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
|
;=========================== begin_copyright_notice ============================
;
; Copyright (C) 2023 Intel Corporation
;
; SPDX-License-Identifier: MIT
;
;============================ end_copyright_notice =============================
; This test checks that i64 data write is translated to v2i32 in VectorProcess.
; Ref: IGC-7889
; Below .ll is generated by the ocloc options
; -device mtl
; -internal_options "-cl-intel-use-bindless-mode -cl-intel-use-bindless-advanced-mode"
; CL Source:
; __kernel void QueryKernelTimestamps(__global ulong* dst) {
; uint gid = get_global_id(0);
; uint currentOffset = gid * 4;
; dst[currentOffset] = 0;
; }
;
; RUN: igc_opt -enable-debugify --igc-vectorprocess -platformmtl -S < %s 2>&1 | FileCheck %s
; ------------------------------------------------
; VectorProcess
; ------------------------------------------------
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024-n8:16:32"
; Debug-info related check
; CHECK-NOT: WARNING
; CHECK: CheckModuleDebugify: PASS
define spir_kernel void @QueryKernelTimestamps(i64 addrspace(1)* %dst, <8 x i32> %r0, <8 x i32> %payloadHeader, <3 x i32> %enqueuedLocalSize, i16 %localIdX, i16 %localIdY, i16 %localIdZ, i32 %bufferOffset) {
entry:
%payloadHeader.scalar = extractelement <8 x i32> %payloadHeader, i32 0
%enqueuedLocalSize.scalar = extractelement <3 x i32> %enqueuedLocalSize, i32 0
%r0.scalar17 = extractelement <8 x i32> %r0, i32 1
%mul.i.i.i = mul i32 %enqueuedLocalSize.scalar, %r0.scalar17
%localIdX2 = zext i16 %localIdX to i32
%add.i.i.i = add i32 %mul.i.i.i, %localIdX2
%add4.i.i.i = add i32 %add.i.i.i, %payloadHeader.scalar
%mul = shl i32 %add4.i.i.i, 5
%0 = addrspacecast i64 addrspace(1)* %dst to i64 addrspace(2490368)*
; CHECK: call void @llvm.genx.GenISA.storerawvector
; CHECK-NOT: i64
; CHECK-SAME: v2i32.v2i32
call void @llvm.genx.GenISA.storeraw.indexed.p2490368i64.i64(i64 addrspace(2490368)* %0, i32 %mul, i64 0, i32 8, i1 false)
ret void
}
declare void @llvm.genx.GenISA.storeraw.indexed.p2490368i64.i64(i64 addrspace(2490368)*, i32, i64, i32, i1)
|
