randomx: optional 2 MB large pages for the cache, dataset and scratchpads

RandomX reads its 256 MB cache, 2 GB dataset and per-VM scratchpads at random,
so most accesses miss the TLB on 4 KB pages. Requesting 2 MB pages recovers
roughly a tenth of the hash rate. Each allocation tries large pages first and
falls back to normal pages when the pool is not reserved, so nothing regresses
when hugepages are unavailable. -nolargepages opts out, and the diagnostics
report which allocations got large pages.

4 files changed, 110 insertions, 11 deletions

diff --git a/src/main_gui.cpp b/src/main_gui.cpp
index 0a961bf..a01bdee 100644
--- a/src/main_gui.cpp
+++ b/src/main_gui.cpp
@@ -127,6 +127,8 @@ static void PrintUsage()
     printf("  /stratumbridgeport=N       (default 3333; listen on 127.0.0.1)\n");
     printf("  /solomine\n");
     printf("  /genproclimit=N            (mining threads; 0 or absent = every core but one)\n");
+    printf("  /nolargepages              (do not ask for 2 MB pages for the RandomX\n");
+    printf("                              cache, dataset and scratchpads)\n");
     printf("  /checkblocks=N             (blocks re-verified at startup, default 288, 0 = all)\n");
     printf("\n");
     printf("Pool operator announcement:\n");
@@ -230,6 +232,11 @@ static void ParseStartupArguments(int argc, char* argv[])
     if (!strCheckBlocks.empty())
         nCheckBlocksOnLoad = atoi(strCheckBlocks.c_str());
 
+    // Must be read before LoadBlockIndex(), which is what first calls
+    // RandomXInit() and fixes how the cache is allocated.
+    if (arg(argc,argv,"/nolargepages") || arg(argc,argv,"-nolargepages"))
+        fRandomXLargePages = false;
+
     // /genproclimit=N -- threads to hash with. 0 or absent means automatic,
     // which is every core but one. Named after Bitcoin's own option so it reads
     // familiarly to anyone who has run one of these before.
diff --git a/src/net.cpp b/src/net.cpp
index ecfd565..287a4bd 100644
--- a/src/net.cpp
+++ b/src/net.cpp
@@ -366,6 +366,7 @@ string GetDiagnosticsText()
                                      nMining,
                                      (RandomXFastReady() ? 2080 : 256) + 2 * nMining).c_str()
                          : ", not mining");
+    str += strprintf("  large pages       %s\n", RandomXLargePagesStatus());
 
     str += SockAccountingText();
 
diff --git a/src/randomx_pow.cpp b/src/randomx_pow.cpp
index 1d058d4..5872090 100644
--- a/src/randomx_pow.cpp
+++ b/src/randomx_pow.cpp
@@ -23,6 +23,39 @@ static std::mutex      g_csInit;
 static bool            g_fInit    = false;
 static bool            g_fFast    = false;
 
+// RandomX reads a 256 MB cache, a 2 GB dataset and a 2 MB scratchpad per VM
+// at random, so most reads miss the TLB on 4 KB pages. 2 MB pages recover
+// roughly a tenth of the hash rate.
+//
+// The pool must be reserved first (vm.nr_hugepages), and RandomX returns NULL
+// instead of falling back, so each allocation below retries without the flag.
+bool fRandomXLargePages = true;   // cleared by -nolargepages
+static bool g_fLargeCache   = false;
+static bool g_fLargeDataset = false;
+static bool g_fLargeAny     = false;
+
+static randomx_flags WithLargePages(randomx_flags f)
+{
+    return (randomx_flags)(f | RANDOMX_FLAG_LARGE_PAGES);
+}
+
+// Ask for a cache in large pages, then plain pages. Sets fLargeOut to say
+// which one answered.
+static randomx_cache* AllocCache(randomx_flags flags, bool& fLargeOut)
+{
+    fLargeOut = false;
+    if (fRandomXLargePages)
+    {
+        randomx_cache* c = randomx_alloc_cache(WithLargePages(flags));
+        if (c)
+        {
+            fLargeOut = true;
+            return c;
+        }
+    }
+    return randomx_alloc_cache(flags);
+}
+
 
 bool RandomXInit()
 {
@@ -36,28 +69,56 @@ bool RandomXInit()
     // Detect the best flags for this CPU (JIT, hardware AES, Argon2).
     g_flags = randomx_get_flags();
 
-    g_cache = randomx_alloc_cache(g_flags);
+    g_cache = AllocCache(g_flags, g_fLargeCache);
     if (!g_cache)
     {
         // Try without JIT as a fallback
         g_flags = RANDOMX_FLAG_DEFAULT;
-        g_cache = randomx_alloc_cache(g_flags);
+        g_cache = AllocCache(g_flags, g_fLargeCache);
         if (!g_cache)
             return error("RandomX: failed to allocate cache (256 MB)");
     }
     randomx_init_cache(g_cache, RANDOMX_KEY, strlen(RANDOMX_KEY));
 
-    // Verification VM in light mode (cache only)
-    g_vmVerify = randomx_create_vm(g_flags, g_cache, NULL);
+    // Verification VM in light mode (cache only). The scratchpad is only 2 MB
+    // and this VM is created once, but it hashes every block the node ever
+    // validates, so it gets the same treatment.
+    bool fLargeVerify = false;
+    if (fRandomXLargePages)
+    {
+        g_vmVerify = randomx_create_vm(WithLargePages(g_flags), g_cache, NULL);
+        fLargeVerify = (g_vmVerify != NULL);
+    }
+    if (!g_vmVerify)
+        g_vmVerify = randomx_create_vm(g_flags, g_cache, NULL);
     if (!g_vmVerify)
         return error("RandomX: failed to create verification VM");
 
+    g_fLargeAny = g_fLargeCache || fLargeVerify;
     g_fInit = true;
-    printf("RandomX: initialized (cache 256 MB, flags=%d)\n", (int)g_flags);
+    printf("RandomX: initialized (cache 256 MB, flags=%d, large pages: %s)\n",
+           (int)g_flags, RandomXLargePagesStatus());
+    if (fRandomXLargePages && !g_fLargeCache)
+        printf("RandomX: large pages unavailable for the cache; "
+               "reserve them with sysctl vm.nr_hugepages=1280 for about 10%% more hash rate\n");
     return true;
 }
 
 
+const char* RandomXLargePagesStatus()
+{
+    if (!fRandomXLargePages)
+        return "off (-nolargepages)";
+    if (g_fLargeCache && g_fLargeDataset)
+        return "cache + dataset";
+    if (g_fLargeCache)
+        return "cache only";
+    if (g_fLargeDataset)
+        return "dataset only";
+    return g_fLargeAny ? "scratchpad only" : "unavailable";
+}
+
+
 // One dataset, once, no matter how many miners ask for it.
 //
 // Every miner thread calls this at startup. The g_fFast guard below is only
@@ -87,8 +148,18 @@ bool RandomXInitDataset(int nThreads)
     if (g_fFast)
         return true;
 
-    randomx_flags fastFlags = (randomx_flags)(g_flags | RANDOMX_FLAG_FULL_MEM);
-    g_dataset = randomx_alloc_dataset(fastFlags);
+    // Only RANDOMX_FLAG_LARGE_PAGES means anything to randomx_alloc_dataset;
+    // the rest of g_flags belongs to the VMs that read it.
+    if (fRandomXLargePages)
+    {
+        g_dataset = randomx_alloc_dataset(WithLargePages(RANDOMX_FLAG_DEFAULT));
+        g_fLargeDataset = (g_dataset != NULL);
+        if (!g_dataset)
+            printf("RandomX: large pages unavailable for the 2 GB dataset; "
+                   "reserve 1280 with sysctl vm.nr_hugepages for about 10%% more hash rate\n");
+    }
+    if (!g_dataset)
+        g_dataset = randomx_alloc_dataset(RANDOMX_FLAG_DEFAULT);
     if (!g_dataset)
     {
         printf("RandomX: not enough memory for the 2 GB dataset, staying in light mode\n");
@@ -142,13 +213,26 @@ void* RandomXCreateMinerVM()
 {
     if (!g_fInit && !RandomXInit())
         return NULL;
+
+    randomx_flags   flags   = g_flags;
+    randomx_cache*  cache   = g_cache;
+    randomx_dataset* dataset = NULL;
     if (g_fFast && g_dataset)
     {
-        randomx_flags fastFlags = (randomx_flags)(g_flags | RANDOMX_FLAG_FULL_MEM);
-        return randomx_create_vm(fastFlags, NULL, g_dataset);
+        flags   = (randomx_flags)(g_flags | RANDOMX_FLAG_FULL_MEM);
+        cache   = NULL;
+        dataset = g_dataset;
+    }
+
+    // Each thread gets its own 2 MB scratchpad, so the large-page pool can run
+    // dry partway through a fleet of miners. Threads that miss out still run.
+    if (fRandomXLargePages)
+    {
+        randomx_vm* vm = randomx_create_vm(WithLargePages(flags), cache, dataset);
+        if (vm)
+            return vm;
     }
-    // Light mode: own VM sharing the cache
-    return randomx_create_vm(g_flags, g_cache, NULL);
+    return randomx_create_vm(flags, cache, dataset);
 }
 
 
diff --git a/src/randomx_pow.h b/src/randomx_pow.h
index 0027571..355b708 100644
--- a/src/randomx_pow.h
+++ b/src/randomx_pow.h
@@ -41,4 +41,11 @@ uint256 RandomXHashWithVM(void* vm, const void* pHeader, size_t nSize);
 // True when the 2 GB dataset is ready (fast mode active).
 bool RandomXFastReady();
 
+// Ask for large (2 MB) pages for the cache, dataset and scratchpads. Set false
+// by -nolargepages. Must be set before the first RandomXInit().
+extern bool fRandomXLargePages;
+
+// Which allocations actually got large pages, for the status display.
+const char* RandomXLargePagesStatus();
+
 #endif