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
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
|
;=========================== begin_copyright_notice ============================
;
; Copyright (C) 2020-2023 Intel Corporation
;
; SPDX-License-Identifier: MIT
;
;============================ end_copyright_notice =============================
; RUN: %opt %use_old_pass_manager% -GenXLowering -march=genx64 -mcpu=Gen9 -mtriple=spir64-unknown-unknown -S < %s | FileCheck %s
; RUN: %opt %use_old_pass_manager% -GenXLowering -march=genx64 -mcpu=XeLP -mtriple=spir64-unknown-unknown -S < %s | FileCheck %s --check-prefix CHECK-NO-IEEE
declare <8 x double> @llvm.sqrt.v8f64(<8 x double>)
declare <8 x float> @llvm.sqrt.v8f32(<8 x float>)
define <8 x double> @test_sqrt_double(<8 x double> %a0) {
; CHECK: [[IEEE_SQRT_DOUBLE1:%.*]] = call <8 x double> @llvm.genx.ieee.sqrt.v8f64(<8 x double> %a0)
; CHECK-NEXT: ret <8 x double> [[IEEE_SQRT_DOUBLE1]]
%1 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> %a0)
ret <8 x double> %1
}
define <8 x double> @test_sqrt_fast_double(<8 x double> %a0) {
; COM: Generating @llvm.genx.ieee.sqrt since @llvm.genx.sqrt cannot take double.
; CHECK: [[IEEE_SQRT_DOUBLE2:%.*]] = call <8 x double> @llvm.genx.ieee.sqrt.v8f64(<8 x double> %a0)
; CHECK-NEXT: ret <8 x double> [[IEEE_SQRT_DOUBLE2]]
%1 = call fast <8 x double> @llvm.sqrt.v8f64(<8 x double> %a0)
ret <8 x double> %1
}
define <8 x double> @test_sqrt_afn_double(<8 x double> %a0) {
; COM: Generating @llvm.genx.ieee.sqrt since @llvm.genx.sqrt cannot take double.
; CHECK: [[IEEE_SQRT_DOUBLE3:%.*]] = call <8 x double> @llvm.genx.ieee.sqrt.v8f64(<8 x double> %a0)
; CHECK-NEXT: ret <8 x double> [[IEEE_SQRT_DOUBLE3]]
%1 = call afn <8 x double> @llvm.sqrt.v8f64(<8 x double> %a0)
ret <8 x double> %1
}
define <8 x float> @test_sqrt_not_afn(<8 x float> %a0) {
; CHECK: [[IEEE_SQRT_FLOAT:%.*]] = call <8 x float> @llvm.genx.ieee.sqrt.v8f32(<8 x float> %a0)
; CHECK-NEXT: ret <8 x float> [[IEEE_SQRT_FLOAT]]
; CHECK-NO-IEEE: [[IEEE_SQRT_FLOAT:%.*]] = call <8 x float> @llvm.genx.sqrt.v8f32(<8 x float> %a0)
; CHECK-NO-IEEE-NEXT: ret <8 x float> [[IEEE_SQRT_FLOAT]]
%1 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %a0)
ret <8 x float> %1
}
define <8 x float> @test_sqrt_fast(<8 x float> %a0) {
; CHECK: [[NATIVE_SQRT:%.*]] = call <8 x float> @llvm.genx.sqrt.v8f32(<8 x float> %a0)
; CHECK-NEXT: ret <8 x float> [[NATIVE_SQRT]]
%1 = call fast <8 x float> @llvm.sqrt.v8f32(<8 x float> %a0)
ret <8 x float> %1
}
define <8 x float> @test_sqrt_afn(<8 x float> %a0) {
; CHECK: [[NATIVE_SQRT2:%.*]] = call <8 x float> @llvm.genx.sqrt.v8f32(<8 x float> %a0)
; CHECK-NEXT: ret <8 x float> [[NATIVE_SQRT2]]
%1 = call afn <8 x float> @llvm.sqrt.v8f32(<8 x float> %a0)
ret <8 x float> %1
}
declare <8 x float> @llvm.ceil.v8f32(<8 x float>)
define <8 x float> @test_ceil(<8 x float> %a0) {
; CHECK: [[NATIVE_CEIL:%.*]] = call <8 x float> @llvm.genx.rndu.v8f32(<8 x float> %a0)
; CHECK-NEXT: ret <8 x float> [[NATIVE_CEIL]]
%1 = call <8 x float> @llvm.ceil.v8f32(<8 x float> %a0)
ret <8 x float> %1
}
|
