Problem 204

Hamming numbers of type 100 not exceeding 10^9.

View problem on Project Euler

Performance comparison

Flow source

# Project Euler 204
# Hamming numbers of type 100 not exceeding 10^9.

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

let mut PRIMES: ptr<i64> = null
let mut PC: i32 = 0
let mut LIMIT: i64 = 1000000000

function search(x: i64, index_min: i32) -> i64 {
    let mut result: i64 = 1
    let mut i: i32 = index_min
    while i < PC {
        let product: i64 = PRIMES[i] * x
        if product > LIMIT { break }
        result = result + search(product, i)
        i = i + 1
    }
    return result
}

function main() -> i32 {
    let max_prime: i32 = 100
    PRIMES = calloc(max_prime as i64, 8)
    if PRIMES == null { return 1 }
    PC = 0
    let mut i: i32 = 2
    while i <= max_prime {
        let mut is_p: bool = true
        let mut j: i32 = 0
        while j < PC {
            if PRIMES[j] * PRIMES[j] > i as i64 { break }
            if (i as i64) % PRIMES[j] == 0 { is_p = false; break }
            j = j + 1
        }
        if is_p {
            PRIMES[PC] = i as i64
            PC = PC + 1
        }
        i = i + 1
    }
    printf("%lld\n", search(1, 0))
    free(PRIMES)
    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 search_i64_i32(int64_t x, int32_t index_min);
int32_t main(void);

/* Module statics */
static int64_t* PRIMES = NULL;
static int32_t PC = 0;
static int64_t LIMIT = 1000000000;



int64_t search_i64_i32(int64_t x, int32_t index_min) {
    int64_t result = 1;
    int32_t i = index_min;
    while (i < PC) {
        int64_t product = (PRIMES[i] * x);
        if (product > LIMIT) {
            break;
        }
        result = (result + search_i64_i32(product, i));
        i = (i + 1);
    }
    return result;
}

int32_t main(void) {
    int32_t max_prime = 100;
    PRIMES = calloc(((int64_t)(max_prime)), 8);
    if (PRIMES == NULL) {
        return 1;
    }
    PC = 0;
    int32_t i = 2;
    while (i <= max_prime) {
        bool is_p = 1;
        int32_t j = 0;
        while (j < PC) {
            if ((PRIMES[j] * PRIMES[j]) > ((int64_t)(i))) {
                break;
            }
            if (FLOW_CHECKED_MOD((((int64_t)(i))), (PRIMES[j])) == 0) {
                is_p = 0;
                break;
            }
            j = (j + 1);
        }
        if (is_p) {
            PRIMES[PC] = ((int64_t)(i));
            PC = (PC + 1);
        }
        i = (i + 1);
    }
    printf("%lld\n", search_i64_i32(1, 0));
    free(PRIMES);
    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) -> ()
  // Module static: PRIMES
  llvm.mlir.global internal @PRIMES() {addr_space = 0 : i32} : !llvm.ptr {
    %0 = llvm.mlir.zero : !llvm.ptr
    llvm.return %0 : !llvm.ptr
  }
  // Module static: PC
  llvm.mlir.global internal @PC(0 : i32) : i32
  // Module static: LIMIT
  llvm.mlir.global internal @LIMIT(1000000000 : i64) : i64
  func.func @search(%arg0: i64, %arg1: i32) -> i64 {
    %1 = arith.constant 1 : i32
    %2 = arith.extsi %1 : i32 to i64
    %3 = llvm.mlir.constant(1 : i64) : i64
    %4 = llvm.alloca %3 x i64 : (i64) -> !llvm.ptr
    llvm.store %2, %4 : i64, !llvm.ptr
    %5 = llvm.mlir.constant(1 : i64) : i64
    %6 = llvm.alloca %5 x i32 : (i64) -> !llvm.ptr
    llvm.store %arg1, %6 : i32, !llvm.ptr
    cf.br ^bb0
    ^bb0:
    %7 = llvm.load %6 : !llvm.ptr -> i32
    %8 = llvm.mlir.addressof @PC : !llvm.ptr
    %9 = llvm.load %8 : !llvm.ptr -> i32
    %10 = arith.cmpi slt, %7, %9 : i32
    cf.cond_br %10, ^bb1, ^bb2
    ^bb1:
      %12 = llvm.mlir.addressof @PRIMES : !llvm.ptr
      %13 = llvm.load %12 : !llvm.ptr -> !llvm.ptr
      %14 = llvm.load %6 : !llvm.ptr -> i32
      %15 = arith.extsi %14 : i32 to i64
      %16 = llvm.getelementptr %13[%15] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %11 = llvm.load %16 : !llvm.ptr -> i64
      %17 = arith.muli %11, %arg0 : i64
      %18 = llvm.mlir.addressof @LIMIT : !llvm.ptr
      %19 = llvm.load %18 : !llvm.ptr -> i64
      %20 = arith.cmpi sgt, %17, %19 : i64
      cf.cond_br %20, ^bb3, ^bb4
      ^bb3:
        cf.br ^bb2
      ^bb4:
        cf.br ^bb5
      ^bb5:
      %21 = llvm.load %4 : !llvm.ptr -> i64
      %23 = llvm.load %6 : !llvm.ptr -> i32
      %22 = func.call @search(%17, %23) : (i64, i32) -> i64
      %24 = arith.addi %21, %22 : i64
      llvm.store %24, %4 : i64, !llvm.ptr
      %25 = llvm.load %6 : !llvm.ptr -> i32
      %26 = arith.constant 1 : i32
      %27 = arith.addi %25, %26 : i32
      llvm.store %27, %6 : i32, !llvm.ptr
      cf.br ^bb0
    ^bb2:
    %28 = llvm.load %4 : !llvm.ptr -> i64
    func.return %28 : i64
  }
  func.func @main() -> i32 {
    %29 = arith.constant 100 : i32
    %31 = arith.extsi %29 : i32 to i64
    %32 = arith.constant 8 : i32
    %33 = arith.extsi %32 : i32 to i64
    %30 = func.call @calloc(%31, %33) : (i64, i64) -> !llvm.ptr
    %34 = llvm.mlir.addressof @PRIMES : !llvm.ptr
    llvm.store %30, %34 : !llvm.ptr, !llvm.ptr
    %35 = llvm.mlir.addressof @PRIMES : !llvm.ptr
    %36 = llvm.load %35 : !llvm.ptr -> !llvm.ptr
    %37 = llvm.mlir.zero : !llvm.ptr
    %38 = llvm.icmp "eq" %36, %37 : !llvm.ptr
    cf.cond_br %38, ^bb6, ^bb7
    ^bb6:
      %39 = arith.constant 1 : i32
      func.return %39 : i32
    ^bb7:
      cf.br ^bb8
    ^bb8:
    %40 = arith.constant 0 : i32
    %41 = llvm.mlir.addressof @PC : !llvm.ptr
    llvm.store %40, %41 : i32, !llvm.ptr
    %42 = arith.constant 2 : i32
    %43 = llvm.mlir.constant(1 : i64) : i64
    %44 = llvm.alloca %43 x i32 : (i64) -> !llvm.ptr
    llvm.store %42, %44 : i32, !llvm.ptr
    cf.br ^bb9
    ^bb9:
    %45 = llvm.load %44 : !llvm.ptr -> i32
    %46 = arith.cmpi sle, %45, %29 : i32
    cf.cond_br %46, ^bb10, ^bb11
    ^bb10:
      %47 = arith.constant 1 : i1
      %48 = llvm.mlir.constant(1 : i64) : i64
      %49 = llvm.alloca %48 x i1 : (i64) -> !llvm.ptr
      llvm.store %47, %49 : i1, !llvm.ptr
      %50 = arith.constant 0 : i32
      %51 = llvm.mlir.constant(1 : i64) : i64
      %52 = llvm.alloca %51 x i32 : (i64) -> !llvm.ptr
      llvm.store %50, %52 : i32, !llvm.ptr
      cf.br ^bb12
      ^bb12:
      %53 = llvm.load %52 : !llvm.ptr -> i32
      %54 = llvm.mlir.addressof @PC : !llvm.ptr
      %55 = llvm.load %54 : !llvm.ptr -> i32
      %56 = arith.cmpi slt, %53, %55 : i32
      cf.cond_br %56, ^bb13, ^bb14
      ^bb13:
        %58 = llvm.mlir.addressof @PRIMES : !llvm.ptr
        %59 = llvm.load %58 : !llvm.ptr -> !llvm.ptr
        %60 = llvm.load %52 : !llvm.ptr -> i32
        %61 = arith.extsi %60 : i32 to i64
        %62 = llvm.getelementptr %59[%61] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        %57 = llvm.load %62 : !llvm.ptr -> i64
        %64 = llvm.mlir.addressof @PRIMES : !llvm.ptr
        %65 = llvm.load %64 : !llvm.ptr -> !llvm.ptr
        %66 = llvm.load %52 : !llvm.ptr -> i32
        %67 = arith.extsi %66 : i32 to i64
        %68 = llvm.getelementptr %65[%67] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        %63 = llvm.load %68 : !llvm.ptr -> i64
        %69 = arith.muli %57, %63 : i64
        %70 = llvm.load %44 : !llvm.ptr -> i32
        %71 = arith.extsi %70 : i32 to i64
        %72 = arith.cmpi sgt, %69, %71 : i64
        cf.cond_br %72, ^bb15, ^bb16
        ^bb15:
          cf.br ^bb14
        ^bb16:
          cf.br ^bb17
        ^bb17:
        %73 = llvm.load %44 : !llvm.ptr -> i32
        %74 = arith.extsi %73 : i32 to i64
        %76 = llvm.mlir.addressof @PRIMES : !llvm.ptr
        %77 = llvm.load %76 : !llvm.ptr -> !llvm.ptr
        %78 = llvm.load %52 : !llvm.ptr -> i32
        %79 = arith.extsi %78 : i32 to i64
        %80 = llvm.getelementptr %77[%79] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        %75 = llvm.load %80 : !llvm.ptr -> i64
        %81 = arith.remsi %74, %75 : i64
        %82 = arith.constant 0 : i32
        %84 = arith.extsi %82 : i32 to i64
        %83 = arith.cmpi eq, %81, %84 : i64
        cf.cond_br %83, ^bb18, ^bb19
        ^bb18:
          %85 = arith.constant 0 : i1
          llvm.store %85, %49 : i1, !llvm.ptr
          cf.br ^bb14
        ^bb19:
          cf.br ^bb20
        ^bb20:
        %86 = llvm.load %52 : !llvm.ptr -> i32
        %87 = arith.constant 1 : i32
        %88 = arith.addi %86, %87 : i32
        llvm.store %88, %52 : i32, !llvm.ptr
        cf.br ^bb12
      ^bb14:
      %89 = llvm.load %49 : !llvm.ptr -> i1
      cf.cond_br %89, ^bb21, ^bb22
      ^bb21:
        %90 = llvm.load %44 : !llvm.ptr -> i32
        %91 = arith.extsi %90 : i32 to i64
        %92 = llvm.mlir.addressof @PRIMES : !llvm.ptr
        %93 = llvm.load %92 : !llvm.ptr -> !llvm.ptr
        %94 = llvm.mlir.addressof @PC : !llvm.ptr
        %95 = llvm.load %94 : !llvm.ptr -> i32
        %96 = arith.extsi %95 : i32 to i64
        %97 = llvm.getelementptr %93[%96] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        llvm.store %91, %97 : i64, !llvm.ptr
        %98 = llvm.mlir.addressof @PC : !llvm.ptr
        %99 = llvm.load %98 : !llvm.ptr -> i32
        %100 = arith.constant 1 : i32
        %101 = arith.addi %99, %100 : i32
        %102 = llvm.mlir.addressof @PC : !llvm.ptr
        llvm.store %101, %102 : i32, !llvm.ptr
        cf.br ^bb23
      ^bb22:
        cf.br ^bb23
      ^bb23:
      %103 = llvm.load %44 : !llvm.ptr -> i32
      %104 = arith.constant 1 : i32
      %105 = arith.addi %103, %104 : i32
      llvm.store %105, %44 : i32, !llvm.ptr
      cf.br ^bb9
    ^bb11:
    %106 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %108 = arith.constant 1 : i32
    %109 = arith.constant 0 : i32
    %110 = arith.extsi %108 : i32 to i64
    %107 = func.call @search(%110, %109) : (i64, i32) -> i64
    %111 = llvm.call @printf(%106, %107) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, i64) -> i32
    %113 = llvm.mlir.addressof @PRIMES : !llvm.ptr
    %114 = llvm.load %113 : !llvm.ptr -> !llvm.ptr
    func.call @free(%114) : (!llvm.ptr) -> ()
    %115 = arith.constant 0 : i32
    func.return %115 : i32
  }
}