Add auto-vectorization of pointwise operators (#1047)

* Enable auto vectorization
* Handle vector types with convert function
* Dont vectorize when it will cause problems with preload

src/cpp_generator.cpp

 #include <migraphx/cpp_generator.hpp>
 #include <migraphx/module.hpp>
 #include <migraphx/operation.hpp>
+#include <migraphx/ranges.hpp>
 #include <migraphx/instruction.hpp>
 #include <migraphx/builtin.hpp>
 #include <migraphx/stringutils.hpp>
@@ -51,6 +52,7 @@ cpp_generator::function& cpp_generator::function::set_types(const module& m)
 cpp_generator::function&
 cpp_generator::function::set_types(const module& m, const std::function<std::string(shape)>& parse)
 {
+    this->params.clear();
     auto pmap = m.get_parameter_shapes();
     std::map<std::string, shape> input_map(pmap.begin(), pmap.end());
     std::transform(
@@ -63,11 +65,30 @@ cpp_generator::function::set_types(const module& m, const std::function<std::str
     return *this;
 }
 
+cpp_generator::function& cpp_generator::function::set_generic_types(const module& m)
+{
+    this->params.clear();
+    auto pmap = m.get_parameter_shapes();
+    std::map<std::string, shape> input_map(pmap.begin(), pmap.end());
+    std::transform(
+        input_map.begin(), input_map.end(), std::back_inserter(this->params), [&](auto&& p) {
+            return param{p.first, "T" + p.first};
+        });
+
+    std::transform(input_map.begin(),
+                   input_map.end(),
+                   std::back_inserter(this->tparams),
+                   [&](auto&& p) { return "class T" + p.first; });
+    this->return_type = "auto";
+    return *this;
+}
+
 struct cpp_generator_impl
 {
     std::stringstream fs{};
-    std::size_t function_count                   = 0;
-    std::function<std::string(std::string)> fmap = nullptr;
+    std::size_t function_count                                = 0;
+    std::function<std::string(std::string)> fmap              = nullptr;
+    std::unordered_map<std::string, std::string> point_op_map = {};
 };
 cpp_generator::cpp_generator() : impl(std::make_unique<cpp_generator_impl>()) {}
 
@@ -83,41 +104,54 @@ cpp_generator::~cpp_generator() noexcept = default;
 
 void cpp_generator::fmap(const std::function<std::string(std::string)>& f) { impl->fmap = f; }
 
+void cpp_generator::add_point_op(const std::string& op_name, const std::string& code)
+{
+    impl->point_op_map[op_name] = code;
+}
+
 std::string cpp_generator::generate_point_op(const operation& op,
                                              const std::vector<std::string>& args)
 {
-    auto v          = op.to_value();
-    auto attributes = op.attributes();
-    if(not attributes.contains("point_op"))
-        MIGRAPHX_THROW("op is missing point_op attribute: " + op.name());
-    return interpolate_string(attributes["point_op"].to<std::string>(),
-                              [&](auto start, auto last) -> std::string {
-                                  auto key = trim({start, last});
-                                  if(key.empty())
-                                      MIGRAPHX_THROW("Empty parameter");
-                                  std::string fselector = "function:";
-                                  if(starts_with(key, fselector))
-                                  {
-                                      auto fname = key.substr(fselector.size());
-                                      if(impl->fmap == nullptr)
-                                          return fname;
-                                      else
-                                          return impl->fmap(fname);
-                                  }
-                                  else if(with_char(::isdigit)(key[0]))
-                                  {
-                                      auto i = std::stoul(key);
-                                      return args.at(i);
-                                  }
-                                  else if(v.contains(key))
-                                  {
-                                      return v[key].template to<std::string>();
-                                  }
-                                  else
-                                  {
-                                      return key;
-                                  }
-                              });
+    auto v = op.to_value();
+    std::string code;
+    if(contains(impl->point_op_map, op.name()))
+    {
+        code = impl->point_op_map.at(op.name());
+    }
+    else
+    {
+        auto attributes = op.attributes();
+        if(not attributes.contains("point_op"))
+            MIGRAPHX_THROW("op is missing point_op attribute: " + op.name());
+        code = attributes["point_op"].to<std::string>();
+    }
+    return interpolate_string(code, [&](auto start, auto last) -> std::string {
+        auto key = trim({start, last});
+        if(key.empty())
+            MIGRAPHX_THROW("Empty parameter");
+        std::string fselector = "function:";
+        if(starts_with(key, fselector))
+        {
+            auto fname = key.substr(fselector.size());
+            if(impl->fmap == nullptr)
+                return fname;
+            else
+                return impl->fmap(fname);
+        }
+        else if(with_char(::isdigit)(key[0]))
+        {
+            auto i = std::stoul(key);
+            return args.at(i);
+        }
+        else if(v.contains(key))
+        {
+            return v[key].template to<std::string>();
+        }
+        else
+        {
+            return key;
+        }
+    });
 }
 
 std::string cpp_generator::str() const { return impl->fs.str(); }
@@ -148,6 +182,8 @@ cpp_generator::function cpp_generator::generate_module(const module& m)
 std::string cpp_generator::create_function(const cpp_generator::function& f)
 {
     impl->function_count++;
+    if(not f.tparams.empty())
+        impl->fs << "template<" << join_strings(f.tparams, ", ") << ">\n";
     std::string name = f.name.empty() ? "f" + std::to_string(impl->function_count) : f.name;
     impl->fs << join_strings(f.attributes, " ") << " " << f.return_type << " " << name;
     char delim = '(';

src/include/migraphx/cpp_generator.hpp

@@ -34,6 +34,7 @@ struct cpp_generator
         std::string return_type             = "void";
         std::string name                    = "";
         std::vector<std::string> attributes = {};
+        std::vector<std::string> tparams    = {};
         function& set_body(const module& m, const generate_module_callback& g);
         function& set_body(const std::string& s)
         {
@@ -52,6 +53,7 @@ struct cpp_generator
         }
         function& set_types(const module& m);
         function& set_types(const module& m, const std::function<std::string(shape)>& parse);
+        function& set_generic_types(const module& m);
     };
 
     cpp_generator();
@@ -66,6 +68,8 @@ struct cpp_generator
 
     void fmap(const std::function<std::string(std::string)>& f);
 
+    void add_point_op(const std::string& op_name, const std::string& code);
+
     std::string generate_point_op(const operation& op, const std::vector<std::string>& args);
 
     std::string str() const;

src/targets/gpu/compile_hip_code_object.cpp

@@ -114,6 +114,7 @@ operation compile_hip_code_object(const std::string& content, hip_compile_option
     options.params += " -DMIGRAPHX_NGLOBAL=" + std::to_string(options.global);
     options.params += " -DMIGRAPHX_NLOCAL=" + std::to_string(options.local);
     options.params += " " + join_strings(compiler_warnings(), " ");
+    options.params += " -ftemplate-backtrace-limit=0";
     options.params += " -Werror";
     auto cos = compile_hip_src(srcs, std::move(options.params), get_device_name());
     if(cos.size() != 1)

src/targets/gpu/compile_pointwise.cpp

@@ -63,8 +63,16 @@ operation compile_pointwise(context& ctx, const std::vector<shape>& inputs, modu
     run_passes(m, {eliminate_common_subexpression{}, dead_code_elimination{}});
     cpp_generator g;
     g.fmap([](const std::string& fname) { return "migraphx::" + fname; });
-    auto name = g.create_function(g.generate_module(m).set_attributes({"__device__"}));
-    return compile_pointwise((ctx), inputs, "&" + name, g.str());
+    g.add_point_op("where", "${function:where}(${0}, ${1}, ${2})");
+    g.add_point_op("prelu", "${function:where}(${0} < 0, ${0} * ${1}, ${0})");
+    g.add_point_op("sign", "${function:where}(${0} > 0, 1, ${function:where}(${0} < 0, -1, 0))");
+    g.add_point_op("equal", "migraphx::abs(${0} == ${1})");
+    g.add_point_op("less", "migraphx::abs(${0} < ${1})");
+    g.add_point_op("greater", "migraphx::abs(${0} > ${1})");
+    g.add_point_op("not", "migraphx::abs(not ${0})");
+    auto name =
+        g.create_function(g.generate_module(m).set_attributes({"__device__"}).set_generic_types(m));
+    return compile_pointwise((ctx), inputs, "MIGRAPHX_LIFT(" + name + ")", g.str());
 }
 
 } // namespace gpu

src/targets/gpu/kernels/include/migraphx/kernels/debug.hpp

@@ -5,6 +5,9 @@
 
 namespace migraphx {
 
+#define MIGRAPHX_STRINGIZE_1(...) #__VA_ARGS__
+#define MIGRAPHX_STRINGIZE(...) MIGRAPHX_STRINGIZE_1(__VA_ARGS__)
+
 // Workaround hip's broken abort on device code
 #ifdef __HIP_DEVICE_COMPILE__
 // NOLINTNEXTLINE
@@ -14,19 +17,67 @@ namespace migraphx {
 #define MIGRAPHX_HIP_NORETURN [[noreturn]]
 #endif
 
+namespace debug {
+struct swallow
+{
+    template <class... Ts>
+    constexpr swallow(Ts&&...)
+    {
+    }
+};
+
+template <size_t N>
+struct print_buffer
+{
+    char buffer[N + 1] = {0};
+    char* pos          = buffer;
+
+    constexpr void append(char c)
+    {
+        if(c == 0)
+            return;
+        if(pos < buffer + N)
+        {
+            *pos = c;
+            pos++;
+        }
+    }
+
+    template <size_t M>
+    constexpr void append(const char (&array)[M])
+    {
+        for(int i = 0; i < M; i++)
+            append(array[i]);
+    }
+};
+
+template <class... Ts>
+__host__ __device__ void print(const Ts&... xs)
+{
+    const auto size = (sizeof(xs) + ...);
+    print_buffer<size> buffer;
+    swallow{(buffer.append(xs), 0)...};
+    printf("%s", buffer.buffer);
+}
+
+} // namespace debug
+
 // noreturn cannot be used on this function because abort in hip is broken
+template <class T1, class T2, class T3, class T4>
 MIGRAPHX_HIP_NORETURN inline __host__ __device__ void
-assert_fail(const char* assertion, const char* file, unsigned int line, const char* function)
+assert_fail(const T1& assertion, const T2& file, const T3& line, const T4& function)
 {
-    printf("%s:%u: %s: assertion '%s' failed.\n", file, line, function, assertion);
+    // printf is broken on hip with more than one argument, so use a simple print functions instead
+    debug::print(file, ":", line, ": ", function, ": assertion '", assertion, "' failed.\n");
+    // printf("%s:%s: %s: assertion '%s' failed.\n", file, line, function, assertion);
     abort();
 }
 
 #ifdef MIGRAPHX_DEBUG
-#define MIGRAPHX_ASSERT(cond)            \
-    ((cond) ? void(0) : [](auto... xs) { \
-        assert_fail(xs...);              \
-    }(#cond, __FILE__, __LINE__, __PRETTY_FUNCTION__))
+#define MIGRAPHX_ASSERT(cond)                               \
+    ((cond) ? void(0) : [](auto&&... private_migraphx_xs) { \
+        assert_fail(private_migraphx_xs...);                \
+    }(#cond, __FILE__, MIGRAPHX_STRINGIZE(__LINE__), __PRETTY_FUNCTION__))
 #else
 #define MIGRAPHX_ASSERT(cond)
 #endif

src/targets/gpu/kernels/include/migraphx/kernels/functional.hpp

@@ -137,12 +137,48 @@ constexpr void each_args(F)
 {
 }
 
+template <class F, class T>
+constexpr auto fold_impl(F&&, T&& x)
+{
+    return static_cast<T&&>(x);
+}
+
+template <class F, class T, class U, class... Ts>
+constexpr auto fold_impl(F&& f, T&& x, U&& y, Ts&&... xs)
+{
+    return fold_impl(f, f(static_cast<T&&>(x), static_cast<U&&>(y)), static_cast<Ts&&>(xs)...);
+}
+
+template <class F>
+constexpr auto fold(F f)
+{
+    return [=](auto&&... xs) { return fold_impl(f, static_cast<decltype(xs)&&>(xs)...); };
+}
+
 template <class... Ts>
-auto pack(Ts... xs)
+constexpr auto pack(Ts... xs)
 {
     return [=](auto f) { return f(xs...); };
 }
 
+template <class Compare, class P1, class P2>
+constexpr auto pack_compare(Compare compare, P1 p1, P2 p2)
+{
+    return p1([&](auto... xs) {
+        return p2([&](auto... ys) {
+            auto c = [&](auto x, auto y) -> int {
+                if(compare(x, y))
+                    return 1;
+                else if(compare(y, x))
+                    return -1;
+                else
+                    return 0;
+            };
+            return fold([](auto x, auto y) { return x ? x : y; })(c(xs, ys)..., 0);
+        });
+    });
+}
+
 template <index_int N>
 constexpr auto arg_c()
 {
@@ -187,7 +223,7 @@ constexpr auto transform_args(F f, Fs... fs)
 
 // NOLINTNEXTLINE
 #define MIGRAPHX_LIFT(...) \
-    ([](auto&&... xs) MIGRAPHX_RETURNS((__VA_ARGS__)(static_cast<decltype(xs)>(xs)...))
+    [](auto&&... xs) MIGRAPHX_RETURNS((__VA_ARGS__)(static_cast<decltype(xs)>(xs)...))
 
 } // namespace migraphx
 #endif // MIGRAPHX_GUARD_KERNELS_FUNCTIONAL_HPP

src/targets/gpu/kernels/include/migraphx/kernels/integral_constant.hpp

@@ -13,6 +13,7 @@ struct integral_constant
     using type               = integral_constant;
     constexpr operator value_type() const noexcept { return value; }
     constexpr value_type operator()() const noexcept { return value; }
+    static constexpr type to() { return {}; }
 };
 
 // NOLINTNEXTLINE

src/targets/gpu/kernels/include/migraphx/kernels/math.hpp

@@ -4,6 +4,7 @@
 #include <migraphx/kernels/types.hpp>
 #include <migraphx/kernels/vec.hpp>
 #include <migraphx/kernels/functional.hpp>
+#include <migraphx/kernels/type_traits.hpp>
 #include <hip/hip_fp16.h>
 #include <hip/math_functions.h>
 
@@ -19,19 +20,30 @@ constexpr T as_float(T x)
 } // namespace math
 
 // NOLINTNEXTLINE
-#define MIGRAPHX_DEVICE_MATH(name, fname) \
-    template <class... Ts>                \
+#define MIGRAPHX_DEVICE_MATH(name, fname)                              \
+    template <class... Ts, MIGRAPHX_REQUIRES(not is_any_vec<Ts...>())> \
     auto __device__ name(Ts... xs) MIGRAPHX_RETURNS(fname(xs...))
 
 // NOLINTNEXTLINE
-#define MIGRAPHX_DEVICE_MATH_FOR(type, name, fname) \
-    template <class... Ts>                          \
-    auto __device__ name(type x, Ts... xs) MIGRAPHX_RETURNS(fname(x, xs...))
+#define MIGRAPHX_DEVICE_MATH_VEC(name)                                       \
+    template <class... Ts, MIGRAPHX_REQUIRES(is_any_vec<Ts...>())>           \
+    auto __device__ name(Ts... xs)                                           \
+    {                                                                        \
+        return vec_transform(xs...)([](auto... ys) { return name(ys...); }); \
+    }
 
 // NOLINTNEXTLINE
-#define MIGRAPHX_DEVICE_MATH_HALF(name, fname)       \
-    template <class... Ts>                           \
-    auto __device__ name(migraphx::half x, Ts... xs) \
+#define MIGRAPHX_DEVICE_MATH_FOR(type, name, fname)                    \
+    template <class... Ts, MIGRAPHX_REQUIRES(not is_any_vec<Ts...>())> \
+    auto __device__ name(type x, Ts... xs)->type                       \
+    {                                                                  \
+        return fname(x, xs...);                                        \
+    }
+
+// NOLINTNEXTLINE
+#define MIGRAPHX_DEVICE_MATH_HALF(name, fname)                         \
+    template <class... Ts, MIGRAPHX_REQUIRES(not is_any_vec<Ts...>())> \
+    auto __device__ name(migraphx::half x, Ts... xs)                   \
         MIGRAPHX_RETURNS(fname(math::as_float(x), math::as_float(xs)...))
 
 MIGRAPHX_DEVICE_MATH(abs, ::abs)
@@ -99,21 +111,51 @@ MIGRAPHX_DEVICE_MATH_HALF(tan, ::tan)
 MIGRAPHX_DEVICE_MATH_HALF(tanh, ::tanh)
 
 template <class T, class U>
-constexpr auto& max(const T& a, const U& b)
+constexpr auto where(bool cond, const T& a, const U& b)
 {
-    return (a < b) ? b : a;
+    return cond ? a : b;
 }
 
+MIGRAPHX_DEVICE_MATH_VEC(abs)
+MIGRAPHX_DEVICE_MATH_VEC(acos)
+MIGRAPHX_DEVICE_MATH_VEC(acosh)
+MIGRAPHX_DEVICE_MATH_VEC(asin)
+MIGRAPHX_DEVICE_MATH_VEC(asinh)
+MIGRAPHX_DEVICE_MATH_VEC(atan)
+MIGRAPHX_DEVICE_MATH_VEC(atanh)
+MIGRAPHX_DEVICE_MATH_VEC(ceil)
+MIGRAPHX_DEVICE_MATH_VEC(cos)
+MIGRAPHX_DEVICE_MATH_VEC(cosh)
+MIGRAPHX_DEVICE_MATH_VEC(erf)
+MIGRAPHX_DEVICE_MATH_VEC(exp)
+MIGRAPHX_DEVICE_MATH_VEC(floor)
+MIGRAPHX_DEVICE_MATH_VEC(log)
+MIGRAPHX_DEVICE_MATH_VEC(pow)
+MIGRAPHX_DEVICE_MATH_VEC(round)
+MIGRAPHX_DEVICE_MATH_VEC(rsqrt)
+MIGRAPHX_DEVICE_MATH_VEC(sin)
+MIGRAPHX_DEVICE_MATH_VEC(sinh)
+MIGRAPHX_DEVICE_MATH_VEC(sqrt)
+MIGRAPHX_DEVICE_MATH_VEC(tan)
+MIGRAPHX_DEVICE_MATH_VEC(tanh)
+MIGRAPHX_DEVICE_MATH_VEC(where)
+
 template <class T, class U>
-constexpr auto& min(const T& a, const U& b)
+constexpr auto max(const T& a, const U& b)
 {
-    return (a > b) ? b : a;
+    return where(a < b, b, a);
 }
 
 template <class T, class U>
-constexpr T convert(U x)
+constexpr auto min(const T& a, const U& b)
 {
-    return x;
+    return where(a > b, b, a);
+}
+
+template <class T, class U>
+constexpr auto convert(U v)
+{
+    return vec_transform(v)([](auto x) -> T { return x; });
 }
 
 } // namespace migraphx

src/targets/gpu/kernels/include/migraphx/kernels/pointwise.hpp

@@ -10,13 +10,38 @@
 
 namespace migraphx {
 
+template <class T>
+struct implicit_conversion_op
+{
+    T x;
+
+    template <index_int N, class U>
+    constexpr operator vec<U, N>() const
+    {
+        static_assert(vec_size<T>() == N, "Vector mismatch size");
+        return __builtin_convertvector(x, vec<U, N>);
+    }
+
+    template <class U>
+    constexpr operator U() const
+    {
+        return x;
+    }
+};
+
+template <class T>
+constexpr implicit_conversion_op<T> implicit_conversion(T x)
+{
+    return {x};
+}
+
 template <class F, class T, class... Ts>
 __device__ void pointwise_tensor(index idx, F f, T out, Ts... xs)
 {
     preload<typename T::type>(idx, xs...)([&](auto... ps) {
         idx.global_stride(out.get_shape().elements(), [&](auto i) {
             auto multi_idx = out.get_shape().multi(i);
-            out[multi_idx] = f(ps[multi_idx]...);
+            out[multi_idx] = implicit_conversion(f(ps[multi_idx]...));
         });
     });
 }
@@ -24,7 +49,7 @@ __device__ void pointwise_tensor(index idx, F f, T out, Ts... xs)
 template <class F, class... Ts>
 __device__ void pointwise(F f, Ts*... ps)
 {
-    auto t = transform_args(make_tensors(), rotate_last());
+    auto t = transform_args(make_tensors(), rotate_last(), auto_vectorize());
     t(ps...)([&](auto... xs) {
         auto idx = make_index();
         pointwise_tensor(idx, f, xs...);

src/targets/gpu/kernels/include/migraphx/kernels/preload.hpp

@@ -29,7 +29,7 @@ constexpr auto traverse_preload(Shapes... ss)
 }
 
 template <class T, class... Shapes>
-constexpr index_int compute_preload_size(Shapes...)
+constexpr index_int compute_preload_size_c(Shapes...)
 {
     index_int size = 0;
     traverse_preload<T>(Shapes{}...)(
@@ -37,6 +37,12 @@ constexpr index_int compute_preload_size(Shapes...)
     return size;
 }
 
+template <class T, class... Shapes>
+constexpr auto compute_preload_size(Shapes...)
+{
+    return _c<compute_preload_size_c<T>(Shapes{}...)>;
+}
+
 template <class F, class T, class... Ts>
 __device__ auto preload_copy(index idx, F f, __shared__ T* buffer, Ts... xs)
 {
@@ -48,11 +54,21 @@ __device__ auto preload_copy(index idx, F f, __shared__ T* buffer, Ts... xs)
         [&](auto x, auto offset, auto copy) {
             if constexpr(copy)
             {
-                auto v = vectorize(x);
-                auto b = as_vec(tensor_vec_size(v), buffer + offset);
-                idx.local_stride(v.get_shape().element_space(),
-                                 [&](auto i) { b[i] = v.data()[i]; });
-                return x.with(buffer + offset);
+                if constexpr(decltype(tensor_vec_size(x)){} == 0)
+                {
+                    auto v = vectorize(x);
+                    auto b = as_vec(tensor_vec_size(v), buffer + offset);
+                    idx.local_stride(v.get_shape().element_space(),
+                                     [&](auto i) { b[i] = v.data()[i]; });
+                    return x.with(buffer + offset);
+                }
+                else
+                {
+                    auto b = as_vec(tensor_vec_size(x), buffer + offset);
+                    idx.local_stride(x.get_shape().element_space(),
+                                     [&](auto i) { b[i] = x.data()[i]; });
+                    return x.with(b);
+                }
             }
             else
             {
@@ -78,7 +94,7 @@ template <class T, class... Ts>
 __device__ auto preload(index idx, Ts... xs)
 {
     using type               = typename remove_vec<T>::type;
-    constexpr auto size      = compute_preload_size<type>(xs.get_shape()...);
+    constexpr auto size      = decltype(compute_preload_size<type>(xs.get_shape()...)){};
     const index_int max_size = 512 * sizeof(type);
     return [=](auto f) {
         if constexpr(size > 0 and size < max_size)

src/targets/gpu/kernels/include/migraphx/kernels/vec.hpp

@@ -3,6 +3,7 @@
 
 #include <migraphx/kernels/types.hpp>
 #include <migraphx/kernels/integral_constant.hpp>
+#include <migraphx/kernels/functional.hpp>
 
 namespace migraphx {
 
@@ -24,6 +25,38 @@ constexpr auto vec_size()
     return decltype(vec_size(T{})){};
 }
 
+template <class... Ts>
+constexpr auto is_any_vec()
+{
+    if constexpr(sizeof...(Ts) == 0)
+        return false_type{};
+    else
+        return bool_constant<((vec_size<Ts>() + ...) > 0)>{};
+}
+
+template <class T, class I>
+constexpr auto vec_at(T x, I i)
+{
+    if constexpr(vec_size<T>() == 0)
+        return x;
+    else
+    {
+        MIGRAPHX_ASSERT(i < vec_size<T>());
+        return x[i];
+    }
+}
+
+template <class... Ts>
+constexpr auto common_vec_size()
+{
+    return fold([](auto x, auto y) {
+        if constexpr(x > y)
+            return x;
+        else
+            return y;
+    })(vec_size<Ts>()...);
+}
+
 template <index_int N, class T>
 __device__ __host__ auto as_vec(T* x)
 {
@@ -33,5 +66,25 @@ __device__ __host__ auto as_vec(T* x)
         return reinterpret_cast<vec<T, N>*>(x);
 }
 
+template <class... Ts>
+constexpr auto vec_transform(Ts... xs)
+{
+    return [=](auto f) {
+        if constexpr(is_any_vec<Ts...>())
+        {
+            using type             = decltype(f(vec_at(xs, 0)...));
+            constexpr auto size    = common_vec_size<Ts...>();
+            vec<type, size> result = {0};
+            for(int i = 0; i < size; i++)
+                result[i] = f(vec_at(xs, i)...);
+            return result;
+        }
+        else
+        {
+            return f(xs...);
+        }
+    };
+}
+
 } // namespace migraphx
 #endif // MIGRAPHX_GUARD_KERNELS_VEC_HPP

src/targets/gpu/kernels/include/migraphx/kernels/vectorize.hpp

@@ -7,40 +7,70 @@
 namespace migraphx {
 
 template <class T>
-constexpr auto tensor_vec_size(T)
+constexpr auto tensor_vec_size()
 {
     return vec_size<typename T::type>();
 }
 
-template <index_int N, class Shape>
-constexpr auto as_vec_shape(Shape s)
+template <class T>
+constexpr auto tensor_vec_size(T)
 {
-    auto lens    = transform(s.lens, s.strides, [](auto len, auto stride) {
-        if(stride == 1)
-            return len / N;
-        else
-            return len;
-    });
-    auto strides = transform(s.strides, [](auto stride) {
-        if(stride == 1)
-            return stride;
-        return stride / N;
+    return tensor_vec_size<T>();
+}
+
+template <index_int N, class Shape, class Axis>
+constexpr auto shape_step(Shape s, Axis)
+{
+    static_assert(N > 0, "Vector size must be non-zero");
+    return sequence(s.lens.size(), [&](auto... is) {
+        auto lens    = transform(s.lens, index_ints<is...>{}, [&](auto i, auto j) {
+            constexpr auto axis = Axis::to();
+            MIGRAPHX_ASSERT(i != 0);
+            MIGRAPHX_ASSERT(j != axis or i % N == 0);
+            if(j == axis)
+                return i / N;
+            else
+                return i;
+        });
+        auto strides = transform(s.strides, index_ints<is...>{}, [&](auto i, auto j) {
+            constexpr auto axis = Axis::to();
+            // If stride of the axis is zero then we dont need to adjust the other strides
+            if(Shape{}.strides[axis] == 0)
+                return i;
+            MIGRAPHX_ASSERT(j == axis or i % N == 0);
+            if(j == axis)
+                return i;
+            else
+                return i / N;
+        });
+        MIGRAPHX_ASSERT(make_shape(lens, strides).elements() * N == s.elements());
+        MIGRAPHX_ASSERT(strides[Axis{}] == 0 or
+                        make_shape(lens, strides).element_space() * N == s.element_space());
+        return make_shape(lens, strides);
     });
-    MIGRAPHX_ASSERT(make_shape(lens, strides).element_space() * N == s.element_space());
-    return make_shape(lens, strides);
 }
 
-template <index_int N, class T>
-__device__ __host__ auto as_vec(T x)
+// Bools can not be used as a vector type so convert it to int8
+template <class T>
+__device__ __host__ T* remove_bool(T* x)
+{
+    return x;
+}
+
+inline __device__ __host__ int8_t* remove_bool(bool* x) { return reinterpret_cast<int8_t*>(x); }
+
+template <index_int N, class T, class Axis>
+__device__ __host__ auto as_vec(T x, Axis axis)
 {
     if constexpr(N == 0)
         return x;
     else
-        return make_tensor_view(as_vec<N>(x.data()), as_vec_shape<N>(x.get_shape()));
+        return make_tensor_view(as_vec<N>(remove_bool(x.data())),
+                                shape_step<N>(x.get_shape(), axis));
 }
 
 template <index_int N, class T, class Axis>
-constexpr auto tensor_step(T x, Axis)
+constexpr auto tensor_step(T x, Axis axis)
 {
     if constexpr(N == 0)
     {
@@ -49,17 +79,8 @@ constexpr auto tensor_step(T x, Axis)
     else
     {
         constexpr auto s = decltype(x.get_shape()){};
-        MIGRAPHX_ASSERT(s.strides[Axis{}] == 0);
-        return sequence(x.get_shape().lens.size(), [&](auto... is) {
-            auto lens = transform(s.lens, index_ints<is...>{}, [&](auto i, auto j) {
-                constexpr auto axis = Axis{};
-                if(j == axis)
-                    return i / N;
-                else
-                    return i;
-            });
-            return make_tensor_view(x.data(), make_shape(lens, s.strides));
-        });
+        MIGRAPHX_ASSERT(s.strides[axis] == 0);
+        return make_tensor_view(x.data(), shape_step<N>(s, axis));
     }
 }
 
@@ -69,45 +90,71 @@ __device__ __host__ auto as_vec(IntegralConstant ic, T&& x)
     return as_vec<ic>(x);
 }
 
-template <class... Shapes>
-constexpr index_int find_vector_axis(Shapes... ss)
+template <class Shape>
+constexpr index_int find_vector_axis_c(Shape s)
 {
+    // Find the fastest axis that is not broadcasted
     index_int axis = 0;
-    bool b         = false;
+    for(index_int i = 1; i < s.lens.size(); i++)
+    {
+        if(s.strides[i] == 0)
+            continue;
+        if(s.strides[axis] == 0 or
+           pack_compare(less{}, pack(s.strides[i], s.lens[i]), pack(s.strides[axis], s.lens[axis])))
+            axis = i;
+    }
+    return axis;
+}
+
+template <class... Shapes>
+constexpr index_int find_vector_axis_c(Shapes... ss)
+{
+    const bool all_broadcasted = (ss.broadcasted() and ...);
+    index_int axis             = 0;
+    bool b                     = false;
     by([&](auto s) {
         if(b)
             return;
-        auto it = find(s.strides.begin(), s.strides.end(), 1);
-        if(it == s.strides.end())
+        // Skip broadcasted shapes if there are shapes not broadcasted
+        if(not all_broadcasted and s.broadcasted())
             return;
-        axis = it - s.strides.begin();
-        b    = true;
+        axis = find_vector_axis_c(s);
+        if(s.strides[axis] == 1)
+            b = true;
     })(ss...);
+    if(not b)
+        return -1;
     return axis;
 }
 
+template <class... Shapes>
+constexpr auto find_vector_axis(Shapes...)
+{
+    return _c<find_vector_axis_c(Shapes{}...)>;
+}
+
 template <index_int N, class Axis, class... Shapes>
-constexpr auto is_vectorizable(Axis axis, Shapes... ss)
+constexpr auto is_vectorizable_c(Axis axis, Shapes... ss)
 {
-    return (((ss.lens[axis] % N) == 0 and ss.strides[axis] == 1) and ...);
+    return ((axis < ss.lens.size() and ss.lens[axis] % N == 0 and
+             // Only vectorize broadcasted types with stride 0, since this causes issues in the
+             // preloader
+             ((not ss.broadcasted() and ss.strides[axis] == 1) or ss.strides[axis] == 0)) and
+            ...);
 }
 
-template <index_int N, class Shape>
-constexpr bool is_vectorizable(Shape s)
+template <index_int N, class Axis, class... Shapes>
+constexpr auto is_vectorizable(Axis, Shapes...)
 {
-    auto it = find(s.strides.begin(), s.strides.end(), 1);
-    if(it == s.strides.end())
-        return false;
-    auto axis = it - s.strides.begin();
-    return (s.lens[axis] % N) == 0 and s.strides[axis] == 1;
+    return _c<is_vectorizable_c<N>(Axis::to(), Shapes{}...)>;
 }
 
 template <class P>
 constexpr auto find_vectorize_size(P pred)
 {
-    if constexpr(pred(_c<4>))
+    if constexpr(decltype(pred(_c<4>)){})
         return _c<4>;
-    else if constexpr(pred(_c<2>))
+    else if constexpr(decltype(pred(_c<2>)){})
         return _c<2>;
     else
         return _c<0>;
@@ -116,11 +163,12 @@ constexpr auto find_vectorize_size(P pred)
 template <class T>
 __host__ __device__ auto vectorize(T x)
 {
-    if constexpr(vec_size<T>() == 0)
+    if constexpr(tensor_vec_size<T>() == 0)
     {
+        constexpr auto axis = find_vector_axis(x.get_shape());
         constexpr auto n =
-            find_vectorize_size([&](auto i) { return _c<is_vectorizable<i>(x.get_shape())>; });
-        return as_vec<n>(x);
+            find_vectorize_size([&](auto i) { return is_vectorizable<i>(axis, x.get_shape()); });
+        return as_vec<n>(x, axis);
     }
     else
     {
@@ -128,34 +176,46 @@ __host__ __device__ auto vectorize(T x)
     }
 }
 
+template <class F, class... Ts>
+inline __device__ __host__ auto auto_vectorize_impl(F f, Ts... xs)
+{
+    // TODO: Just check there a single axis of 1
+    constexpr bool packed_or_broadcasted =
+        ((xs.get_shape().packed() or xs.get_shape().broadcasted()) and ...);
+    if constexpr(packed_or_broadcasted)
+    {
+        constexpr auto axis = decltype(find_vector_axis(xs.get_shape()...)){};
+        constexpr auto n    = find_vectorize_size(
+            [&](auto i) { return is_vectorizable<i>(axis, xs.get_shape()...); });
+        by(
+            [&](auto x) {
+                constexpr auto s = decltype(x.get_shape()){};
+                if constexpr(axis < s.strides.size())
+                {
+                    MIGRAPHX_ASSERT(s.strides[axis] == 0 or s.strides[axis] == 1);
+                    MIGRAPHX_ASSERT(s.lens[axis] > 0);
+                    MIGRAPHX_ASSERT(n == 0 or s.lens[axis] % n == 0);
+                    if constexpr(s.strides[axis] == 0)
+                        return tensor_step<n>(x, axis);
+                    else
+                        return as_vec<n>(x, axis);
+                }
+                else
+                {
+                    return x;
+                }
+            },
+            f)(xs...);
+    }
+    else
+    {
+        f(xs...);
+    }
+}
+
 inline __device__ __host__ auto auto_vectorize()
 {
-    return [](auto... xs) {
-        return [=](auto f) {
-            // TODO: Just check there a single axis of 1
-            constexpr bool packed_or_broadcasted =
-                ((xs.get_shape().packed() or xs.get_shape().broadcasted()) and ...);
-            if constexpr(packed_or_broadcasted)
-            {
-                constexpr auto axis = find_vector_axis(xs.get_shape()...);
-                constexpr auto n    = find_vectorize_size(
-                    [&](auto i) { return _c<is_vectorizable<i>(axis, xs.get_shape()...)>; });
-                by(
-                    [&](auto x) {
-                        constexpr auto s = x.get_shape();
-                        if constexpr(s.strides[axis] == 0)
-                            return tensor_step<n>(x, axis);
-                        else
-                            return as_vec<n>(x);
-                    },
-                    f)(xs...);
-            }
-            else
-            {
-                f(xs...);
-            }
-        };
-    };
+    return [](auto... xs) { return [=](auto f) { auto_vectorize_impl(f, xs...); }; };
 }
 
 } // namespace migraphx