clang/test/CodeGen/AArch64/sve-acle-__ARM_FEATURE_SVE_VECTOR_OPERATORS.c

   1 // RUN: %clang_cc1 -triple aarch64 -target-feature +sve -O1 -Werror -Wall -emit-llvm -o - %s -mvscale-min=1 -mvscale-max=1 | FileCheck %s -D#VBITS=128 --check-prefixes=CHECK128
   2 // RUN: %clang_cc1 -triple aarch64 -target-feature +sve -O1 -Werror -Wall -emit-llvm -o - %s -mvscale-min=2 -mvscale-max=2 | FileCheck %s -D#VBITS=256 --check-prefixes=CHECK,CHECK256
   3 // RUN: %clang_cc1 -triple aarch64 -target-feature +sve -O1 -Werror -Wall -emit-llvm -o - %s -mvscale-min=4 -mvscale-max=4 | FileCheck %s -D#VBITS=512 --check-prefixes=CHECK,CHECK512
   4 // RUN: %clang_cc1 -triple aarch64 -target-feature +sve -O1 -Werror -Wall -emit-llvm -o - %s -mvscale-min=8 -mvscale-max=8 | FileCheck %s -D#VBITS=1024 --check-prefixes=CHECK,CHECK1024
   5 // RUN: %clang_cc1 -triple aarch64 -target-feature +sve -O1 -Werror -Wall -emit-llvm -o - %s -mvscale-min=16 -mvscale-max=16 | FileCheck %s -D#VBITS=2048 --check-prefixes=CHECK,CHECK2048
   6
   7 // REQUIRES: aarch64-registered-target
   8
   9 // Examples taken from section "3.7.3.3 Behavior specific to SVE
  10 // vectors" of the SVE ACLE (Version 00bet6) that can be found at
  11 // https://developer.arm.com/documentation/100987/latest
  12 //
  13 // Example has been expanded to work with mutiple values of
  14 // -mvscale-{min,max}.
  15
  16 #include <arm_sve.h>
  17
  18 // Page 27, item 1
  19 #if __ARM_FEATURE_SVE_BITS == 256 && __ARM_FEATURE_SVE_VECTOR_OPERATORS
  20 // CHECK256-LABEL: @x256 ={{.*}} local_unnamed_addr global <4 x i64> <i64 0, i64 1, i64 2, i64 3>, align 16
  21 typedef svint64_t vec256 __attribute__((arm_sve_vector_bits(256)));
  22 vec256 x256 = {0, 1, 2, 3};
  23 #endif
  24
  25 #if __ARM_FEATURE_SVE_BITS == 512 && __ARM_FEATURE_SVE_VECTOR_OPERATORS
  26 // CHECK512-LABEL: @x512 ={{.*}} local_unnamed_addr global <8 x i64> <i64 0, i64 1, i64 2, i64 3, i64 3, i64 2, i64 1, i64 0>, align 16
  27 typedef svint64_t vec512 __attribute__((arm_sve_vector_bits(512)));
  28 vec512 x512 = {0, 1, 2, 3, 3 , 2 , 1, 0};
  29 #endif
  30
  31 #if __ARM_FEATURE_SVE_BITS == 1024 && __ARM_FEATURE_SVE_VECTOR_OPERATORS
  32 // CHECK1024-LABEL: @x1024 ={{.*}} local_unnamed_addr global <16 x i64> <i64 0, i64 1, i64 2, i64 3, i64 3, i64 2, i64 1, i64 0, i64 0, i64 1, i64 2, i64 3, i64 3, i64 2, i64 1, i64 0>, align 16
  33 typedef svint64_t vec1024 __attribute__((arm_sve_vector_bits(1024)));
  34 vec1024 x1024 = {0, 1, 2, 3, 3 , 2 , 1, 0, 0, 1, 2, 3, 3 , 2 , 1, 0};
  35 #endif
  36
  37 #if __ARM_FEATURE_SVE_BITS == 2048 && __ARM_FEATURE_SVE_VECTOR_OPERATORS
  38 // CHECK2048-LABEL: @x2048 ={{.*}} local_unnamed_addr global <32 x i64> <i64 0, i64 1, i64 2, i64 3, i64 3, i64 2, i64 1, i64 0, i64 0, i64 1, i64 2, i64 3, i64 3, i64 2, i64 1, i64 0, i64 0, i64 1, i64 2, i64 3, i64 3, i64 2, i64 1, i64 0, i64 0, i64 1, i64 2, i64 3, i64 3, i64 2, i64 1, i64 0>, align 16
  39 typedef svint64_t vec2048 __attribute__((arm_sve_vector_bits(2048)));
  40 vec2048 x2048 = {0, 1, 2, 3, 3 , 2 , 1, 0, 0, 1, 2, 3, 3 , 2 , 1, 0,
  41                  0, 1, 2, 3, 3 , 2 , 1, 0, 0, 1, 2, 3, 3 , 2 , 1, 0};
  42 #endif
  43
  44 // Page 27, item 2. We can not change the ABI of existing vector
  45 // types, including vec_int8.  That's why in the SVE ACLE, VLST is
  46 // distinct from, but mostly interchangeable with, the corresponding
  47 // GNUT. VLST is treated for ABI purposes like an SVE type but GNUT
  48 // continues to be a normal GNU vector type, with base Armv8-A PCS
  49 // rules.
  50 #if __ARM_FEATURE_SVE_BITS && __ARM_FEATURE_SVE_VECTOR_OPERATORS
  51 #define N __ARM_FEATURE_SVE_BITS
  52 typedef int8_t vec_int8 __attribute__((vector_size(N / 8)));
  53 // CHECK128-LABEL: define{{.*}} <16 x i8> @f2(<16 x i8> noundef %x)
  54 // CHECK128-NEXT:  entry:
  55 // CHECK128-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
  56 // CHECK128-NEXT:    [[CASTSCALABLESVE:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.insert.nxv16i8.v16i8(<vscale x 16 x i8> poison, <16 x i8> [[X:%.*]], i64 0)
  57 // CHECK128-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.asrd.nxv16i8(<vscale x 16 x i1> [[TMP0]], <vscale x 16 x i8> [[CASTSCALABLESVE]], i32 1)
  58 // CHECK128-NEXT:    [[CASTFIXEDSVE:%.*]] = tail call <16 x i8> @llvm.vector.extract.v16i8.nxv16i8(<vscale x 16 x i8> [[TMP1]], i64 0)
  59 // CHECK128-NEXT:    ret <16 x i8> [[CASTFIXEDSVE]]
  60
  61 // CHECK-LABEL: define{{.*}} void @f2(
  62 // CHECK-SAME:   ptr dead_on_unwind noalias nocapture writable writeonly sret(<[[#div(VBITS,8)]] x i8>) align 16 initializes((0, [[#div(VBITS,8)]])) %agg.result, ptr nocapture noundef readonly %0)
  63 // CHECK-NEXT: entry:
  64 // CHECK-NEXT:   [[X:%.*]] = load <[[#div(VBITS,8)]] x i8>, ptr [[TMP0:%.*]], align 16, [[TBAA6:!tbaa !.*]]
  65 // CHECK-NEXT:   [[TMP1:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
  66 // CHECK-NEXT:   [[CASTSCALABLESVE:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.insert.nxv16i8.v[[#div(VBITS,8)]]i8(<vscale x 16 x i8> poison, <[[#div(VBITS,8)]] x i8> [[X]], i64 0)
  67 // CHECK-NEXT:   [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.asrd.nxv16i8(<vscale x 16 x i1> [[TMP1]], <vscale x 16 x i8> [[CASTSCALABLESVE]], i32 1)
  68 // CHECK-NEXT:   [[CASTFIXEDSVE:%.*]] = tail call <[[#div(VBITS,8)]] x i8> @llvm.vector.extract.v[[#div(VBITS,8)]]i8.nxv16i8(<vscale x 16 x i8> [[TMP2]], i64 0)
  69 // CHECK-NEXT:   store <[[#div(VBITS,8)]] x i8> [[CASTFIXEDSVE]], ptr [[AGG_RESULT:%.*]], align 16, [[TBAA6]]
  70 // CHECK-NEXT:   ret void
  71 vec_int8 f2(vec_int8 x) { return svasrd_x(svptrue_b8(), x, 1); }
  72 #endif
  73
  74 // Page 27, item 3.
  75 #if __ARM_FEATURE_SVE_BITS && __ARM_FEATURE_SVE_VECTOR_OPERATORS
  76 #define N __ARM_FEATURE_SVE_BITS
  77 typedef int8_t vec1 __attribute__((vector_size(N / 8)));
  78 void f3(vec1);
  79 typedef svint8_t vec2 __attribute__((arm_sve_vector_bits(N)));
  80
  81 // CHECK128-LABEL: define{{.*}} void @g(<vscale x 16 x i8> noundef %x.coerce)
  82 // CHECK128-NEXT: entry:
  83 // CHECK128-NEXT:  [[X:%.*]] = tail call <16 x i8> @llvm.vector.extract.v16i8.nxv16i8(<vscale x 16 x i8> [[X_COERCE:%.*]], i64 0)
  84 // CHECK128-NEXT:    call void @f3(<16 x i8> noundef [[X]]) [[ATTR5:#.*]]
  85 // CHECK128-NEXT:    ret void
  86
  87 // CHECK-LABEL: define{{.*}} void @g(<vscale x 16 x i8> noundef %x.coerce)
  88 // CHECK-NEXT: entry:
  89 // CHECK-NEXT:   [[INDIRECT_ARG_TEMP:%.*]] = alloca <[[#div(VBITS,8)]] x i8>, align 16
  90 // CHECK-NEXT:   [[X:%.*]] = tail call <[[#div(VBITS,8)]] x i8> @llvm.vector.extract.v[[#div(VBITS,8)]]i8.nxv16i8(<vscale x 16 x i8> [[X_COERCE:%.*]], i64 0)
  91 // CHECK-NEXT:   call void @llvm.lifetime.start.p0(i64 [[SIZE:[0-9]+]], ptr nonnull [[INDIRECT_ARG_TEMP]]) #[[ATTR6:[0-9]+]]
  92 // CHECK-NEXT:   store <[[#div(VBITS,8)]] x i8> [[X]], ptr [[INDIRECT_ARG_TEMP]], align 16, [[TBAA6]]
  93 // CHECK-NEXT:   call void @f3(ptr noundef nonnull [[INDIRECT_ARG_TEMP]]) [[ATTR5:#.*]]
  94 // CHECK-NEXT:   call void @llvm.lifetime.end.p0(i64 [[SIZE]], ptr nonnull [[INDIRECT_ARG_TEMP]]) #[[ATTR6:[0-9]+]]
  95 // CHECK-NEXT:   ret void
  96
  97 // CHECK128-LABEL: declare void @f3(<16 x i8> noundef)
  98
  99 // CHECK-LABEL: declare void @f3(
 100 // CHECK-SAME:   ptr noundef)
 101 void g(vec2 x) { f3(x); } // OK
 102 #endif