From d1a6b293e9afe087948f396fba99719265cb6f7e Mon Sep 17 00:00:00 2001
From: Andrew Noyes <andrew@weaselab.dev>
Date: Thu, 7 Mar 2024 12:41:34 -0800
Subject: [PATCH] Revert "Add getChildNodeGeq, use in nextLogical"

This reverts commit 53bc36f62830a16aef2b51cd5ec0d2bb2c02ef0e.

Apparently this used more instructions. Not sure I understand.
---
 ConflictSet.cpp | 58 +++----------------------------------------------
 1 file changed, 3 insertions(+), 55 deletions(-)

diff --git a/ConflictSet.cpp b/ConflictSet.cpp
index eff4935..0c4d81b 100644
--- a/ConflictSet.cpp
+++ b/ConflictSet.cpp
@@ -484,58 +484,6 @@ int getChildGeq(Node *self, int child) {
   return -1;
 }
 
-Node *getChildNodeGeq(Node *self, int child) {
-  if (child > 255) {
-    return nullptr;
-  }
-  if (self->type <= Type::Node16) {
-    auto *self16 = static_cast<Node16 *>(self);
-#ifdef HAS_AVX
-    __m128i key_vec = _mm_set1_epi8(child);
-    __m128i indices;
-    memcpy(&indices, self16->index, sizeof(self16->index));
-    __m128i results = _mm_cmpeq_epi8(key_vec, _mm_min_epu8(key_vec, indices));
-    int mask = (1 << self16->numChildren) - 1;
-    uint32_t bitfield = _mm_movemask_epi8(results) & mask;
-    return bitfield == 0 ? nullptr
-                         : self16->children[std::countr_zero(bitfield)].child;
-#elif defined(HAS_ARM_NEON)
-    uint8x16_t indices;
-    memcpy(&indices, self16->index, sizeof(self16->index));
-    // 0xff for each leq
-    auto results = vcleq_u8(vdupq_n_u8(child), indices);
-    uint64_t mask = self->numChildren == 16
-                        ? uint64_t(-1)
-                        : (uint64_t(1) << (self->numChildren * 4)) - 1;
-    // 0xf for each 0xff (within mask)
-    uint64_t bitfield =
-        vget_lane_u64(
-            vreinterpret_u64_u8(vshrn_n_u16(vreinterpretq_u16_u8(results), 4)),
-            0) &
-        mask;
-    return bitfield == 0
-               ? nullptr
-               : self16->children[std::countr_zero(bitfield) / 4].child;
-#else
-    for (int i = 0; i < self->numChildren; ++i) {
-      if (self16->index[i] >= child) {
-        return self16->children[i].child;
-      }
-    }
-    return nullptr;
-#endif
-  } else if (self->type == Type::Node48) {
-    auto *self48 = static_cast<Node48 *>(self);
-    int c = self48->bitSet.firstSetGeq(child);
-    return c >= 0 ? self48->children[self48->index[c]].child : nullptr;
-  } else {
-    assert(self->type == Type::Node256);
-    auto *self256 = static_cast<Node256 *>(self);
-    int c = self256->bitSet.firstSetGeq(child);
-    return c >= 0 ? self256->children[c].child : nullptr;
-  }
-}
-
 void setChildrenParents(Node4 *n) {
   for (int i = 0; i < n->numChildren; ++i) {
     n->children[i].child->parent = n;
@@ -750,9 +698,9 @@ void eraseChild(Node *self, uint8_t index, NodeAllocators *allocators) {
 Node *nextPhysical(Node *node) {
   int index = -1;
   for (;;) {
-    Node *nextChild = getChildNodeGeq(node, index + 1);
-    if (nextChild != nullptr) {
-      return nextChild;
+    auto nextChild = getChildGeq(node, index + 1);
+    if (nextChild >= 0) {
+      return getChildExists(node, nextChild);
     }
     index = node->parentsIndex;
     node = node->parent;