forked from OSchip/llvm-project
[AArch64][SVE] Implement pfirst and pnext intrinsics
Reviewers: sdesmalen, efriedma, dancgr, mgudim, cameron.mcinally Reviewed By: cameron.mcinally Subscribers: tschuett, kristof.beyls, hiraditya, rkruppe, psnobl, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71472
This commit is contained in:
parent
49199465a3
commit
eca0c97a6b
|
@ -804,6 +804,12 @@ let TargetPrefix = "aarch64" in { // All intrinsics start with "llvm.aarch64.".
|
|||
llvm_i32_ty],
|
||||
[IntrNoMem]>;
|
||||
|
||||
class AdvSIMD_Pred1VectorArg_Intrinsic
|
||||
: Intrinsic<[llvm_anyvector_ty],
|
||||
[LLVMScalarOrSameVectorWidth<0, llvm_i1_ty>,
|
||||
LLVMMatchType<0>],
|
||||
[IntrNoMem]>;
|
||||
|
||||
class AdvSIMD_Pred2VectorArg_Intrinsic
|
||||
: Intrinsic<[llvm_anyvector_ty],
|
||||
[LLVMScalarOrSameVectorWidth<0, llvm_i1_ty>,
|
||||
|
@ -1441,6 +1447,8 @@ def int_aarch64_sve_ptrue : AdvSIMD_SVE_PTRUE_Intrinsic;
|
|||
// Predicate operations
|
||||
//
|
||||
|
||||
def int_aarch64_sve_pfirst : AdvSIMD_Pred1VectorArg_Intrinsic;
|
||||
def int_aarch64_sve_pnext : AdvSIMD_Pred1VectorArg_Intrinsic;
|
||||
def int_aarch64_sve_punpkhi : AdvSIMD_SVE_PUNPKHI_Intrinsic;
|
||||
def int_aarch64_sve_punpklo : AdvSIMD_SVE_PUNPKHI_Intrinsic;
|
||||
|
||||
|
|
|
@ -295,8 +295,8 @@ let Predicates = [HasSVE] in {
|
|||
|
||||
def PTEST_PP : sve_int_ptest<0b010000, "ptest">;
|
||||
def PFALSE : sve_int_pfalse<0b000000, "pfalse">;
|
||||
defm PFIRST : sve_int_pfirst<0b00000, "pfirst">;
|
||||
defm PNEXT : sve_int_pnext<0b00110, "pnext">;
|
||||
defm PFIRST : sve_int_pfirst<0b00000, "pfirst", int_aarch64_sve_pfirst>;
|
||||
defm PNEXT : sve_int_pnext<0b00110, "pnext", int_aarch64_sve_pnext>;
|
||||
|
||||
defm AND_PPzPP : sve_int_pred_log<0b0000, "and", int_aarch64_sve_and>;
|
||||
defm BIC_PPzPP : sve_int_pred_log<0b0001, "bic", int_aarch64_sve_bic>;
|
||||
|
|
|
@ -419,15 +419,22 @@ class sve_int_pfirst_next<bits<2> sz8_64, bits<5> opc, string asm,
|
|||
let Defs = [NZCV];
|
||||
}
|
||||
|
||||
multiclass sve_int_pfirst<bits<5> opc, string asm> {
|
||||
def : sve_int_pfirst_next<0b01, opc, asm, PPR8>;
|
||||
multiclass sve_int_pfirst<bits<5> opc, string asm, SDPatternOperator op> {
|
||||
def _B : sve_int_pfirst_next<0b01, opc, asm, PPR8>;
|
||||
|
||||
def : SVE_2_Op_Pat<nxv16i1, op, nxv16i1, nxv16i1, !cast<Instruction>(NAME # _B)>;
|
||||
}
|
||||
|
||||
multiclass sve_int_pnext<bits<5> opc, string asm> {
|
||||
multiclass sve_int_pnext<bits<5> opc, string asm, SDPatternOperator op> {
|
||||
def _B : sve_int_pfirst_next<0b00, opc, asm, PPR8>;
|
||||
def _H : sve_int_pfirst_next<0b01, opc, asm, PPR16>;
|
||||
def _S : sve_int_pfirst_next<0b10, opc, asm, PPR32>;
|
||||
def _D : sve_int_pfirst_next<0b11, opc, asm, PPR64>;
|
||||
|
||||
def : SVE_2_Op_Pat<nxv16i1, op, nxv16i1, nxv16i1, !cast<Instruction>(NAME # _B)>;
|
||||
def : SVE_2_Op_Pat<nxv8i1, op, nxv8i1, nxv8i1, !cast<Instruction>(NAME # _H)>;
|
||||
def : SVE_2_Op_Pat<nxv4i1, op, nxv4i1, nxv4i1, !cast<Instruction>(NAME # _S)>;
|
||||
def : SVE_2_Op_Pat<nxv2i1, op, nxv2i1, nxv2i1, !cast<Instruction>(NAME # _D)>;
|
||||
}
|
||||
|
||||
//===----------------------------------------------------------------------===//
|
||||
|
|
|
@ -1,5 +1,63 @@
|
|||
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
|
||||
|
||||
;
|
||||
; PFIRST
|
||||
;
|
||||
|
||||
define <vscale x 16 x i1> @pfirst_b8(<vscale x 16 x i1> %pg, <vscale x 16 x i1> %a) {
|
||||
; CHECK-LABEL: pfirst_b8:
|
||||
; CHECK: pfirst p1.b, p0, p1.b
|
||||
; CHECK-NEXT: mov p0.b, p1.b
|
||||
; CHECK-NEXT: ret
|
||||
%out = call <vscale x 16 x i1> @llvm.aarch64.sve.pfirst.nxv16i1(<vscale x 16 x i1> %pg,
|
||||
<vscale x 16 x i1> %a)
|
||||
ret <vscale x 16 x i1> %out
|
||||
}
|
||||
|
||||
;
|
||||
; PNEXT
|
||||
;
|
||||
|
||||
define <vscale x 16 x i1> @pnext_b8(<vscale x 16 x i1> %pg, <vscale x 16 x i1> %a) {
|
||||
; CHECK-LABEL: pnext_b8:
|
||||
; CHECK: pnext p1.b, p0, p1.b
|
||||
; CHECK-NEXT: mov p0.b, p1.b
|
||||
; CHECK-NEXT: ret
|
||||
%out = call <vscale x 16 x i1> @llvm.aarch64.sve.pnext.nxv16i1(<vscale x 16 x i1> %pg,
|
||||
<vscale x 16 x i1> %a)
|
||||
ret <vscale x 16 x i1> %out
|
||||
}
|
||||
|
||||
define <vscale x 8 x i1> @pnext_b16(<vscale x 8 x i1> %pg, <vscale x 8 x i1> %a) {
|
||||
; CHECK-LABEL: pnext_b16:
|
||||
; CHECK: pnext p1.h, p0, p1.h
|
||||
; CHECK-NEXT: mov p0.b, p1.b
|
||||
; CHECK-NEXT: ret
|
||||
%out = call <vscale x 8 x i1> @llvm.aarch64.sve.pnext.nxv8i1(<vscale x 8 x i1> %pg,
|
||||
<vscale x 8 x i1> %a)
|
||||
ret <vscale x 8 x i1> %out
|
||||
}
|
||||
|
||||
define <vscale x 4 x i1> @pnext_b32(<vscale x 4 x i1> %pg, <vscale x 4 x i1> %a) {
|
||||
; CHECK-LABEL: pnext_b32:
|
||||
; CHECK: pnext p1.s, p0, p1.s
|
||||
; CHECK-NEXT: mov p0.b, p1.b
|
||||
; CHECK-NEXT: ret
|
||||
%out = call <vscale x 4 x i1> @llvm.aarch64.sve.pnext.nxv4i1(<vscale x 4 x i1> %pg,
|
||||
<vscale x 4 x i1> %a)
|
||||
ret <vscale x 4 x i1> %out
|
||||
}
|
||||
|
||||
define <vscale x 2 x i1> @pnext_b64(<vscale x 2 x i1> %pg, <vscale x 2 x i1> %a) {
|
||||
; CHECK-LABEL: pnext_b64:
|
||||
; CHECK: pnext p1.d, p0, p1.d
|
||||
; CHECK-NEXT: mov p0.b, p1.b
|
||||
; CHECK-NEXT: ret
|
||||
%out = call <vscale x 2 x i1> @llvm.aarch64.sve.pnext.nxv2i1(<vscale x 2 x i1> %pg,
|
||||
<vscale x 2 x i1> %a)
|
||||
ret <vscale x 2 x i1> %out
|
||||
}
|
||||
|
||||
;
|
||||
; PUNPKHI
|
||||
;
|
||||
|
@ -56,6 +114,13 @@ define <vscale x 2 x i1> @punpklo_b4(<vscale x 4 x i1> %a) {
|
|||
ret <vscale x 2 x i1> %res
|
||||
}
|
||||
|
||||
declare <vscale x 16 x i1> @llvm.aarch64.sve.pfirst.nxv16i1(<vscale x 16 x i1>, <vscale x 16 x i1>)
|
||||
|
||||
declare <vscale x 16 x i1> @llvm.aarch64.sve.pnext.nxv16i1(<vscale x 16 x i1>, <vscale x 16 x i1>)
|
||||
declare <vscale x 8 x i1> @llvm.aarch64.sve.pnext.nxv8i1(<vscale x 8 x i1>, <vscale x 8 x i1>)
|
||||
declare <vscale x 4 x i1> @llvm.aarch64.sve.pnext.nxv4i1(<vscale x 4 x i1>, <vscale x 4 x i1>)
|
||||
declare <vscale x 2 x i1> @llvm.aarch64.sve.pnext.nxv2i1(<vscale x 2 x i1>, <vscale x 2 x i1>)
|
||||
|
||||
declare <vscale x 8 x i1> @llvm.aarch64.sve.punpkhi.nxv8i1(<vscale x 16 x i1>)
|
||||
declare <vscale x 4 x i1> @llvm.aarch64.sve.punpkhi.nxv4i1(<vscale x 8 x i1>)
|
||||
declare <vscale x 2 x i1> @llvm.aarch64.sve.punpkhi.nxv2i1(<vscale x 4 x i1>)
|
||||
|
|
Loading…
Reference in New Issue