Home | History | Annotate | Download | only in CodeGen
      1 // RUN: %clang_cc1 -O3 -triple arm64-apple-ios7 -target-feature +neon -ffreestanding -S -o - %s | FileCheck %s
      2 // RUN: %clang_cc1 -O3 -triple arm64-apple-ios7 -target-feature +neon -ffreestanding -S -o - -emit-llvm %s | \
      3 // RUN:   FileCheck -check-prefix=CHECK-IR %s
      4 // REQUIRES: aarch64-registered-target
      5 
      6 /// Test vdupq_n_f64 and vmovq_nf64 ARM64 intrinsics
      7 // <rdar://problem/11778405> ARM64: vdupq_n_f64 and vdupq_lane_f64 intrinsics
      8 // missing
      9 
     10 
     11 #include <arm_neon.h>
     12 
     13 // vdupq_n_f64 -> dup.2d v0, v0[0]
     14 //
     15 float64x2_t test_vdupq_n_f64(float64_t w)
     16 {
     17     return vdupq_n_f64(w);
     18   // CHECK-LABEL: test_vdupq_n_f64:
     19   // CHECK: dup.2d v0, v0[0]
     20   // CHECK-NEXT: ret
     21 }
     22 
     23 // might as well test this while we're here
     24 // vdupq_n_f32 -> dup.4s v0, v0[0]
     25 float32x4_t test_vdupq_n_f32(float32_t w)
     26 {
     27     return vdupq_n_f32(w);
     28   // CHECK-LABEL: test_vdupq_n_f32:
     29   // CHECK: dup.4s v0, v0[0]
     30   // CHECK-NEXT: ret
     31 }
     32 
     33 // vdupq_lane_f64 -> dup.2d v0, v0[0]
     34 // this was in <rdar://problem/11778405>, but had already been implemented,
     35 // test anyway
     36 float64x2_t test_vdupq_lane_f64(float64x1_t V)
     37 {
     38     return vdupq_lane_f64(V, 0);
     39   // CHECK-LABEL: test_vdupq_lane_f64:
     40   // CHECK: dup.2d v0, v0[0]
     41   // CHECK-NEXT: ret
     42 }
     43 
     44 // vmovq_n_f64 -> dup Vd.2d,X0
     45 // this wasn't in <rdar://problem/11778405>, but it was between the vdups
     46 float64x2_t test_vmovq_n_f64(float64_t w)
     47 {
     48   return vmovq_n_f64(w);
     49   // CHECK-LABEL: test_vmovq_n_f64:
     50   // CHECK: dup.2d v0, v0[0]
     51   // CHECK-NEXT: ret
     52 }
     53 
     54 float16x4_t test_vmov_n_f16(float16_t *a1)
     55 {
     56   // CHECK-IR-LABEL: test_vmov_n_f16
     57   return vmov_n_f16(*a1);
     58   // CHECK-IR: insertelement {{.*}} i32 0{{ *$}}
     59   // CHECK-IR: insertelement {{.*}} i32 1{{ *$}}
     60   // CHECK-IR: insertelement {{.*}} i32 2{{ *$}}
     61   // CHECK-IR: insertelement {{.*}} i32 3{{ *$}}
     62 }
     63 
     64 // Disable until scalar problem in backend is fixed. Change CHECK-IR@ to
     65 // CHECK-IR<colon>
     66 /*
     67 float64x1_t test_vmov_n_f64(float64_t a1)
     68 {
     69   // CHECK-IR@ test_vmov_n_f64
     70   return vmov_n_f64(a1);
     71   // CHECK-IR@ insertelement {{.*}} i32 0{{ *$}}
     72 }
     73 */
     74 
     75 float16x8_t test_vmovq_n_f16(float16_t *a1)
     76 {
     77   // CHECK-IR-LABEL: test_vmovq_n_f16
     78   return vmovq_n_f16(*a1);
     79   // CHECK-IR: insertelement {{.*}} i32 0{{ *$}}
     80   // CHECK-IR: insertelement {{.*}} i32 1{{ *$}}
     81   // CHECK-IR: insertelement {{.*}} i32 2{{ *$}}
     82   // CHECK-IR: insertelement {{.*}} i32 3{{ *$}}
     83   // CHECK-IR: insertelement {{.*}} i32 4{{ *$}}
     84   // CHECK-IR: insertelement {{.*}} i32 5{{ *$}}
     85   // CHECK-IR: insertelement {{.*}} i32 6{{ *$}}
     86   // CHECK-IR: insertelement {{.*}} i32 7{{ *$}}
     87 }
     88 
     89