|  | // RUN: %clang_cc1 -triple thumbv7-apple-darwin10 -target-cpu cortex-a8 -fsyntax-only -Wvector-conversion -ffreestanding -verify %s | 
|  | // REQUIRES: aarch64-registered-target || arm-registered-target | 
|  |  | 
|  | #ifndef INCLUDE | 
|  |  | 
|  | #include <arm_neon.h> | 
|  |  | 
|  | // Should not report incompatible vector types. | 
|  | int32x2_t test(int32x2_t x) { | 
|  | return vshr_n_s32(x, 31); | 
|  | } | 
|  |  | 
|  | // ...but should warn when the types really do not match. | 
|  | float32x2_t test2(uint32x2_t x) { | 
|  | return vcvt_n_f32_s32(x, 9); // expected-warning {{incompatible vector types}} | 
|  | } | 
|  |  | 
|  | // Check immediate range for vcvt_n intrinsics is 1 to 32. | 
|  | float32x2_t test3(uint32x2_t x) { | 
|  | // FIXME: The "incompatible result type" error is due to pr10112 and should be | 
|  | // removed when that is fixed. | 
|  | return vcvt_n_f32_u32(x, 0); // expected-error-re {{argument value {{.*}} is outside the valid range}} | 
|  | } | 
|  |  | 
|  | typedef signed int vSInt32 __attribute__((__vector_size__(16))); | 
|  | int32x4_t test4(int32x4_t a, vSInt32 b) { | 
|  | a += b; | 
|  | b += a; | 
|  | return b += a; | 
|  | } | 
|  |  | 
|  | // Warn for incompatible pointer types used with vld/vst intrinsics. | 
|  | int16x8_t test5(int *p) { | 
|  | return vld1q_s16(p); // expected-warning {{incompatible pointer types}} | 
|  | } | 
|  | void test6(float *p, int32x2_t v) { | 
|  | return vst1_s32(p, v); // expected-warning {{incompatible pointer types}} | 
|  | } | 
|  |  | 
|  | #define INCLUDE | 
|  | #include "arm-neon-types.c" | 
|  | #else | 
|  |  | 
|  | // Make sure we don't get a warning about using a static function in an | 
|  | // extern inline function from a header. | 
|  | extern inline uint8x8_t test7(uint8x8_t a, uint8x8_t b) { | 
|  | return vadd_u8(a, b); | 
|  | } | 
|  | #endif |