| 1 |
; RUN: llc < %s -mtriple=ve -mattr=+vpu | FileCheck %s |
| 2 |
|
| 3 |
;;; Test vector sum intrinsic instructions |
| 4 |
;;; |
| 5 |
;;; Note: |
| 6 |
;;; We test VSUM*vl and VSUM*vml instructions. |
| 7 |
|
| 8 |
; Function Attrs: nounwind readnone |
| 9 |
define fastcc <256 x double> @vsumwsx_vvl(<256 x double> %0) { |
| 10 |
; CHECK-LABEL: vsumwsx_vvl: |
| 11 |
; CHECK: # %bb.0: |
| 12 |
; CHECK-NEXT: lea %s0, 256 |
| 13 |
; CHECK-NEXT: lvl %s0 |
| 14 |
; CHECK-NEXT: vsum.w.sx %v0, %v0 |
| 15 |
; CHECK-NEXT: b.l.t (, %s10) |
| 16 |
%2 = tail call fast <256 x double> @llvm.ve.vl.vsumwsx.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.vsumwsx.vvl(<256 x double>, i32) |
| 22 |
|
| 23 |
; Function Attrs: nounwind readnone |
| 24 |
define fastcc <256 x double> @vsumwsx_vvml(<256 x double> %0, <256 x i1> %1) { |
| 25 |
; CHECK-LABEL: vsumwsx_vvml: |
| 26 |
; CHECK: # %bb.0: |
| 27 |
; CHECK-NEXT: lea %s0, 256 |
| 28 |
; CHECK-NEXT: lvl %s0 |
| 29 |
; CHECK-NEXT: vsum.w.sx %v0, %v0, %vm1 |
| 30 |
; CHECK-NEXT: b.l.t (, %s10) |
| 31 |
%3 = tail call fast <256 x double> @llvm.ve.vl.vsumwsx.vvml(<256 x double> %0, <256 x i1> %1, i32 256) |
| 32 |
ret <256 x double> %3 |
| 33 |
} |
| 34 |
|
| 35 |
; Function Attrs: nounwind readnone |
| 36 |
declare <256 x double> @llvm.ve.vl.vsumwsx.vvml(<256 x double>, <256 x i1>, i32) |
| 37 |
|
| 38 |
; Function Attrs: nounwind readnone |
| 39 |
define fastcc <256 x double> @vsumwzx_vvl(<256 x double> %0) { |
| 40 |
; CHECK-LABEL: vsumwzx_vvl: |
| 41 |
; CHECK: # %bb.0: |
| 42 |
; CHECK-NEXT: lea %s0, 256 |
| 43 |
; CHECK-NEXT: lvl %s0 |
| 44 |
; CHECK-NEXT: vsum.w.zx %v0, %v0 |
| 45 |
; CHECK-NEXT: b.l.t (, %s10) |
| 46 |
%2 = tail call fast <256 x double> @llvm.ve.vl.vsumwzx.vvl(<256 x double> %0, i32 256) |
| 47 |
ret <256 x double> %2 |
| 48 |
} |
| 49 |
|
| 50 |
; Function Attrs: nounwind readnone |
| 51 |
declare <256 x double> @llvm.ve.vl.vsumwzx.vvl(<256 x double>, i32) |
| 52 |
|
| 53 |
; Function Attrs: nounwind readnone |
| 54 |
define fastcc <256 x double> @vsumwzx_vvml(<256 x double> %0, <256 x i1> %1) { |
| 55 |
; CHECK-LABEL: vsumwzx_vvml: |
| 56 |
; CHECK: # %bb.0: |
| 57 |
; CHECK-NEXT: lea %s0, 256 |
| 58 |
; CHECK-NEXT: lvl %s0 |
| 59 |
; CHECK-NEXT: vsum.w.zx %v0, %v0, %vm1 |
| 60 |
; CHECK-NEXT: b.l.t (, %s10) |
| 61 |
%3 = tail call fast <256 x double> @llvm.ve.vl.vsumwzx.vvml(<256 x double> %0, <256 x i1> %1, i32 256) |
| 62 |
ret <256 x double> %3 |
| 63 |
} |
| 64 |
|
| 65 |
; Function Attrs: nounwind readnone |
| 66 |
declare <256 x double> @llvm.ve.vl.vsumwzx.vvml(<256 x double>, <256 x i1>, i32) |
| 67 |
|
| 68 |
; Function Attrs: nounwind readnone |
| 69 |
define fastcc <256 x double> @vsuml_vvl(<256 x double> %0) { |
| 70 |
; CHECK-LABEL: vsuml_vvl: |
| 71 |
; CHECK: # %bb.0: |
| 72 |
; CHECK-NEXT: lea %s0, 256 |
| 73 |
; CHECK-NEXT: lvl %s0 |
| 74 |
; CHECK-NEXT: vsum.l %v0, %v0 |
| 75 |
; CHECK-NEXT: b.l.t (, %s10) |
| 76 |
%2 = tail call fast <256 x double> @llvm.ve.vl.vsuml.vvl(<256 x double> %0, i32 256) |
| 77 |
ret <256 x double> %2 |
| 78 |
} |
| 79 |
|
| 80 |
; Function Attrs: nounwind readnone |
| 81 |
declare <256 x double> @llvm.ve.vl.vsuml.vvl(<256 x double>, i32) |
| 82 |
|
| 83 |
; Function Attrs: nounwind readnone |
| 84 |
define fastcc <256 x double> @vsuml_vvml(<256 x double> %0, <256 x i1> %1) { |
| 85 |
; CHECK-LABEL: vsuml_vvml: |
| 86 |
; CHECK: # %bb.0: |
| 87 |
; CHECK-NEXT: lea %s0, 256 |
| 88 |
; CHECK-NEXT: lvl %s0 |
| 89 |
; CHECK-NEXT: vsum.l %v0, %v0, %vm1 |
| 90 |
; CHECK-NEXT: b.l.t (, %s10) |
| 91 |
%3 = tail call fast <256 x double> @llvm.ve.vl.vsuml.vvml(<256 x double> %0, <256 x i1> %1, i32 256) |
| 92 |
ret <256 x double> %3 |
| 93 |
} |
| 94 |
|
| 95 |
; Function Attrs: nounwind readnone |
| 96 |
declare <256 x double> @llvm.ve.vl.vsuml.vvml(<256 x double>, <256 x i1>, i32) |