1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-02-01 05:01:59 +01:00
llvm-mirror/test/CodeGen/AArch64/sve-gather-scatter-dag-combine.ll
Andrzej Warzynski c6e3a07bc6 [AArch64][SVE] Extend int_aarch64_sve_ld1_gather_imm
The ACLE distinguishes between the following addressing modes for gather
loads:
  * "scalar base, vector offset", and
  * "vector base, scalar offset".
For the "vector base, scalar offset" case, the
`int_aarch64_sve_ld1_gather_imm` intrinsic was added in 79f2422d.
Currently, that intrinsic assumes that the scalar offset is passed as an
immediate.  As a result, it does not cater for cases where scalar offset
is stored in a register.

In this patch `int_aarch64_sve_ld1_gather_imm` is extended so that all
cases are covered:
* `int_aarch64_sve_ld1_gather_imm` is renamed as
  `int_aarch64_sve_ld1_gather_scalar_offset`
* new DAG combine rules are added for GLD1_IMM for scenarios where the
  offset is a non-immediate scalar or an out-of-range immediate
* sve-intrinsics-gather-loads-vector-base.ll is renamed as
  sve-intrinsics-gather-loads-vector-base-imm-offset.ll
* sve-intrinsics-gather-loads-vector-base-scalar-offset.ll is added to test
  file for non-immediate offsets

Similar changes are made for scatter store intrinsics.

Reviewed By: sdesmalen, efriedma

Differential Revision: https://reviews.llvm.org/D71773
2020-01-20 12:19:18 +00:00

73 lines
3.8 KiB
LLVM

; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
; Verify that DAG combine rules for LD1 + sext/zext don't apply when the
; result of LD1 has multiple uses
define <vscale x 2 x i64> @no_dag_combine_zext_sext(<vscale x 2 x i1> %pg,
<vscale x 2 x i64> %base,
<vscale x 2 x i8>* %res_out,
<vscale x 2 x i1> %pred) {
; CHECK-LABEL: no_dag_combine_zext_sext
; CHECK: ld1b { z0.d }, p0/z, [z0.d, #16]
; CHECK-NEXT: st1b { z0.d }, p1, [x0]
; CHECK-NEXT: and z0.d, z0.d, #0xff
; CHECK-NEXT: ret
%load = call <vscale x 2 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i8.nxv2i64(<vscale x 2 x i1> %pg,
<vscale x 2 x i64> %base,
i64 16)
%res1 = zext <vscale x 2 x i8> %load to <vscale x 2 x i64>
%res2 = sext <vscale x 2 x i8> %load to <vscale x 2 x i64>
call void @llvm.masked.store.nxv2i8(<vscale x 2 x i8> %load,
<vscale x 2 x i8> *%res_out,
i32 8,
<vscale x 2 x i1> %pred)
ret <vscale x 2 x i64> %res1
}
define <vscale x 2 x i64> @no_dag_combine_sext(<vscale x 2 x i1> %pg,
<vscale x 2 x i64> %base,
<vscale x 2 x i8>* %res_out,
<vscale x 2 x i1> %pred) {
; CHECK-LABEL: no_dag_combine_sext
; CHECK: ld1b { z1.d }, p0/z, [z0.d, #16]
; CHECK-NEXT: ptrue p0.d
; CHECK-NEXT: sxtb z0.d, p0/m, z1.d
; CHECK-NEXT: st1b { z1.d }, p1, [x0]
; CHECK-NEXT: ret
%load = call <vscale x 2 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i8.nxv2i64(<vscale x 2 x i1> %pg,
<vscale x 2 x i64> %base,
i64 16)
%res = sext <vscale x 2 x i8> %load to <vscale x 2 x i64>
call void @llvm.masked.store.nxv2i8(<vscale x 2 x i8> %load,
<vscale x 2 x i8> *%res_out,
i32 8,
<vscale x 2 x i1> %pred)
ret <vscale x 2 x i64> %res
}
define <vscale x 2 x i64> @no_dag_combine_zext(<vscale x 2 x i1> %pg,
<vscale x 2 x i64> %base,
<vscale x 2 x i8>* %res_out,
<vscale x 2 x i1> %pred) {
; CHECK-LABEL: no_dag_combine_zext
; CHECK: ld1b { z0.d }, p0/z, [z0.d, #16]
; CHECK-NEXT: st1b { z0.d }, p1, [x0]
; CHECK-NEXT: and z0.d, z0.d, #0xff
; CHECK-NEXT: ret
%load = call <vscale x 2 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i8.nxv2i64(<vscale x 2 x i1> %pg,
<vscale x 2 x i64> %base,
i64 16)
%res = zext <vscale x 2 x i8> %load to <vscale x 2 x i64>
call void @llvm.masked.store.nxv2i8(<vscale x 2 x i8> %load,
<vscale x 2 x i8> *%res_out,
i32 8,
<vscale x 2 x i1> %pred)
ret <vscale x 2 x i64> %res
}
declare <vscale x 2 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i8.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, i64)
declare void @llvm.masked.store.nxv2i8(<vscale x 2 x i8>, <vscale x 2 x i8>*, i32, <vscale x 2 x i1>)