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
234 lines
7.2 KiB
LLVM
234 lines
7.2 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
|
|
; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx %s -o - | FileCheck %s --check-prefix=LA32
|
|
; RUN: llc --mtriple=loongarch64 --mattr=+lsx %s -o - | FileCheck %s --check-prefix=LA64
|
|
|
|
define void @load_trunc_2i64_to_2i32(ptr %ptr, ptr %dst) nounwind {
|
|
; LA32-LABEL: load_trunc_2i64_to_2i32:
|
|
; LA32: # %bb.0:
|
|
; LA32-NEXT: vld $vr0, $a0, 0
|
|
; LA32-NEXT: vpickve2gr.w $a0, $vr0, 2
|
|
; LA32-NEXT: st.w $a0, $a1, 4
|
|
; LA32-NEXT: vpickve2gr.w $a0, $vr0, 0
|
|
; LA32-NEXT: st.w $a0, $a1, 0
|
|
; LA32-NEXT: ret
|
|
;
|
|
; LA64-LABEL: load_trunc_2i64_to_2i32:
|
|
; LA64: # %bb.0:
|
|
; LA64-NEXT: vld $vr0, $a0, 0
|
|
; LA64-NEXT: vshuf4i.w $vr0, $vr0, 8
|
|
; LA64-NEXT: vstelm.d $vr0, $a1, 0, 0
|
|
; LA64-NEXT: ret
|
|
%a = load <2 x i64>, ptr %ptr
|
|
%trunc = trunc <2 x i64> %a to <2 x i32>
|
|
store <2 x i32> %trunc, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
define void @load_trunc_2i64_to_2i16(ptr %ptr, ptr %dst) nounwind {
|
|
; LA32-LABEL: load_trunc_2i64_to_2i16:
|
|
; LA32: # %bb.0:
|
|
; LA32-NEXT: vld $vr0, $a0, 0
|
|
; LA32-NEXT: vextrins.h $vr0, $vr0, 20
|
|
; LA32-NEXT: vpickve2gr.w $a0, $vr0, 0
|
|
; LA32-NEXT: st.w $a0, $a1, 0
|
|
; LA32-NEXT: ret
|
|
;
|
|
; LA64-LABEL: load_trunc_2i64_to_2i16:
|
|
; LA64: # %bb.0:
|
|
; LA64-NEXT: vld $vr0, $a0, 0
|
|
; LA64-NEXT: vextrins.h $vr0, $vr0, 20
|
|
; LA64-NEXT: vstelm.w $vr0, $a1, 0, 0
|
|
; LA64-NEXT: ret
|
|
%a = load <2 x i64>, ptr %ptr
|
|
%trunc = trunc <2 x i64> %a to <2 x i16>
|
|
store <2 x i16> %trunc, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
define void @load_trunc_2i64_to_2i8(ptr %ptr, ptr %dst) nounwind {
|
|
; LA32-LABEL: load_trunc_2i64_to_2i8:
|
|
; LA32: # %bb.0:
|
|
; LA32-NEXT: vld $vr0, $a0, 0
|
|
; LA32-NEXT: vextrins.b $vr0, $vr0, 24
|
|
; LA32-NEXT: vstelm.h $vr0, $a1, 0, 0
|
|
; LA32-NEXT: ret
|
|
;
|
|
; LA64-LABEL: load_trunc_2i64_to_2i8:
|
|
; LA64: # %bb.0:
|
|
; LA64-NEXT: vld $vr0, $a0, 0
|
|
; LA64-NEXT: vextrins.b $vr0, $vr0, 24
|
|
; LA64-NEXT: vstelm.h $vr0, $a1, 0, 0
|
|
; LA64-NEXT: ret
|
|
%a = load <2 x i64>, ptr %ptr
|
|
%trunc = trunc <2 x i64> %a to <2 x i8>
|
|
store <2 x i8> %trunc, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
define void @load_trunc_4i32_to_4i16(ptr %ptr, ptr %dst) nounwind {
|
|
; LA32-LABEL: load_trunc_4i32_to_4i16:
|
|
; LA32: # %bb.0:
|
|
; LA32-NEXT: vld $vr0, $a0, 0
|
|
; LA32-NEXT: vpickev.h $vr0, $vr0, $vr0
|
|
; LA32-NEXT: vpickve2gr.w $a0, $vr0, 1
|
|
; LA32-NEXT: st.w $a0, $a1, 4
|
|
; LA32-NEXT: vpickve2gr.w $a0, $vr0, 0
|
|
; LA32-NEXT: st.w $a0, $a1, 0
|
|
; LA32-NEXT: ret
|
|
;
|
|
; LA64-LABEL: load_trunc_4i32_to_4i16:
|
|
; LA64: # %bb.0:
|
|
; LA64-NEXT: vld $vr0, $a0, 0
|
|
; LA64-NEXT: vpickev.h $vr0, $vr0, $vr0
|
|
; LA64-NEXT: vstelm.d $vr0, $a1, 0, 0
|
|
; LA64-NEXT: ret
|
|
%a = load <4 x i32>, ptr %ptr
|
|
%trunc = trunc <4 x i32> %a to <4 x i16>
|
|
store <4 x i16> %trunc, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
define void @load_trunc_4i32_to_4i8(ptr %ptr, ptr %dst) nounwind {
|
|
; LA32-LABEL: load_trunc_4i32_to_4i8:
|
|
; LA32: # %bb.0:
|
|
; LA32-NEXT: vld $vr0, $a0, 0
|
|
; LA32-NEXT: pcalau12i $a0, %pc_hi20(.LCPI4_0)
|
|
; LA32-NEXT: vld $vr1, $a0, %pc_lo12(.LCPI4_0)
|
|
; LA32-NEXT: vshuf.b $vr0, $vr0, $vr0, $vr1
|
|
; LA32-NEXT: vpickve2gr.w $a0, $vr0, 0
|
|
; LA32-NEXT: st.w $a0, $a1, 0
|
|
; LA32-NEXT: ret
|
|
;
|
|
; LA64-LABEL: load_trunc_4i32_to_4i8:
|
|
; LA64: # %bb.0:
|
|
; LA64-NEXT: vld $vr0, $a0, 0
|
|
; LA64-NEXT: pcalau12i $a0, %pc_hi20(.LCPI4_0)
|
|
; LA64-NEXT: vld $vr1, $a0, %pc_lo12(.LCPI4_0)
|
|
; LA64-NEXT: vshuf.b $vr0, $vr0, $vr0, $vr1
|
|
; LA64-NEXT: vstelm.w $vr0, $a1, 0, 0
|
|
; LA64-NEXT: ret
|
|
%a = load <4 x i32>, ptr %ptr
|
|
%trunc = trunc <4 x i32> %a to <4 x i8>
|
|
store <4 x i8> %trunc, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
define void @load_trunc_8i16_to_8i8(ptr %ptr, ptr %dst) nounwind {
|
|
; LA32-LABEL: load_trunc_8i16_to_8i8:
|
|
; LA32: # %bb.0:
|
|
; LA32-NEXT: vld $vr0, $a0, 0
|
|
; LA32-NEXT: vpickev.b $vr0, $vr0, $vr0
|
|
; LA32-NEXT: vpickve2gr.w $a0, $vr0, 1
|
|
; LA32-NEXT: st.w $a0, $a1, 4
|
|
; LA32-NEXT: vpickve2gr.w $a0, $vr0, 0
|
|
; LA32-NEXT: st.w $a0, $a1, 0
|
|
; LA32-NEXT: ret
|
|
;
|
|
; LA64-LABEL: load_trunc_8i16_to_8i8:
|
|
; LA64: # %bb.0:
|
|
; LA64-NEXT: vld $vr0, $a0, 0
|
|
; LA64-NEXT: vpickev.b $vr0, $vr0, $vr0
|
|
; LA64-NEXT: vstelm.d $vr0, $a1, 0, 0
|
|
; LA64-NEXT: ret
|
|
%a = load <8 x i16>, ptr %ptr
|
|
%trunc = trunc <8 x i16> %a to <8 x i8>
|
|
store <8 x i8> %trunc, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
define void @load_trunc_2i32_to_2i16(ptr %ptr, ptr %dst) nounwind {
|
|
; LA32-LABEL: load_trunc_2i32_to_2i16:
|
|
; LA32: # %bb.0:
|
|
; LA32-NEXT: ld.w $a2, $a0, 0
|
|
; LA32-NEXT: ld.w $a0, $a0, 4
|
|
; LA32-NEXT: vinsgr2vr.w $vr0, $a2, 0
|
|
; LA32-NEXT: vinsgr2vr.w $vr0, $a0, 1
|
|
; LA32-NEXT: vshuf4i.h $vr0, $vr0, 8
|
|
; LA32-NEXT: vpickve2gr.w $a0, $vr0, 0
|
|
; LA32-NEXT: st.w $a0, $a1, 0
|
|
; LA32-NEXT: ret
|
|
;
|
|
; LA64-LABEL: load_trunc_2i32_to_2i16:
|
|
; LA64: # %bb.0:
|
|
; LA64-NEXT: ld.d $a0, $a0, 0
|
|
; LA64-NEXT: vinsgr2vr.d $vr0, $a0, 0
|
|
; LA64-NEXT: vshuf4i.h $vr0, $vr0, 8
|
|
; LA64-NEXT: vstelm.w $vr0, $a1, 0, 0
|
|
; LA64-NEXT: ret
|
|
%a = load <2 x i32>, ptr %ptr
|
|
%trunc = trunc <2 x i32> %a to <2 x i16>
|
|
store <2 x i16> %trunc, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
define void @load_trunc_2i32_to_2i8(ptr %ptr, ptr %dst) nounwind {
|
|
; LA32-LABEL: load_trunc_2i32_to_2i8:
|
|
; LA32: # %bb.0:
|
|
; LA32-NEXT: ld.w $a2, $a0, 0
|
|
; LA32-NEXT: ld.w $a0, $a0, 4
|
|
; LA32-NEXT: vinsgr2vr.w $vr0, $a2, 0
|
|
; LA32-NEXT: vinsgr2vr.w $vr0, $a0, 1
|
|
; LA32-NEXT: vextrins.b $vr0, $vr0, 20
|
|
; LA32-NEXT: vstelm.h $vr0, $a1, 0, 0
|
|
; LA32-NEXT: ret
|
|
;
|
|
; LA64-LABEL: load_trunc_2i32_to_2i8:
|
|
; LA64: # %bb.0:
|
|
; LA64-NEXT: ld.d $a0, $a0, 0
|
|
; LA64-NEXT: vinsgr2vr.d $vr0, $a0, 0
|
|
; LA64-NEXT: vextrins.b $vr0, $vr0, 20
|
|
; LA64-NEXT: vstelm.h $vr0, $a1, 0, 0
|
|
; LA64-NEXT: ret
|
|
%a = load <2 x i32>, ptr %ptr
|
|
%trunc = trunc <2 x i32> %a to <2 x i8>
|
|
store <2 x i8> %trunc, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
define void @load_trunc_4i16_to_4i8(ptr %ptr, ptr %dst) nounwind {
|
|
; LA32-LABEL: load_trunc_4i16_to_4i8:
|
|
; LA32: # %bb.0:
|
|
; LA32-NEXT: ld.w $a2, $a0, 0
|
|
; LA32-NEXT: ld.w $a0, $a0, 4
|
|
; LA32-NEXT: vinsgr2vr.w $vr0, $a2, 0
|
|
; LA32-NEXT: vinsgr2vr.w $vr0, $a0, 1
|
|
; LA32-NEXT: vpickev.b $vr0, $vr0, $vr0
|
|
; LA32-NEXT: vpickve2gr.w $a0, $vr0, 0
|
|
; LA32-NEXT: st.w $a0, $a1, 0
|
|
; LA32-NEXT: ret
|
|
;
|
|
; LA64-LABEL: load_trunc_4i16_to_4i8:
|
|
; LA64: # %bb.0:
|
|
; LA64-NEXT: ld.d $a0, $a0, 0
|
|
; LA64-NEXT: vinsgr2vr.d $vr0, $a0, 0
|
|
; LA64-NEXT: vpickev.b $vr0, $vr0, $vr0
|
|
; LA64-NEXT: vstelm.w $vr0, $a1, 0, 0
|
|
; LA64-NEXT: ret
|
|
%a = load <4 x i16>, ptr %ptr
|
|
%trunc = trunc <4 x i16> %a to <4 x i8>
|
|
store <4 x i8> %trunc, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
define void @load_trunc_2i16_to_2i8(ptr %ptr, ptr %dst) nounwind {
|
|
; LA32-LABEL: load_trunc_2i16_to_2i8:
|
|
; LA32: # %bb.0:
|
|
; LA32-NEXT: ld.w $a0, $a0, 0
|
|
; LA32-NEXT: vinsgr2vr.w $vr0, $a0, 0
|
|
; LA32-NEXT: vshuf4i.b $vr0, $vr0, 8
|
|
; LA32-NEXT: vstelm.h $vr0, $a1, 0, 0
|
|
; LA32-NEXT: ret
|
|
;
|
|
; LA64-LABEL: load_trunc_2i16_to_2i8:
|
|
; LA64: # %bb.0:
|
|
; LA64-NEXT: ld.w $a0, $a0, 0
|
|
; LA64-NEXT: vinsgr2vr.w $vr0, $a0, 0
|
|
; LA64-NEXT: vshuf4i.b $vr0, $vr0, 8
|
|
; LA64-NEXT: vstelm.h $vr0, $a1, 0, 0
|
|
; LA64-NEXT: ret
|
|
%a = load <2 x i16>, ptr %ptr
|
|
%trunc = trunc <2 x i16> %a to <2 x i8>
|
|
store <2 x i8> %trunc, ptr %dst
|
|
ret void
|
|
}
|