Merge triton-mlir branch - Complete rewrite of the backend from scratch (#1004)

This PR merges the `triton-mlir` branch, in which we have been quietly
rewriting the Triton backend from scratch to increase maintainability,
stability and ultimately performance. Changes to the runtime are
minimal, and this new version aims to remain backward-compatible with
the previous commit. The legacy backend is now officially deprecated,
but can still be accessed via the `legacy-backend` tag.

Co-authored-by: Keren Zhou <kerenzhou@openai.com>
Co-authored-by: Yan Chunwei <yanchunwei@outlook.com>
Co-authored-by: goostavz <109190422+goostavz@users.noreply.github.com>
Co-authored-by: Shintaro Iwasaki <siwasaki@fb.com>
Co-authored-by: Yan Da <dyanab@connect.ust.hk>
Co-authored-by: Jun Yang <yangjunpro@gmail.com>
Co-authored-by: Ian Bearman <ianb@microsoft.com>
Co-authored-by: Jason Ansel <jansel@jansel.net>
Co-authored-by: Qingyi Liu <qingyil@nvidia.com>
Co-authored-by: ben-zhang-609 <110140741+ben-zhang-609@users.noreply.github.com>
Co-authored-by: Chenggang Zhao <lyricz@yeah.net>
Co-authored-by: ben-zhang-609 <benzh609@gmail.com>
Co-authored-by: dongdongl <dongdongl@nvidia.com>

test/lib/Analysis/TestAxisInfo.cpp

@@ -0,0 +1,69 @@
+#include "mlir/Pass/Pass.h"
+#include "triton/Analysis/AxisInfo.h"
+
+using namespace mlir;
+
+namespace {
+
+struct TestAxisInfoPass
+    : public PassWrapper<TestAxisInfoPass, OperationPass<FuncOp>> {
+
+  // LLVM15+
+  // MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(TestAlignmentPass);
+
+  void print(const std::string &name, raw_ostream &os, ArrayRef<int> vals) {
+    os << name << ": [";
+    for (size_t d = 0; d < vals.size(); d++) {
+      if (d != 0)
+        os << ", ";
+      os << vals[d];
+    }
+    os << "]";
+  }
+
+  StringRef getArgument() const final { return "test-print-alignment"; }
+  StringRef getDescription() const final {
+    return "print the result of the alignment analysis pass";
+  }
+
+  void runOnOperation() override {
+    Operation *operation = getOperation();
+    auto &os = llvm::errs();
+    os << "Testing: " << operation->getName() << "\n";
+    AxisInfoAnalysis analysis(&getContext());
+    analysis.run(operation);
+    operation->walk([&](Operation *op) {
+      if (op->getNumResults() < 1)
+        return;
+      for (Value result : op->getResults()) {
+        // std::ostringstream oss;
+        // result.print(oss);
+        // os << " => ";
+        LatticeElement<AxisInfo> *latticeElement =
+            analysis.lookupLatticeElement(result);
+        if (!latticeElement) {
+          os << "None\n";
+          return;
+        }
+        AxisInfo &info = latticeElement->getValue();
+        print("Contiguity", os, info.getContiguity());
+        os << " ; ";
+        print("Divisibility", os, info.getDivisibility());
+        os << " ; ";
+        print("Constancy", os, info.getConstancy());
+        os << " ( ";
+        result.print(os);
+        os << " ) ";
+        os << "\n";
+      }
+    });
+  }
+};
+
+} // namespace
+
+namespace mlir {
+namespace test {
+void registerTestAlignmentPass() { PassRegistration<TestAxisInfoPass>(); }
+} // namespace test
+} // namespace mlir