Problem 230

Fibonacci word digits at special indices.

View problem on Project Euler

Performance comparison

Flow source

# Project Euler 230
# Fibonacci word digits at special indices.

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

function digit_at(A: ptr<i8>, B: ptr<i8>, block: i64, index: i64, fib: ptr<i64>, fibn: i32) -> i32 {
    let mut idx: i64 = index
    let mut n: i32 = 0
    while n < fibn && fib[n] <= idx {
        n = n + 1
    }
    # n is smallest with fib[n] > idx? actually find generation
    while true {
        if n == 0 {
            return (A[idx] - 48) as i32
        }
        if n == 1 {
            return (B[idx] - 48) as i32
        }
        if idx < fib[n - 2] {
            n = n - 2
        } else {
            idx = idx - fib[n - 2]
            n = n - 1
        }
    }
    return 0
}

function main() -> i32 {
    let A_str: ptr<i8> = "1415926535897932384626433832795028841971693993751058209749445923078164062862089986280348253421170679"
    let B_str: ptr<i8> = "8214808651328230664709384460955058223172535940812848111745028410270193852110555964462294895493038196"
    let block: i64 = 100
    let fib: ptr<i64> = calloc(100, 8)
    if fib == null { return 1 }
    fib[0] = block
    fib[1] = block
    let mut n: i32 = 2
    while n < 90 {
        fib[n] = fib[n - 1] + fib[n - 2]
        if fib[n] < 0 { break }  # overflow guard
        n = n + 1
    }
    let fibn: i32 = n
    let mut result: i64 = 0
    let mut nn: i32 = 17
    while nn >= 0 {
        let mut index: i64 = 127 + 19 * (nn as i64)
        let mut t: i32 = 1
        while t <= nn {
            index = index * 7
            t = t + 1
        }
        index = index - 1
        let d: i32 = digit_at(A_str, B_str, block, index, fib, fibn)
        result = result * 10 + (d as i64)
        nn = nn - 1
    }
    printf("%lld\n", result)
    free(fib)
    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 digit_at_ptr_i8_ptr_i8_i64_i64_ptr_i64_i32(int8_t* A, int8_t* B, int64_t block, int64_t index, int64_t* fib, int32_t fibn);
int32_t main(void);



int32_t digit_at_ptr_i8_ptr_i8_i64_i64_ptr_i64_i32(int8_t* A, int8_t* B, int64_t block, int64_t index, int64_t* fib, int32_t fibn) {
    int64_t idx = index;
    int32_t n = 0;
    while ((n < fibn && fib[n] <= idx)) {
        n = (n + 1);
    }
    while (1) {
        if (n == 0) {
            return ((int32_t)((A[idx] - 48)));
        }
        if (n == 1) {
            return ((int32_t)((B[idx] - 48)));
        }
        if (idx < fib[(n - 2)]) {
            n = (n - 2);
        } else {
            idx = (idx - fib[(n - 2)]);
            n = (n - 1);
        }
    }
    return 0;
}

int32_t main(void) {
    int8_t* A_str = (int8_t*)("1415926535897932384626433832795028841971693993751058209749445923078164062862089986280348253421170679");
    int8_t* B_str = (int8_t*)("8214808651328230664709384460955058223172535940812848111745028410270193852110555964462294895493038196");
    int64_t block = 100;
    int64_t* fib = (int64_t*)(calloc(100, 8));
    if (fib == NULL) {
        return 1;
    }
    fib[0] = block;
    fib[1] = block;
    int32_t n = 2;
    while (n < 90) {
        fib[n] = (fib[(n - 1)] + fib[(n - 2)]);
        if (fib[n] < 0) {
            break;
        }
        n = (n + 1);
    }
    int32_t fibn = n;
    int64_t result = 0;
    int32_t nn = 17;
    while (nn >= 0) {
        int64_t index = (127 + (19 * ((int64_t)(nn))));
        int32_t t = 1;
        while (t <= nn) {
            index = (index * 7);
            t = (t + 1);
        }
        index = (index - 1);
        int32_t d = digit_at_ptr_i8_ptr_i8_i64_i64_ptr_i64_i32(A_str, B_str, block, index, fib, fibn);
        result = ((result * 10) + ((int64_t)(d)));
        nn = (nn - 1);
    }
    printf("%lld\n", result);
    free(fib);
    return 0;
}

Generated MLIR

module {
  llvm.func @printf(!llvm.ptr, ...) -> i32
  llvm.mlir.global internal constant @str_0("1415926535897932384626433832795028841971693993751058209749445923078164062862089986280348253421170679\00") {addr_space = 0 : i32} : !llvm.array<101 x i8>
  llvm.mlir.global internal constant @str_1("8214808651328230664709384460955058223172535940812848111745028410270193852110555964462294895493038196\00") {addr_space = 0 : i32} : !llvm.array<101 x i8>
  llvm.mlir.global internal constant @str_2("%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 @digit_at(%arg0: !llvm.ptr, %arg1: !llvm.ptr, %arg2: i64, %arg3: i64, %arg4: !llvm.ptr, %arg5: i32) -> i32 {
    %0 = llvm.mlir.constant(1 : i64) : i64
    %1 = llvm.alloca %0 x i64 : (i64) -> !llvm.ptr
    llvm.store %arg3, %1 : i64, !llvm.ptr
    %2 = arith.constant 0 : i32
    %3 = llvm.mlir.constant(1 : i64) : i64
    %4 = llvm.alloca %3 x i32 : (i64) -> !llvm.ptr
    llvm.store %2, %4 : i32, !llvm.ptr
    cf.br ^bb0
    ^bb0:
    %5 = llvm.load %4 : !llvm.ptr -> i32
    %6 = arith.cmpi slt, %5, %arg5 : i32
    %7 = scf.if %6 -> (i1) {
      %9 = llvm.load %4 : !llvm.ptr -> i32
      %10 = arith.extsi %9 : i32 to i64
      %11 = llvm.getelementptr %arg4[%10] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %8 = llvm.load %11 : !llvm.ptr -> i64
      %12 = llvm.load %1 : !llvm.ptr -> i64
      %13 = arith.cmpi sle, %8, %12 : i64
      scf.yield %13 : i1
    } else {
      %14 = arith.constant false
      scf.yield %14 : i1
    }
    cf.cond_br %7, ^bb1, ^bb2
    ^bb1:
      %15 = llvm.load %4 : !llvm.ptr -> i32
      %16 = arith.constant 1 : i32
      %17 = arith.addi %15, %16 : i32
      llvm.store %17, %4 : i32, !llvm.ptr
      cf.br ^bb0
    ^bb2:
    cf.br ^bb3
    ^bb3:
    %18 = arith.constant 1 : i1
    cf.cond_br %18, ^bb4, ^bb5
    ^bb4:
      %19 = llvm.load %4 : !llvm.ptr -> i32
      %20 = arith.constant 0 : i32
      %21 = arith.cmpi eq, %19, %20 : i32
      cf.cond_br %21, ^bb6, ^bb7
      ^bb6:
        %23 = llvm.load %1 : !llvm.ptr -> i64
        %24 = llvm.getelementptr %arg0[%23] : (!llvm.ptr, i64) -> !llvm.ptr, i8
        %22 = llvm.load %24 : !llvm.ptr -> i8
        %25 = arith.constant 48 : i32
        %27 = arith.extsi %22 : i8 to i32
        %26 = arith.subi %27, %25 : i32
        func.return %26 : i32
      ^bb7:
        cf.br ^bb8
      ^bb8:
      %28 = llvm.load %4 : !llvm.ptr -> i32
      %29 = arith.constant 1 : i32
      %30 = arith.cmpi eq, %28, %29 : i32
      cf.cond_br %30, ^bb9, ^bb10
      ^bb9:
        %32 = llvm.load %1 : !llvm.ptr -> i64
        %33 = llvm.getelementptr %arg1[%32] : (!llvm.ptr, i64) -> !llvm.ptr, i8
        %31 = llvm.load %33 : !llvm.ptr -> i8
        %34 = arith.constant 48 : i32
        %36 = arith.extsi %31 : i8 to i32
        %35 = arith.subi %36, %34 : i32
        func.return %35 : i32
      ^bb10:
        cf.br ^bb11
      ^bb11:
      %37 = llvm.load %1 : !llvm.ptr -> i64
      %39 = llvm.load %4 : !llvm.ptr -> i32
      %40 = arith.constant 2 : i32
      %41 = arith.subi %39, %40 : i32
      %42 = arith.extsi %41 : i32 to i64
      %43 = llvm.getelementptr %arg4[%42] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %38 = llvm.load %43 : !llvm.ptr -> i64
      %44 = arith.cmpi slt, %37, %38 : i64
      cf.cond_br %44, ^bb12, ^bb13
      ^bb12:
        %45 = llvm.load %4 : !llvm.ptr -> i32
        %46 = arith.constant 2 : i32
        %47 = arith.subi %45, %46 : i32
        llvm.store %47, %4 : i32, !llvm.ptr
        cf.br ^bb14
      ^bb13:
        %48 = llvm.load %1 : !llvm.ptr -> i64
        %50 = llvm.load %4 : !llvm.ptr -> i32
        %51 = arith.constant 2 : i32
        %52 = arith.subi %50, %51 : i32
        %53 = arith.extsi %52 : i32 to i64
        %54 = llvm.getelementptr %arg4[%53] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        %49 = llvm.load %54 : !llvm.ptr -> i64
        %55 = arith.subi %48, %49 : i64
        llvm.store %55, %1 : i64, !llvm.ptr
        %56 = llvm.load %4 : !llvm.ptr -> i32
        %57 = arith.constant 1 : i32
        %58 = arith.subi %56, %57 : i32
        llvm.store %58, %4 : i32, !llvm.ptr
        cf.br ^bb14
      ^bb14:
      cf.br ^bb3
    ^bb5:
    %59 = arith.constant 0 : i32
    func.return %59 : i32
  }
  func.func @main() -> i32 {
    %60 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %61 = llvm.mlir.addressof @str_1 : !llvm.ptr
    %62 = arith.constant 100 : i32
    %63 = arith.extsi %62 : i32 to i64
    %65 = arith.constant 100 : i32
    %66 = arith.constant 8 : i32
    %67 = arith.extsi %65 : i32 to i64
    %68 = arith.extsi %66 : i32 to i64
    %64 = func.call @calloc(%67, %68) : (i64, i64) -> !llvm.ptr
    %69 = llvm.mlir.zero : !llvm.ptr
    %70 = llvm.icmp "eq" %64, %69 : !llvm.ptr
    cf.cond_br %70, ^bb15, ^bb16
    ^bb15:
      %71 = arith.constant 1 : i32
      func.return %71 : i32
    ^bb16:
      cf.br ^bb17
    ^bb17:
    %72 = arith.constant 0 : i32
    %73 = arith.extsi %72 : i32 to i64
    %74 = llvm.getelementptr %64[%73] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %63, %74 : i64, !llvm.ptr
    %75 = arith.constant 1 : i32
    %76 = arith.extsi %75 : i32 to i64
    %77 = llvm.getelementptr %64[%76] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %63, %77 : i64, !llvm.ptr
    %78 = arith.constant 2 : i32
    %79 = llvm.mlir.constant(1 : i64) : i64
    %80 = llvm.alloca %79 x i32 : (i64) -> !llvm.ptr
    llvm.store %78, %80 : i32, !llvm.ptr
    cf.br ^bb18
    ^bb18:
    %81 = llvm.load %80 : !llvm.ptr -> i32
    %82 = arith.constant 90 : i32
    %83 = arith.cmpi slt, %81, %82 : i32
    cf.cond_br %83, ^bb19, ^bb20
    ^bb19:
      %85 = llvm.load %80 : !llvm.ptr -> i32
      %86 = arith.constant 1 : i32
      %87 = arith.subi %85, %86 : i32
      %88 = arith.extsi %87 : i32 to i64
      %89 = llvm.getelementptr %64[%88] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %84 = llvm.load %89 : !llvm.ptr -> i64
      %91 = llvm.load %80 : !llvm.ptr -> i32
      %92 = arith.constant 2 : i32
      %93 = arith.subi %91, %92 : i32
      %94 = arith.extsi %93 : i32 to i64
      %95 = llvm.getelementptr %64[%94] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %90 = llvm.load %95 : !llvm.ptr -> i64
      %96 = arith.addi %84, %90 : i64
      %97 = llvm.load %80 : !llvm.ptr -> i32
      %98 = arith.extsi %97 : i32 to i64
      %99 = llvm.getelementptr %64[%98] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      llvm.store %96, %99 : i64, !llvm.ptr
      %101 = llvm.load %80 : !llvm.ptr -> i32
      %102 = arith.extsi %101 : i32 to i64
      %103 = llvm.getelementptr %64[%102] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %100 = llvm.load %103 : !llvm.ptr -> i64
      %104 = arith.constant 0 : i32
      %106 = arith.extsi %104 : i32 to i64
      %105 = arith.cmpi slt, %100, %106 : i64
      cf.cond_br %105, ^bb21, ^bb22
      ^bb21:
        cf.br ^bb20
      ^bb22:
        cf.br ^bb23
      ^bb23:
      %107 = llvm.load %80 : !llvm.ptr -> i32
      %108 = arith.constant 1 : i32
      %109 = arith.addi %107, %108 : i32
      llvm.store %109, %80 : i32, !llvm.ptr
      cf.br ^bb18
    ^bb20:
    %110 = llvm.load %80 : !llvm.ptr -> i32
    %111 = arith.constant 0 : i32
    %112 = arith.extsi %111 : i32 to i64
    %113 = llvm.mlir.constant(1 : i64) : i64
    %114 = llvm.alloca %113 x i64 : (i64) -> !llvm.ptr
    llvm.store %112, %114 : i64, !llvm.ptr
    %115 = arith.constant 17 : i32
    %116 = llvm.mlir.constant(1 : i64) : i64
    %117 = llvm.alloca %116 x i32 : (i64) -> !llvm.ptr
    llvm.store %115, %117 : i32, !llvm.ptr
    cf.br ^bb24
    ^bb24:
    %118 = llvm.load %117 : !llvm.ptr -> i32
    %119 = arith.constant 0 : i32
    %120 = arith.cmpi sge, %118, %119 : i32
    cf.cond_br %120, ^bb25, ^bb26
    ^bb25:
      %121 = arith.constant 127 : i32
      %122 = arith.constant 19 : i32
      %123 = llvm.load %117 : !llvm.ptr -> i32
      %124 = arith.extsi %123 : i32 to i64
      %126 = arith.extsi %122 : i32 to i64
      %125 = arith.muli %126, %124 : i64
      %128 = arith.extsi %121 : i32 to i64
      %127 = arith.addi %128, %125 : i64
      %129 = llvm.mlir.constant(1 : i64) : i64
      %130 = llvm.alloca %129 x i64 : (i64) -> !llvm.ptr
      llvm.store %127, %130 : i64, !llvm.ptr
      %131 = arith.constant 1 : i32
      %132 = llvm.mlir.constant(1 : i64) : i64
      %133 = llvm.alloca %132 x i32 : (i64) -> !llvm.ptr
      llvm.store %131, %133 : i32, !llvm.ptr
      cf.br ^bb27
      ^bb27:
      %134 = llvm.load %133 : !llvm.ptr -> i32
      %135 = llvm.load %117 : !llvm.ptr -> i32
      %136 = arith.cmpi sle, %134, %135 : i32
      cf.cond_br %136, ^bb28, ^bb29
      ^bb28:
        %137 = llvm.load %130 : !llvm.ptr -> i64
        %138 = arith.constant 7 : i32
        %140 = arith.extsi %138 : i32 to i64
        %139 = arith.muli %137, %140 : i64
        llvm.store %139, %130 : i64, !llvm.ptr
        %141 = llvm.load %133 : !llvm.ptr -> i32
        %142 = arith.constant 1 : i32
        %143 = arith.addi %141, %142 : i32
        llvm.store %143, %133 : i32, !llvm.ptr
        cf.br ^bb27
      ^bb29:
      %144 = llvm.load %130 : !llvm.ptr -> i64
      %145 = arith.constant 1 : i32
      %147 = arith.extsi %145 : i32 to i64
      %146 = arith.subi %144, %147 : i64
      llvm.store %146, %130 : i64, !llvm.ptr
      %149 = llvm.load %130 : !llvm.ptr -> i64
      %148 = func.call @digit_at(%60, %61, %63, %149, %64, %110) : (!llvm.ptr, !llvm.ptr, i64, i64, !llvm.ptr, i32) -> i32
      %150 = llvm.load %114 : !llvm.ptr -> i64
      %151 = arith.constant 10 : i32
      %153 = arith.extsi %151 : i32 to i64
      %152 = arith.muli %150, %153 : i64
      %154 = arith.extsi %148 : i32 to i64
      %155 = arith.addi %152, %154 : i64
      llvm.store %155, %114 : i64, !llvm.ptr
      %156 = llvm.load %117 : !llvm.ptr -> i32
      %157 = arith.constant 1 : i32
      %158 = arith.subi %156, %157 : i32
      llvm.store %158, %117 : i32, !llvm.ptr
      cf.br ^bb24
    ^bb26:
    %159 = llvm.mlir.addressof @str_2 : !llvm.ptr
    %160 = llvm.load %114 : !llvm.ptr -> i64
    %161 = llvm.call @printf(%159, %160) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, i64) -> i32
    func.call @free(%64) : (!llvm.ptr) -> ()
    %163 = arith.constant 0 : i32
    func.return %163 : i32
  }
}