Skip to content

Commit

Permalink
Fix up tests committed in 57a6bb3
Browse files Browse the repository at this point in the history 
I committed an older version of the patch without the test updates.
This patch uses the latest versions on
https://reviews.llvm.org/D142904
  • Loading branch information
@david-arm
david-arm committed Feb 3, 2023
1 parent 57a6bb3 commit 6e1ebb9
Show file tree
Hide file tree
Showing 3 changed files with 228 additions and 174 deletions.
106 changes: 71 additions & 35 deletions llvm/test/CodeGen/AArch64/sve2p1-intrinsics-uzpx2.ll
View file
Expand Up @@ -3,88 +3,115 @@

; == 8 to 64-bit elements ==

define { <vscale x 16 x i8>, <vscale x 16 x i8> } @uzp_x2_i8(<vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) nounwind {
define <vscale x 16 x i8> @uzp_x2_i8(<vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) nounwind {
; CHECK-LABEL: uzp_x2_i8:
; CHECK: // %bb.0:
; CHECK-NEXT: uzp { z0.b, z1.b }, z0.b, z1.b
; CHECK-NEXT: uzp { z2.b, z3.b }, z0.b, z1.b
; CHECK-NEXT: add z0.b, z2.b, z0.b
; CHECK-NEXT: ret
%res = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.uzp.x2.nxv16i8(<vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
ret { <vscale x 16 x i8>, <vscale x 16 x i8> } %res
%uzp = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.uzp.x2.nxv16i8(<vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
%uzp0 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %uzp, 0
%add = add <vscale x 16 x i8> %uzp0, %zn
ret <vscale x 16 x i8> %add
}

define { <vscale x 8 x i16>, <vscale x 8 x i16> } @uzp_x2_i16(<vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm) nounwind {
define <vscale x 8 x i16> @uzp_x2_i16(<vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm) nounwind {
; CHECK-LABEL: uzp_x2_i16:
; CHECK: // %bb.0:
; CHECK-NEXT: uzp { z0.h, z1.h }, z0.h, z1.h
; CHECK-NEXT: uzp { z2.h, z3.h }, z0.h, z1.h
; CHECK-NEXT: add z0.h, z2.h, z0.h
; CHECK-NEXT: ret
%res = call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.uzp.x2.nxv8i16(<vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm)
ret { <vscale x 8 x i16>, <vscale x 8 x i16> } %res
%uzp = call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.uzp.x2.nxv8i16(<vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm)
%uzp0 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %uzp, 0
%add = add <vscale x 8 x i16> %uzp0, %zn
ret <vscale x 8 x i16> %add
}

define { <vscale x 8 x half>, <vscale x 8 x half> } @uzp_x2_f16(<vscale x 8 x half> %zn, <vscale x 8 x half> %zm) nounwind {
define <vscale x 8 x half> @uzp_x2_f16(<vscale x 8 x half> %zn, <vscale x 8 x half> %zm) nounwind {
; CHECK-LABEL: uzp_x2_f16:
; CHECK: // %bb.0:
; CHECK-NEXT: uzp { z0.h, z1.h }, z0.h, z1.h
; CHECK-NEXT: uzp { z2.h, z3.h }, z0.h, z1.h
; CHECK-NEXT: fadd z0.h, z2.h, z0.h
; CHECK-NEXT: ret
%res = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.uzp.x2.nxv8f16(<vscale x 8 x half> %zn, <vscale x 8 x half> %zm)
ret { <vscale x 8 x half>, <vscale x 8 x half> } %res
%uzp = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.uzp.x2.nxv8f16(<vscale x 8 x half> %zn, <vscale x 8 x half> %zm)
%uzp0 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } %uzp, 0
%add = fadd <vscale x 8 x half> %uzp0, %zn
ret <vscale x 8 x half> %add
}

define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @uzp_x2_bf16(<vscale x 8 x bfloat> %zn, <vscale x 8 x bfloat> %zm) nounwind {
; CHECK-LABEL: uzp_x2_bf16:
; CHECK: // %bb.0:
; CHECK-NEXT: uzp { z0.h, z1.h }, z0.h, z1.h
; CHECK-NEXT: ret
%res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.uzp.x2.nxv8bf16(<vscale x 8 x bfloat> %zn, <vscale x 8 x bfloat> %zm)
ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res
%uzp = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.uzp.x2.nxv8bf16(<vscale x 8 x bfloat> %zn, <vscale x 8 x bfloat> %zm)
ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %uzp
}

define { <vscale x 4 x i32>, <vscale x 4 x i32> } @uzp_x2_i32(<vscale x 4 x i32> %zn, <vscale x 4 x i32> %zm) nounwind {
define <vscale x 4 x i32> @uzp_x2_i32(<vscale x 4 x i32> %zn, <vscale x 4 x i32> %zm) nounwind {
; CHECK-LABEL: uzp_x2_i32:
; CHECK: // %bb.0:
; CHECK-NEXT: uzp { z0.s, z1.s }, z0.s, z1.s
; CHECK-NEXT: uzp { z2.s, z3.s }, z0.s, z1.s
; CHECK-NEXT: add z0.s, z2.s, z0.s
; CHECK-NEXT: ret
%res = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.uzp.x2.nxv4i32(<vscale x 4 x i32> %zn, <vscale x 4 x i32> %zm)
ret { <vscale x 4 x i32>, <vscale x 4 x i32> } %res
%uzp = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.uzp.x2.nxv4i32(<vscale x 4 x i32> %zn, <vscale x 4 x i32> %zm)
%uzp0 = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } %uzp, 0
%add = add <vscale x 4 x i32> %uzp0, %zn
ret <vscale x 4 x i32> %add
}

define { <vscale x 4 x float>, <vscale x 4 x float> } @uzp_x2_f32(<vscale x 4 x float> %zn, <vscale x 4 x float> %zm) nounwind {
define <vscale x 4 x float> @uzp_x2_f32(<vscale x 4 x float> %zn, <vscale x 4 x float> %zm) nounwind {
; CHECK-LABEL: uzp_x2_f32:
; CHECK: // %bb.0:
; CHECK-NEXT: uzp { z0.s, z1.s }, z0.s, z1.s
; CHECK-NEXT: uzp { z2.s, z3.s }, z0.s, z1.s
; CHECK-NEXT: fadd z0.s, z2.s, z0.s
; CHECK-NEXT: ret
%res = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.uzp.x2.nxv4f32(<vscale x 4 x float> %zn, <vscale x 4 x float> %zm)
ret { <vscale x 4 x float>, <vscale x 4 x float> } %res
%uzp = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.uzp.x2.nxv4f32(<vscale x 4 x float> %zn, <vscale x 4 x float> %zm)
%uzp0 = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } %uzp, 0
%add = fadd <vscale x 4 x float> %uzp0, %zn
ret <vscale x 4 x float> %add
}

define { <vscale x 2 x i64>, <vscale x 2 x i64> } @uzp_x2_i64(<vscale x 2 x i64> %zn, <vscale x 2 x i64> %zm) nounwind {
define <vscale x 2 x i64> @uzp_x2_i64(<vscale x 2 x i64> %zn, <vscale x 2 x i64> %zm) nounwind {
; CHECK-LABEL: uzp_x2_i64:
; CHECK: // %bb.0:
; CHECK-NEXT: uzp { z0.d, z1.d }, z0.d, z1.d
; CHECK-NEXT: uzp { z2.d, z3.d }, z0.d, z1.d
; CHECK-NEXT: add z0.d, z2.d, z0.d
; CHECK-NEXT: ret
%res = call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.uzp.x2.nxv2i64(<vscale x 2 x i64> %zn, <vscale x 2 x i64> %zm)
ret { <vscale x 2 x i64>, <vscale x 2 x i64> } %res
%uzp = call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.uzp.x2.nxv2i64(<vscale x 2 x i64> %zn, <vscale x 2 x i64> %zm)
%uzp0 = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } %uzp, 0
%add = add <vscale x 2 x i64> %uzp0, %zn
ret <vscale x 2 x i64> %add
}

define { <vscale x 2 x double>, <vscale x 2 x double> } @uzp_x2_f64(<vscale x 2 x double> %zn, <vscale x 2 x double> %zm) nounwind {
define <vscale x 2 x double> @uzp_x2_f64(<vscale x 2 x double> %zn, <vscale x 2 x double> %zm) nounwind {
; CHECK-LABEL: uzp_x2_f64:
; CHECK: // %bb.0:
; CHECK-NEXT: uzp { z0.d, z1.d }, z0.d, z1.d
; CHECK-NEXT: uzp { z2.d, z3.d }, z0.d, z1.d
; CHECK-NEXT: fadd z0.d, z2.d, z0.d
; CHECK-NEXT: ret
%res = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.uzp.x2.nxv2f64(<vscale x 2 x double> %zn, <vscale x 2 x double> %zm)
ret { <vscale x 2 x double>, <vscale x 2 x double> } %res
%uzp = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.uzp.x2.nxv2f64(<vscale x 2 x double> %zn, <vscale x 2 x double> %zm)
%uzp0 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } %uzp, 0
%add = fadd <vscale x 2 x double> %uzp0, %zn
ret <vscale x 2 x double> %add
}


; == 128-bit elements ==

define { <vscale x 16 x i8>, <vscale x 16 x i8> } @uzpq_x2_i8(<vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) nounwind {
; NOTE: For the 128-bit case we only need to check the <vscale x 16 x i8> to
; ensure the tuple result starts at the correct register multiple. The other
; variants all test the same code path.
define <vscale x 16 x i8> @uzpq_x2_i8(<vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) nounwind {
; CHECK-LABEL: uzpq_x2_i8:
; CHECK: // %bb.0:
; CHECK-NEXT: uzp { z0.q, z1.q }, z0.q, z1.q
; CHECK-NEXT: uzp { z2.q, z3.q }, z0.q, z1.q
; CHECK-NEXT: add z0.b, z2.b, z0.b
; CHECK-NEXT: ret
%res = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.uzpq.x2.nxv16i8(<vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
ret { <vscale x 16 x i8>, <vscale x 16 x i8> } %res
%uzp = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.uzpq.x2.nxv16i8(<vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
%uzp0 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %uzp, 0
%add = add <vscale x 16 x i8> %uzp0, %zn
ret <vscale x 16 x i8> %add
}

define { <vscale x 8 x i16>, <vscale x 8 x i16> } @uzpq_x2_i16(<vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm) nounwind {
Expand Down Expand Up @@ -150,9 +177,18 @@ define { <vscale x 2 x double>, <vscale x 2 x double> } @uzpq_x2_f64(<vscale x 2
ret { <vscale x 2 x double>, <vscale x 2 x double> } %res
}

define { <vscale x 16 x i8>, <vscale x 16 x i8> } @uzpq_x2_i8_not_tied(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) nounwind {
; CHECK-LABEL: uzpq_x2_i8_not_tied:
; CHECK: // %bb.0:
; CHECK-NEXT: uzp { z0.q, z1.q }, z1.q, z2.q
; CHECK-NEXT: ret
%res = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.uzpq.x2.nxv16i8(<vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
ret { <vscale x 16 x i8>, <vscale x 16 x i8> } %res
}


; == 8 to 64-bit elements ==
declare { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.uzp.x2.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
declare { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.uzp.x2.nxv16i8(<vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
declare { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.uzp.x2.nxv8i16(<vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm)
declare { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.uzp.x2.nxv4i32(<vscale x 4 x i32> %zn, <vscale x 4 x i32> %zm)
declare { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.uzp.x2.nxv2i64(<vscale x 2 x i64> %zn, <vscale x 2 x i64> %zm)
Expand Down

0 comments on commit 6e1ebb9

Please sign in to comment.