mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2025-01-31 20:51:52 +01:00
54d477baad
Summary:
This patch adds the following LLVM IR intrinsics for SVE:
1. non-temporal gather loads
* @llvm.aarch64.sve.ldnt1.gather
* @llvm.aarch64.sve.ldnt1.gather.uxtw
* @llvm.aarch64.sve.ldnt1.gather.scalar.offset
2. non-temporal scatter stores
* @llvm.aarch64.sve.stnt1.scatter
* @llvm.aarch64.sve.ldnt1.gather.uxtw
* @llvm.aarch64.sve.ldnt1.gather.scalar.offset
These intrinsic are mapped to the corresponding SVE instructions
(example for half-words, zero-extending):
* ldnt1h { z0.s }, p0/z, [z0.s, x0]
* stnt1h { z0.s }, p0/z, [z0.s, x0]
Note that for non-temporal gathers/scatters, the SVE spec defines only
one instruction type: "vector + scalar". For this reason, we swap the
arguments when processing intrinsics that implement the "scalar +
vector" addressing mode:
* @llvm.aarch64.sve.ldnt1.gather
* @llvm.aarch64.sve.ldnt1.gather.uxtw
* @llvm.aarch64.sve.stnt1.scatter
* @llvm.aarch64.sve.ldnt1.gather.uxtw
In other words, all intrinsics for gather-loads and scatter-stores
implemented in this patch are mapped to the same load and store
instruction, respectively.
The sve2_mem_gldnt_vs multiclass (and it's counterpart for scatter
stores) from SVEInstrFormats.td was split into:
* sve2_mem_gldnt_vec_vs_32_ptrs (32bit wide base addresses)
* sve2_mem_gldnt_vec_vs_62_ptrs (64bit wide base addresses)
This is consistent with what we did for
@llvm.aarch64.sve.ld1.scalar.offset and highlights the actual split in
the spec and the implementation.
Reviewed by: sdesmalen
Differential Revision: https://reviews.llvm.org/D74858
97 lines
4.7 KiB
LLVM
97 lines
4.7 KiB
LLVM
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve2 < %s | FileCheck %s
|
|
|
|
;
|
|
; LDNT1B, LDNT1W, LDNT1H, LDNT1D: base + 32-bit unscaled offsets, zero (uxtw)
|
|
; extended to 64 bits.
|
|
; e.g. ldnt1h { z0.s }, p0/z, [z0.s, x0]
|
|
;
|
|
|
|
; LDNT1B
|
|
define <vscale x 4 x i32> @gldnt1b_s_uxtw(<vscale x 4 x i1> %pg, i8* %base, <vscale x 4 x i32> %b) {
|
|
; CHECK-LABEL: gldnt1b_s_uxtw:
|
|
; CHECK: ldnt1b { z0.s }, p0/z, [z0.s, x0]
|
|
; CHECK-NEXT: ret
|
|
%load = call <vscale x 4 x i8> @llvm.aarch64.sve.ldnt1.gather.uxtw.nxv4i8(<vscale x 4 x i1> %pg,
|
|
i8* %base,
|
|
<vscale x 4 x i32> %b)
|
|
%res = zext <vscale x 4 x i8> %load to <vscale x 4 x i32>
|
|
ret <vscale x 4 x i32> %res
|
|
}
|
|
|
|
; LDNT1H
|
|
define <vscale x 4 x i32> @gldnt1h_s_uxtw(<vscale x 4 x i1> %pg, i16* %base, <vscale x 4 x i32> %b) {
|
|
; CHECK-LABEL: gldnt1h_s_uxtw:
|
|
; CHECK: ldnt1h { z0.s }, p0/z, [z0.s, x0]
|
|
; CHECK-NEXT: ret
|
|
%load = call <vscale x 4 x i16> @llvm.aarch64.sve.ldnt1.gather.uxtw.nxv4i16(<vscale x 4 x i1> %pg,
|
|
i16* %base,
|
|
<vscale x 4 x i32> %b)
|
|
%res = zext <vscale x 4 x i16> %load to <vscale x 4 x i32>
|
|
ret <vscale x 4 x i32> %res
|
|
}
|
|
|
|
; LDNT1W
|
|
define <vscale x 4 x i32> @gldnt1w_s_uxtw(<vscale x 4 x i1> %pg, i32* %base, <vscale x 4 x i32> %b) {
|
|
; CHECK-LABEL: gldnt1w_s_uxtw:
|
|
; CHECK: ldnt1w { z0.s }, p0/z, [z0.s, x0]
|
|
; CHECK-NEXT: ret
|
|
%load = call <vscale x 4 x i32> @llvm.aarch64.sve.ldnt1.gather.uxtw.nxv4i32(<vscale x 4 x i1> %pg,
|
|
i32* %base,
|
|
<vscale x 4 x i32> %b)
|
|
ret <vscale x 4 x i32> %load
|
|
}
|
|
|
|
define <vscale x 4 x float> @gldnt1w_s_uxtw_float(<vscale x 4 x i1> %pg, float* %base, <vscale x 4 x i32> %b) {
|
|
; CHECK-LABEL: gldnt1w_s_uxtw_float:
|
|
; CHECK: ldnt1w { z0.s }, p0/z, [z0.s, x0]
|
|
; CHECK-NEXT: ret
|
|
%load = call <vscale x 4 x float> @llvm.aarch64.sve.ldnt1.gather.uxtw.nxv4f32(<vscale x 4 x i1> %pg,
|
|
float* %base,
|
|
<vscale x 4 x i32> %b)
|
|
ret <vscale x 4 x float> %load
|
|
}
|
|
|
|
; LDNT1SB, LDNT1SW, LDNT1SH: base + 32-bit unscaled offsets, zero (uxtw)
|
|
; extended to 64 bits.
|
|
; e.g. ldnt1sh { z0.s }, p0/z, [z0.s, x0]
|
|
;
|
|
|
|
; LDNT1SB
|
|
define <vscale x 4 x i32> @gldnt1sb_s_uxtw(<vscale x 4 x i1> %pg, i8* %base, <vscale x 4 x i32> %b) {
|
|
; CHECK-LABEL: gldnt1sb_s_uxtw:
|
|
; CHECK: ldnt1sb { z0.s }, p0/z, [z0.s, x0]
|
|
; CHECK-NEXT: ret
|
|
%load = call <vscale x 4 x i8> @llvm.aarch64.sve.ldnt1.gather.uxtw.nxv4i8(<vscale x 4 x i1> %pg,
|
|
i8* %base,
|
|
<vscale x 4 x i32> %b)
|
|
%res = sext <vscale x 4 x i8> %load to <vscale x 4 x i32>
|
|
ret <vscale x 4 x i32> %res
|
|
}
|
|
|
|
; LDNT1SH
|
|
define <vscale x 4 x i32> @gldnt1sh_s_uxtw(<vscale x 4 x i1> %pg, i16* %base, <vscale x 4 x i32> %b) {
|
|
; CHECK-LABEL: gldnt1sh_s_uxtw:
|
|
; CHECK: ldnt1sh { z0.s }, p0/z, [z0.s, x0]
|
|
; CHECK-NEXT: ret
|
|
%load = call <vscale x 4 x i16> @llvm.aarch64.sve.ldnt1.gather.uxtw.nxv4i16(<vscale x 4 x i1> %pg,
|
|
i16* %base,
|
|
<vscale x 4 x i32> %b)
|
|
%res = sext <vscale x 4 x i16> %load to <vscale x 4 x i32>
|
|
ret <vscale x 4 x i32> %res
|
|
}
|
|
|
|
; LDNT1B/LDNT1SB
|
|
declare <vscale x 4 x i8> @llvm.aarch64.sve.ldnt1.gather.uxtw.nxv4i8(<vscale x 4 x i1>, i8*, <vscale x 4 x i32>)
|
|
declare <vscale x 4 x i8> @llvm.aarch64.sve.ldnt1.gather.sxtw.nxv4i8(<vscale x 4 x i1>, i8*, <vscale x 4 x i32>)
|
|
|
|
; LDNT1H/LDNT1SH
|
|
declare <vscale x 4 x i16> @llvm.aarch64.sve.ldnt1.gather.sxtw.nxv4i16(<vscale x 4 x i1>, i16*, <vscale x 4 x i32>)
|
|
declare <vscale x 4 x i16> @llvm.aarch64.sve.ldnt1.gather.uxtw.nxv4i16(<vscale x 4 x i1>, i16*, <vscale x 4 x i32>)
|
|
|
|
; LDNT1W/LDNT1SW
|
|
declare <vscale x 4 x i32> @llvm.aarch64.sve.ldnt1.gather.sxtw.nxv4i32(<vscale x 4 x i1>, i32*, <vscale x 4 x i32>)
|
|
declare <vscale x 4 x i32> @llvm.aarch64.sve.ldnt1.gather.uxtw.nxv4i32(<vscale x 4 x i1>, i32*, <vscale x 4 x i32>)
|
|
|
|
declare <vscale x 4 x float> @llvm.aarch64.sve.ldnt1.gather.sxtw.nxv4f32(<vscale x 4 x i1>, float*, <vscale x 4 x i32>)
|
|
declare <vscale x 4 x float> @llvm.aarch64.sve.ldnt1.gather.uxtw.nxv4f32(<vscale x 4 x i1>, float*, <vscale x 4 x i32>)
|