[flang-commits] [flang] [flang][runtime] Allow 1023 active asynchronous IDs (PR #82446)

[Peter Klausler via flang-commits]

https://github.com/klausler created https://github.com/llvm/llvm-project/pull/82446

The present limit of 63 is too low for some tests; bump it up to 1023 by using an array of bit-sets.

>From 15443e03aa494d039e84e8c6f39b3039d2507501 Mon Sep 17 00:00:00 2001
From: Peter Klausler <pklausler at nvidia.com>
Date: Tue, 20 Feb 2024 16:40:00 -0800
Subject: [PATCH] [flang][runtime] Allow 1023 active asynchronous IDs

The present limit of 63 is too low for some tests; bump it up to
1023 by using an array of bit-sets.
---
 flang/runtime/unit.cpp | 21 +++++++++++++--------
 flang/runtime/unit.h   | 10 +++++++---
 2 files changed, 20 insertions(+), 11 deletions(-)

diff --git a/flang/runtime/unit.cpp b/flang/runtime/unit.cpp
index 58ca313d9e4454..09782d2f849278 100644
--- a/flang/runtime/unit.cpp
+++ b/flang/runtime/unit.cpp
@@ -1001,25 +1001,30 @@ int ExternalFileUnit::GetAsynchronousId(IoErrorHandler &handler) {
   if (!mayAsynchronous()) {
     handler.SignalError(IostatBadAsynchronous);
     return -1;
-  } else if (auto least{asyncIdAvailable_.LeastElement()}) {
-    asyncIdAvailable_.reset(*least);
-    return static_cast<int>(*least);
   } else {
+    for (int j{0}; 64 * j < maxAsyncIds; ++j) {
+      if (auto least{asyncIdAvailable_[j].LeastElement()}) {
+        asyncIdAvailable_[j].reset(*least);
+        return 64 * j + static_cast<int>(*least);
+      }
+    }
     handler.SignalError(IostatTooManyAsyncOps);
     return -1;
   }
 }
 
 bool ExternalFileUnit::Wait(int id) {
-  if (static_cast<std::size_t>(id) >= asyncIdAvailable_.size() ||
-      asyncIdAvailable_.test(id)) {
+  if (static_cast<std::size_t>(id) >= maxAsyncIds ||
+      asyncIdAvailable_[id / 64].test(id % 64)) {
     return false;
   } else {
     if (id == 0) { // means "all IDs"
-      asyncIdAvailable_.set();
-      asyncIdAvailable_.reset(0);
+      for (int j{0}; 64 * j < maxAsyncIds; ++j) {
+        asyncIdAvailable_[j].set();
+      }
+      asyncIdAvailable_[0].reset(0);
     } else {
-      asyncIdAvailable_.set(id);
+      asyncIdAvailable_[id / 64].set(id % 64);
     }
     return true;
   }
diff --git a/flang/runtime/unit.h b/flang/runtime/unit.h
index 140fda3c4d2a81..e3c8757645bb9e 100644
--- a/flang/runtime/unit.h
+++ b/flang/runtime/unit.h
@@ -36,10 +36,14 @@ class ExternalFileUnit : public ConnectionState,
                          public OpenFile,
                          public FileFrame<ExternalFileUnit> {
 public:
+  static constexpr int maxAsyncIds{64 * 16};
+
   explicit ExternalFileUnit(int unitNumber) : unitNumber_{unitNumber} {
     isUTF8 = executionEnvironment.defaultUTF8;
-    asyncIdAvailable_.set();
-    asyncIdAvailable_.reset(0);
+    for (int j{0}; 64 * j < maxAsyncIds; ++j) {
+      asyncIdAvailable_[j].set();
+    }
+    asyncIdAvailable_[0].reset(0);
   }
   ~ExternalFileUnit() {}
 
@@ -150,7 +154,7 @@ class ExternalFileUnit : public ConnectionState,
   std::size_t recordOffsetInFrame_{0}; // of currentRecordNumber
   bool swapEndianness_{false};
   bool createdForInternalChildIo_{false};
-  common::BitSet<64> asyncIdAvailable_;
+  common::BitSet<64> asyncIdAvailable_[maxAsyncIds / 64];
 
   // When a synchronous I/O statement is in progress on this unit, holds its
   // state.