Convert batchnorm to multiply and add

src/fwd_conv_batchnorm_rewrite.cpp

@@ -4,6 +4,7 @@
 #include <migraphx/op/batch_norm.hpp>
 #include <migraphx/op/broadcast.hpp>
 #include <migraphx/op/add.hpp>
+#include <migraphx/op/mul.hpp>
 #include <migraphx/iterator_for.hpp>
 #include <migraphx/ranges.hpp>
 #include <migraphx/dfor.hpp>
@@ -25,46 +26,39 @@ void fwd_conv_batchnorm_rewrite::apply(program& p) const
         if(any_of({gamma, bias, mean, variance}, [](auto arg) { return arg.empty(); }))
             continue;
 
-        auto conv_ins = ins->inputs()[0];
-        if(conv_ins->name() != "convolution")
-            continue;
-        // Get convolution weights
-        auto weights = conv_ins->inputs()[1]->eval();
-        if(weights.empty())
-            continue;
+        auto s = shape{ins->get_shape().type(), {ins->get_shape().lens()[1]}};
         // Get epsilon
         auto bn_op   = any_cast<op::batch_norm_inference>(ins->get_operator());
         auto epsilon = bn_op.epsilon;
-        // Get convolution op
-        auto conv_op      = conv_ins->get_operator();
-        auto weights_lens = weights.get_shape().lens();
-        auto conv_lens    = conv_ins->get_shape().lens();
-        argument new_weights{weights.get_shape()};
-        argument new_bias{{bias.get_shape().type(), {bias.get_shape().elements()}}};
-        visit_all(weights, gamma, bias, mean, variance, new_weights, new_bias)(
-            [&](auto weights2,
-                auto gamma2,
+
+        argument a{s};
+        argument b{s};
+        visit_all(gamma, bias, mean, variance, a, b)(
+        [&](auto gamma2,
             auto bias2,
             auto mean2,
             auto variance2,
-                auto new_weights2,
-                auto new_bias2) {
-                dfor(weights_lens[0], weights_lens[1], weights_lens[2], weights_lens[3])(
-                    [&](std::size_t k, std::size_t c, std::size_t h, std::size_t w) {
-                        new_weights2(k, c, h, w) =
-                            gamma2[k] / std::sqrt(variance2[k] + epsilon) * weights2(k, c, h, w);
+            auto a2,
+            auto b2) {
+            dfor(a.get_shape().elements())(
+                [&](std::size_t c) {
+                    a2[c] =
+                        gamma2[c] / std::sqrt(variance2[c] + epsilon);
                 });
-                dfor(new_bias.get_shape().elements())([&](std::size_t c) {
-                    new_bias2[c] =
+            dfor(b.get_shape().elements())([&](std::size_t c) {
+                b2[c] =
                     bias2[c] - (gamma2[c] * mean2[c] / std::sqrt(variance2[c] + epsilon));
             });
         });
-        // Replace convolution instruction with updated weights
-        auto l_weights = p.add_literal({weights.get_shape(), new_weights.data()});
-        auto l_bias    = p.add_literal({new_bias.get_shape(), new_bias.data()});
-        auto c = p.replace_instruction(conv_ins, conv_op, {conv_ins->inputs()[0], l_weights});
-        auto b = p.insert_instruction(ins, op::broadcast{1, c->get_shape().lens()}, l_bias);
-        p.replace_instruction(ins, op::add{}, {c, b});
+
+        auto broadcast = op::broadcast{1, ins->get_shape().lens()};
+        auto a_ins = p.add_literal({a.get_shape(), a.data()});
+        auto a_broadcast = p.insert_instruction(ins, broadcast, a_ins);
+        auto mul = p.insert_instruction(ins, op::mul{}, ins->inputs().front(), a_broadcast);
+        auto b_ins = p.add_literal({b.get_shape(), b.data()});
+        auto b_broadcast = p.insert_instruction(ins, broadcast, b_ins);
+        auto add = p.insert_instruction(ins, op::add{}, mul, b_broadcast);
+        p.replace_instruction(ins, add);
     }
 }
 

test/fwd_conv_batchnorm_rewrite_test.cpp

@@ -96,7 +96,7 @@ TEST_CASE(non_literal)
     migraphx::fwd_conv_batchnorm_rewrite opt;
     opt.apply(p2);
     EXPECT(any_of(p1, &is_batch_norm));
-    EXPECT(any_of(p2, &is_batch_norm));
+    EXPECT(none_of(p2, &is_batch_norm));
 }
 
 TEST_CASE(as_literal)