diff --git a/clang/include/clang/Basic/arm_neon.td b/clang/include/clang/Basic/arm_neon.td
diff --git a/clang/include/clang/Basic/arm_neon_incl.td b/clang/include/clang/Basic/arm_neon_incl.td
@@ -198,10 +198,8 @@ def OP_UNAVAILABLE : Operation {
 //
 // The prototype is a string that defines the return type of the intrinsic
 // and the type of each argument. The return type and every argument gets a
-// "modifier" that can change in some way the "base type" of the intrinsic.
-//
-// The modifier 'd' means "default" and does not modify the base type in any
-// way. The available modifiers are given below.
+// set of "modifiers" that can change in some way the "base type" of the
+// intrinsic.
 //
 // Typespecs
 // ---------
@@ -226,41 +224,34 @@ def OP_UNAVAILABLE : Operation {
 // -------------------
 // prototype: return (arg, arg, ...)
 //
-// v: void
-// t: best-fit integer (int/poly args)
-// x: signed integer   (int/float args)
-// u: unsigned integer (int/float args)
-// f: float (int args)
-// F: double (int args)
-// H: half (int args)
-// 0: half (int args), ignore 'Q' size modifier.
-// 1: half (int args), force 'Q' size modifier.
-// d: default
-// g: default, ignore 'Q' size modifier.
-// j: default, force 'Q' size modifier.
-// w: double width elements, same num elts
-// n: double width elements, half num elts
-// h: half width elements, double num elts
-// q: half width elements, quad num elts
-// e: half width elements, double num elts, unsigned
-// m: half width elements, same num elts
-// i: constant int
-// l: constant uint64
-// s: scalar of element type
-// z: scalar of half width element type, signed
-// r: scalar of double width element type, signed
-// b: scalar of unsigned integer/long type (int/float args)
-// $: scalar of signed integer/long type (int/float args)
-// y: scalar of float
-// o: scalar of double
-// k: default elt width, double num elts
-// 2,3,4: array of default vectors
-// B,C,D: array of default elts, force 'Q' size modifier.
-// p: pointer type
-// c: const pointer type
-// 7: vector of 8-bit elements, ignore 'Q' size modifier
-// 8: vector of 8-bit elements, same width as default type
-// 9: vector of 8-bit elements, force 'Q' size modifier
+// Each type modifier is either a single character, or a group surrounded by
+// parentheses.
+//
+// .: default
+// v: change to void category.
+// S: change to signed integer category.
+// U: change to unsigned integer category.
+// F: change to floating category.
+// P: change to polynomial category.
+// p: change polynomial to equivalent integer category. Otherwise nop.
+//
+// >: double element width (vector size unchanged).
+// <: half element width (vector size unchanged).
+//
+// 1: change to scalar.
+// 2: change to struct of two vectors.
+// 3: change to struct of three vectors.
+// 4: change to struct of four vectors.
+//
+// *: make a pointer argument.
+// c: make a constant argument (for pointers).
+//
+// Q: force 128-bit width.
+// q: force 64-bit width.
+//
+// I: make 32-bit signed scalar immediate
+// !: make this the key type passed to CGBuiltin.cpp in a polymorphic call.
+
 
 // Every intrinsic subclasses Inst.
 class Inst <string n, string p, string t, Operation o> {

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -5456,6 +5456,11 @@ Value *CodeGenFunction::EmitCommonNeonBuiltinExpr(
     llvm::Type *Tys[2] = { Ty, GetFloatNeonType(this, Type) };
     return EmitNeonCall(CGM.getIntrinsic(LLVMIntrinsic, Tys), Ops, NameHint);
   }
+  case NEON::BI__builtin_neon_vcvtx_f32_v: {
+    llvm::Type *Tys[2] = { VTy->getTruncatedElementVectorType(VTy), Ty};
+    return EmitNeonCall(CGM.getIntrinsic(LLVMIntrinsic, Tys), Ops, NameHint);
+
+  }
   case NEON::BI__builtin_neon_vext_v:
   case NEON::BI__builtin_neon_vextq_v: {
     int CV = cast<ConstantInt>(Ops[2])->getSExtValue();

diff --git a/clang/test/CodeGen/aarch64-neon-intrinsics.c b/clang/test/CodeGen/aarch64-neon-intrinsics.c
@@ -17756,8 +17756,6 @@ float32_t test_vminnmv_f32(float32x2_t a) {
 }
 
 // CHECK-LABEL: @test_vpaddq_s64(
-// CHECK:   [[TMP0:%.*]] = bitcast <2 x i64> %a to <16 x i8>
-// CHECK:   [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
 // CHECK:   [[VPADDQ_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.addp.v2i64(<2 x i64> %a, <2 x i64> %b)
 // CHECK:   [[VPADDQ_V3_I:%.*]] = bitcast <2 x i64> [[VPADDQ_V2_I]] to <16 x i8>
 // CHECK:   ret <2 x i64> [[VPADDQ_V2_I]]
@@ -17766,8 +17764,6 @@ int64x2_t test_vpaddq_s64(int64x2_t a, int64x2_t b) {
 }
 
 // CHECK-LABEL: @test_vpaddq_u64(
-// CHECK:   [[TMP0:%.*]] = bitcast <2 x i64> %a to <16 x i8>
-// CHECK:   [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
 // CHECK:   [[VPADDQ_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.addp.v2i64(<2 x i64> %a, <2 x i64> %b)
 // CHECK:   [[VPADDQ_V3_I:%.*]] = bitcast <2 x i64> [[VPADDQ_V2_I]] to <16 x i8>
 // CHECK:   ret <2 x i64> [[VPADDQ_V2_I]]

diff --git a/clang/test/CodeGen/aarch64-neon-scalar-x-indexed-elem.c b/clang/test/CodeGen/aarch64-neon-scalar-x-indexed-elem.c
@@ -407,12 +407,10 @@ int64_t test_vqdmlsls_laneq_s32(int64_t a, int32_t b, int32x4_t c) {
 }
 
 // CHECK-LABEL: define <1 x double> @test_vmulx_lane_f64_0() #0 {
-// CHECK:   [[TMP0:%.*]] = bitcast i64 4599917171378402754 to <1 x double>
-// CHECK:   [[TMP1:%.*]] = bitcast i64 4606655882138939123 to <1 x double>
-// CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x double> [[TMP0]], i32 0
-// CHECK:   [[VGET_LANE7:%.*]] = extractelement <1 x double> [[TMP1]], i32 0
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x double> <double 0x3FD6304BC43AB5C2>, i32 0
+// CHECK:   [[VGET_LANE7:%.*]] = extractelement <1 x double> <double 0x3FEE211E215AEEF3>, i32 0
 // CHECK:   [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGET_LANE7]])
-// CHECK:   [[VSET_LANE:%.*]] = insertelement <1 x double> [[TMP0]], double [[VMULXD_F64_I]], i32 0
+// CHECK:   [[VSET_LANE:%.*]] = insertelement <1 x double> <double 0x3FD6304BC43AB5C2>, double [[VMULXD_F64_I]], i32 0
 // CHECK:   ret <1 x double> [[VSET_LANE]]
 float64x1_t test_vmulx_lane_f64_0() {
       float64x1_t arg1;
@@ -426,13 +424,11 @@ float64x1_t test_vmulx_lane_f64_0() {
 }
 
 // CHECK-LABEL: define <1 x double> @test_vmulx_laneq_f64_2() #1 {
-// CHECK:   [[TMP0:%.*]] = bitcast i64 4599917171378402754 to <1 x double>
-// CHECK:   [[TMP1:%.*]] = bitcast i64 4606655882138939123 to <1 x double>
-// CHECK:   [[SHUFFLE_I:%.*]] = shufflevector <1 x double> [[TMP0]], <1 x double> [[TMP1]], <2 x i32> <i32 0, i32 1>
-// CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x double> [[TMP0]], i32 0
+// CHECK:   [[SHUFFLE_I:%.*]] = shufflevector <1 x double> <double 0x3FD6304BC43AB5C2>, <1 x double> <double 0x3FEE211E215AEEF3>, <2 x i32> <i32 0, i32 1>
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x double> <double 0x3FD6304BC43AB5C2>, i32 0
 // CHECK:   [[VGETQ_LANE:%.*]] = extractelement <2 x double> [[SHUFFLE_I]], i32 1
 // CHECK:   [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGETQ_LANE]])
-// CHECK:   [[VSET_LANE:%.*]] = insertelement <1 x double> [[TMP0]], double [[VMULXD_F64_I]], i32 0
+// CHECK:   [[VSET_LANE:%.*]] = insertelement <1 x double> <double 0x3FD6304BC43AB5C2>, double [[VMULXD_F64_I]], i32 0
 // CHECK:   ret <1 x double> [[VSET_LANE]]
 float64x1_t test_vmulx_laneq_f64_2() {
       float64x1_t arg1;

diff --git a/clang/utils/TableGen/NeonEmitter.cpp b/clang/utils/TableGen/NeonEmitter.cpp
diff --git a/clang/utils/convert_arm_neon.py b/clang/utils/convert_arm_neon.py
@@ -0,0 +1,172 @@
+#!/usr/bin/env python3
+
+# This script was committed on 20/11/2019 and it would probably make sense to remove
+# it after the next release branches.
+
+# This script is pipe based and converts an arm_neon.td (or arm_fp16.td) file
+# using the old single-char type modifiers to an equivalent new-style form where
+# each modifier is orthogonal and they can be composed.
+#
+# It was used to directly generate the .td files on master, so if you have any
+# local additions I would suggest implementing any modifiers here, and running
+# it over your entire pre-merge .td files rather than trying to resolve any
+# conflicts manually.
+
+import re, sys
+MOD_MAP = {
+    'v': 'v',
+    'x': 'S',
+    'u': 'U',
+    'd': '.',
+    'g': 'q',
+    'j': 'Q',
+    'w': '>Q',
+    'n': '>',
+    'h': '<',
+    'q': '<Q',
+    'e': '<U',
+    'm': '<q',
+    'i': 'I',
+    'l': 'IU>',
+    's': '1',
+    'z': '1<',
+    'r': '1>',
+    'b': '1U',
+    '$': '1S',
+    'k': 'Q',
+    '2': '2',
+    '3': '3',
+    '4': '4',
+    'B': '2Q',
+    'C': '3Q',
+    'D': '4Q',
+    'p': '*',
+    'c': 'c*',
+    '7': '<<q',
+    '8': '<<',
+    '9': '<<Q',
+    't': 'p'
+    }
+
+
+def typespec_elt_size(typespec):
+    if 'c' in typespec:
+        return 8
+    elif 's' in typespec or 'h' in typespec:
+        return 16
+    elif 'i' in typespec or 'f' in typespec:
+        return 32
+    elif 'l' in typespec or 'd' in typespec:
+        return 64
+    elif 'k' in typespec:
+        return 128
+
+def get_resize(cur, desired):
+    res = ''
+    while cur < desired:
+        res += '>'
+        cur *= 2
+    while cur > desired:
+        res += '<'
+        cur /= 2
+    return res
+
+
+def remap_protocol(proto, typespec, name):
+    key_type = 0
+
+    # Conversions like to see the integer type so they know signedness.
+    if 'vcvt' in name and '_f' in name and name != 'vcvt_f32_f64' and name != 'vcvt_f64_f32':
+        key_type = 1
+    default_width = typespec_elt_size(typespec)
+    inconsistent_width = False
+    for elt in typespec:
+        new_width = typespec_elt_size(elt)
+        if new_width and new_width != default_width:
+            inconsistent_width = True
+
+    res = ''
+    for i, c in enumerate(proto):
+        # void and pointers make for bad discriminators in CGBuiltin.cpp.
+        if c in 'vcp':
+                key_type += 1
+
+        if c in MOD_MAP:
+            cur_mod = MOD_MAP[c]
+        elif inconsistent_width:
+            # Otherwise it's a fixed output width modifier.
+            sys.stderr.write(f'warning: {name} uses fixed output size but has inconsistent input widths: {proto} {typespec}\n')
+
+        if c == 'Y':
+            # y: scalar of half float
+            resize = get_resize(default_width, 16)
+            cur_mod = f'1F{resize}'
+        elif c == 'y':
+            # y: scalar of float
+            resize = get_resize(default_width, 32)
+            cur_mod = f'1F{resize}'
+        elif c == 'o':
+            # o: scalar of double
+            resize = get_resize(default_width, 64)
+            cur_mod = f'1F{resize}'
+        elif c == 'I':
+            # I: scalar of 32-bit signed
+            resize = get_resize(default_width, 32)
+            cur_mod = f'1S{resize}'
+        elif c == 'L':
+            # L: scalar of 64-bit signed
+            resize = get_resize(default_width, 64)
+            cur_mod = f'1S{resize}'
+        elif c == 'U':
+            # I: scalar of 32-bit unsigned
+            resize = get_resize(default_width, 32)
+            cur_mod = f'1U{resize}'
+        elif c == 'O':
+            # O: scalar of 64-bit unsigned
+            resize = get_resize(default_width, 64)
+            cur_mod = f'1U{resize}'
+        elif c == 'f':
+            # f: float (int args)
+            resize = get_resize(default_width, 32)
+            cur_mod = f'F{resize}'
+        elif c == 'F':
+            # F: double (int args)
+            resize = get_resize(default_width, 64)
+            cur_mod = f'F{resize}'
+        elif c == 'H':
+            # H: half (int args)
+            resize = get_resize(default_width, 16)
+            cur_mod = f'F{resize}'
+        elif c == '0':
+            # 0: half (int args), ignore 'Q' size modifier.
+            resize = get_resize(default_width, 16)
+            cur_mod = f'Fq{resize}'
+        elif c == '1':
+            # 1: half (int args), force 'Q' size modifier.
+            resize = get_resize(default_width, 16)
+            cur_mod = f'FQ{resize}'
+
+        if len(cur_mod) == 0:
+            raise Exception(f'WTF: {c} in {name}')
+
+        if key_type != 0 and key_type == i:
+            cur_mod += '!'
+
+        if len(cur_mod) == 1:
+            res += cur_mod
+        else:
+            res += '(' + cur_mod + ')'
+
+    return res
+
+def replace_insts(m):
+    start, end = m.span('proto')
+    start -= m.start()
+    end -= m.start()
+    new_proto = remap_protocol(m['proto'], m['kinds'], m['name'])
+    return m.group()[:start] + new_proto + m.group()[end:]
+
+INST = re.compile(r'Inst<"(?P<name>.*?)",\s*"(?P<proto>.*?)",\s*"(?P<kinds>.*?)"')
+
+new_td = INST.sub(replace_insts, sys.stdin.read())
+sys.stdout.write(new_td)