Problem 003

Largest prime factor of 600851475143. Trial division with an early even strip and odd steps — Flow's i64 and while syntax keep the algorithm obvious without sacrificing speed.

View problem on Project Euler

Performance comparison

Flow source

# Project Euler 003
# Largest prime factor of 600851475143.
#
# Trial division with an early even strip and odd steps — Flow's i64
# and while syntax keep the algorithm obvious without sacrificing speed.

function largest_prime_factor(n0: i64) -> i64 {
    let mut n: i64 = n0
    let mut last: i64 = 1

    while n % 2 == 0 {
        last = 2
        n = n / 2
    }

    let mut p: i64 = 3
    while p * p <= n {
        while n % p == 0 {
            last = p
            n = n / p
        }
        p = p + 2
    }

    if n > 1 {
        last = n
    }
    return last
}

function main() -> i32 {
    printf("%lld\n", largest_prime_factor(600851475143))
    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 largest_prime_factor_i64(int64_t n0);
int32_t main(void);

int64_t largest_prime_factor_i64(int64_t n0) {
    int64_t n = n0;
    int64_t last = 1;
    while (FLOW_CHECKED_MOD((n), (2)) == 0) {
        last = 2;
        n = FLOW_CHECKED_DIV((n), (2));
    }
    int64_t p = 3;
    while ((p * p) <= n) {
        while (FLOW_CHECKED_MOD((n), (p)) == 0) {
            last = p;
            n = FLOW_CHECKED_DIV((n), (p));
        }
        p = (p + 2);
    }
    if (n > 1) {
        last = n;
    }
    return last;
}

int32_t main(void) {
    printf("%lld\n", largest_prime_factor_i64(600851475143));
    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 @largest_prime_factor(%arg0: i64) -> i64 {
    %0 = llvm.mlir.constant(1 : i64) : i64
    %1 = llvm.alloca %0 x i64 : (i64) -> !llvm.ptr
    llvm.store %arg0, %1 : i64, !llvm.ptr
    %2 = arith.constant 1 : i32
    %3 = arith.extsi %2 : i32 to i64
    %4 = llvm.mlir.constant(1 : i64) : i64
    %5 = llvm.alloca %4 x i64 : (i64) -> !llvm.ptr
    llvm.store %3, %5 : i64, !llvm.ptr
    cf.br ^bb0
    ^bb0:
    %6 = llvm.load %1 : !llvm.ptr -> i64
    %7 = arith.constant 2 : i32
    %9 = arith.extsi %7 : i32 to i64
    %8 = arith.remsi %6, %9 : i64
    %10 = arith.constant 0 : i32
    %12 = arith.extsi %10 : i32 to i64
    %11 = arith.cmpi eq, %8, %12 : i64
    cf.cond_br %11, ^bb1, ^bb2
    ^bb1:
      %13 = arith.constant 2 : i32
      %14 = arith.extsi %13 : i32 to i64
      llvm.store %14, %5 : i64, !llvm.ptr
      %15 = llvm.load %1 : !llvm.ptr -> i64
      %16 = arith.constant 2 : i32
      %18 = arith.extsi %16 : i32 to i64
      %17 = arith.divsi %15, %18 : i64
      llvm.store %17, %1 : i64, !llvm.ptr
      cf.br ^bb0
    ^bb2:
    %19 = arith.constant 3 : i32
    %20 = arith.extsi %19 : i32 to i64
    %21 = llvm.mlir.constant(1 : i64) : i64
    %22 = llvm.alloca %21 x i64 : (i64) -> !llvm.ptr
    llvm.store %20, %22 : i64, !llvm.ptr
    cf.br ^bb3
    ^bb3:
    %23 = llvm.load %22 : !llvm.ptr -> i64
    %24 = llvm.load %22 : !llvm.ptr -> i64
    %25 = arith.muli %23, %24 : i64
    %26 = llvm.load %1 : !llvm.ptr -> i64
    %27 = arith.cmpi sle, %25, %26 : i64
    cf.cond_br %27, ^bb4, ^bb5
    ^bb4:
      cf.br ^bb6
      ^bb6:
      %28 = llvm.load %1 : !llvm.ptr -> i64
      %29 = llvm.load %22 : !llvm.ptr -> i64
      %30 = arith.remsi %28, %29 : i64
      %31 = arith.constant 0 : i32
      %33 = arith.extsi %31 : i32 to i64
      %32 = arith.cmpi eq, %30, %33 : i64
      cf.cond_br %32, ^bb7, ^bb8
      ^bb7:
        %34 = llvm.load %22 : !llvm.ptr -> i64
        llvm.store %34, %5 : i64, !llvm.ptr
        %35 = llvm.load %1 : !llvm.ptr -> i64
        %36 = llvm.load %22 : !llvm.ptr -> i64
        %37 = arith.divsi %35, %36 : i64
        llvm.store %37, %1 : i64, !llvm.ptr
        cf.br ^bb6
      ^bb8:
      %38 = llvm.load %22 : !llvm.ptr -> i64
      %39 = arith.constant 2 : i32
      %41 = arith.extsi %39 : i32 to i64
      %40 = arith.addi %38, %41 : i64
      llvm.store %40, %22 : i64, !llvm.ptr
      cf.br ^bb3
    ^bb5:
    %42 = llvm.load %1 : !llvm.ptr -> i64
    %43 = arith.constant 1 : i32
    %45 = arith.extsi %43 : i32 to i64
    %44 = arith.cmpi sgt, %42, %45 : i64
    cf.cond_br %44, ^bb9, ^bb10
    ^bb9:
      %46 = llvm.load %1 : !llvm.ptr -> i64
      llvm.store %46, %5 : i64, !llvm.ptr
      cf.br ^bb11
    ^bb10:
      cf.br ^bb11
    ^bb11:
    %47 = llvm.load %5 : !llvm.ptr -> i64
    func.return %47 : i64
  }
  func.func @main() -> i32 {
    %48 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %50 = arith.constant 596556507847 : i32
    %51 = arith.extsi %50 : i32 to i64
    %49 = func.call @largest_prime_factor(%51) : (i64) -> i64
    %52 = llvm.call @printf(%48, %49) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, i64) -> i32
    %53 = arith.constant 0 : i32
    func.return %53 : i32
  }
}