Problem 509

Divisor Nim — g(n)=v2(n); winning triples <= 123456787654321 mod 1234567890.

View problem on Project Euler

Performance comparison

Flow source

# Project Euler 509
# Divisor Nim — g(n)=v2(n); winning triples <= 123456787654321 mod 1234567890.

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

function main() -> i32 {
    let MOD: i64 = 1234567890
    let N: i64 = 123456787654321
    let f: ptr<i64> = calloc(64, 8)
    if f == null { return 1 }
    let mut k: i64 = 0
    while k < 63 {
        f[k] = (N >> k) - (N >> (k + 1))
        k = k + 1
    }
    let mut losing: i64 = 0
    let mut i: i64 = 0
    while i < 64 {
        let ci: i64 = f[i] % MOD
        if ci != 0 {
            let mut j: i64 = 0
            while j < 64 {
                let cj: i64 = f[j] % MOD
                if cj != 0 {
                    let x: i64 = i ^ j
                    let cx: i64 = f[x] % MOD
                    losing = (losing + ((ci * cj) % MOD) * cx) % MOD
                }
                j = j + 1
            }
        }
        i = i + 1
    }
    let nmod: i64 = N % MOD
    let total: i64 = (((nmod * nmod) % MOD) * nmod) % MOD
    let mut winning: i64 = (total - losing) % MOD
    if winning < 0 {
        winning = winning + MOD
    }
    printf("%lld\n", winning)
    free(f)
    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; }

int32_t main(void);



int32_t main(void) {
    int64_t MOD = 1234567890;
    int64_t N = 123456787654321;
    int64_t* f = (int64_t*)(calloc(64, 8));
    if (f == NULL) {
        return 1;
    }
    int64_t k = 0;
    while (k < 63) {
        f[k] = (FLOW_CHECKED_SHR((N), (k)) - FLOW_CHECKED_SHR((N), ((k + 1))));
        k = (k + 1);
    }
    int64_t losing = 0;
    int64_t i = 0;
    while (i < 64) {
        int64_t ci = FLOW_CHECKED_MOD((f[i]), (MOD));
        if (ci != 0) {
            int64_t j = 0;
            while (j < 64) {
                int64_t cj = FLOW_CHECKED_MOD((f[j]), (MOD));
                if (cj != 0) {
                    int64_t x = (i ^ j);
                    int64_t cx = FLOW_CHECKED_MOD((f[x]), (MOD));
                    losing = FLOW_CHECKED_MOD(((losing + (FLOW_CHECKED_MOD(((ci * cj)), (MOD)) * cx))), (MOD));
                }
                j = (j + 1);
            }
        }
        i = (i + 1);
    }
    int64_t nmod = FLOW_CHECKED_MOD((N), (MOD));
    int64_t total = FLOW_CHECKED_MOD(((FLOW_CHECKED_MOD(((nmod * nmod)), (MOD)) * nmod)), (MOD));
    int64_t winning = FLOW_CHECKED_MOD(((total - losing)), (MOD));
    if (winning < 0) {
        winning = (winning + MOD);
    }
    printf("%lld\n", winning);
    free(f);
    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 @main() -> i32 {
    %0 = arith.constant 1234567890 : i32
    %1 = arith.extsi %0 : i32 to i64
    %2 = arith.constant 123452492687025 : i32
    %3 = arith.extsi %2 : i32 to i64
    %5 = arith.constant 64 : i32
    %6 = arith.constant 8 : i32
    %7 = arith.extsi %5 : i32 to i64
    %8 = arith.extsi %6 : i32 to i64
    %4 = func.call @calloc(%7, %8) : (i64, i64) -> !llvm.ptr
    %9 = llvm.mlir.zero : !llvm.ptr
    %10 = llvm.icmp "eq" %4, %9 : !llvm.ptr
    cf.cond_br %10, ^bb0, ^bb1
    ^bb0:
      %11 = arith.constant 1 : i32
      func.return %11 : i32
    ^bb1:
      cf.br ^bb2
    ^bb2:
    %12 = arith.constant 0 : i32
    %13 = arith.extsi %12 : i32 to i64
    %14 = llvm.mlir.constant(1 : i64) : i64
    %15 = llvm.alloca %14 x i64 : (i64) -> !llvm.ptr
    llvm.store %13, %15 : i64, !llvm.ptr
    cf.br ^bb3
    ^bb3:
    %16 = llvm.load %15 : !llvm.ptr -> i64
    %17 = arith.constant 63 : i32
    %19 = arith.extsi %17 : i32 to i64
    %18 = arith.cmpi slt, %16, %19 : i64
    cf.cond_br %18, ^bb4, ^bb5
    ^bb4:
      %20 = llvm.load %15 : !llvm.ptr -> i64
      %21 = arith.shrsi %3, %20 : i64
      %22 = llvm.load %15 : !llvm.ptr -> i64
      %23 = arith.constant 1 : i32
      %25 = arith.extsi %23 : i32 to i64
      %24 = arith.addi %22, %25 : i64
      %26 = arith.shrsi %3, %24 : i64
      %27 = arith.subi %21, %26 : i64
      %28 = llvm.load %15 : !llvm.ptr -> i64
      %29 = llvm.getelementptr %4[%28] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      llvm.store %27, %29 : i64, !llvm.ptr
      %30 = llvm.load %15 : !llvm.ptr -> i64
      %31 = arith.constant 1 : i32
      %33 = arith.extsi %31 : i32 to i64
      %32 = arith.addi %30, %33 : i64
      llvm.store %32, %15 : i64, !llvm.ptr
      cf.br ^bb3
    ^bb5:
    %34 = arith.constant 0 : i32
    %35 = arith.extsi %34 : i32 to i64
    %36 = llvm.mlir.constant(1 : i64) : i64
    %37 = llvm.alloca %36 x i64 : (i64) -> !llvm.ptr
    llvm.store %35, %37 : i64, !llvm.ptr
    %38 = arith.constant 0 : i32
    %39 = arith.extsi %38 : i32 to i64
    %40 = llvm.mlir.constant(1 : i64) : i64
    %41 = llvm.alloca %40 x i64 : (i64) -> !llvm.ptr
    llvm.store %39, %41 : i64, !llvm.ptr
    cf.br ^bb6
    ^bb6:
    %42 = llvm.load %41 : !llvm.ptr -> i64
    %43 = arith.constant 64 : i32
    %45 = arith.extsi %43 : i32 to i64
    %44 = arith.cmpi slt, %42, %45 : i64
    cf.cond_br %44, ^bb7, ^bb8
    ^bb7:
      %47 = llvm.load %41 : !llvm.ptr -> i64
      %48 = llvm.getelementptr %4[%47] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %46 = llvm.load %48 : !llvm.ptr -> i64
      %49 = arith.remsi %46, %1 : i64
      %50 = arith.constant 0 : i32
      %52 = arith.extsi %50 : i32 to i64
      %51 = arith.cmpi ne, %49, %52 : i64
      cf.cond_br %51, ^bb9, ^bb10
      ^bb9:
        %53 = arith.constant 0 : i32
        %54 = arith.extsi %53 : i32 to i64
        %55 = llvm.mlir.constant(1 : i64) : i64
        %56 = llvm.alloca %55 x i64 : (i64) -> !llvm.ptr
        llvm.store %54, %56 : i64, !llvm.ptr
        cf.br ^bb12
        ^bb12:
        %57 = llvm.load %56 : !llvm.ptr -> i64
        %58 = arith.constant 64 : i32
        %60 = arith.extsi %58 : i32 to i64
        %59 = arith.cmpi slt, %57, %60 : i64
        cf.cond_br %59, ^bb13, ^bb14
        ^bb13:
          %62 = llvm.load %56 : !llvm.ptr -> i64
          %63 = llvm.getelementptr %4[%62] : (!llvm.ptr, i64) -> !llvm.ptr, i64
          %61 = llvm.load %63 : !llvm.ptr -> i64
          %64 = arith.remsi %61, %1 : i64
          %65 = arith.constant 0 : i32
          %67 = arith.extsi %65 : i32 to i64
          %66 = arith.cmpi ne, %64, %67 : i64
          cf.cond_br %66, ^bb15, ^bb16
          ^bb15:
            %68 = llvm.load %41 : !llvm.ptr -> i64
            %69 = llvm.load %56 : !llvm.ptr -> i64
            %70 = arith.xori %68, %69 : i64
            %72 = llvm.getelementptr %4[%70] : (!llvm.ptr, i64) -> !llvm.ptr, i64
            %71 = llvm.load %72 : !llvm.ptr -> i64
            %73 = arith.remsi %71, %1 : i64
            %74 = llvm.load %37 : !llvm.ptr -> i64
            %75 = arith.muli %49, %64 : i64
            %76 = arith.remsi %75, %1 : i64
            %77 = arith.muli %76, %73 : i64
            %78 = arith.addi %74, %77 : i64
            %79 = arith.remsi %78, %1 : i64
            llvm.store %79, %37 : i64, !llvm.ptr
            cf.br ^bb17
          ^bb16:
            cf.br ^bb17
          ^bb17:
          %80 = llvm.load %56 : !llvm.ptr -> i64
          %81 = arith.constant 1 : i32
          %83 = arith.extsi %81 : i32 to i64
          %82 = arith.addi %80, %83 : i64
          llvm.store %82, %56 : i64, !llvm.ptr
          cf.br ^bb12
        ^bb14:
        cf.br ^bb11
      ^bb10:
        cf.br ^bb11
      ^bb11:
      %84 = llvm.load %41 : !llvm.ptr -> i64
      %85 = arith.constant 1 : i32
      %87 = arith.extsi %85 : i32 to i64
      %86 = arith.addi %84, %87 : i64
      llvm.store %86, %41 : i64, !llvm.ptr
      cf.br ^bb6
    ^bb8:
    %88 = arith.remsi %3, %1 : i64
    %89 = arith.muli %88, %88 : i64
    %90 = arith.remsi %89, %1 : i64
    %91 = arith.muli %90, %88 : i64
    %92 = arith.remsi %91, %1 : i64
    %93 = llvm.load %37 : !llvm.ptr -> i64
    %94 = arith.subi %92, %93 : i64
    %95 = arith.remsi %94, %1 : i64
    %96 = llvm.mlir.constant(1 : i64) : i64
    %97 = llvm.alloca %96 x i64 : (i64) -> !llvm.ptr
    llvm.store %95, %97 : i64, !llvm.ptr
    %98 = llvm.load %97 : !llvm.ptr -> i64
    %99 = arith.constant 0 : i32
    %101 = arith.extsi %99 : i32 to i64
    %100 = arith.cmpi slt, %98, %101 : i64
    cf.cond_br %100, ^bb18, ^bb19
    ^bb18:
      %102 = llvm.load %97 : !llvm.ptr -> i64
      %103 = arith.addi %102, %1 : i64
      llvm.store %103, %97 : i64, !llvm.ptr
      cf.br ^bb20
    ^bb19:
      cf.br ^bb20
    ^bb20:
    %104 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %105 = llvm.load %97 : !llvm.ptr -> i64
    %106 = llvm.call @printf(%104, %105) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, i64) -> i32
    func.call @free(%4) : (!llvm.ptr) -> ()
    %108 = arith.constant 0 : i32
    func.return %108 : i32
  }
}