[ARM][AArch64] Implement __arm_rsrf, __arm_rsrf64, __arm_wsrf & __arm_wsrf64

[vhscampos]

Summary

Adding support for ACLE intrinsics.

Patch by Michael Platings.

[vhscampos]

Run clang-format

[chill]

I'm a bit unsure how much we're willing to pollute the namespace: in this particular case looking at the four __bitcast_* functions.
I appreciate that getting rid of them would require emitting LLVM IR bitcasts, so a bit more effort compared to the macro approach.

Comments?

[efriedma]

We have __builtin_bit_cast now.

[vhscampos]

Use __builtin_bit_cast to perform the relevant bitcasts.

clang/lib/Headers/arm_acle.h

 /* 10.1 Special register intrinsics */
 #define __arm_rsr(sysreg) __builtin_arm_rsr(sysreg)
 #define __arm_rsr64(sysreg) __builtin_arm_rsr64(sysreg)
 #define __arm_rsrp(sysreg) __builtin_arm_rsrp(sysreg)
 #define __arm_wsr(sysreg, v) __builtin_arm_wsr(sysreg, v)
 #define __arm_wsr64(sysreg, v) __builtin_arm_wsr64(sysreg, v)
 #define __arm_wsrp(sysreg, v) __builtin_arm_wsrp(sysreg, v)
 
+static __inline__ float __attribute__((__always_inline__, __nodebug__))
+__bit_cast_to_float_from_uint32(uint32_t __from) {
+  float __to;
+  __builtin_memcpy(&__to, &__from, sizeof(__to));
+  return __to;
+}
+static __inline__ uint32_t __attribute__((__always_inline__, __nodebug__))
+__bit_cast_to_uint32_from_float(float __from) {
+  uint32_t __to;
+  __builtin_memcpy(&__to, &__from, sizeof(__to));
+  return __to;
+}
+static __inline__ double __attribute__((__always_inline__, __nodebug__))
+__bit_cast_to_double_from_uint64(uint64_t __from) {
+  double __to;
+  __builtin_memcpy(&__to, &__from, sizeof(__to));
+  return __to;
+}
+static __inline__ uint64_t __attribute__((__always_inline__, __nodebug__))
+__bit_cast_to_uint64_from_double(double __from) {
+  uint64_t __to;
+  __builtin_memcpy(&__to, &__from, sizeof(__to));
+  return __to;
+}
+#define __arm_rsrf(sysreg) __bit_cast_to_float_from_uint32(__arm_rsr(sysreg))
+#define __arm_wsrf(sysreg, v)                                                  \
+  __arm_wsr(sysreg, __bit_cast_to_uint32_from_float(v))
+#define __arm_rsrf64(sysreg)                                                   \
+  __bit_cast_to_double_from_uint64(__arm_rsr64(sysreg))
+#define __arm_wsrf64(sysreg, v)                                                \
+  __arm_wsr64(sysreg, __bit_cast_to_uint64_from_double(v))
+
 /* Memory Tagging Extensions (MTE) Intrinsics */
 #if __ARM_FEATURE_MEMORY_TAGGING
 #define __arm_mte_create_random_tag(__ptr, __mask)  __builtin_arm_irg(__ptr, __mask)
 #define __arm_mte_increment_tag(__ptr, __tag_offset)  __builtin_arm_addg(__ptr, __tag_offset)
 #define __arm_mte_exclude_tag(__ptr, __excluded)  __builtin_arm_gmi(__ptr, __excluded)
 #define __arm_mte_get_tag(__ptr) __builtin_arm_ldg(__ptr)
 #define __arm_mte_set_tag(__ptr) __builtin_arm_stg(__ptr)
 #define __arm_mte_ptrdiff(__ptra, __ptrb) __builtin_arm_subp(__ptra, __ptrb)

clang/test/CodeGen/arm_acle.c

 
 // ARM-LABEL: test_wsrp
 // AArch64: call void @llvm.write_register.i64(metadata ![[M1:[0-9]]], i64 %{{.*}})
 // AArch32: call void @llvm.write_register.i32(metadata ![[M4:[0-9]]], i32 %{{.*}})
 void test_wsrp(void *v) {
   __arm_wsrp("sysreg", v);
 }
 
+// ARM-LABEL: test_rsrf
+// AArch64: call i64 @llvm.read_register.i64(metadata ![[M0:[0-9]]])
+// AArch32: call i32 @llvm.read_register.i32(metadata ![[M2:[0-9]]])
+// ARM-NOT: uitofp
+// ARM: bitcast
+float test_rsrf() {
+#ifdef __ARM_32BIT_STATE
+  return __arm_rsrf("cp1:2:c3:c4:5");
+#else
+  return __arm_rsrf("1:2:3:4:5");
+#endif
+}
+// ARM-LABEL: test_rsrf64
+// AArch64: call i64 @llvm.read_register.i64(metadata ![[M0:[0-9]]])
+// AArch32: call i64 @llvm.read_register.i64(metadata ![[M3:[0-9]]])
+// ARM-NOT: uitofp
+// ARM: bitcast
+double test_rsrf64() {
+#ifdef __ARM_32BIT_STATE
+  return __arm_rsrf64("cp1:2:c3");
+#else
+  return __arm_rsrf64("1:2:3:4:5");
+#endif
+}
+// ARM-LABEL: test_wsrf
+// ARM-NOT: fptoui
+// ARM: bitcast
+// AArch64: call void @llvm.write_register.i64(metadata ![[M0:[0-9]]], i64 %{{.*}})
+// AArch32: call void @llvm.write_register.i32(metadata ![[M2:[0-9]]], i32 %{{.*}})
+void test_wsrf(float v) {
+#ifdef __ARM_32BIT_STATE
+  __arm_wsrf("cp1:2:c3:c4:5", v);
+#else
+  __arm_wsrf("1:2:3:4:5", v);
+#endif
+}
+// ARM-LABEL: test_wsrf64
+// ARM-NOT: fptoui
+// ARM: bitcast
+// AArch64: call void @llvm.write_register.i64(metadata ![[M0:[0-9]]], i64 %{{.*}})
+// AArch32: call void @llvm.write_register.i64(metadata ![[M3:[0-9]]], i64 %{{.*}})
+void test_wsrf64(double v) {
+#ifdef __ARM_32BIT_STATE
+  __arm_wsrf64("cp1:2:c3", v);
+#else
+  __arm_wsrf64("1:2:3:4:5", v);
+#endif
+}
+
 // AArch32: ![[M2]] = !{!"cp1:2:c3:c4:5"}
 // AArch32: ![[M3]] = !{!"cp1:2:c3"}
 // AArch32: ![[M4]] = !{!"sysreg"}
 
 // AArch64: ![[M0]] = !{!"1:2:3:4:5"}
 // AArch64: ![[M1]] = !{!"sysreg"}
 
 // AArch64-v8_3-LABEL: @test_jcvt(
 // AArch64-v8_3: call i32 @llvm.aarch64.fjcvtzs
 #ifdef __ARM_64BIT_STATE
 int32_t test_jcvt(double v) {
   return __jcvt(v);
 }
 #endif