Problem 883

Remarkable Triangles: count triangles on the hexagonal lattice with lattice-point incenter and inradius <= 10^6.

Answer14854003484704
Output14854003484704
StatusPASS
Native helperno
Runtime320 ms
Peak memory33232 KB
Time complexityO(n^2) (estimated)
Space complexityO(n^2) (estimated)

Performance comparison

MetricOur solutionBest known
Time complexityO(n^2)O(n^2)
Space complexityO(n^2)O(n^2)
ApproachFlow solutionCombinatorial or DP counting
VerdictOptimal

Flow source

# Project Euler 883
# Remarkable Triangles: count triangles on the hexagonal lattice with
# lattice-point incenter and inradius <= 10^6.

extern {
    function calloc(n: i64, size: i64) -> ptr<void>
    function free(p: ptr<void>) -> void
    function sqrt(x: f64) -> f64
}

# Cache for count_hex_points_leq
const CACHE_CAP: i64 = 1048576
const CACHE_MASK: i64 = 1048575
let mut g_cache_key: ptr<i64> = null
let mut g_cache_val: ptr<i64> = null
let mut g_cache_used: ptr<i8> = null

function isqrt_u64(n: i64) -> i64 {
    if n == 0 { return 0 }
    let mut x: i64 = sqrt(n as f64) as i64
    while x * x > n { x = x - 1 }
    while (x + 1) * (x + 1) <= n { x = x + 1 }
    return x
}

function gcd_u64(a0: i64, b0: i64) -> i64 {
    let mut a: i64 = a0
    let mut b: i64 = b0
    while b != 0 {
        let t: i64 = a % b
        a = b
        b = t
    }
    return a
}

function cache_hash(key: i64) -> i64 {
    let h: i64 = key ^ (key >> 16)
    let h2: i64 = h * 2654435761
    let h3: i64 = h2 ^ (h2 >> 16)
    return h3 & CACHE_MASK
}

function cache_lookup(key: i64, found: ptr<i32>) -> i64 {
    let mut h: i64 = cache_hash(key)
    while g_cache_used[h] != 0 {
        if g_cache_key[h] == key {
            found[0] = 1
            return g_cache_val[h]
        }
        h = (h + 1) & CACHE_MASK
    }
    found[0] = 0
    return 0
}

function cache_insert(key: i64, val: i64) -> void {
    let mut h: i64 = cache_hash(key)
    while g_cache_used[h] != 0 {
        if g_cache_key[h] == key {
            g_cache_val[h] = val
            return
        }
        h = (h + 1) & CACHE_MASK
    }
    g_cache_used[h] = 1
    g_cache_key[h] = key
    g_cache_val[h] = val
}

function chi_prefix(m: i64) -> i64 {
    return (m + 2) / 3 - (m + 1) / 3
}

function count_hex_points_leq(B: i64) -> i64 {
    if B == 0 { return 1 }
    let found: ptr<i32> = calloc(1, 4)
    let hit: i64 = cache_lookup(B, found)
    if found[0] == 1 {
        free(found)
        return hit
    }
    free(found)

    let mut total: i64 = 0
    let n: i64 = B
    let mut i: i64 = 1
    while i <= n {
        let q: i64 = n / i
        let j: i64 = n / q
        total = total + q * (chi_prefix(j) - chi_prefix(i - 1))
        i = j + 1
    }

    let res: i64 = 1 + 6 * total
    cache_insert(B, res)
    return res
}

function remarkable_triangles(R_num: i64, R_den: i64) -> i64 {
    let C_num: i64 = 12 * R_num * R_num
    let C_den: i64 = R_den * R_den
    let Dmax: i64 = isqrt_u64(C_num / C_den)

    let M: ptr<i32> = calloc(Dmax + 1, 4)

    # Family 1: d = 3*t, multiplicity = 2^{omega(t)-1} where t % 3 != 1
    let Nmax: i64 = Dmax / 3
    let omega: ptr<i8> = calloc(Nmax + 1, 1)
    let mut p: i64 = 2
    while p <= Nmax {
        if omega[p] == 0 {
            let mut k: i64 = p
            while k <= Nmax {
                omega[k] = omega[k] + 1
                k = k + p
            }
        }
        p = p + 1
    }
    let mut t: i64 = 2
    while t <= Nmax {
        if t % 3 != 1 {
            M[3 * t] = M[3 * t] + (1 << (omega[t] - 1))
        }
        t = t + 1
    }
    free(omega)

    # Family 2: d = a*(a+3v), gcd(u,v)=1, u mod 3 != v mod 3
    let D: i64 = Dmax
    let vmax: i64 = D / 3
    let mut v: i64 = 1
    while v <= vmax {
        let disc: i64 = 9 * v * v + 4 * D
        let sq: i64 = isqrt_u64(disc)
        let amax: i64 = (sq - 3 * v) / 2
        if amax <= 0 { v = v + 1; continue }
        let mut a: i64 = 1
        while a <= amax {
            let u: i64 = v + a
            if u % 3 != v % 3 {
                if gcd_u64(u, v) == 1 {
                    let d: i64 = a * (a + 3 * v)
                    M[d] = M[d] + 1
                }
            }
            a = a + 1
        }
        v = v + 1
    }

    # Sum contributions over d
    let mut total_scalene: i64 = 0
    let mut d: i64 = 1
    while d <= Dmax {
        let mult: i32 = M[d]
        if mult == 0 { d = d + 1; continue }
        let B: i64 = C_num / (C_den * d * d)
        if B == 0 { d = d + 1; continue }
        let pts: i64 = 0
        if d % 3 == 0 {
            pts = count_hex_points_leq(B) - 1
        } else {
            pts = count_hex_points_leq(B / 3) - 1
        }
        total_scalene = total_scalene + 2 * (mult as i64) * pts
        d = d + 1
    }

    # Equilateral triangles
    let Beq: i64 = (4 * R_num * R_num) / (R_den * R_den)
    let equi_points: i64 = count_hex_points_leq(Beq) - 1
    let total_equilateral: i64 = equi_points / 3

    free(M)
    return total_scalene + total_equilateral
}

function main() -> i32 {
    g_cache_key = calloc(CACHE_CAP, 8)
    g_cache_val = calloc(CACHE_CAP, 8)
    g_cache_used = calloc(CACHE_CAP, 1)

    let ans: i64 = remarkable_triangles(1000000, 1)
    printf("%lld\n", ans)

    free(g_cache_key)
    free(g_cache_val)
    free(g_cache_used)
    return 0
}

Generated C

#include <stdint.h>
#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

/* Flow runtime helpers */
typedef struct flow_temp_node { struct flow_temp_node* next; } flow_temp_node;
static flow_temp_node* flow_temp_head = NULL;
static int flow_temp_atexit_set = 0;
__attribute__((unused)) static void flow_temp_free_all(void) {
    while (flow_temp_head) {
        flow_temp_node* n = flow_temp_head;
        flow_temp_head = n->next;
        free(n);
    }
}
__attribute__((unused)) static void* flow_temp_alloc(size_t nbytes) {
    flow_temp_node* node = (flow_temp_node*)malloc(sizeof(flow_temp_node) + nbytes);
    if (!node) return NULL;
    node->next = flow_temp_head;
    flow_temp_head = node;
    if (!flow_temp_atexit_set) {
        flow_temp_atexit_set = 1;
        atexit(flow_temp_free_all);
    }
    return (void*)(node + 1);
}
#ifndef FLOW_DIAG
#define FLOW_DIAG(msg) fprintf(stderr, "%s", (msg))
#endif
#ifndef FLOW_LOG
#define FLOW_LOG(fmt, ...) printf(fmt, __VA_ARGS__)
#endif
#ifndef FLOW_LOG_EMPTY
#define FLOW_LOG_EMPTY(fmt) printf(fmt)
#endif
static char* flow_strcat(const char* a, const char* b) {
    size_t la = strlen(a ? a : ""), lb = strlen(b ? b : "");
    char* r = (char*)flow_temp_alloc(la + lb + 1);
    if (!r) return NULL;
    if (la) memcpy(r, a, la);
    if (lb) memcpy(r + la, b, lb);
    r[la + lb] = '\0';
    return r;
}

#define __flow_in_arr(arr, val) __extension__ ({ \
    int _found = 0; \
    size_t _n = sizeof(arr)/sizeof((arr)[0]); \
    for (size_t _i = 0; _i < _n; _i++) { \
        if ((arr)[_i] == (val)) { _found = 1; break; } \
    } _found; })

/* Unified fault handler (MISRA #279) — override with -DFLOW_FAULT_HANDLER=fn */
#ifndef FLOW_FAULT_HANDLER
__attribute__((unused)) static inline void flow_fault_handler(const char* msg) {
    fprintf(stderr, "flow: %s\n", msg ? msg : "fault");
    abort();
#if defined(__GNUC__) || defined(__clang__)
    __builtin_unreachable();
#endif
}
#else
#define flow_fault_handler FLOW_FAULT_HANDLER
#endif
#define flow_div_by_zero_handler() flow_fault_handler("division by zero")
#define flow_shift_ub_handler() flow_fault_handler("invalid shift (amount out of range or left-shift of negative)")

#ifndef FLOW_CHECKED_DIV
#define FLOW_CHECKED_DIV(L, R) (((R) != 0) ? ((L) / (R)) : (flow_div_by_zero_handler(), (L) * 0))
#endif
#ifndef FLOW_CHECKED_MOD
#define FLOW_CHECKED_MOD(L, R) (((R) != 0) ? ((L) % (R)) : (flow_div_by_zero_handler(), (L) * 0))
#endif
#ifndef FLOW_CHECKED_SHL
#define FLOW_CHECKED_SHL(L, R) ((((R) >= 0) && ((unsigned long long)(R) < (sizeof(L) * 8ull)) && ((L) >= 0)) ? ((L) << (R)) : (flow_shift_ub_handler(), (L) * 0))
#endif
#ifndef FLOW_CHECKED_SHR
#define FLOW_CHECKED_SHR(L, R) ((((R) >= 0) && ((unsigned long long)(R) < (sizeof(L) * 8ull))) ? ((L) >> (R)) : (flow_shift_ub_handler(), (L) * 0))
#endif

#include <math.h>

void* _ui_state = NULL;

static inline float i32_to_f32(int32_t v) { return (float)v; }

/* Host stub for @gpu kernels (device codegen replaces this). */
static inline int32_t gpu_thread_id(void) { return 0; }

int64_t isqrt_u64_i64(int64_t n);
int64_t gcd_u64_i64_i64(int64_t a0, int64_t b0);
int64_t cache_hash_i64(int64_t key);
int64_t cache_lookup_i64_ptr_i32(int64_t key, int32_t* found);
void cache_insert_i64_i64(int64_t key, int64_t val);
int64_t chi_prefix_i64(int64_t m);
int64_t count_hex_points_leq_i64(int64_t B);
int64_t remarkable_triangles_i64_i64(int64_t R_num, int64_t R_den);
int32_t main(void);

static const int64_t CACHE_CAP = 1048576;
static const int64_t CACHE_MASK = 1048575;

/* Module statics */
static int64_t* g_cache_key = NULL;
static int64_t* g_cache_val = NULL;
static int8_t* g_cache_used = NULL;




int64_t isqrt_u64_i64(int64_t n) {
    if (n == 0) {
        return 0;
    }
    int64_t x = ((int64_t)(sqrt(((double)(n)))));
    while ((x * x) > n) {
        x = (x - 1);
    }
    while (((x + 1) * (x + 1)) <= n) {
        x = (x + 1);
    }
    return x;
}

int64_t gcd_u64_i64_i64(int64_t a0, int64_t b0) {
    int64_t a = a0;
    int64_t b = b0;
    while (b != 0) {
        int64_t t = FLOW_CHECKED_MOD((a), (b));
        a = b;
        b = t;
    }
    return a;
}

int64_t cache_hash_i64(int64_t key) {
    int64_t h = (key ^ FLOW_CHECKED_SHR((key), (16)));
    int64_t h2 = (h * 2654435761);
    int64_t h3 = (h2 ^ FLOW_CHECKED_SHR((h2), (16)));
    return (h3 & CACHE_MASK);
}

int64_t cache_lookup_i64_ptr_i32(int64_t key, int32_t* found) {
    int64_t h = cache_hash_i64(key);
    while (g_cache_used[h] != 0) {
        if (g_cache_key[h] == key) {
            found[0] = 1;
            return g_cache_val[h];
        }
        h = ((h + 1) & CACHE_MASK);
    }
    found[0] = 0;
    return 0;
}

void cache_insert_i64_i64(int64_t key, int64_t val) {
    int64_t h = cache_hash_i64(key);
    while (g_cache_used[h] != 0) {
        if (g_cache_key[h] == key) {
            g_cache_val[h] = val;
            return;
        }
        h = ((h + 1) & CACHE_MASK);
    }
    g_cache_used[h] = 1;
    g_cache_key[h] = key;
    g_cache_val[h] = val;
}

int64_t chi_prefix_i64(int64_t m) {
    return (FLOW_CHECKED_DIV(((m + 2)), (3)) - FLOW_CHECKED_DIV(((m + 1)), (3)));
}

int64_t count_hex_points_leq_i64(int64_t B) {
    if (B == 0) {
        return 1;
    }
    int32_t* found = (int32_t*)(calloc(1, 4));
    int64_t hit = cache_lookup_i64_ptr_i32(B, found);
    if (found[0] == 1) {
        free(found);
        return hit;
    }
    free(found);
    int64_t total = 0;
    int64_t n = B;
    int64_t i = 1;
    while (i <= n) {
        int64_t q = FLOW_CHECKED_DIV((n), (i));
        int64_t j = FLOW_CHECKED_DIV((n), (q));
        total = (total + (q * (chi_prefix_i64(j) - chi_prefix_i64((i - 1)))));
        i = (j + 1);
    }
    int64_t res = (1 + (6 * total));
    cache_insert_i64_i64(B, res);
    return res;
}

int64_t remarkable_triangles_i64_i64(int64_t R_num, int64_t R_den) {
    int64_t C_num = ((12 * R_num) * R_num);
    int64_t C_den = (R_den * R_den);
    int64_t Dmax = isqrt_u64_i64(FLOW_CHECKED_DIV((C_num), (C_den)));
    int32_t* M = (int32_t*)(calloc((Dmax + 1), 4));
    int64_t Nmax = FLOW_CHECKED_DIV((Dmax), (3));
    int8_t* omega = (int8_t*)(calloc((Nmax + 1), 1));
    int64_t p = 2;
    while (p <= Nmax) {
        if (omega[p] == 0) {
            int64_t k = p;
            while (k <= Nmax) {
                omega[k] = (omega[k] + 1);
                k = (k + p);
            }
        }
        p = (p + 1);
    }
    int64_t t = 2;
    while (t <= Nmax) {
        if (FLOW_CHECKED_MOD((t), (3)) != 1) {
            M[(3 * t)] = (M[(3 * t)] + FLOW_CHECKED_SHL((1), ((omega[t] - 1))));
        }
        t = (t + 1);
    }
    free(omega);
    int64_t D = Dmax;
    int64_t vmax = FLOW_CHECKED_DIV((D), (3));
    int64_t v = 1;
    while (v <= vmax) {
        int64_t disc = (((9 * v) * v) + (4 * D));
        int64_t sq = isqrt_u64_i64(disc);
        int64_t amax = FLOW_CHECKED_DIV(((sq - (3 * v))), (2));
        if (amax <= 0) {
            v = (v + 1);
            continue;
        }
        int64_t a = 1;
        while (a <= amax) {
            int64_t u = (v + a);
            if (FLOW_CHECKED_MOD((u), (3)) != FLOW_CHECKED_MOD((v), (3))) {
                if (gcd_u64_i64_i64(u, v) == 1) {
                    int64_t d = (a * (a + (3 * v)));
                    M[d] = (M[d] + 1);
                }
            }
            a = (a + 1);
        }
        v = (v + 1);
    }
    int64_t total_scalene = 0;
    int64_t d = 1;
    while (d <= Dmax) {
        int32_t mult = M[d];
        if (mult == 0) {
            d = (d + 1);
            continue;
        }
        int64_t B = FLOW_CHECKED_DIV((C_num), (((C_den * d) * d)));
        if (B == 0) {
            d = (d + 1);
            continue;
        }
        int64_t pts = 0;
        if (FLOW_CHECKED_MOD((d), (3)) == 0) {
            pts = (count_hex_points_leq_i64(B) - 1);
        } else {
            pts = (count_hex_points_leq_i64(FLOW_CHECKED_DIV((B), (3))) - 1);
        }
        total_scalene = (total_scalene + ((2 * ((int64_t)(mult))) * pts));
        d = (d + 1);
    }
    int64_t Beq = FLOW_CHECKED_DIV((((4 * R_num) * R_num)), ((R_den * R_den)));
    int64_t equi_points = (count_hex_points_leq_i64(Beq) - 1);
    int64_t total_equilateral = FLOW_CHECKED_DIV((equi_points), (3));
    free(M);
    return (total_scalene + total_equilateral);
}

int32_t main(void) {
    g_cache_key = calloc(CACHE_CAP, 8);
    g_cache_val = calloc(CACHE_CAP, 8);
    g_cache_used = calloc(CACHE_CAP, 1);
    int64_t ans = remarkable_triangles_i64_i64(1000000, 1);
    printf("%lld\n", ans);
    free(g_cache_key);
    free(g_cache_val);
    free(g_cache_used);
    return 0;
}

Generated MLIR

module {
  llvm.func @printf(!llvm.ptr, ...) -> i32
  llvm.mlir.global internal constant @str_0("%lld\n\00") {addr_space = 0 : i32} : !llvm.array<6 x i8>
  func.func private @calloc(i64, i64) -> !llvm.ptr
  func.func private @free(!llvm.ptr) -> ()
  func.func private @sqrt(f64) -> f64
  // Constant: CACHE_CAP
  llvm.mlir.global internal constant @CACHE_CAP(1048576 : i64) : i64
  // Constant: CACHE_MASK
  llvm.mlir.global internal constant @CACHE_MASK(1048575 : i64) : i64
  // Module static: g_cache_key
  llvm.mlir.global internal @g_cache_key() {addr_space = 0 : i32} : !llvm.ptr {
    %0 = llvm.mlir.zero : !llvm.ptr
    llvm.return %0 : !llvm.ptr
  }
  // Module static: g_cache_val
  llvm.mlir.global internal @g_cache_val() {addr_space = 0 : i32} : !llvm.ptr {
    %1 = llvm.mlir.zero : !llvm.ptr
    llvm.return %1 : !llvm.ptr
  }
  // Module static: g_cache_used
  llvm.mlir.global internal @g_cache_used() {addr_space = 0 : i32} : !llvm.ptr {
    %2 = llvm.mlir.zero : !llvm.ptr
    llvm.return %2 : !llvm.ptr
  }
  func.func @isqrt_u64(%arg0: i64) -> i64 {
    %3 = arith.constant 0 : i32
    %5 = arith.extsi %3 : i32 to i64
    %4 = arith.cmpi eq, %arg0, %5 : i64
    cf.cond_br %4, ^bb0, ^bb1
    ^bb0:
      %6 = arith.constant 0 : i32
      %7 = arith.extsi %6 : i32 to i64
      func.return %7 : i64
    ^bb1:
      cf.br ^bb2
    ^bb2:
    %8 = arith.sitofp %arg0 : i64 to f64
    %9 = math.sqrt %8 : f64
    %10 = arith.fptosi %9 : f64 to i64
    %11 = llvm.mlir.constant(1 : i64) : i64
    %12 = llvm.alloca %11 x i64 : (i64) -> !llvm.ptr
    llvm.store %10, %12 : i64, !llvm.ptr
    cf.br ^bb3
    ^bb3:
    %13 = llvm.load %12 : !llvm.ptr -> i64
    %14 = llvm.load %12 : !llvm.ptr -> i64
    %15 = arith.muli %13, %14 : i64
    %16 = arith.cmpi sgt, %15, %arg0 : i64
    cf.cond_br %16, ^bb4, ^bb5
    ^bb4:
      %17 = llvm.load %12 : !llvm.ptr -> i64
      %18 = arith.constant 1 : i32
      %20 = arith.extsi %18 : i32 to i64
      %19 = arith.subi %17, %20 : i64
      llvm.store %19, %12 : i64, !llvm.ptr
      cf.br ^bb3
    ^bb5:
    cf.br ^bb6
    ^bb6:
    %21 = llvm.load %12 : !llvm.ptr -> i64
    %22 = arith.constant 1 : i32
    %24 = arith.extsi %22 : i32 to i64
    %23 = arith.addi %21, %24 : i64
    %25 = llvm.load %12 : !llvm.ptr -> i64
    %26 = arith.constant 1 : i32
    %28 = arith.extsi %26 : i32 to i64
    %27 = arith.addi %25, %28 : i64
    %29 = arith.muli %23, %27 : i64
    %30 = arith.cmpi sle, %29, %arg0 : i64
    cf.cond_br %30, ^bb7, ^bb8
    ^bb7:
      %31 = llvm.load %12 : !llvm.ptr -> i64
      %32 = arith.constant 1 : i32
      %34 = arith.extsi %32 : i32 to i64
      %33 = arith.addi %31, %34 : i64
      llvm.store %33, %12 : i64, !llvm.ptr
      cf.br ^bb6
    ^bb8:
    %35 = llvm.load %12 : !llvm.ptr -> i64
    func.return %35 : i64
  }
  func.func @gcd_u64(%arg0: i64, %arg1: i64) -> i64 {
    %36 = llvm.mlir.constant(1 : i64) : i64
    %37 = llvm.alloca %36 x i64 : (i64) -> !llvm.ptr
    llvm.store %arg0, %37 : i64, !llvm.ptr
    %38 = llvm.mlir.constant(1 : i64) : i64
    %39 = llvm.alloca %38 x i64 : (i64) -> !llvm.ptr
    llvm.store %arg1, %39 : i64, !llvm.ptr
    cf.br ^bb9
    ^bb9:
    %40 = llvm.load %39 : !llvm.ptr -> i64
    %41 = arith.constant 0 : i32
    %43 = arith.extsi %41 : i32 to i64
    %42 = arith.cmpi ne, %40, %43 : i64
    cf.cond_br %42, ^bb10, ^bb11
    ^bb10:
      %44 = llvm.load %37 : !llvm.ptr -> i64
      %45 = llvm.load %39 : !llvm.ptr -> i64
      %46 = arith.remsi %44, %45 : i64
      %47 = llvm.load %39 : !llvm.ptr -> i64
      llvm.store %47, %37 : i64, !llvm.ptr
      llvm.store %46, %39 : i64, !llvm.ptr
      cf.br ^bb9
    ^bb11:
    %48 = llvm.load %37 : !llvm.ptr -> i64
    func.return %48 : i64
  }
  func.func @cache_hash(%arg0: i64) -> i64 {
    %49 = arith.constant 16 : i32
    %51 = arith.extsi %49 : i32 to i64
    %50 = arith.shrsi %arg0, %51 : i64
    %52 = arith.xori %arg0, %50 : i64
    %53 = arith.constant -1640531535 : i32
    %55 = arith.extsi %53 : i32 to i64
    %54 = arith.muli %52, %55 : i64
    %56 = arith.constant 16 : i32
    %58 = arith.extsi %56 : i32 to i64
    %57 = arith.shrsi %54, %58 : i64
    %59 = arith.xori %54, %57 : i64
    %60 = llvm.mlir.addressof @CACHE_MASK : !llvm.ptr
    %61 = llvm.load %60 : !llvm.ptr -> i64
    %62 = arith.andi %59, %61 : i64
    func.return %62 : i64
  }
  func.func @cache_lookup(%arg0: i64, %arg1: !llvm.ptr) -> i64 {
    %63 = func.call @cache_hash(%arg0) : (i64) -> i64
    %64 = llvm.mlir.constant(1 : i64) : i64
    %65 = llvm.alloca %64 x i64 : (i64) -> !llvm.ptr
    llvm.store %63, %65 : i64, !llvm.ptr
    cf.br ^bb12
    ^bb12:
    %67 = llvm.mlir.addressof @g_cache_used : !llvm.ptr
    %68 = llvm.load %67 : !llvm.ptr -> !llvm.ptr
    %69 = llvm.load %65 : !llvm.ptr -> i64
    %70 = llvm.getelementptr %68[%69] : (!llvm.ptr, i64) -> !llvm.ptr, i8
    %66 = llvm.load %70 : !llvm.ptr -> i8
    %71 = arith.constant 0 : i32
    %73 = arith.extsi %66 : i8 to i32
    %72 = arith.cmpi ne, %73, %71 : i32
    cf.cond_br %72, ^bb13, ^bb14
    ^bb13:
      %75 = llvm.mlir.addressof @g_cache_key : !llvm.ptr
      %76 = llvm.load %75 : !llvm.ptr -> !llvm.ptr
      %77 = llvm.load %65 : !llvm.ptr -> i64
      %78 = llvm.getelementptr %76[%77] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %74 = llvm.load %78 : !llvm.ptr -> i64
      %79 = arith.cmpi eq, %74, %arg0 : i64
      cf.cond_br %79, ^bb15, ^bb16
      ^bb15:
        %80 = arith.constant 1 : i32
        %81 = arith.constant 0 : i32
        %82 = arith.extsi %81 : i32 to i64
        %83 = llvm.getelementptr %arg1[%82] : (!llvm.ptr, i64) -> !llvm.ptr, i32
        llvm.store %80, %83 : i32, !llvm.ptr
        %85 = llvm.mlir.addressof @g_cache_val : !llvm.ptr
        %86 = llvm.load %85 : !llvm.ptr -> !llvm.ptr
        %87 = llvm.load %65 : !llvm.ptr -> i64
        %88 = llvm.getelementptr %86[%87] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        %84 = llvm.load %88 : !llvm.ptr -> i64
        func.return %84 : i64
      ^bb16:
        cf.br ^bb17
      ^bb17:
      %89 = llvm.load %65 : !llvm.ptr -> i64
      %90 = arith.constant 1 : i32
      %92 = arith.extsi %90 : i32 to i64
      %91 = arith.addi %89, %92 : i64
      %93 = llvm.mlir.addressof @CACHE_MASK : !llvm.ptr
      %94 = llvm.load %93 : !llvm.ptr -> i64
      %95 = arith.andi %91, %94 : i64
      llvm.store %95, %65 : i64, !llvm.ptr
      cf.br ^bb12
    ^bb14:
    %96 = arith.constant 0 : i32
    %97 = arith.constant 0 : i32
    %98 = arith.extsi %97 : i32 to i64
    %99 = llvm.getelementptr %arg1[%98] : (!llvm.ptr, i64) -> !llvm.ptr, i32
    llvm.store %96, %99 : i32, !llvm.ptr
    %100 = arith.constant 0 : i32
    %101 = arith.extsi %100 : i32 to i64
    func.return %101 : i64
  }
  func.func @cache_insert(%arg0: i64, %arg1: i64) -> () {
    %102 = func.call @cache_hash(%arg0) : (i64) -> i64
    %103 = llvm.mlir.constant(1 : i64) : i64
    %104 = llvm.alloca %103 x i64 : (i64) -> !llvm.ptr
    llvm.store %102, %104 : i64, !llvm.ptr
    cf.br ^bb18
    ^bb18:
    %106 = llvm.mlir.addressof @g_cache_used : !llvm.ptr
    %107 = llvm.load %106 : !llvm.ptr -> !llvm.ptr
    %108 = llvm.load %104 : !llvm.ptr -> i64
    %109 = llvm.getelementptr %107[%108] : (!llvm.ptr, i64) -> !llvm.ptr, i8
    %105 = llvm.load %109 : !llvm.ptr -> i8
    %110 = arith.constant 0 : i32
    %112 = arith.extsi %105 : i8 to i32
    %111 = arith.cmpi ne, %112, %110 : i32
    cf.cond_br %111, ^bb19, ^bb20
    ^bb19:
      %114 = llvm.mlir.addressof @g_cache_key : !llvm.ptr
      %115 = llvm.load %114 : !llvm.ptr -> !llvm.ptr
      %116 = llvm.load %104 : !llvm.ptr -> i64
      %117 = llvm.getelementptr %115[%116] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %113 = llvm.load %117 : !llvm.ptr -> i64
      %118 = arith.cmpi eq, %113, %arg0 : i64
      cf.cond_br %118, ^bb21, ^bb22
      ^bb21:
        %119 = llvm.mlir.addressof @g_cache_val : !llvm.ptr
        %120 = llvm.load %119 : !llvm.ptr -> !llvm.ptr
        %121 = llvm.load %104 : !llvm.ptr -> i64
        %122 = llvm.getelementptr %120[%121] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        llvm.store %arg1, %122 : i64, !llvm.ptr
        func.return
      ^bb22:
        cf.br ^bb23
      ^bb23:
      %123 = llvm.load %104 : !llvm.ptr -> i64
      %124 = arith.constant 1 : i32
      %126 = arith.extsi %124 : i32 to i64
      %125 = arith.addi %123, %126 : i64
      %127 = llvm.mlir.addressof @CACHE_MASK : !llvm.ptr
      %128 = llvm.load %127 : !llvm.ptr -> i64
      %129 = arith.andi %125, %128 : i64
      llvm.store %129, %104 : i64, !llvm.ptr
      cf.br ^bb18
    ^bb20:
    %130 = arith.constant 1 : i32
    %131 = llvm.mlir.addressof @g_cache_used : !llvm.ptr
    %132 = llvm.load %131 : !llvm.ptr -> !llvm.ptr
    %133 = llvm.load %104 : !llvm.ptr -> i64
    %134 = arith.trunci %130 : i32 to i8
    %135 = llvm.getelementptr %132[%133] : (!llvm.ptr, i64) -> !llvm.ptr, i8
    llvm.store %134, %135 : i8, !llvm.ptr
    %136 = llvm.mlir.addressof @g_cache_key : !llvm.ptr
    %137 = llvm.load %136 : !llvm.ptr -> !llvm.ptr
    %138 = llvm.load %104 : !llvm.ptr -> i64
    %139 = llvm.getelementptr %137[%138] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %arg0, %139 : i64, !llvm.ptr
    %140 = llvm.mlir.addressof @g_cache_val : !llvm.ptr
    %141 = llvm.load %140 : !llvm.ptr -> !llvm.ptr
    %142 = llvm.load %104 : !llvm.ptr -> i64
    %143 = llvm.getelementptr %141[%142] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %arg1, %143 : i64, !llvm.ptr
    func.return
  }
  func.func @chi_prefix(%arg0: i64) -> i64 {
    %144 = arith.constant 2 : i32
    %146 = arith.extsi %144 : i32 to i64
    %145 = arith.addi %arg0, %146 : i64
    %147 = arith.constant 3 : i32
    %149 = arith.extsi %147 : i32 to i64
    %148 = arith.divsi %145, %149 : i64
    %150 = arith.constant 1 : i32
    %152 = arith.extsi %150 : i32 to i64
    %151 = arith.addi %arg0, %152 : i64
    %153 = arith.constant 3 : i32
    %155 = arith.extsi %153 : i32 to i64
    %154 = arith.divsi %151, %155 : i64
    %156 = arith.subi %148, %154 : i64
    func.return %156 : i64
  }
  func.func @count_hex_points_leq(%arg0: i64) -> i64 {
    %157 = arith.constant 0 : i32
    %159 = arith.extsi %157 : i32 to i64
    %158 = arith.cmpi eq, %arg0, %159 : i64
    cf.cond_br %158, ^bb24, ^bb25
    ^bb24:
      %160 = arith.constant 1 : i32
      %161 = arith.extsi %160 : i32 to i64
      func.return %161 : i64
    ^bb25:
      cf.br ^bb26
    ^bb26:
    %163 = arith.constant 1 : i32
    %164 = arith.constant 4 : i32
    %165 = arith.extsi %163 : i32 to i64
    %166 = arith.extsi %164 : i32 to i64
    %162 = func.call @calloc(%165, %166) : (i64, i64) -> !llvm.ptr
    %167 = func.call @cache_lookup(%arg0, %162) : (i64, !llvm.ptr) -> i64
    %169 = arith.constant 0 : i32
    %170 = arith.extsi %169 : i32 to i64
    %171 = llvm.getelementptr %162[%170] : (!llvm.ptr, i64) -> !llvm.ptr, i32
    %168 = llvm.load %171 : !llvm.ptr -> i32
    %172 = arith.constant 1 : i32
    %173 = arith.cmpi eq, %168, %172 : i32
    cf.cond_br %173, ^bb27, ^bb28
    ^bb27:
      func.call @free(%162) : (!llvm.ptr) -> ()
      func.return %167 : i64
    ^bb28:
      cf.br ^bb29
    ^bb29:
    func.call @free(%162) : (!llvm.ptr) -> ()
    %176 = arith.constant 0 : i32
    %177 = arith.extsi %176 : i32 to i64
    %178 = llvm.mlir.constant(1 : i64) : i64
    %179 = llvm.alloca %178 x i64 : (i64) -> !llvm.ptr
    llvm.store %177, %179 : i64, !llvm.ptr
    %180 = arith.constant 1 : i32
    %181 = arith.extsi %180 : i32 to i64
    %182 = llvm.mlir.constant(1 : i64) : i64
    %183 = llvm.alloca %182 x i64 : (i64) -> !llvm.ptr
    llvm.store %181, %183 : i64, !llvm.ptr
    cf.br ^bb30
    ^bb30:
    %184 = llvm.load %183 : !llvm.ptr -> i64
    %185 = arith.cmpi sle, %184, %arg0 : i64
    cf.cond_br %185, ^bb31, ^bb32
    ^bb31:
      %186 = llvm.load %183 : !llvm.ptr -> i64
      %187 = arith.divsi %arg0, %186 : i64
      %188 = arith.divsi %arg0, %187 : i64
      %189 = llvm.load %179 : !llvm.ptr -> i64
      %190 = func.call @chi_prefix(%188) : (i64) -> i64
      %192 = llvm.load %183 : !llvm.ptr -> i64
      %193 = arith.constant 1 : i32
      %195 = arith.extsi %193 : i32 to i64
      %194 = arith.subi %192, %195 : i64
      %191 = func.call @chi_prefix(%194) : (i64) -> i64
      %196 = arith.subi %190, %191 : i64
      %197 = arith.muli %187, %196 : i64
      %198 = arith.addi %189, %197 : i64
      llvm.store %198, %179 : i64, !llvm.ptr
      %199 = arith.constant 1 : i32
      %201 = arith.extsi %199 : i32 to i64
      %200 = arith.addi %188, %201 : i64
      llvm.store %200, %183 : i64, !llvm.ptr
      cf.br ^bb30
    ^bb32:
    %202 = arith.constant 1 : i32
    %203 = arith.constant 6 : i32
    %204 = llvm.load %179 : !llvm.ptr -> i64
    %206 = arith.extsi %203 : i32 to i64
    %205 = arith.muli %206, %204 : i64
    %208 = arith.extsi %202 : i32 to i64
    %207 = arith.addi %208, %205 : i64
    func.call @cache_insert(%arg0, %207) : (i64, i64) -> ()
    func.return %207 : i64
  }
  func.func @remarkable_triangles(%arg0: i64, %arg1: i64) -> i64 {
    %210 = arith.constant 12 : i32
    %212 = arith.extsi %210 : i32 to i64
    %211 = arith.muli %212, %arg0 : i64
    %213 = arith.muli %211, %arg0 : i64
    %214 = arith.muli %arg1, %arg1 : i64
    %216 = arith.divsi %213, %214 : i64
    %215 = func.call @isqrt_u64(%216) : (i64) -> i64
    %218 = arith.constant 1 : i32
    %220 = arith.extsi %218 : i32 to i64
    %219 = arith.addi %215, %220 : i64
    %221 = arith.constant 4 : i32
    %222 = arith.extsi %221 : i32 to i64
    %217 = func.call @calloc(%219, %222) : (i64, i64) -> !llvm.ptr
    %223 = arith.constant 3 : i32
    %225 = arith.extsi %223 : i32 to i64
    %224 = arith.divsi %215, %225 : i64
    %227 = arith.constant 1 : i32
    %229 = arith.extsi %227 : i32 to i64
    %228 = arith.addi %224, %229 : i64
    %230 = arith.constant 1 : i32
    %231 = arith.extsi %230 : i32 to i64
    %226 = func.call @calloc(%228, %231) : (i64, i64) -> !llvm.ptr
    %232 = arith.constant 2 : i32
    %233 = arith.extsi %232 : i32 to i64
    %234 = llvm.mlir.constant(1 : i64) : i64
    %235 = llvm.alloca %234 x i64 : (i64) -> !llvm.ptr
    llvm.store %233, %235 : i64, !llvm.ptr
    cf.br ^bb33
    ^bb33:
    %236 = llvm.load %235 : !llvm.ptr -> i64
    %237 = arith.cmpi sle, %236, %224 : i64
    cf.cond_br %237, ^bb34, ^bb35
    ^bb34:
      %239 = llvm.load %235 : !llvm.ptr -> i64
      %240 = llvm.getelementptr %226[%239] : (!llvm.ptr, i64) -> !llvm.ptr, i8
      %238 = llvm.load %240 : !llvm.ptr -> i8
      %241 = arith.constant 0 : i32
      %243 = arith.extsi %238 : i8 to i32
      %242 = arith.cmpi eq, %243, %241 : i32
      cf.cond_br %242, ^bb36, ^bb37
      ^bb36:
        %244 = llvm.load %235 : !llvm.ptr -> i64
        %245 = llvm.mlir.constant(1 : i64) : i64
        %246 = llvm.alloca %245 x i64 : (i64) -> !llvm.ptr
        llvm.store %244, %246 : i64, !llvm.ptr
        cf.br ^bb39
        ^bb39:
        %247 = llvm.load %246 : !llvm.ptr -> i64
        %248 = arith.cmpi sle, %247, %224 : i64
        cf.cond_br %248, ^bb40, ^bb41
        ^bb40:
          %250 = llvm.load %246 : !llvm.ptr -> i64
          %251 = llvm.getelementptr %226[%250] : (!llvm.ptr, i64) -> !llvm.ptr, i8
          %249 = llvm.load %251 : !llvm.ptr -> i8
          %252 = arith.constant 1 : i32
          %254 = arith.extsi %249 : i8 to i32
          %253 = arith.addi %254, %252 : i32
          %255 = llvm.load %246 : !llvm.ptr -> i64
          %256 = arith.trunci %253 : i32 to i8
          %257 = llvm.getelementptr %226[%255] : (!llvm.ptr, i64) -> !llvm.ptr, i8
          llvm.store %256, %257 : i8, !llvm.ptr
          %258 = llvm.load %246 : !llvm.ptr -> i64
          %259 = llvm.load %235 : !llvm.ptr -> i64
          %260 = arith.addi %258, %259 : i64
          llvm.store %260, %246 : i64, !llvm.ptr
          cf.br ^bb39
        ^bb41:
        cf.br ^bb38
      ^bb37:
        cf.br ^bb38
      ^bb38:
      %261 = llvm.load %235 : !llvm.ptr -> i64
      %262 = arith.constant 1 : i32
      %264 = arith.extsi %262 : i32 to i64
      %263 = arith.addi %261, %264 : i64
      llvm.store %263, %235 : i64, !llvm.ptr
      cf.br ^bb33
    ^bb35:
    %265 = arith.constant 2 : i32
    %266 = arith.extsi %265 : i32 to i64
    %267 = llvm.mlir.constant(1 : i64) : i64
    %268 = llvm.alloca %267 x i64 : (i64) -> !llvm.ptr
    llvm.store %266, %268 : i64, !llvm.ptr
    cf.br ^bb42
    ^bb42:
    %269 = llvm.load %268 : !llvm.ptr -> i64
    %270 = arith.cmpi sle, %269, %224 : i64
    cf.cond_br %270, ^bb43, ^bb44
    ^bb43:
      %271 = llvm.load %268 : !llvm.ptr -> i64
      %272 = arith.constant 3 : i32
      %274 = arith.extsi %272 : i32 to i64
      %273 = arith.remsi %271, %274 : i64
      %275 = arith.constant 1 : i32
      %277 = arith.extsi %275 : i32 to i64
      %276 = arith.cmpi ne, %273, %277 : i64
      cf.cond_br %276, ^bb45, ^bb46
      ^bb45:
        %279 = arith.constant 3 : i32
        %280 = llvm.load %268 : !llvm.ptr -> i64
        %282 = arith.extsi %279 : i32 to i64
        %281 = arith.muli %282, %280 : i64
        %283 = llvm.getelementptr %217[%281] : (!llvm.ptr, i64) -> !llvm.ptr, i32
        %278 = llvm.load %283 : !llvm.ptr -> i32
        %284 = arith.constant 1 : i32
        %286 = llvm.load %268 : !llvm.ptr -> i64
        %287 = llvm.getelementptr %226[%286] : (!llvm.ptr, i64) -> !llvm.ptr, i8
        %285 = llvm.load %287 : !llvm.ptr -> i8
        %288 = arith.constant 1 : i32
        %290 = arith.extsi %285 : i8 to i32
        %289 = arith.subi %290, %288 : i32
        %291 = arith.shli %284, %289 : i32
        %292 = arith.addi %278, %291 : i32
        %293 = arith.constant 3 : i32
        %294 = llvm.load %268 : !llvm.ptr -> i64
        %296 = arith.extsi %293 : i32 to i64
        %295 = arith.muli %296, %294 : i64
        %297 = llvm.getelementptr %217[%295] : (!llvm.ptr, i64) -> !llvm.ptr, i32
        llvm.store %292, %297 : i32, !llvm.ptr
        cf.br ^bb47
      ^bb46:
        cf.br ^bb47
      ^bb47:
      %298 = llvm.load %268 : !llvm.ptr -> i64
      %299 = arith.constant 1 : i32
      %301 = arith.extsi %299 : i32 to i64
      %300 = arith.addi %298, %301 : i64
      llvm.store %300, %268 : i64, !llvm.ptr
      cf.br ^bb42
    ^bb44:
    func.call @free(%226) : (!llvm.ptr) -> ()
    %303 = arith.constant 3 : i32
    %305 = arith.extsi %303 : i32 to i64
    %304 = arith.divsi %215, %305 : i64
    %306 = arith.constant 1 : i32
    %307 = arith.extsi %306 : i32 to i64
    %308 = llvm.mlir.constant(1 : i64) : i64
    %309 = llvm.alloca %308 x i64 : (i64) -> !llvm.ptr
    llvm.store %307, %309 : i64, !llvm.ptr
    cf.br ^bb48
    ^bb48:
    %310 = llvm.load %309 : !llvm.ptr -> i64
    %311 = arith.cmpi sle, %310, %304 : i64
    cf.cond_br %311, ^bb49, ^bb50
    ^bb49:
      %312 = arith.constant 9 : i32
      %313 = llvm.load %309 : !llvm.ptr -> i64
      %315 = arith.extsi %312 : i32 to i64
      %314 = arith.muli %315, %313 : i64
      %316 = llvm.load %309 : !llvm.ptr -> i64
      %317 = arith.muli %314, %316 : i64
      %318 = arith.constant 4 : i32
      %320 = arith.extsi %318 : i32 to i64
      %319 = arith.muli %320, %215 : i64
      %321 = arith.addi %317, %319 : i64
      %322 = func.call @isqrt_u64(%321) : (i64) -> i64
      %323 = arith.constant 3 : i32
      %324 = llvm.load %309 : !llvm.ptr -> i64
      %326 = arith.extsi %323 : i32 to i64
      %325 = arith.muli %326, %324 : i64
      %327 = arith.subi %322, %325 : i64
      %328 = arith.constant 2 : i32
      %330 = arith.extsi %328 : i32 to i64
      %329 = arith.divsi %327, %330 : i64
      %331 = arith.constant 0 : i32
      %333 = arith.extsi %331 : i32 to i64
      %332 = arith.cmpi sle, %329, %333 : i64
      cf.cond_br %332, ^bb51, ^bb52
      ^bb51:
        %334 = llvm.load %309 : !llvm.ptr -> i64
        %335 = arith.constant 1 : i32
        %337 = arith.extsi %335 : i32 to i64
        %336 = arith.addi %334, %337 : i64
        llvm.store %336, %309 : i64, !llvm.ptr
        cf.br ^bb48
      ^bb52:
        cf.br ^bb53
      ^bb53:
      %338 = arith.constant 1 : i32
      %339 = arith.extsi %338 : i32 to i64
      %340 = llvm.mlir.constant(1 : i64) : i64
      %341 = llvm.alloca %340 x i64 : (i64) -> !llvm.ptr
      llvm.store %339, %341 : i64, !llvm.ptr
      cf.br ^bb54
      ^bb54:
      %342 = llvm.load %341 : !llvm.ptr -> i64
      %343 = arith.cmpi sle, %342, %329 : i64
      cf.cond_br %343, ^bb55, ^bb56
      ^bb55:
        %344 = llvm.load %309 : !llvm.ptr -> i64
        %345 = llvm.load %341 : !llvm.ptr -> i64
        %346 = arith.addi %344, %345 : i64
        %347 = arith.constant 3 : i32
        %349 = arith.extsi %347 : i32 to i64
        %348 = arith.remsi %346, %349 : i64
        %350 = llvm.load %309 : !llvm.ptr -> i64
        %351 = arith.constant 3 : i32
        %353 = arith.extsi %351 : i32 to i64
        %352 = arith.remsi %350, %353 : i64
        %354 = arith.cmpi ne, %348, %352 : i64
        cf.cond_br %354, ^bb57, ^bb58
        ^bb57:
          %356 = llvm.load %309 : !llvm.ptr -> i64
          %355 = func.call @gcd_u64(%346, %356) : (i64, i64) -> i64
          %357 = arith.constant 1 : i32
          %359 = arith.extsi %357 : i32 to i64
          %358 = arith.cmpi eq, %355, %359 : i64
          cf.cond_br %358, ^bb60, ^bb61
          ^bb60:
            %360 = llvm.load %341 : !llvm.ptr -> i64
            %361 = llvm.load %341 : !llvm.ptr -> i64
            %362 = arith.constant 3 : i32
            %363 = llvm.load %309 : !llvm.ptr -> i64
            %365 = arith.extsi %362 : i32 to i64
            %364 = arith.muli %365, %363 : i64
            %366 = arith.addi %361, %364 : i64
            %367 = arith.muli %360, %366 : i64
            %369 = llvm.getelementptr %217[%367] : (!llvm.ptr, i64) -> !llvm.ptr, i32
            %368 = llvm.load %369 : !llvm.ptr -> i32
            %370 = arith.constant 1 : i32
            %371 = arith.addi %368, %370 : i32
            %372 = llvm.getelementptr %217[%367] : (!llvm.ptr, i64) -> !llvm.ptr, i32
            llvm.store %371, %372 : i32, !llvm.ptr
            cf.br ^bb62
          ^bb61:
            cf.br ^bb62
          ^bb62:
          cf.br ^bb59
        ^bb58:
          cf.br ^bb59
        ^bb59:
        %373 = llvm.load %341 : !llvm.ptr -> i64
        %374 = arith.constant 1 : i32
        %376 = arith.extsi %374 : i32 to i64
        %375 = arith.addi %373, %376 : i64
        llvm.store %375, %341 : i64, !llvm.ptr
        cf.br ^bb54
      ^bb56:
      %377 = llvm.load %309 : !llvm.ptr -> i64
      %378 = arith.constant 1 : i32
      %380 = arith.extsi %378 : i32 to i64
      %379 = arith.addi %377, %380 : i64
      llvm.store %379, %309 : i64, !llvm.ptr
      cf.br ^bb48
    ^bb50:
    %381 = arith.constant 0 : i32
    %382 = arith.extsi %381 : i32 to i64
    %383 = llvm.mlir.constant(1 : i64) : i64
    %384 = llvm.alloca %383 x i64 : (i64) -> !llvm.ptr
    llvm.store %382, %384 : i64, !llvm.ptr
    %385 = arith.constant 1 : i32
    %386 = arith.extsi %385 : i32 to i64
    %387 = llvm.mlir.constant(1 : i64) : i64
    %388 = llvm.alloca %387 x i64 : (i64) -> !llvm.ptr
    llvm.store %386, %388 : i64, !llvm.ptr
    cf.br ^bb63
    ^bb63:
    %389 = llvm.load %388 : !llvm.ptr -> i64
    %390 = arith.cmpi sle, %389, %215 : i64
    cf.cond_br %390, ^bb64, ^bb65
    ^bb64:
      %392 = llvm.load %388 : !llvm.ptr -> i64
      %393 = llvm.getelementptr %217[%392] : (!llvm.ptr, i64) -> !llvm.ptr, i32
      %391 = llvm.load %393 : !llvm.ptr -> i32
      %394 = arith.constant 0 : i32
      %395 = arith.cmpi eq, %391, %394 : i32
      cf.cond_br %395, ^bb66, ^bb67
      ^bb66:
        %396 = llvm.load %388 : !llvm.ptr -> i64
        %397 = arith.constant 1 : i32
        %399 = arith.extsi %397 : i32 to i64
        %398 = arith.addi %396, %399 : i64
        llvm.store %398, %388 : i64, !llvm.ptr
        cf.br ^bb63
      ^bb67:
        cf.br ^bb68
      ^bb68:
      %400 = llvm.load %388 : !llvm.ptr -> i64
      %401 = arith.muli %214, %400 : i64
      %402 = llvm.load %388 : !llvm.ptr -> i64
      %403 = arith.muli %401, %402 : i64
      %404 = arith.divsi %213, %403 : i64
      %405 = arith.constant 0 : i32
      %407 = arith.extsi %405 : i32 to i64
      %406 = arith.cmpi eq, %404, %407 : i64
      cf.cond_br %406, ^bb69, ^bb70
      ^bb69:
        %408 = llvm.load %388 : !llvm.ptr -> i64
        %409 = arith.constant 1 : i32
        %411 = arith.extsi %409 : i32 to i64
        %410 = arith.addi %408, %411 : i64
        llvm.store %410, %388 : i64, !llvm.ptr
        cf.br ^bb63
      ^bb70:
        cf.br ^bb71
      ^bb71:
      %412 = arith.constant 0 : i32
      %413 = arith.extsi %412 : i32 to i64
      %414 = llvm.load %388 : !llvm.ptr -> i64
      %415 = arith.constant 3 : i32
      %417 = arith.extsi %415 : i32 to i64
      %416 = arith.remsi %414, %417 : i64
      %418 = arith.constant 0 : i32
      %420 = arith.extsi %418 : i32 to i64
      %419 = arith.cmpi eq, %416, %420 : i64
      %421 = scf.if %419 -> (i64) {
        %422 = func.call @count_hex_points_leq(%404) : (i64) -> i64
        %423 = arith.constant 1 : i32
        %425 = arith.extsi %423 : i32 to i64
        %424 = arith.subi %422, %425 : i64
        scf.yield %424 : i64
      } else {
        %427 = arith.constant 3 : i32
        %429 = arith.extsi %427 : i32 to i64
        %428 = arith.divsi %404, %429 : i64
        %426 = func.call @count_hex_points_leq(%428) : (i64) -> i64
        %430 = arith.constant 1 : i32
        %432 = arith.extsi %430 : i32 to i64
        %431 = arith.subi %426, %432 : i64
        scf.yield %431 : i64
      }
      %433 = llvm.load %384 : !llvm.ptr -> i64
      %434 = arith.constant 2 : i32
      %435 = arith.extsi %391 : i32 to i64
      %437 = arith.extsi %434 : i32 to i64
      %436 = arith.muli %437, %435 : i64
      %438 = arith.muli %436, %421 : i64
      %439 = arith.addi %433, %438 : i64
      llvm.store %439, %384 : i64, !llvm.ptr
      %440 = llvm.load %388 : !llvm.ptr -> i64
      %441 = arith.constant 1 : i32
      %443 = arith.extsi %441 : i32 to i64
      %442 = arith.addi %440, %443 : i64
      llvm.store %442, %388 : i64, !llvm.ptr
      cf.br ^bb63
    ^bb65:
    %444 = arith.constant 4 : i32
    %446 = arith.extsi %444 : i32 to i64
    %445 = arith.muli %446, %arg0 : i64
    %447 = arith.muli %445, %arg0 : i64
    %448 = arith.muli %arg1, %arg1 : i64
    %449 = arith.divsi %447, %448 : i64
    %450 = func.call @count_hex_points_leq(%449) : (i64) -> i64
    %451 = arith.constant 1 : i32
    %453 = arith.extsi %451 : i32 to i64
    %452 = arith.subi %450, %453 : i64
    %454 = arith.constant 3 : i32
    %456 = arith.extsi %454 : i32 to i64
    %455 = arith.divsi %452, %456 : i64
    func.call @free(%217) : (!llvm.ptr) -> ()
    %458 = llvm.load %384 : !llvm.ptr -> i64
    %459 = arith.addi %458, %455 : i64
    func.return %459 : i64
  }
  func.func @main() -> i32 {
    %461 = llvm.mlir.addressof @CACHE_CAP : !llvm.ptr
    %462 = llvm.load %461 : !llvm.ptr -> i64
    %463 = arith.constant 8 : i32
    %464 = arith.extsi %463 : i32 to i64
    %460 = func.call @calloc(%462, %464) : (i64, i64) -> !llvm.ptr
    %465 = llvm.mlir.addressof @g_cache_key : !llvm.ptr
    llvm.store %460, %465 : !llvm.ptr, !llvm.ptr
    %467 = llvm.mlir.addressof @CACHE_CAP : !llvm.ptr
    %468 = llvm.load %467 : !llvm.ptr -> i64
    %469 = arith.constant 8 : i32
    %470 = arith.extsi %469 : i32 to i64
    %466 = func.call @calloc(%468, %470) : (i64, i64) -> !llvm.ptr
    %471 = llvm.mlir.addressof @g_cache_val : !llvm.ptr
    llvm.store %466, %471 : !llvm.ptr, !llvm.ptr
    %473 = llvm.mlir.addressof @CACHE_CAP : !llvm.ptr
    %474 = llvm.load %473 : !llvm.ptr -> i64
    %475 = arith.constant 1 : i32
    %476 = arith.extsi %475 : i32 to i64
    %472 = func.call @calloc(%474, %476) : (i64, i64) -> !llvm.ptr
    %477 = llvm.mlir.addressof @g_cache_used : !llvm.ptr
    llvm.store %472, %477 : !llvm.ptr, !llvm.ptr
    %479 = arith.constant 1000000 : i32
    %480 = arith.constant 1 : i32
    %481 = arith.extsi %479 : i32 to i64
    %482 = arith.extsi %480 : i32 to i64
    %478 = func.call @remarkable_triangles(%481, %482) : (i64, i64) -> i64
    %483 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %484 = llvm.call @printf(%483, %478) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, i64) -> i32
    %486 = llvm.mlir.addressof @g_cache_key : !llvm.ptr
    %487 = llvm.load %486 : !llvm.ptr -> !llvm.ptr
    func.call @free(%487) : (!llvm.ptr) -> ()
    %489 = llvm.mlir.addressof @g_cache_val : !llvm.ptr
    %490 = llvm.load %489 : !llvm.ptr -> !llvm.ptr
    func.call @free(%490) : (!llvm.ptr) -> ()
    %492 = llvm.mlir.addressof @g_cache_used : !llvm.ptr
    %493 = llvm.load %492 : !llvm.ptr -> !llvm.ptr
    func.call @free(%493) : (!llvm.ptr) -> ()
    %494 = arith.constant 0 : i32
    func.return %494 : i32
  }
}