1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
|
;=========================== begin_copyright_notice ============================
;
; Copyright (C) 2020-2023 Intel Corporation
;
; SPDX-License-Identifier: MIT
;
;============================ end_copyright_notice =============================
; RUN: %opt %use_old_pass_manager% -GenXPromoteArray -march=genx64 -mcpu=Gen9 -S < %s | FileCheck %s
target datalayout = "e-p:64:64-i64:64-n8:16:32"
%homogen.struct = type { i32, { i32, i32, [4 x i32] }, [2 x [2 x [2 x i32]]] }
define dllexport void @f_f(i64 %offset_a, i64 %offset_b, i64 %offset_c) {
%buf = alloca %homogen.struct, align 64
; CHECK: %[[BUF:[^ ]+]] = alloca <15 x i32>
; COM: %homogen.struct = type { i32, { i32, i32, [4 x i32] }, [2 x [2 x [2 x i32]]] }
; COM: ^
; COM: this one
%buf_offset_b = getelementptr inbounds %homogen.struct, %homogen.struct * %buf, i64 0, i32 1, i32 2, i64 %offset_a
store i32 43, i32* %buf_offset_b, align 64
; CHECK: %[[TRUNC_OFF_A:[^ ]+]] = trunc i64 %offset_a to i32
; CHECK: %[[SCALE_OFF_A:[^ ]+]] = mul i32 %[[TRUNC_OFF_A]], 1
; CHECK: %[[ADD_INIT_OFF_A:[^ ]+]] = add i32 3, %[[SCALE_OFF_A]]
; CHECK: %[[USELESS_ADD_OFF_A:[^ ]+]] = add i32 %[[ADD_INIT_OFF_A]], 0
; CHECK: %[[VB:[^ ]+]] = load <15 x i32>, <15 x i32>* %[[BUF]]
; CHECK: %[[NEW_VB:[^ ]+]] = insertelement <15 x i32> %[[VB]], i32 43, i32 %[[USELESS_ADD_OFF_A]]
; CHECK: store <15 x i32> %[[NEW_VB]], <15 x i32>* %[[BUF]]
; COM: %homogen.struct = type { i32, { i32, i32, [4 x i32] }, [2 x [2 x [2 x i32]]] }
; COM: ^
; COM: this one
%buf_offset_c = getelementptr inbounds %homogen.struct, %homogen.struct * %buf, i64 0, i32 2, i64 %offset_a, i64 %offset_b, i64 %offset_c
store i32 44, i32* %buf_offset_c, align 64
; CHECK: %[[C_TRUNC_OFF_A:[^ ]+]] = trunc i64 %offset_a to i32
; CHECK: %[[C_SCALE_OFF_A:[^ ]+]] = mul i32 %[[C_TRUNC_OFF_A]], 4
; CHECK: %[[C_ADD_INIT_OFF_A:[^ ]+]] = add i32 7, %[[C_SCALE_OFF_A]]
; CHECK: %[[C_TRUNC_OFF_B:[^ ]+]] = trunc i64 %offset_b to i32
; CHECK: %[[C_SCALE_OFF_B:[^ ]+]] = mul i32 %[[C_TRUNC_OFF_B]], 2
; CHECK: %[[C_ADD_PREV_OFF_B:[^ ]+]] = add i32 %[[C_ADD_INIT_OFF_A]], %[[C_SCALE_OFF_B]]
; CHECK: %[[C_TRUNC_OFF_C:[^ ]+]] = trunc i64 %offset_c to i32
; CHECK: %[[C_SCALE_OFF_C:[^ ]+]] = mul i32 %[[C_TRUNC_OFF_C]], 1
; CHECK: %[[C_ADD_PREV_OFF_C:[^ ]+]] = add i32 %[[C_ADD_PREV_OFF_B]], %[[C_SCALE_OFF_C]]
; CHECK: %[[C_USELESS_ADD_OFF_C:[^ ]+]] = add i32 %[[C_ADD_PREV_OFF_C]], 0
; CHECK: %[[VECC:[^ ]+]] = load <15 x i32>, <15 x i32>* %[[BUF]]
; CHECK: %[[NEW_VECC:[^ ]+]] = insertelement <15 x i32> %[[VECC]], i32 44, i32 %[[C_USELESS_ADD_OFF_C]]
; CHECK: store <15 x i32> %[[NEW_VECC]], <15 x i32>* %[[BUF]]
ret void
}
|
