]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #50]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #46]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #44]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #42]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #38]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #36]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #34]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x half>, ptr %a
|
|
%res = call <16 x half> @llvm.rint.v16f16(<16 x half> %op)
|
|
store <16 x half> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x float> @frintx_v2f32(<2 x float> %op) {
|
|
; CHECK-LABEL: frintx_v2f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: frintx z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintx_v2f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintx s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x float> @llvm.rint.v2f32(<2 x float> %op)
|
|
ret <2 x float> %res
|
|
}
|
|
|
|
define <4 x float> @frintx_v4f32(<4 x float> %op) {
|
|
; CHECK-LABEL: frintx_v4f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frintx z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintx_v4f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frintx s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintx s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x float> @llvm.rint.v4f32(<4 x float> %op)
|
|
ret <4 x float> %res
|
|
}
|
|
|
|
define void @frintx_v8f32(ptr %a) {
|
|
; CHECK-LABEL: frintx_v8f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frintx z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: frintx z1.s, p0/m, z1.s
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintx_v8f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: frintx s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: frintx s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frintx s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintx s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintx s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x float>, ptr %a
|
|
%res = call <8 x float> @llvm.rint.v8f32(<8 x float> %op)
|
|
store <8 x float> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <1 x double> @frintx_v1f64(<1 x double> %op) {
|
|
; CHECK-LABEL: frintx_v1f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: frintx d0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintx_v1f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: frintx d0, d0
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <1 x double> @llvm.rint.v1f64(<1 x double> %op)
|
|
ret <1 x double> %res
|
|
}
|
|
|
|
define <2 x double> @frintx_v2f64(<2 x double> %op) {
|
|
; CHECK-LABEL: frintx_v2f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frintx z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintx_v2f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintx d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintx d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x double> @llvm.rint.v2f64(<2 x double> %op)
|
|
ret <2 x double> %res
|
|
}
|
|
|
|
define void @frintx_v4f64(ptr %a) {
|
|
; CHECK-LABEL: frintx_v4f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frintx z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: frintx z1.d, p0/m, z1.d
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintx_v4f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frintx d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frintx d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintx d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintx d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x double>, ptr %a
|
|
%res = call <4 x double> @llvm.rint.v4f64(<4 x double> %op)
|
|
store <4 x double> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; ROUND -> FRINTA
|
|
;
|
|
|
|
define <2 x half> @frinta_v2f16(<2 x half> %op) {
|
|
; CHECK-LABEL: frinta_v2f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frinta z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinta_v2f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x half> @llvm.round.v2f16(<2 x half> %op)
|
|
ret <2 x half> %res
|
|
}
|
|
|
|
define <4 x half> @frinta_v4f16(<4 x half> %op) {
|
|
; CHECK-LABEL: frinta_v4f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frinta z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinta_v4f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x half> @llvm.round.v4f16(<4 x half> %op)
|
|
ret <4 x half> %res
|
|
}
|
|
|
|
define <8 x half> @frinta_v8f16(<8 x half> %op) {
|
|
; CHECK-LABEL: frinta_v8f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frinta z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinta_v8f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <8 x half> @llvm.round.v8f16(<8 x half> %op)
|
|
ret <8 x half> %res
|
|
}
|
|
|
|
define void @frinta_v16f16(ptr %a) {
|
|
; CHECK-LABEL: frinta_v16f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frinta z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: frinta z1.h, p0/m, z1.h
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinta_v16f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #62]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #60]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #58]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #54]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #52]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #50]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #46]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #44]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #42]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #38]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #36]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #34]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x half>, ptr %a
|
|
%res = call <16 x half> @llvm.round.v16f16(<16 x half> %op)
|
|
store <16 x half> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x float> @frinta_v2f32(<2 x float> %op) {
|
|
; CHECK-LABEL: frinta_v2f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: frinta z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinta_v2f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frinta s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x float> @llvm.round.v2f32(<2 x float> %op)
|
|
ret <2 x float> %res
|
|
}
|
|
|
|
define <4 x float> @frinta_v4f32(<4 x float> %op) {
|
|
; CHECK-LABEL: frinta_v4f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frinta z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinta_v4f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frinta s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frinta s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x float> @llvm.round.v4f32(<4 x float> %op)
|
|
ret <4 x float> %res
|
|
}
|
|
|
|
define void @frinta_v8f32(ptr %a) {
|
|
; CHECK-LABEL: frinta_v8f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frinta z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: frinta z1.s, p0/m, z1.s
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinta_v8f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: frinta s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: frinta s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frinta s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frinta s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frinta s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x float>, ptr %a
|
|
%res = call <8 x float> @llvm.round.v8f32(<8 x float> %op)
|
|
store <8 x float> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <1 x double> @frinta_v1f64(<1 x double> %op) {
|
|
; CHECK-LABEL: frinta_v1f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: frinta d0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinta_v1f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: frinta d0, d0
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <1 x double> @llvm.round.v1f64(<1 x double> %op)
|
|
ret <1 x double> %res
|
|
}
|
|
|
|
define <2 x double> @frinta_v2f64(<2 x double> %op) {
|
|
; CHECK-LABEL: frinta_v2f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frinta z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinta_v2f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frinta d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frinta d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x double> @llvm.round.v2f64(<2 x double> %op)
|
|
ret <2 x double> %res
|
|
}
|
|
|
|
define void @frinta_v4f64(ptr %a) {
|
|
; CHECK-LABEL: frinta_v4f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frinta z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: frinta z1.d, p0/m, z1.d
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frinta_v4f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frinta d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frinta d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frinta d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frinta d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x double>, ptr %a
|
|
%res = call <4 x double> @llvm.round.v4f64(<4 x double> %op)
|
|
store <4 x double> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; ROUNDEVEN -> FRINTN
|
|
;
|
|
|
|
define <2 x half> @frintn_v2f16(<2 x half> %op) {
|
|
; CHECK-LABEL: frintn_v2f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frintn z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintn_v2f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x half> @llvm.roundeven.v2f16(<2 x half> %op)
|
|
ret <2 x half> %res
|
|
}
|
|
|
|
define <4 x half> @frintn_v4f16(<4 x half> %op) {
|
|
; CHECK-LABEL: frintn_v4f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frintn z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintn_v4f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x half> @llvm.roundeven.v4f16(<4 x half> %op)
|
|
ret <4 x half> %res
|
|
}
|
|
|
|
define <8 x half> @frintn_v8f16(<8 x half> %op) {
|
|
; CHECK-LABEL: frintn_v8f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frintn z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintn_v8f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <8 x half> @llvm.roundeven.v8f16(<8 x half> %op)
|
|
ret <8 x half> %res
|
|
}
|
|
|
|
define void @frintn_v16f16(ptr %a) {
|
|
; CHECK-LABEL: frintn_v16f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frintn z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: frintn z1.h, p0/m, z1.h
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintn_v16f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #62]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #60]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #58]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #54]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #52]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #50]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #46]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #44]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #42]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #38]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #36]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #34]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x half>, ptr %a
|
|
%res = call <16 x half> @llvm.roundeven.v16f16(<16 x half> %op)
|
|
store <16 x half> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x float> @frintn_v2f32(<2 x float> %op) {
|
|
; CHECK-LABEL: frintn_v2f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: frintn z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintn_v2f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintn s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x float> @llvm.roundeven.v2f32(<2 x float> %op)
|
|
ret <2 x float> %res
|
|
}
|
|
|
|
define <4 x float> @frintn_v4f32(<4 x float> %op) {
|
|
; CHECK-LABEL: frintn_v4f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frintn z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintn_v4f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frintn s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintn s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x float> @llvm.roundeven.v4f32(<4 x float> %op)
|
|
ret <4 x float> %res
|
|
}
|
|
|
|
define void @frintn_v8f32(ptr %a) {
|
|
; CHECK-LABEL: frintn_v8f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frintn z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: frintn z1.s, p0/m, z1.s
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintn_v8f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: frintn s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: frintn s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frintn s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintn s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintn s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x float>, ptr %a
|
|
%res = call <8 x float> @llvm.roundeven.v8f32(<8 x float> %op)
|
|
store <8 x float> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <1 x double> @frintn_v1f64(<1 x double> %op) {
|
|
; CHECK-LABEL: frintn_v1f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: frintn d0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintn_v1f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: frintn d0, d0
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <1 x double> @llvm.roundeven.v1f64(<1 x double> %op)
|
|
ret <1 x double> %res
|
|
}
|
|
|
|
define <2 x double> @frintn_v2f64(<2 x double> %op) {
|
|
; CHECK-LABEL: frintn_v2f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frintn z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintn_v2f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintn d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintn d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x double> @llvm.roundeven.v2f64(<2 x double> %op)
|
|
ret <2 x double> %res
|
|
}
|
|
|
|
define void @frintn_v4f64(ptr %a) {
|
|
; CHECK-LABEL: frintn_v4f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frintn z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: frintn z1.d, p0/m, z1.d
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintn_v4f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frintn d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frintn d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintn d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintn d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x double>, ptr %a
|
|
%res = call <4 x double> @llvm.roundeven.v4f64(<4 x double> %op)
|
|
store <4 x double> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; TRUNC -> FRINTZ
|
|
;
|
|
|
|
define <2 x half> @frintz_v2f16(<2 x half> %op) {
|
|
; CHECK-LABEL: frintz_v2f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frintz z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintz_v2f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x half> @llvm.trunc.v2f16(<2 x half> %op)
|
|
ret <2 x half> %res
|
|
}
|
|
|
|
define <4 x half> @frintz_v4f16(<4 x half> %op) {
|
|
; CHECK-LABEL: frintz_v4f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: frintz z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintz_v4f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x half> @llvm.trunc.v4f16(<4 x half> %op)
|
|
ret <4 x half> %res
|
|
}
|
|
|
|
define <8 x half> @frintz_v8f16(<8 x half> %op) {
|
|
; CHECK-LABEL: frintz_v8f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frintz z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintz_v8f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <8 x half> @llvm.trunc.v8f16(<8 x half> %op)
|
|
ret <8 x half> %res
|
|
}
|
|
|
|
define void @frintz_v16f16(ptr %a) {
|
|
; CHECK-LABEL: frintz_v16f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: frintz z0.h, p0/m, z0.h
|
|
; CHECK-NEXT: frintz z1.h, p0/m, z1.h
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintz_v16f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #62]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #60]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #58]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #54]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #52]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #50]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #46]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #44]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #42]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #38]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #36]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #34]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x half>, ptr %a
|
|
%res = call <16 x half> @llvm.trunc.v16f16(<16 x half> %op)
|
|
store <16 x half> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x float> @frintz_v2f32(<2 x float> %op) {
|
|
; CHECK-LABEL: frintz_v2f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: frintz z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintz_v2f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintz s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x float> @llvm.trunc.v2f32(<2 x float> %op)
|
|
ret <2 x float> %res
|
|
}
|
|
|
|
define <4 x float> @frintz_v4f32(<4 x float> %op) {
|
|
; CHECK-LABEL: frintz_v4f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frintz z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintz_v4f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frintz s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintz s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x float> @llvm.trunc.v4f32(<4 x float> %op)
|
|
ret <4 x float> %res
|
|
}
|
|
|
|
define void @frintz_v8f32(ptr %a) {
|
|
; CHECK-LABEL: frintz_v8f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: frintz z0.s, p0/m, z0.s
|
|
; CHECK-NEXT: frintz z1.s, p0/m, z1.s
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintz_v8f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: frintz s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: frintz s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: frintz s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: frintz s1, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintz s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x float>, ptr %a
|
|
%res = call <8 x float> @llvm.trunc.v8f32(<8 x float> %op)
|
|
store <8 x float> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <1 x double> @frintz_v1f64(<1 x double> %op) {
|
|
; CHECK-LABEL: frintz_v1f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: frintz d0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintz_v1f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: frintz d0, d0
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <1 x double> @llvm.trunc.v1f64(<1 x double> %op)
|
|
ret <1 x double> %res
|
|
}
|
|
|
|
define <2 x double> @frintz_v2f64(<2 x double> %op) {
|
|
; CHECK-LABEL: frintz_v2f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frintz z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintz_v2f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintz d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintz d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x double> @llvm.trunc.v2f64(<2 x double> %op)
|
|
ret <2 x double> %res
|
|
}
|
|
|
|
define void @frintz_v4f64(ptr %a) {
|
|
; CHECK-LABEL: frintz_v4f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: frintz z0.d, p0/m, z0.d
|
|
; CHECK-NEXT: frintz z1.d, p0/m, z1.d
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: frintz_v4f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: frintz d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: frintz d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: frintz d1, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: frintz d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x double>, ptr %a
|
|
%res = call <4 x double> @llvm.trunc.v4f64(<4 x double> %op)
|
|
store <4 x double> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
;
|
|
; FCANONICALIZE -> FMINNM
|
|
;
|
|
|
|
define <2 x half> @fcanonicalize_v2f16(<2 x half> %op) {
|
|
; CHECK-LABEL: fcanonicalize_v2f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: fminnm z0.h, p0/m, z0.h, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: fcanonicalize_v2f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %op)
|
|
ret <2 x half> %res
|
|
}
|
|
|
|
define <4 x half> @fcanonicalize_v4f16(<4 x half> %op) {
|
|
; CHECK-LABEL: fcanonicalize_v4f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl4
|
|
; CHECK-NEXT: fminnm z0.h, p0/m, z0.h, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: fcanonicalize_v4f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> %op)
|
|
ret <4 x half> %res
|
|
}
|
|
|
|
define <8 x half> @fcanonicalize_v8f16(<8 x half> %op) {
|
|
; CHECK-LABEL: fcanonicalize_v8f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: fminnm z0.h, p0/m, z0.h, z0.h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: fcanonicalize_v8f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <8 x half> @llvm.canonicalize.v8f16(<8 x half> %op)
|
|
ret <8 x half> %res
|
|
}
|
|
|
|
define void @fcanonicalize_v16f16(ptr %a) {
|
|
; CHECK-LABEL: fcanonicalize_v16f16:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.h, vl8
|
|
; CHECK-NEXT: fminnm z0.h, p0/m, z0.h, z0.h
|
|
; CHECK-NEXT: fminnm z1.h, p0/m, z1.h, z1.h
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: fcanonicalize_v16f16:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #30]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #62]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #60]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #26]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #58]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #54]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #52]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #50]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #46]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #44]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #42]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #38]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #36]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #34]
|
|
; NONEON-NOSVE-NEXT: ldr h0, [sp]
|
|
; NONEON-NOSVE-NEXT: fcvt s0, h0
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: fcvt h0, s0
|
|
; NONEON-NOSVE-NEXT: str h0, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <16 x half>, ptr %a
|
|
%res = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> %op)
|
|
store <16 x half> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <2 x float> @fcanonicalize_v2f32(<2 x float> %op) {
|
|
; CHECK-LABEL: fcanonicalize_v2f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl2
|
|
; CHECK-NEXT: fminnm z0.s, p0/m, z0.s, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: fcanonicalize_v2f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str d0, [sp, #-16]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fminnm s1, s0, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x float> @llvm.canonicalize.v2f32(<2 x float> %op)
|
|
ret <2 x float> %res
|
|
}
|
|
|
|
define <4 x float> @fcanonicalize_v4f32(<4 x float> %op) {
|
|
; CHECK-LABEL: fcanonicalize_v4f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: fminnm z0.s, p0/m, z0.s, z0.s
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: fcanonicalize_v4f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fminnm s1, s0, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fminnm s1, s0, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <4 x float> @llvm.canonicalize.v4f32(<4 x float> %op)
|
|
ret <4 x float> %res
|
|
}
|
|
|
|
define void @fcanonicalize_v8f32(ptr %a) {
|
|
; CHECK-LABEL: fcanonicalize_v8f32:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.s, vl4
|
|
; CHECK-NEXT: fminnm z0.s, p0/m, z0.s, z0.s
|
|
; CHECK-NEXT: fminnm z1.s, p0/m, z1.s, z1.s
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: fcanonicalize_v8f32:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #28]
|
|
; NONEON-NOSVE-NEXT: fminnm s1, s0, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #56]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #20]
|
|
; NONEON-NOSVE-NEXT: fminnm s1, s0, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
|
|
; NONEON-NOSVE-NEXT: fminnm s1, s0, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #40]
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
|
|
; NONEON-NOSVE-NEXT: fminnm s1, s0, s0
|
|
; NONEON-NOSVE-NEXT: ldr s0, [sp]
|
|
; NONEON-NOSVE-NEXT: fminnm s0, s0, s0
|
|
; NONEON-NOSVE-NEXT: stp s0, s1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <8 x float>, ptr %a
|
|
%res = call <8 x float> @llvm.canonicalize.v8f32(<8 x float> %op)
|
|
store <8 x float> %res, ptr %a
|
|
ret void
|
|
}
|
|
|
|
define <1 x double> @fcanonicalize_v1f64(<1 x double> %op) {
|
|
; CHECK-LABEL: fcanonicalize_v1f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: fminnm d0, d0, d0
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: fcanonicalize_v1f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: sub sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
|
|
; NONEON-NOSVE-NEXT: fminnm d0, d0, d0
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #16
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <1 x double> @llvm.canonicalize.v1f64(<1 x double> %op)
|
|
ret <1 x double> %res
|
|
}
|
|
|
|
define <2 x double> @fcanonicalize_v2f64(<2 x double> %op) {
|
|
; CHECK-LABEL: fcanonicalize_v2f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: fminnm z0.d, p0/m, z0.d, z0.d
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: fcanonicalize_v2f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: str q0, [sp, #-32]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fminnm d1, d0, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: fminnm d0, d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #32
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%res = call <2 x double> @llvm.canonicalize.v2f64(<2 x double> %op)
|
|
ret <2 x double> %res
|
|
}
|
|
|
|
define void @fcanonicalize_v4f64(ptr %a) {
|
|
; CHECK-LABEL: fcanonicalize_v4f64:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: ldp q0, q1, [x0]
|
|
; CHECK-NEXT: ptrue p0.d, vl2
|
|
; CHECK-NEXT: fminnm z0.d, p0/m, z0.d, z0.d
|
|
; CHECK-NEXT: fminnm z1.d, p0/m, z1.d, z1.d
|
|
; CHECK-NEXT: stp q0, q1, [x0]
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; NONEON-NOSVE-LABEL: fcanonicalize_v4f64:
|
|
; NONEON-NOSVE: // %bb.0:
|
|
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
|
|
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
|
|
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
|
|
; NONEON-NOSVE-NEXT: fminnm d1, d0, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #16]
|
|
; NONEON-NOSVE-NEXT: fminnm d0, d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #48]
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
|
|
; NONEON-NOSVE-NEXT: fminnm d1, d0, d0
|
|
; NONEON-NOSVE-NEXT: ldr d0, [sp]
|
|
; NONEON-NOSVE-NEXT: fminnm d0, d0, d0
|
|
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
|
|
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
|
|
; NONEON-NOSVE-NEXT: add sp, sp, #64
|
|
; NONEON-NOSVE-NEXT: ret
|
|
%op = load <4 x double>, ptr %a
|
|
%res = call <4 x double> @llvm.canonicalize.v4f64(<4 x double> %op)
|
|
store <4 x double> %res, ptr %a
|
|
ret void
|
|
}
|