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
106 lines
4.1 KiB
LLVM
106 lines
4.1 KiB
LLVM
; RUN: llc < %s -mtriple=ve -mattr=+vpu | FileCheck %s
|
|
|
|
; Test for correct placement of 'lvl' instructions
|
|
|
|
; Function Attrs: nounwind readonly
|
|
declare <256 x double> @llvm.ve.vl.vld.vssl(i64, ptr, i32)
|
|
declare void @llvm.ve.vl.vst.vssl(<256 x double>, i64, ptr, i32)
|
|
|
|
; Check that the backend can handle constant VL as well as parametric VL
|
|
; sources.
|
|
|
|
; Function Attrs: nounwind
|
|
define void @switching_vl(i32 %evl, i32 %evl2, ptr %P, ptr %Q) {
|
|
; CHECK-LABEL: switching_vl:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: lea %s4, 256
|
|
; CHECK-NEXT: lvl %s4
|
|
; CHECK-NEXT: vld %v0, 8, %s2
|
|
; CHECK-NEXT: and %s0, %s0, (32)0
|
|
; CHECK-NEXT: lvl %s0
|
|
; CHECK-NEXT: vst %v0, 16, %s3
|
|
; CHECK-NEXT: lea %s4, 128
|
|
; CHECK-NEXT: lvl %s4
|
|
; CHECK-NEXT: vld %v0, 16, %s2
|
|
; CHECK-NEXT: and %s1, %s1, (32)0
|
|
; CHECK-NEXT: lvl %s1
|
|
; CHECK-NEXT: vst %v0, 16, %s3
|
|
; CHECK-NEXT: lvl %s4
|
|
; CHECK-NEXT: vld %v0, 8, %s2
|
|
; CHECK-NEXT: lvl %s0
|
|
; CHECK-NEXT: vst %v0, 16, %s3
|
|
; CHECK-NEXT: b.l.t (, %s10)
|
|
%l0 = tail call <256 x double> @llvm.ve.vl.vld.vssl(i64 8, ptr %P, i32 256)
|
|
tail call void @llvm.ve.vl.vst.vssl(<256 x double> %l0, i64 16, ptr %Q, i32 %evl)
|
|
%l1 = tail call <256 x double> @llvm.ve.vl.vld.vssl(i64 16, ptr %P, i32 128)
|
|
tail call void @llvm.ve.vl.vst.vssl(<256 x double> %l1, i64 16, ptr %Q, i32 %evl2)
|
|
%l2 = tail call <256 x double> @llvm.ve.vl.vld.vssl(i64 8, ptr %P, i32 128)
|
|
tail call void @llvm.ve.vl.vst.vssl(<256 x double> %l2, i64 16, ptr %Q, i32 %evl)
|
|
ret void
|
|
}
|
|
|
|
; Check that no redundant 'lvl' is inserted when vector length does not change
|
|
; in a basic block.
|
|
|
|
; Function Attrs: nounwind
|
|
define void @stable_vl(i32 %evl, ptr %P, ptr %Q) {
|
|
; CHECK-LABEL: stable_vl:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: and %s0, %s0, (32)0
|
|
; CHECK-NEXT: lvl %s0
|
|
; CHECK-NEXT: vld %v0, 8, %s1
|
|
; CHECK-NEXT: vst %v0, 16, %s2
|
|
; CHECK-NEXT: vld %v0, 16, %s1
|
|
; CHECK-NEXT: vst %v0, 16, %s2
|
|
; CHECK-NEXT: vld %v0, 8, %s1
|
|
; CHECK-NEXT: vst %v0, 16, %s2
|
|
; CHECK-NEXT: b.l.t (, %s10)
|
|
%l0 = tail call <256 x double> @llvm.ve.vl.vld.vssl(i64 8, ptr %P, i32 %evl)
|
|
tail call void @llvm.ve.vl.vst.vssl(<256 x double> %l0, i64 16, ptr %Q, i32 %evl)
|
|
%l1 = tail call <256 x double> @llvm.ve.vl.vld.vssl(i64 16, ptr %P, i32 %evl)
|
|
tail call void @llvm.ve.vl.vst.vssl(<256 x double> %l1, i64 16, ptr %Q, i32 %evl)
|
|
%l2 = tail call <256 x double> @llvm.ve.vl.vld.vssl(i64 8, ptr %P, i32 %evl)
|
|
tail call void @llvm.ve.vl.vst.vssl(<256 x double> %l2, i64 16, ptr %Q, i32 %evl)
|
|
ret void
|
|
}
|
|
|
|
;;; Check the case we have a call in the middle of vector instructions.
|
|
|
|
; Function Attrs: nounwind
|
|
define void @call_invl(i32 %evl, ptr %P, ptr %Q) {
|
|
; CHECK-LABEL: call_invl:
|
|
; CHECK: .LBB{{[0-9]+}}_2:
|
|
; CHECK-NEXT: st %s18, 288(, %s11) # 8-byte Folded Spill
|
|
; CHECK-NEXT: st %s19, 296(, %s11) # 8-byte Folded Spill
|
|
; CHECK-NEXT: st %s20, 304(, %s11) # 8-byte Folded Spill
|
|
; CHECK-NEXT: or %s18, 0, %s1
|
|
; CHECK-NEXT: and %s20, %s0, (32)0
|
|
; CHECK-NEXT: lvl %s20
|
|
; CHECK-NEXT: vld %v0, 8, %s1
|
|
; CHECK-NEXT: or %s19, 0, %s2
|
|
; CHECK-NEXT: vst %v0, 16, %s2
|
|
; CHECK-NEXT: lea %s0, fun@lo
|
|
; CHECK-NEXT: and %s0, %s0, (32)0
|
|
; CHECK-NEXT: lea.sl %s12, fun@hi(, %s0)
|
|
; CHECK-NEXT: bsic %s10, (, %s12)
|
|
; CHECK-NEXT: lvl %s20
|
|
; CHECK-NEXT: vld %v0, 16, %s18
|
|
; CHECK-NEXT: vst %v0, 16, %s19
|
|
; CHECK-NEXT: vld %v0, 8, %s18
|
|
; CHECK-NEXT: vst %v0, 16, %s19
|
|
; CHECK-NEXT: ld %s20, 304(, %s11) # 8-byte Folded Reload
|
|
; CHECK-NEXT: ld %s19, 296(, %s11) # 8-byte Folded Reload
|
|
; CHECK-NEXT: ld %s18, 288(, %s11) # 8-byte Folded Reload
|
|
; CHECK-NEXT: or %s11, 0, %s9
|
|
%l0 = tail call <256 x double> @llvm.ve.vl.vld.vssl(i64 8, ptr %P, i32 %evl)
|
|
tail call void @llvm.ve.vl.vst.vssl(<256 x double> %l0, i64 16, ptr %Q, i32 %evl)
|
|
call void @fun()
|
|
%l1 = tail call <256 x double> @llvm.ve.vl.vld.vssl(i64 16, ptr %P, i32 %evl)
|
|
tail call void @llvm.ve.vl.vst.vssl(<256 x double> %l1, i64 16, ptr %Q, i32 %evl)
|
|
%l2 = tail call <256 x double> @llvm.ve.vl.vld.vssl(i64 8, ptr %P, i32 %evl)
|
|
tail call void @llvm.ve.vl.vst.vssl(<256 x double> %l2, i64 16, ptr %Q, i32 %evl)
|
|
ret void
|
|
}
|
|
|
|
declare void @fun()
|