1 ; RUN: llc -O0 -mtriple=spirv-unknown-unknown %s -o - | FileCheck %s
3 ; CHECK-DAG: %[[#float_64:]] = OpTypeFloat 64
4 ; CHECK-DAG: %[[#float_32:]] = OpTypeFloat 32
5 ; CHECK-DAG: %[[#float_16:]] = OpTypeFloat 16
6 ; CHECK-DAG: %[[#vec2_float_16:]] = OpTypeVector %[[#float_16]] 2
7 ; CHECK-DAG: %[[#vec2_float_32:]] = OpTypeVector %[[#float_32]] 2
8 ; CHECK-DAG: %[[#vec2_float_64:]] = OpTypeVector %[[#float_64]] 2
9 ; CHECK-DAG: %[[#vec3_float_16:]] = OpTypeVector %[[#float_16]] 3
10 ; CHECK-DAG: %[[#vec3_float_32:]] = OpTypeVector %[[#float_32]] 3
11 ; CHECK-DAG: %[[#vec3_float_64:]] = OpTypeVector %[[#float_64]] 3
12 ; CHECK-DAG: %[[#vec4_float_16:]] = OpTypeVector %[[#float_16]] 4
13 ; CHECK-DAG: %[[#vec4_float_32:]] = OpTypeVector %[[#float_32]] 4
14 ; CHECK-DAG: %[[#vec4_float_64:]] = OpTypeVector %[[#float_64]] 4
15 ; CHECK-DAG: %[[#const_f64_1:]] = OpConstant %[[#float_64]] 1
16 ; CHECK-DAG: %[[#const_f32_1:]] = OpConstant %[[#float_32]] 1
17 ; CHECK-DAG: %[[#const_f16_1:]] = OpConstant %[[#float_16]] 1
19 ; CHECK-DAG: %[[#vec2_const_ones_f16:]] = OpConstantComposite %[[#vec2_float_16]] %[[#const_f16_1]] %[[#const_f16_1]]
20 ; CHECK-DAG: %[[#vec3_const_ones_f16:]] = OpConstantComposite %[[#vec3_float_16]] %[[#const_f16_1]] %[[#const_f16_1]] %[[#const_f16_1]]
21 ; CHECK-DAG: %[[#vec4_const_ones_f16:]] = OpConstantComposite %[[#vec4_float_16]] %[[#const_f16_1]] %[[#const_f16_1]] %[[#const_f16_1]] %[[#const_f16_1]]
23 ; CHECK-DAG: %[[#vec2_const_ones_f32:]] = OpConstantComposite %[[#vec2_float_32]] %[[#const_f32_1]] %[[#const_f32_1]]
24 ; CHECK-DAG: %[[#vec3_const_ones_f32:]] = OpConstantComposite %[[#vec3_float_32]] %[[#const_f32_1]] %[[#const_f32_1]] %[[#const_f32_1]]
25 ; CHECK-DAG: %[[#vec4_const_ones_f32:]] = OpConstantComposite %[[#vec4_float_32]] %[[#const_f32_1]] %[[#const_f32_1]] %[[#const_f32_1]] %[[#const_f32_1]]
27 ; CHECK-DAG: %[[#vec2_const_ones_f64:]] = OpConstantComposite %[[#vec2_float_64]] %[[#const_f64_1]] %[[#const_f64_1]]
28 ; CHECK-DAG: %[[#vec3_const_ones_f64:]] = OpConstantComposite %[[#vec3_float_64]] %[[#const_f64_1]] %[[#const_f64_1]] %[[#const_f64_1]]
29 ; CHECK-DAG: %[[#vec4_const_ones_f64:]] = OpConstantComposite %[[#vec4_float_64]] %[[#const_f64_1]] %[[#const_f64_1]] %[[#const_f64_1]] %[[#const_f64_1]]
32 define spir_func noundef half @test_rcp_half(half noundef %p0) #0 {
34 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#float_16]]
35 ; CHECK: OpFDiv %[[#float_16]] %[[#const_f16_1]] %[[#arg0]]
36 %hlsl.rcp = fdiv half 0xH3C00, %p0
40 define spir_func noundef <2 x half> @test_rcp_half2(<2 x half> noundef %p0) #0 {
42 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec2_float_16]]
43 ; CHECK: OpFDiv %[[#vec2_float_16]] %[[#vec2_const_ones_f16]] %[[#arg0]]
44 %hlsl.rcp = fdiv <2 x half> <half 0xH3C00, half 0xH3C00>, %p0
45 ret <2 x half> %hlsl.rcp
48 define spir_func noundef <3 x half> @test_rcp_half3(<3 x half> noundef %p0) #0 {
50 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec3_float_16]]
51 ; CHECK: OpFDiv %[[#vec3_float_16]] %[[#vec3_const_ones_f16]] %[[#arg0]]
52 %hlsl.rcp = fdiv <3 x half> <half 0xH3C00, half 0xH3C00, half 0xH3C00>, %p0
53 ret <3 x half> %hlsl.rcp
56 define spir_func noundef <4 x half> @test_rcp_half4(<4 x half> noundef %p0) #0 {
58 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec4_float_16]]
59 ; CHECK: OpFDiv %[[#vec4_float_16]] %[[#vec4_const_ones_f16]] %[[#arg0]]
60 %hlsl.rcp = fdiv <4 x half> <half 0xH3C00, half 0xH3C00, half 0xH3C00, half 0xH3C00>, %p0
61 ret <4 x half> %hlsl.rcp
64 define spir_func noundef float @test_rcp_float(float noundef %p0) #0 {
66 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#float_32]]
67 ; CHECK: OpFDiv %[[#float_32]] %[[#const_f32_1]] %[[#arg0]]
68 %hlsl.rcp = fdiv float 1.000000e+00, %p0
72 define spir_func noundef <2 x float> @test_rcp_float2(<2 x float> noundef %p0) #0 {
74 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec2_float_32]]
75 ; CHECK: OpFDiv %[[#vec2_float_32]] %[[#vec2_const_ones_f32]] %[[#arg0]]
76 %hlsl.rcp = fdiv <2 x float> <float 1.000000e+00, float 1.000000e+00>, %p0
77 ret <2 x float> %hlsl.rcp
80 define spir_func noundef <3 x float> @test_rcp_float3(<3 x float> noundef %p0) #0 {
82 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec3_float_32]]
83 ; CHECK: OpFDiv %[[#vec3_float_32]] %[[#vec3_const_ones_f32]] %[[#arg0]]
84 %hlsl.rcp = fdiv <3 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, %p0
85 ret <3 x float> %hlsl.rcp
88 define spir_func noundef <4 x float> @test_rcp_float4(<4 x float> noundef %p0) #0 {
90 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec4_float_32]]
91 ; CHECK: OpFDiv %[[#vec4_float_32]] %[[#vec4_const_ones_f32]] %[[#arg0]]
92 %hlsl.rcp = fdiv <4 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, %p0
93 ret <4 x float> %hlsl.rcp
96 define spir_func noundef double @test_rcp_double(double noundef %p0) #0 {
98 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#float_64]]
99 ; CHECK: OpFDiv %[[#float_64]] %[[#const_f64_1]] %[[#arg0]]
100 %hlsl.rcp = fdiv double 1.000000e+00, %p0
104 define spir_func noundef <2 x double> @test_rcp_double2(<2 x double> noundef %p0) #0 {
106 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec2_float_64:]]
107 ; CHECK: OpFDiv %[[#vec2_float_64]] %[[#vec2_const_ones_f64]] %[[#arg0]]
108 %hlsl.rcp = fdiv <2 x double> <double 1.000000e+00, double 1.000000e+00>, %p0
109 ret <2 x double> %hlsl.rcp
112 define spir_func noundef <3 x double> @test_rcp_double3(<3 x double> noundef %p0) #0 {
114 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec3_float_64:]]
115 ; CHECK: OpFDiv %[[#vec3_float_64]] %[[#vec3_const_ones_f64]] %[[#arg0]]
116 %hlsl.rcp = fdiv <3 x double> <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>, %p0
117 ret <3 x double> %hlsl.rcp
120 define spir_func noundef <4 x double> @test_rcp_double4(<4 x double> noundef %p0) #0 {
122 ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec4_float_64]]
123 ; CHECK: OpFDiv %[[#vec4_float_64]] %[[#vec4_const_ones_f64]] %[[#arg0]]
124 %hlsl.rcp = fdiv <4 x double> <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>, %p0
125 ret <4 x double> %hlsl.rcp