llvm-project/clang/test/Sema/ext_vector_casts.c

53 lines
2.0 KiB
C

// RUN: %clang_cc1 -fsyntax-only -verify -fno-lax-vector-conversions %s
typedef __attribute__(( ext_vector_type(2) )) float float2;
typedef __attribute__(( ext_vector_type(4) )) int int4;
typedef __attribute__(( ext_vector_type(8) )) short short8;
typedef __attribute__(( ext_vector_type(4) )) float float4;
typedef float t3 __attribute__ ((vector_size (16)));
static void test() {
float2 vec2;
float4 vec4, vec4_2;
int4 ivec4;
short8 ish8;
t3 vec4_3;
int *ptr;
int i;
vec4 = 5.0f;
vec4 = (float4)5.0f;
vec4 = (float4)5;
vec4 = (float4)vec4_3;
ivec4 = (int4)5.0f;
ivec4 = (int4)5;
ivec4 = (int4)vec4_3;
i = (int)ivec4; // expected-error {{invalid conversion between vector type 'int4' and integer type 'int' of different size}}
i = ivec4; // expected-error {{assigning to 'int' from incompatible type 'int4'}}
ivec4 = (int4)ptr; // expected-error {{invalid conversion between vector type 'int4' and scalar type 'int *'}}
vec4 = (float4)vec2; // expected-error {{invalid conversion between ext-vector type 'float4' and 'float2'}}
ish8 += 5; // expected-error {{can't convert between vector values of different size ('short8' and 'int')}}
ish8 += (short)5;
ivec4 *= 5;
vec4 /= 5.2f;
vec4 %= 4; // expected-error {{invalid operands to binary expression ('float4' and 'int')}}
ivec4 %= 4;
ivec4 += vec4; // expected-error {{can't convert between vector values of different size ('int4' and 'float4')}}
ivec4 += (int4)vec4;
ivec4 -= ivec4;
ivec4 |= ivec4;
ivec4 += ptr; // expected-error {{can't convert between vector values of different size ('int4' and 'int *')}}
}
typedef __attribute__(( ext_vector_type(2) )) float2 vecfloat2; // expected-error{{invalid vector element type 'float2'}}
void inc(float2 f2) {
f2++; // expected-error{{cannot increment value of type 'float2'}}
__real f2; // expected-error{{invalid type 'float2' to __real operator}}
}