| 1 |
; RUN: llc < %s -mtriple=ve -mattr=+vpu | FileCheck %s |
| 2 |
|
| 3 |
;;; Test vector floating square root intrinsic instructions |
| 4 |
;;; |
| 5 |
;;; Note: |
| 6 |
;;; We test VFSQRT*vl and VFSQRT*vl_v instructions. |
| 7 |
|
| 8 |
; Function Attrs: nounwind readnone |
| 9 |
define fastcc <256 x double> @vfsqrtd_vvl(<256 x double> %0) { |
| 10 |
; CHECK-LABEL: vfsqrtd_vvl: |
| 11 |
; CHECK: # %bb.0: |
| 12 |
; CHECK-NEXT: lea %s0, 256 |
| 13 |
; CHECK-NEXT: lvl %s0 |
| 14 |
; CHECK-NEXT: vfsqrt.d %v0, %v0 |
| 15 |
; CHECK-NEXT: b.l.t (, %s10) |
| 16 |
%2 = tail call fast <256 x double> @llvm.ve.vl.vfsqrtd.vvl(<256 x double> %0, i32 256) |
| 17 |
ret <256 x double> %2 |
| 18 |
} |
| 19 |
|
| 20 |
; Function Attrs: nounwind readnone |
| 21 |
declare <256 x double> @llvm.ve.vl.vfsqrtd.vvl(<256 x double>, i32) |
| 22 |
|
| 23 |
; Function Attrs: nounwind readnone |
| 24 |
define fastcc <256 x double> @vfsqrtd_vvvl(<256 x double> %0, <256 x double> %1) { |
| 25 |
; CHECK-LABEL: vfsqrtd_vvvl: |
| 26 |
; CHECK: # %bb.0: |
| 27 |
; CHECK-NEXT: lea %s0, 128 |
| 28 |
; CHECK-NEXT: lvl %s0 |
| 29 |
; CHECK-NEXT: vfsqrt.d %v1, %v0 |
| 30 |
; CHECK-NEXT: lea %s16, 256 |
| 31 |
; CHECK-NEXT: lvl %s16 |
| 32 |
; CHECK-NEXT: vor %v0, (0)1, %v1 |
| 33 |
; CHECK-NEXT: b.l.t (, %s10) |
| 34 |
%3 = tail call fast <256 x double> @llvm.ve.vl.vfsqrtd.vvvl(<256 x double> %0, <256 x double> %1, i32 128) |
| 35 |
ret <256 x double> %3 |
| 36 |
} |
| 37 |
|
| 38 |
; Function Attrs: nounwind readnone |
| 39 |
declare <256 x double> @llvm.ve.vl.vfsqrtd.vvvl(<256 x double>, <256 x double>, i32) |
| 40 |
|
| 41 |
; Function Attrs: nounwind readnone |
| 42 |
define fastcc <256 x double> @vfsqrts_vvl(<256 x double> %0) { |
| 43 |
; CHECK-LABEL: vfsqrts_vvl: |
| 44 |
; CHECK: # %bb.0: |
| 45 |
; CHECK-NEXT: lea %s0, 256 |
| 46 |
; CHECK-NEXT: lvl %s0 |
| 47 |
; CHECK-NEXT: vfsqrt.s %v0, %v0 |
| 48 |
; CHECK-NEXT: b.l.t (, %s10) |
| 49 |
%2 = tail call fast <256 x double> @llvm.ve.vl.vfsqrts.vvl(<256 x double> %0, i32 256) |
| 50 |
ret <256 x double> %2 |
| 51 |
} |
| 52 |
|
| 53 |
; Function Attrs: nounwind readnone |
| 54 |
declare <256 x double> @llvm.ve.vl.vfsqrts.vvl(<256 x double>, i32) |
| 55 |
|
| 56 |
; Function Attrs: nounwind readnone |
| 57 |
define fastcc <256 x double> @vfsqrts_vvvl(<256 x double> %0, <256 x double> %1) { |
| 58 |
; CHECK-LABEL: vfsqrts_vvvl: |
| 59 |
; CHECK: # %bb.0: |
| 60 |
; CHECK-NEXT: lea %s0, 128 |
| 61 |
; CHECK-NEXT: lvl %s0 |
| 62 |
; CHECK-NEXT: vfsqrt.s %v1, %v0 |
| 63 |
; CHECK-NEXT: lea %s16, 256 |
| 64 |
; CHECK-NEXT: lvl %s16 |
| 65 |
; CHECK-NEXT: vor %v0, (0)1, %v1 |
| 66 |
; CHECK-NEXT: b.l.t (, %s10) |
| 67 |
%3 = tail call fast <256 x double> @llvm.ve.vl.vfsqrts.vvvl(<256 x double> %0, <256 x double> %1, i32 128) |
| 68 |
ret <256 x double> %3 |
| 69 |
} |
| 70 |
|
| 71 |
; Function Attrs: nounwind readnone |
| 72 |
declare <256 x double> @llvm.ve.vl.vfsqrts.vvvl(<256 x double>, <256 x double>, i32) |