Problem 717

Summation of a Modular Formula: G(10^7). For each odd prime p, compute the formula using modular exponentiation with i128 intermediates to avoid overflow.

Answer1603036763131
Output1603036763131
StatusPASS
Native helperno
Runtime840 ms
Peak memory10896 KB
Time complexityO(n^2) (estimated)
Space complexityO(n) (estimated)

Performance comparison

MetricOur solutionBest known
Time complexityO(n^2)O(log n)
Space complexityO(n)O(1)
ApproachFlow solutionModular exponentiation
VerdictSuboptimal

Flow source

# Project Euler 717
# Summation of a Modular Formula: G(10^7).
# For each odd prime p, compute the formula using modular exponentiation
# with i128 intermediates to avoid overflow.

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

const N: i64 = 10000000

function modpow(b: i64, e: i64, m: i64) -> i64 {
    let mut r: i64 = 1 % m
    let mut bb: i64 = b % m
    let mut ee: i64 = e
    while ee > 0 {
        if ee % 2 == 1 {
            r = ((r as i128) * (bb as i128) % (m as i128)) as i64
        }
        bb = ((bb as i128) * (bb as i128) % (m as i128)) as i64
        ee = ee / 2
    }
    return r
}

function main() -> i32 {
    let comp: ptr<i8> = calloc(N + 1, 1)
    if comp == null {
        return 1
    }
    let mut i: i64 = 2
    while i * i <= N {
        if comp[i] == 0 {
            let mut j: i64 = i * i
            while j <= N {
                comp[j] = 1
                j = j + i
            }
        }
        i = i + 1
    }
    let mut total: i64 = 0
    let mut p: i64 = 3
    while p <= N {
        if comp[p] == 0 {
            let x: i64 = modpow(2, p, p - 1)
            let r: i64 = modpow(2, x, p)
            let n: i64 = r * (p - 1) / (2 * p) + 1
            let modpp: i64 = p * p
            let pp: i64 = modpow(2, p - 1, modpp)
            let va: i128 = (r as i128) * ((p - 1) as i128) % (modpp as i128)
            let vb: i128 = 0 - va
            let vc: i128 = vb * (pp as i128) % (modpp as i128)
            let mut t: i64 = ((vc + 1) % (modpp as i128)) as i64
            if t < 0 {
                t = t + modpp
            }
            let fp: i64 = t / p
            total = total + (fp + 2 * n) % p
        }
        p = p + 1
    }
    free(comp as ptr<void>)
    printf("%lld\n", total)
    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 modpow_i64_i64_i64(int64_t b, int64_t e, int64_t m);
int32_t main(void);

static const int64_t N = 10000000;



int64_t modpow_i64_i64_i64(int64_t b, int64_t e, int64_t m) {
    int64_t r = FLOW_CHECKED_MOD((1), (m));
    int64_t bb = FLOW_CHECKED_MOD((b), (m));
    int64_t ee = e;
    while (ee > 0) {
        if (FLOW_CHECKED_MOD((ee), (2)) == 1) {
            r = ((int64_t)(FLOW_CHECKED_MOD(((((__int128)(r)) * ((__int128)(bb)))), (((__int128)(m))))));
        }
        bb = ((int64_t)(FLOW_CHECKED_MOD(((((__int128)(bb)) * ((__int128)(bb)))), (((__int128)(m))))));
        ee = FLOW_CHECKED_DIV((ee), (2));
    }
    return r;
}

int32_t main(void) {
    int8_t* comp = (int8_t*)(calloc((N + 1), 1));
    if (comp == NULL) {
        return 1;
    }
    int64_t i = 2;
    while ((i * i) <= N) {
        if (comp[i] == 0) {
            int64_t j = (i * i);
            while (j <= N) {
                comp[j] = 1;
                j = (j + i);
            }
        }
        i = (i + 1);
    }
    int64_t total = 0;
    int64_t p = 3;
    while (p <= N) {
        if (comp[p] == 0) {
            int64_t x = modpow_i64_i64_i64(2, p, (p - 1));
            int64_t r = modpow_i64_i64_i64(2, x, p);
            int64_t n = (FLOW_CHECKED_DIV(((r * (p - 1))), ((2 * p))) + 1);
            int64_t modpp = (p * p);
            int64_t pp = modpow_i64_i64_i64(2, (p - 1), modpp);
            __int128 va = FLOW_CHECKED_MOD(((((__int128)(r)) * ((__int128)((p - 1))))), (((__int128)(modpp))));
            __int128 vb = (0 - va);
            __int128 vc = FLOW_CHECKED_MOD(((vb * ((__int128)(pp)))), (((__int128)(modpp))));
            int64_t t = ((int64_t)(FLOW_CHECKED_MOD(((vc + 1)), (((__int128)(modpp))))));
            if (t < 0) {
                t = (t + modpp);
            }
            int64_t fp = FLOW_CHECKED_DIV((t), (p));
            total = (total + FLOW_CHECKED_MOD(((fp + (2 * n))), (p)));
        }
        p = (p + 1);
    }
    free(((void*)(comp)));
    printf("%lld\n", total);
    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) -> ()
  // Constant: N
  llvm.mlir.global internal constant @N(10000000 : i64) : i64
  func.func @modpow(%arg0: i64, %arg1: i64, %arg2: i64) -> i64 {
    %0 = arith.constant 1 : i32
    %2 = arith.extsi %0 : i32 to i64
    %1 = arith.remsi %2, %arg2 : i64
    %3 = llvm.mlir.constant(1 : i64) : i64
    %4 = llvm.alloca %3 x i64 : (i64) -> !llvm.ptr
    llvm.store %1, %4 : i64, !llvm.ptr
    %5 = arith.remsi %arg0, %arg2 : i64
    %6 = llvm.mlir.constant(1 : i64) : i64
    %7 = llvm.alloca %6 x i64 : (i64) -> !llvm.ptr
    llvm.store %5, %7 : i64, !llvm.ptr
    %8 = llvm.mlir.constant(1 : i64) : i64
    %9 = llvm.alloca %8 x i64 : (i64) -> !llvm.ptr
    llvm.store %arg1, %9 : i64, !llvm.ptr
    cf.br ^bb0
    ^bb0:
    %10 = llvm.load %9 : !llvm.ptr -> i64
    %11 = arith.constant 0 : i32
    %13 = arith.extsi %11 : i32 to i64
    %12 = arith.cmpi sgt, %10, %13 : i64
    cf.cond_br %12, ^bb1, ^bb2
    ^bb1:
      %14 = llvm.load %9 : !llvm.ptr -> i64
      %15 = arith.constant 2 : i32
      %17 = arith.extsi %15 : i32 to i64
      %16 = arith.remsi %14, %17 : i64
      %18 = arith.constant 1 : i32
      %20 = arith.extsi %18 : i32 to i64
      %19 = arith.cmpi eq, %16, %20 : i64
      cf.cond_br %19, ^bb3, ^bb4
      ^bb3:
        %21 = llvm.load %4 : !llvm.ptr -> i64
        %22 = arith.extsi %21 : i64 to i128
        %23 = llvm.load %7 : !llvm.ptr -> i64
        %24 = arith.extsi %23 : i64 to i128
        %26 = arith.trunci %22 : i128 to i64
        %27 = arith.trunci %24 : i128 to i64
        %25 = arith.muli %26, %27 : i64
        %28 = arith.extsi %arg2 : i64 to i128
        %30 = arith.trunci %28 : i128 to i64
        %29 = arith.remsi %25, %30 : i64
        llvm.store %29, %4 : i64, !llvm.ptr
        cf.br ^bb5
      ^bb4:
        cf.br ^bb5
      ^bb5:
      %31 = llvm.load %7 : !llvm.ptr -> i64
      %32 = arith.extsi %31 : i64 to i128
      %33 = llvm.load %7 : !llvm.ptr -> i64
      %34 = arith.extsi %33 : i64 to i128
      %36 = arith.trunci %32 : i128 to i64
      %37 = arith.trunci %34 : i128 to i64
      %35 = arith.muli %36, %37 : i64
      %38 = arith.extsi %arg2 : i64 to i128
      %40 = arith.trunci %38 : i128 to i64
      %39 = arith.remsi %35, %40 : i64
      llvm.store %39, %7 : i64, !llvm.ptr
      %41 = llvm.load %9 : !llvm.ptr -> i64
      %42 = arith.constant 2 : i32
      %44 = arith.extsi %42 : i32 to i64
      %43 = arith.divsi %41, %44 : i64
      llvm.store %43, %9 : i64, !llvm.ptr
      cf.br ^bb0
    ^bb2:
    %45 = llvm.load %4 : !llvm.ptr -> i64
    func.return %45 : i64
  }
  func.func @main() -> i32 {
    %47 = llvm.mlir.addressof @N : !llvm.ptr
    %48 = llvm.load %47 : !llvm.ptr -> i64
    %49 = arith.constant 1 : i32
    %51 = arith.extsi %49 : i32 to i64
    %50 = arith.addi %48, %51 : i64
    %52 = arith.constant 1 : i32
    %53 = arith.extsi %52 : i32 to i64
    %46 = func.call @calloc(%50, %53) : (i64, i64) -> !llvm.ptr
    %54 = llvm.mlir.zero : !llvm.ptr
    %55 = llvm.icmp "eq" %46, %54 : !llvm.ptr
    cf.cond_br %55, ^bb6, ^bb7
    ^bb6:
      %56 = arith.constant 1 : i32
      func.return %56 : i32
    ^bb7:
      cf.br ^bb8
    ^bb8:
    %57 = arith.constant 2 : i32
    %58 = arith.extsi %57 : i32 to i64
    %59 = llvm.mlir.constant(1 : i64) : i64
    %60 = llvm.alloca %59 x i64 : (i64) -> !llvm.ptr
    llvm.store %58, %60 : i64, !llvm.ptr
    cf.br ^bb9
    ^bb9:
    %61 = llvm.load %60 : !llvm.ptr -> i64
    %62 = llvm.load %60 : !llvm.ptr -> i64
    %63 = arith.muli %61, %62 : i64
    %64 = llvm.mlir.addressof @N : !llvm.ptr
    %65 = llvm.load %64 : !llvm.ptr -> i64
    %66 = arith.cmpi sle, %63, %65 : i64
    cf.cond_br %66, ^bb10, ^bb11
    ^bb10:
      %68 = llvm.load %60 : !llvm.ptr -> i64
      %69 = llvm.getelementptr %46[%68] : (!llvm.ptr, i64) -> !llvm.ptr, i8
      %67 = llvm.load %69 : !llvm.ptr -> i8
      %70 = arith.constant 0 : i32
      %72 = arith.extsi %67 : i8 to i32
      %71 = arith.cmpi eq, %72, %70 : i32
      cf.cond_br %71, ^bb12, ^bb13
      ^bb12:
        %73 = llvm.load %60 : !llvm.ptr -> i64
        %74 = llvm.load %60 : !llvm.ptr -> i64
        %75 = arith.muli %73, %74 : i64
        %76 = llvm.mlir.constant(1 : i64) : i64
        %77 = llvm.alloca %76 x i64 : (i64) -> !llvm.ptr
        llvm.store %75, %77 : i64, !llvm.ptr
        cf.br ^bb15
        ^bb15:
        %78 = llvm.load %77 : !llvm.ptr -> i64
        %79 = llvm.mlir.addressof @N : !llvm.ptr
        %80 = llvm.load %79 : !llvm.ptr -> i64
        %81 = arith.cmpi sle, %78, %80 : i64
        cf.cond_br %81, ^bb16, ^bb17
        ^bb16:
          %82 = arith.constant 1 : i32
          %83 = llvm.load %77 : !llvm.ptr -> i64
          %84 = arith.trunci %82 : i32 to i8
          %85 = llvm.getelementptr %46[%83] : (!llvm.ptr, i64) -> !llvm.ptr, i8
          llvm.store %84, %85 : i8, !llvm.ptr
          %86 = llvm.load %77 : !llvm.ptr -> i64
          %87 = llvm.load %60 : !llvm.ptr -> i64
          %88 = arith.addi %86, %87 : i64
          llvm.store %88, %77 : i64, !llvm.ptr
          cf.br ^bb15
        ^bb17:
        cf.br ^bb14
      ^bb13:
        cf.br ^bb14
      ^bb14:
      %89 = llvm.load %60 : !llvm.ptr -> i64
      %90 = arith.constant 1 : i32
      %92 = arith.extsi %90 : i32 to i64
      %91 = arith.addi %89, %92 : i64
      llvm.store %91, %60 : i64, !llvm.ptr
      cf.br ^bb9
    ^bb11:
    %93 = arith.constant 0 : i32
    %94 = arith.extsi %93 : i32 to i64
    %95 = llvm.mlir.constant(1 : i64) : i64
    %96 = llvm.alloca %95 x i64 : (i64) -> !llvm.ptr
    llvm.store %94, %96 : i64, !llvm.ptr
    %97 = arith.constant 3 : i32
    %98 = arith.extsi %97 : i32 to i64
    %99 = llvm.mlir.constant(1 : i64) : i64
    %100 = llvm.alloca %99 x i64 : (i64) -> !llvm.ptr
    llvm.store %98, %100 : i64, !llvm.ptr
    cf.br ^bb18
    ^bb18:
    %101 = llvm.load %100 : !llvm.ptr -> i64
    %102 = llvm.mlir.addressof @N : !llvm.ptr
    %103 = llvm.load %102 : !llvm.ptr -> i64
    %104 = arith.cmpi sle, %101, %103 : i64
    cf.cond_br %104, ^bb19, ^bb20
    ^bb19:
      %106 = llvm.load %100 : !llvm.ptr -> i64
      %107 = llvm.getelementptr %46[%106] : (!llvm.ptr, i64) -> !llvm.ptr, i8
      %105 = llvm.load %107 : !llvm.ptr -> i8
      %108 = arith.constant 0 : i32
      %110 = arith.extsi %105 : i8 to i32
      %109 = arith.cmpi eq, %110, %108 : i32
      cf.cond_br %109, ^bb21, ^bb22
      ^bb21:
        %112 = arith.constant 2 : i32
        %113 = llvm.load %100 : !llvm.ptr -> i64
        %114 = llvm.load %100 : !llvm.ptr -> i64
        %115 = arith.constant 1 : i32
        %117 = arith.extsi %115 : i32 to i64
        %116 = arith.subi %114, %117 : i64
        %118 = arith.extsi %112 : i32 to i64
        %111 = func.call @modpow(%118, %113, %116) : (i64, i64, i64) -> i64
        %120 = arith.constant 2 : i32
        %121 = llvm.load %100 : !llvm.ptr -> i64
        %122 = arith.extsi %120 : i32 to i64
        %119 = func.call @modpow(%122, %111, %121) : (i64, i64, i64) -> i64
        %123 = llvm.load %100 : !llvm.ptr -> i64
        %124 = arith.constant 1 : i32
        %126 = arith.extsi %124 : i32 to i64
        %125 = arith.subi %123, %126 : i64
        %127 = arith.muli %119, %125 : i64
        %128 = arith.constant 2 : i32
        %129 = llvm.load %100 : !llvm.ptr -> i64
        %131 = arith.extsi %128 : i32 to i64
        %130 = arith.muli %131, %129 : i64
        %132 = arith.divsi %127, %130 : i64
        %133 = arith.constant 1 : i32
        %135 = arith.extsi %133 : i32 to i64
        %134 = arith.addi %132, %135 : i64
        %136 = llvm.load %100 : !llvm.ptr -> i64
        %137 = llvm.load %100 : !llvm.ptr -> i64
        %138 = arith.muli %136, %137 : i64
        %140 = arith.constant 2 : i32
        %141 = llvm.load %100 : !llvm.ptr -> i64
        %142 = arith.constant 1 : i32
        %144 = arith.extsi %142 : i32 to i64
        %143 = arith.subi %141, %144 : i64
        %145 = arith.extsi %140 : i32 to i64
        %139 = func.call @modpow(%145, %143, %138) : (i64, i64, i64) -> i64
        %146 = arith.extsi %119 : i64 to i128
        %147 = llvm.load %100 : !llvm.ptr -> i64
        %148 = arith.constant 1 : i32
        %150 = arith.extsi %148 : i32 to i64
        %149 = arith.subi %147, %150 : i64
        %151 = arith.extsi %149 : i64 to i128
        %153 = arith.trunci %146 : i128 to i64
        %154 = arith.trunci %151 : i128 to i64
        %152 = arith.muli %153, %154 : i64
        %155 = arith.extsi %138 : i64 to i128
        %157 = arith.trunci %155 : i128 to i64
        %156 = arith.remsi %152, %157 : i64
        %158 = arith.extsi %156 : i64 to i128
        %159 = arith.constant 0 : i32
        %161 = arith.extsi %159 : i32 to i64
        %162 = arith.trunci %158 : i128 to i64
        %160 = arith.subi %161, %162 : i64
        %163 = arith.extsi %160 : i64 to i128
        %164 = arith.extsi %139 : i64 to i128
        %166 = arith.trunci %163 : i128 to i64
        %167 = arith.trunci %164 : i128 to i64
        %165 = arith.muli %166, %167 : i64
        %168 = arith.extsi %138 : i64 to i128
        %170 = arith.trunci %168 : i128 to i64
        %169 = arith.remsi %165, %170 : i64
        %171 = arith.extsi %169 : i64 to i128
        %172 = arith.constant 1 : i32
        %174 = arith.trunci %171 : i128 to i64
        %175 = arith.extsi %172 : i32 to i64
        %173 = arith.addi %174, %175 : i64
        %176 = arith.extsi %138 : i64 to i128
        %178 = arith.trunci %176 : i128 to i64
        %177 = arith.remsi %173, %178 : i64
        %179 = llvm.mlir.constant(1 : i64) : i64
        %180 = llvm.alloca %179 x i64 : (i64) -> !llvm.ptr
        llvm.store %177, %180 : i64, !llvm.ptr
        %181 = llvm.load %180 : !llvm.ptr -> i64
        %182 = arith.constant 0 : i32
        %184 = arith.extsi %182 : i32 to i64
        %183 = arith.cmpi slt, %181, %184 : i64
        cf.cond_br %183, ^bb24, ^bb25
        ^bb24:
          %185 = llvm.load %180 : !llvm.ptr -> i64
          %186 = arith.addi %185, %138 : i64
          llvm.store %186, %180 : i64, !llvm.ptr
          cf.br ^bb26
        ^bb25:
          cf.br ^bb26
        ^bb26:
        %187 = llvm.load %180 : !llvm.ptr -> i64
        %188 = llvm.load %100 : !llvm.ptr -> i64
        %189 = arith.divsi %187, %188 : i64
        %190 = llvm.load %96 : !llvm.ptr -> i64
        %191 = arith.constant 2 : i32
        %193 = arith.extsi %191 : i32 to i64
        %192 = arith.muli %193, %134 : i64
        %194 = arith.addi %189, %192 : i64
        %195 = llvm.load %100 : !llvm.ptr -> i64
        %196 = arith.remsi %194, %195 : i64
        %197 = arith.addi %190, %196 : i64
        llvm.store %197, %96 : i64, !llvm.ptr
        cf.br ^bb23
      ^bb22:
        cf.br ^bb23
      ^bb23:
      %198 = llvm.load %100 : !llvm.ptr -> i64
      %199 = arith.constant 1 : i32
      %201 = arith.extsi %199 : i32 to i64
      %200 = arith.addi %198, %201 : i64
      llvm.store %200, %100 : i64, !llvm.ptr
      cf.br ^bb18
    ^bb20:
    func.call @free(%46) : (!llvm.ptr) -> ()
    %203 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %204 = llvm.load %96 : !llvm.ptr -> i64
    %205 = llvm.call @printf(%203, %204) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, i64) -> i32
    %206 = arith.constant 0 : i32
    func.return %206 : i32
  }
}