Do not try to use clock_gettime() on OSX
[hashcat.git] / src / hashcat.c
index 162b649..ab38fe2 100644 (file)
@@ -1,4 +1,4 @@
-/**
+ /**
  * Authors.....: Jens Steube <jens.steube@gmail.com>
  *               Gabriele Gristina <matrix@hashcat.net>
  *               magnum <john.magnum@hushmail.com>
@@ -38,7 +38,7 @@ double TARGET_MS_PROFILE[4]     = { 2, 12, 96, 480 };
 #define RESTORE_DISABLE         0
 #define STATUS                  0
 #define STATUS_TIMER            10
-#define STATUS_AUTOMAT          0
+#define MACHINE_READABLE        0
 #define LOOPBACK                0
 #define WEAK_HASH_THRESHOLD     100
 #define SHOW                    0
@@ -76,7 +76,7 @@ double TARGET_MS_PROFILE[4]     = { 2, 12, 96, 480 };
 #define BITMAP_MAX              24
 #define GPU_TEMP_DISABLE        0
 #define GPU_TEMP_ABORT          90
-#define GPU_TEMP_RETAIN         80
+#define GPU_TEMP_RETAIN         0
 #define WORKLOAD_PROFILE        2
 #define KERNEL_ACCEL            0
 #define KERNEL_LOOPS            0
@@ -150,6 +150,8 @@ double TARGET_MS_PROFILE[4]     = { 2, 12, 96, 480 };
 
 #define NUM_DEFAULT_BENCHMARK_ALGORITHMS 143
 
+#define NVIDIA_100PERCENTCPU_WORKAROUND 100
+
 #define global_free(attr)       \
 {                               \
   myfree ((void *) data.attr);  \
@@ -164,6 +166,12 @@ double TARGET_MS_PROFILE[4]     = { 2, 12, 96, 480 };
   attr = NULL;            \
 }
 
+#if defined(_WIN32) || defined(__WIN32__) || defined(__CYGWIN__)
+#define HC_API_CALL __stdcall
+#else
+#define HC_API_CALL
+#endif
+
 static uint default_benchmark_algorithms[NUM_DEFAULT_BENCHMARK_ALGORITHMS] =
 {
   900,
@@ -349,9 +357,7 @@ const char *USAGE_BIG[] =
   "",
   "Usage: %s [options]... hash|hashfile|hccapfile [dictionary|mask|directory]...",
   "",
-  "###########",
-  "# Options #",
-  "###########",
+  "- [ Options ] -",
   "",
   " Options Short / Long          | Type | Description                                          | Example",
   "===============================|======|======================================================|=======================",
@@ -366,7 +372,7 @@ const char *USAGE_BIG[] =
   "     --force                   |      | Ignore warnings                                      |",
   "     --status                  |      | Enable automatic update of the status-screen         |",
   "     --status-timer            | Num  | Sets seconds between status-screen update to X       | --status-timer=1",
-  "     --status-automat          |      | Display the status view in a machine readable format |",
+  "     --machine-readable        |      | Display the status view in a machine readable format |",
   "     --loopback                |      | Add new plains to induct directory                   |",
   "     --weak-hash-threshold     | Num  | Threshold X when to stop checking for weak hashes    | --weak=0",
   "     --markov-hcstat           | File | Specify hcstat file to use                           | --markov-hc=my.hcstat",
@@ -413,9 +419,7 @@ const char *USAGE_BIG[] =
   #ifdef HAVE_HWMON
   "     --gpu-temp-abort          | Num  | Abort if GPU temperature reaches X degrees celsius   | --gpu-temp-abort=100",
   "     --gpu-temp-retain         | Num  | Try to retain GPU temperature at X degrees celsius   | --gpu-temp-retain=95",
-  #ifdef HAVE_ADL
-  "     --powertune-enable        |      | Enable automatic power tuning (AMD OverDrive 6 only) |",
-  #endif
+  "     --powertune-enable        |      | Enable power tuning, restores settings when finished |",
   #endif
   "     --scrypt-tmto             | Num  | Manually override TMTO value for scrypt to X         | --scrypt-tmto=3",
   " -s, --skip                    | Num  | Skip X words from the start                          | -s 1000000",
@@ -436,12 +440,10 @@ const char *USAGE_BIG[] =
   "     --increment-min           | Num  | Start mask incrementing at X                         | --increment-min=4",
   "     --increment-max           | Num  | Stop mask incrementing at X                          | --increment-max=8",
   "",
-  "################",
-  "## Hash modes ##",
-  "################",
+  "- [ Hash modes ] -",
   "",
   "      # | Name                                             | Category",
-  "  ------+--------------------------------------------------+--------------------------------------",
+  "  ======+==================================================+======================================",
   "    900 | MD4                                              | Raw Hash",
   "      0 | MD5                                              | Raw Hash",
   "   5100 | Half MD5                                         | Raw Hash",
@@ -576,6 +578,7 @@ const char *USAGE_BIG[] =
   "     22 | Juniper Netscreen/SSG (ScreenOS)                 | Operating-Systems",
   "    501 | Juniper IVE                                      | Operating-Systems",
   "   5800 | Android PIN                                      | Operating-Systems",
+  "  13800 | Windows 8+ phone PIN/Password                    | Operating-Systems",
   "   8100 | Citrix Netscaler                                 | Operating-Systems",
   "   8500 | RACF                                             | Operating-Systems",
   "   7200 | GRUB 2                                           | Operating-Systems",
@@ -654,12 +657,10 @@ const char *USAGE_BIG[] =
   "  12700 | Blockchain, My Wallet                            | Password Managers",
   "  13400 | Keepass 1 (AES/Twofish) and Keepass 2 (AES)      | Password Managers",
   "",
-  "#####################",
-  "## Outfile Formats ##",
-  "#####################",
+  "- [ Outfile Formats ] -",
   "",
   "  # | Format",
-  " ---+--------",
+  " ===+========",
   "  1 | hash[:salt]",
   "  2 | plain",
   "  3 | hash[:salt]:plain",
@@ -676,35 +677,29 @@ const char *USAGE_BIG[] =
   " 14 | plain:hex_plain:crack_pos",
   " 15 | hash[:salt]:plain:hex_plain:crack_pos",
   "",
-  "##########################",
-  "## Rule Debugging Modes ##",
-  "##########################",
+  "- [ Rule Debugging Modes ] -",
   "",
   "  # | Format",
-  " ---+--------",
+  " ===+========",
   "  1 | Finding-Rule",
   "  2 | Original-Word",
   "  3 | Original-Word:Finding-Rule",
   "  4 | Original-Word:Finding-Rule:Processed-Word",
   "",
-  "##################",
-  "## Attack Modes ##",
-  "##################",
+  "- [ Attack Modes ] -",
   "",
   "  # | Mode",
-  " ---+------",
+  " ===+======",
   "  0 | Straight",
   "  1 | Combination",
   "  3 | Brute-force",
   "  6 | Hybrid Wordlist + Mask",
   "  7 | Hybrid Mask + Wordlist",
   "",
-  "#######################",
-  "## Built-in Charsets ##",
-  "#######################",
+  "- [ Built-in Charsets ] -",
   "",
   "  ? | Charset",
-  " ---+---------",
+  " ===+=========",
   "  l | abcdefghijklmnopqrstuvwxyz",
   "  u | ABCDEFGHIJKLMNOPQRSTUVWXYZ",
   "  d | 0123456789",
@@ -712,22 +707,18 @@ const char *USAGE_BIG[] =
   "  a | ?l?u?d?s",
   "  b | 0x00 - 0xff",
   "",
-  "#########################",
-  "## OpenCL Device Types ##",
-  "#########################",
+  "- [ OpenCL Device Types ] -",
   "",
   "  # | Device Type",
-  " ---+-------------",
+  " ===+=============",
   "  1 | CPU",
   "  2 | GPU",
   "  3 | FPGA, DSP, Co-Processor",
   "",
-  "#######################",
-  "## Workload Profiles ##",
-  "#######################",
+  "- [ Workload Profiles ] -",
   "",
   "  # | Performance | Runtime | Power Consumption | Desktop Impact",
-  " ---+-------------+---------+-------------------+----------------",
+  " ===+=============+=========+===================+=================",
   "  1 | Low         |   2 ms  | Low               | Minimal",
   "  2 | Default     |  12 ms  | Economic          | Noticeable",
   "  3 | High        |  96 ms  | High              | Unresponsive",
@@ -772,7 +763,7 @@ static double get_avg_exec_time (hc_device_param_t *device_param, const int last
   return exec_ms_sum / exec_ms_cnt;
 }
 
-void status_display_automat ()
+void status_display_machine_readable ()
 {
   FILE *out = stdout;
 
@@ -930,9 +921,9 @@ void status_display ()
   if (data.devices_status == STATUS_STARTING) return;
   if (data.devices_status == STATUS_BYPASS)   return;
 
-  if (data.status_automat == 1)
+  if (data.machine_readable == 1)
   {
-    status_display_automat ();
+    status_display_machine_readable ();
 
     return;
   }
@@ -1549,6 +1540,12 @@ void status_display ()
   }
 
   #ifdef HAVE_HWMON
+
+  if (data.devices_status == STATUS_EXHAUSTED)  return;
+  if (data.devices_status == STATUS_CRACKED)    return;
+  if (data.devices_status == STATUS_ABORTED)    return;
+  if (data.devices_status == STATUS_QUIT)       return;
+
   if (data.gpu_temp_disable == 0)
   {
     hc_thread_mutex_lock (mux_adl);
@@ -1559,46 +1556,84 @@ void status_display ()
 
       if (device_param->skipped) continue;
 
-      #define HM_STR_BUF_SIZE 255
+      const int num_temperature = hm_get_temperature_with_device_id (device_id);
+      const int num_fanspeed    = hm_get_fanspeed_with_device_id    (device_id);
+      const int num_utilization = hm_get_utilization_with_device_id (device_id);
+      const int num_corespeed   = hm_get_corespeed_with_device_id   (device_id);
+      const int num_memoryspeed = hm_get_memoryspeed_with_device_id (device_id);
+      const int num_buslanes    = hm_get_buslanes_with_device_id    (device_id);
+      const int num_throttle    = hm_get_throttle_with_device_id    (device_id);
+
+      char output_buf[256] = { 0 };
 
-      if (data.hm_device[device_id].fan_supported == 1)
+      int output_len = 0;
+
+      if (num_temperature >= 0)
       {
-        char utilization[HM_STR_BUF_SIZE] = { 0 };
-        char temperature[HM_STR_BUF_SIZE] = { 0 };
-        char fanspeed[HM_STR_BUF_SIZE] = { 0 };
+        snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Temp:%3uc", num_temperature);
 
-        hm_device_val_to_str ((char *) utilization, HM_STR_BUF_SIZE, "%", hm_get_utilization_with_device_id (device_id));
-        hm_device_val_to_str ((char *) temperature, HM_STR_BUF_SIZE, "c", hm_get_temperature_with_device_id (device_id));
+        output_len = strlen (output_buf);
+      }
 
-        if (device_param->vendor_id == VENDOR_ID_AMD)
-        {
-          hm_device_val_to_str ((char *) fanspeed, HM_STR_BUF_SIZE, "%", hm_get_fanspeed_with_device_id (device_id));
-        }
-        else if (device_param->vendor_id == VENDOR_ID_NV)
-        {
-          hm_device_val_to_str ((char *) fanspeed, HM_STR_BUF_SIZE, "%", hm_get_fanspeed_with_device_id (device_id));
-        }
+      if (num_fanspeed >= 0)
+      {
+        snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Fan:%3u%%", num_fanspeed);
 
-        log_info ("HWMon.GPU.#%d...: %s Util, %s Temp, %s Fan", device_id + 1, utilization, temperature, fanspeed);
+        output_len = strlen (output_buf);
       }
-      else
+
+      if (num_utilization >= 0)
+      {
+        snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Util:%3u%%", num_utilization);
+
+        output_len = strlen (output_buf);
+      }
+
+      if (num_corespeed >= 0)
+      {
+        snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Core:%4uMhz", num_corespeed);
+
+        output_len = strlen (output_buf);
+      }
+
+      if (num_memoryspeed >= 0)
+      {
+        snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Mem:%4uMhz", num_memoryspeed);
+
+        output_len = strlen (output_buf);
+      }
+
+      if (num_buslanes >= 0)
+      {
+        snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " Lanes:%u", num_buslanes);
+
+        output_len = strlen (output_buf);
+      }
+
+      if (num_throttle == 1)
       {
-        char utilization[HM_STR_BUF_SIZE] = { 0 };
-        char temperature[HM_STR_BUF_SIZE] = { 0 };
+        snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " *Throttled*");
+
+        output_len = strlen (output_buf);
+      }
 
-        hm_device_val_to_str ((char *) utilization, HM_STR_BUF_SIZE, "%", hm_get_utilization_with_device_id (device_id));
-        hm_device_val_to_str ((char *) temperature, HM_STR_BUF_SIZE, "c", hm_get_temperature_with_device_id (device_id));
+      if (output_len == 0)
+      {
+        snprintf (output_buf + output_len, sizeof (output_buf) - output_len, " N/A");
 
-        log_info ("HWMon.GPU.#%d...: %s Util, %s Temp, N/A Fan", device_id + 1, utilization, temperature);
+        output_len = strlen (output_buf);
       }
+
+      log_info ("HWMon.Dev.#%d...:%s", device_id + 1, output_buf);
     }
 
     hc_thread_mutex_unlock (mux_adl);
   }
+
   #endif // HAVE_HWMON
 }
 
-static void status_benchmark_automat ()
+static void status_benchmark_automate ()
 {
   u64    speed_cnt[DEVICES_MAX] = { 0 };
   double speed_ms[DEVICES_MAX]  = { 0 };
@@ -1645,9 +1680,9 @@ static void status_benchmark ()
   if (data.devices_status == STATUS_STARTING) return;
   if (data.devices_status == STATUS_BYPASS)   return;
 
-  if (data.status_automat == 1)
+  if (data.machine_readable == 1)
   {
-    status_benchmark_automat ();
+    status_benchmark_automate ();
 
     return;
   }
@@ -2116,6 +2151,7 @@ static void check_hash (hc_device_param_t *device_param, plain_t *plain)
 
       out_fp = stdout;
     }
+
     lock_file (out_fp);
   }
   else
@@ -2369,42 +2405,6 @@ static void save_hash ()
   unlink (old_hashfile);
 }
 
-static float find_kernel_power_div (const u64 total_left, const uint kernel_power_all)
-{
-  // function called only in case kernel_power_all > words_left
-
-  float kernel_power_div = (float) (total_left) / kernel_power_all;
-
-  kernel_power_div += kernel_power_div / 100;
-
-  u32 kernel_power_new = (u32) (kernel_power_all * kernel_power_div);
-
-  while (kernel_power_new < total_left)
-  {
-    kernel_power_div += kernel_power_div / 100;
-
-    kernel_power_new = (u32) (kernel_power_all * kernel_power_div);
-  }
-
-  if (data.quiet == 0)
-  {
-    clear_prompt ();
-
-    //log_info ("");
-
-    log_info ("INFO: approaching final keyspace, workload adjusted");
-    log_info ("");
-
-    fprintf (stdout, "%s", PROMPT);
-
-    fflush (stdout);
-  }
-
-  if ((kernel_power_all * kernel_power_div) < 8) return 1;
-
-  return kernel_power_div;
-}
-
 static void run_kernel (const uint kern_run, hc_device_param_t *device_param, const uint num, const uint event_update)
 {
   uint num_elements = num;
@@ -2609,11 +2609,59 @@ static void run_kernel_amp (hc_device_param_t *device_param, const uint num)
   hc_clFinish (data.ocl, device_param->command_queue);
 }
 
+static void run_kernel_memset (hc_device_param_t *device_param, cl_mem buf, const uint value, const uint num)
+{
+  const u32 num16d = num / 16;
+  const u32 num16m = num % 16;
+
+  if (num16d)
+  {
+    device_param->kernel_params_memset_buf32[1] = value;
+    device_param->kernel_params_memset_buf32[2] = num16d;
+
+    uint kernel_threads = device_param->kernel_threads;
+
+    uint num_elements = num16d;
+
+    while (num_elements % kernel_threads) num_elements++;
+
+    cl_kernel kernel = device_param->kernel_memset;
+
+    hc_clSetKernelArg (data.ocl, kernel, 0, sizeof (cl_mem),  (void *) &buf);
+    hc_clSetKernelArg (data.ocl, kernel, 1, sizeof (cl_uint), device_param->kernel_params_memset[1]);
+    hc_clSetKernelArg (data.ocl, kernel, 2, sizeof (cl_uint), device_param->kernel_params_memset[2]);
+
+    const size_t global_work_size[3] = { num_elements,   1, 1 };
+    const size_t local_work_size[3]  = { kernel_threads, 1, 1 };
+
+    hc_clEnqueueNDRangeKernel (data.ocl, device_param->command_queue, kernel, 1, NULL, global_work_size, local_work_size, 0, NULL, NULL);
+
+    hc_clFlush (data.ocl, device_param->command_queue);
+
+    hc_clFinish (data.ocl, device_param->command_queue);
+  }
+
+  if (num16m)
+  {
+    u32 tmp[4];
+
+    tmp[0] = value;
+    tmp[1] = value;
+    tmp[2] = value;
+    tmp[3] = value;
+
+    hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, buf, CL_TRUE, num16d * 16, num16m, tmp, 0, NULL, NULL);
+  }
+}
+
 static void run_kernel_bzero (hc_device_param_t *device_param, cl_mem buf, const size_t size)
 {
+  run_kernel_memset (device_param, buf, 0, size);
+
+  /*
   int rc = -1;
 
-  if (device_param->opencl_v12 && device_param->vendor_id == VENDOR_ID_AMD)
+  if (device_param->opencl_v12 && device_param->platform_vendor_id == VENDOR_ID_AMD)
   {
     // So far tested, amd is the only supporting this OpenCL 1.2 function without segfaulting
 
@@ -2644,6 +2692,7 @@ static void run_kernel_bzero (hc_device_param_t *device_param, cl_mem buf, const
 
     myfree (tmp);
   }
+  */
 }
 
 static void choose_kernel (hc_device_param_t *device_param, const uint attack_exec, const uint attack_mode, const uint opts_type, const salt_t *salt_buf, const uint highest_pw_len, const uint pws_cnt)
@@ -2836,7 +2885,7 @@ static void run_copy (hc_device_param_t *device_param, const uint pws_cnt)
 
 static double try_run (hc_device_param_t *device_param, const u32 kernel_accel, const u32 kernel_loops)
 {
-  const u32 kernel_power = device_param->device_processors * device_param->kernel_threads * kernel_accel;
+  const u32 kernel_power_try = device_param->device_processors * device_param->kernel_threads * kernel_accel;
 
   device_param->kernel_params_buf32[25] = 0;
   device_param->kernel_params_buf32[26] = kernel_loops; // not a bug, both need to be set
@@ -2844,11 +2893,11 @@ static double try_run (hc_device_param_t *device_param, const u32 kernel_accel,
 
   if (data.attack_exec == ATTACK_EXEC_INSIDE_KERNEL)
   {
-    run_kernel (KERN_RUN_1, device_param, kernel_power, true);
+    run_kernel (KERN_RUN_1, device_param, kernel_power_try, true);
   }
   else
   {
-    run_kernel (KERN_RUN_2, device_param, kernel_power, true);
+    run_kernel (KERN_RUN_2, device_param, kernel_power_try, true);
   }
 
   const double exec_ms_prev = get_avg_exec_time (device_param, 1);
@@ -2895,16 +2944,30 @@ static void autotune (hc_device_param_t *device_param)
 
   const u32 kernel_power_max = device_param->device_processors * device_param->kernel_threads * kernel_accel_max;
 
-  for (u32 i = 0; i < kernel_power_max; i++)
+  if (data.attack_kern == ATTACK_KERN_BF)
   {
-    device_param->pws_buf[i].i[0]   = i;
-    device_param->pws_buf[i].i[1]   = 0x01234567;
-    device_param->pws_buf[i].pw_len = 7;
+    run_kernel_memset (device_param, device_param->d_pws_buf, 7, kernel_power_max * sizeof (pw_t));
   }
+  else
+  {
+    for (u32 i = 0; i < kernel_power_max; i++)
+    {
+      device_param->pws_buf[i].i[0]   = i;
+      device_param->pws_buf[i].i[1]   = 0x01234567;
+      device_param->pws_buf[i].pw_len = 7 + (i & 7);
+    }
 
-  hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_pws_buf, CL_TRUE, 0, kernel_power_max * sizeof (pw_t), device_param->pws_buf, 0, NULL, NULL);
+    hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_pws_buf, CL_TRUE, 0, kernel_power_max * sizeof (pw_t), device_param->pws_buf, 0, NULL, NULL);
+  }
 
-  if (data.attack_exec == ATTACK_EXEC_OUTSIDE_KERNEL)
+  if (data.attack_exec == ATTACK_EXEC_INSIDE_KERNEL)
+  {
+    if (data.kernel_rules_cnt > 1)
+    {
+      hc_clEnqueueCopyBuffer (data.ocl, device_param->command_queue, device_param->d_rules, device_param->d_rules_c, 0, 0, MIN (kernel_loops_max, KERNEL_RULES) * sizeof (kernel_rule_t), 0, NULL, NULL);
+    }
+  }
+  else
   {
     run_kernel_amp (device_param, kernel_power_max);
   }
@@ -2975,16 +3038,27 @@ static void autotune (hc_device_param_t *device_param)
     exec_ms_pre_final = MIN (exec_ms_pre_final, exec_ms_pre_final_v);
   }
 
+  u32 diff = kernel_loops - kernel_accel;
+
   if ((kernel_loops_min < kernel_loops_max) && (kernel_accel_min < kernel_accel_max))
   {
-    for (u32 f = 2; f < 1024; f++)
+    u32 kernel_accel_orig = kernel_accel;
+    u32 kernel_loops_orig = kernel_loops;
+
+    for (u32 f = 1; f < 1024; f++)
     {
-      const u32 kernel_accel_try = kernel_accel * f;
-      const u32 kernel_loops_try = kernel_loops / f;
+      const u32 kernel_accel_try = (float) kernel_accel_orig * f;
+      const u32 kernel_loops_try = (float) kernel_loops_orig / f;
 
       if (kernel_accel_try > kernel_accel_max) break;
       if (kernel_loops_try < kernel_loops_min) break;
 
+      u32 diff_new = kernel_loops_try - kernel_accel_try;
+
+      if (diff_new > diff) break;
+
+      diff_new = diff;
+
       double exec_ms = try_run (device_param, kernel_accel_try, kernel_loops_try);
 
       for (int i = 0; i < VERIFIER_CNT; i++)
@@ -3019,10 +3093,19 @@ static void autotune (hc_device_param_t *device_param)
 
   // reset them fake words
 
+  /*
   memset (device_param->pws_buf, 0, kernel_power_max * sizeof (pw_t));
 
   hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_pws_buf,     CL_TRUE, 0, kernel_power_max * sizeof (pw_t), device_param->pws_buf, 0, NULL, NULL);
   hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_pws_amp_buf, CL_TRUE, 0, kernel_power_max * sizeof (pw_t), device_param->pws_buf, 0, NULL, NULL);
+  */
+
+  run_kernel_memset (device_param, device_param->d_pws_buf, 0, kernel_power_max * sizeof (pw_t));
+
+  if (data.attack_exec == ATTACK_EXEC_OUTSIDE_KERNEL)
+  {
+    run_kernel_memset (device_param, device_param->d_pws_amp_buf, 0, kernel_power_max * sizeof (pw_t));
+  }
 
   // reset timer
 
@@ -3319,7 +3402,10 @@ static void run_cracker (hc_device_param_t *device_param, const uint pws_cnt)
        * result
        */
 
-      check_cracked (device_param, salt_pos);
+      if (data.benchmark == 0)
+      {
+        check_cracked (device_param, salt_pos);
+      }
 
       /**
        * progress
@@ -3713,9 +3799,11 @@ static void *thread_monitor (void *p)
   uint status_left  = data.status_timer;
 
   #ifdef HAVE_HWMON
-  uint hwmon_check   = 0;
+  uint hwmon_check = 0;
+
+  int slowdown_warnings = 0;
 
-  // these variables are mainly used for fan control (AMD only)
+  // these variables are mainly used for fan control
 
   int *fan_speed_chgd = (int *) mycalloc (data.devices_cnt, sizeof (int));
 
@@ -3724,12 +3812,10 @@ static void *thread_monitor (void *p)
   int *temp_diff_old = (int *) mycalloc (data.devices_cnt, sizeof (int));
   int *temp_diff_sum = (int *) mycalloc (data.devices_cnt, sizeof (int));
 
-  #ifdef HAVE_ADL
   int temp_threshold = 1; // degrees celcius
 
   int fan_speed_min =  15; // in percentage
   int fan_speed_max = 100;
-  #endif // HAVE_ADL
 
   time_t last_temp_check_time;
   #endif // HAVE_HWMON
@@ -3780,6 +3866,63 @@ static void *thread_monitor (void *p)
     if (data.devices_status != STATUS_RUNNING) continue;
 
     #ifdef HAVE_HWMON
+
+    if (hwmon_check == 1)
+    {
+      hc_thread_mutex_lock (mux_adl);
+
+      for (uint device_id = 0; device_id < data.devices_cnt; device_id++)
+      {
+        hc_device_param_t *device_param = &data.devices_param[device_id];
+
+        if (device_param->skipped) continue;
+
+        if (device_param->device_vendor_id == VENDOR_ID_NV)
+        {
+          if (data.hm_nvapi)
+          {
+            NV_GPU_PERF_POLICIES_INFO_PARAMS_V1   perfPolicies_info   = { 0 };
+            NV_GPU_PERF_POLICIES_STATUS_PARAMS_V1 perfPolicies_status = { 0 };
+
+            perfPolicies_info.version   = MAKE_NVAPI_VERSION (NV_GPU_PERF_POLICIES_INFO_PARAMS_V1, 1);
+            perfPolicies_status.version = MAKE_NVAPI_VERSION (NV_GPU_PERF_POLICIES_STATUS_PARAMS_V1, 1);
+
+            hm_NvAPI_GPU_GetPerfPoliciesInfo (data.hm_nvapi, data.hm_device[device_id].nvapi, &perfPolicies_info);
+
+            perfPolicies_status.info_value = perfPolicies_info.info_value;
+
+            hm_NvAPI_GPU_GetPerfPoliciesStatus (data.hm_nvapi, data.hm_device[device_id].nvapi, &perfPolicies_status);
+
+            if (perfPolicies_status.throttle & 2)
+            {
+              if (slowdown_warnings < 3)
+              {
+                if (data.quiet == 0) clear_prompt ();
+
+                log_info ("WARNING: Drivers temperature threshold hit on GPU #%d, expect performance to drop...", device_id + 1);
+
+                if (slowdown_warnings == 2)
+                {
+                  log_info ("");
+                }
+
+                if (data.quiet == 0) fprintf (stdout, "%s", PROMPT);
+                if (data.quiet == 0) fflush (stdout);
+
+                slowdown_warnings++;
+              }
+            }
+            else
+            {
+              slowdown_warnings = 0;
+            }
+          }
+        }
+      }
+
+      hc_thread_mutex_unlock (mux_adl);
+    }
+
     if (hwmon_check == 1)
     {
       hc_thread_mutex_lock (mux_adl);
@@ -3811,12 +3954,11 @@ static void *thread_monitor (void *p)
           break;
         }
 
-        #ifdef HAVE_ADL
         const int gpu_temp_retain = data.gpu_temp_retain;
 
-        if (gpu_temp_retain) // VENDOR_ID_AMD implied
+        if (gpu_temp_retain)
         {
-          if (data.hm_device[device_id].fan_supported == 1)
+          if (data.hm_device[device_id].fan_set_supported == 1)
           {
             int temp_cur = temperature;
 
@@ -3856,7 +3998,14 @@ static void *thread_monitor (void *p)
 
                 if ((freely_change_fan_speed == 1) || (fan_speed_must_change == 1))
                 {
-                  hm_set_fanspeed_with_device_id_amd (device_id, fan_speed_new);
+                  if (device_param->device_vendor_id == VENDOR_ID_AMD)
+                  {
+                    hm_set_fanspeed_with_device_id_adl (device_id, fan_speed_new, 1);
+                  }
+                  else if (device_param->device_vendor_id == VENDOR_ID_NV)
+                  {
+
+                  }
 
                   fan_speed_chgd[device_id] = 1;
                 }
@@ -3866,7 +4015,6 @@ static void *thread_monitor (void *p)
             }
           }
         }
-        #endif // HAVE_ADL
       }
 
       hc_thread_mutex_unlock (mux_adl);
@@ -4256,8 +4404,8 @@ static void *thread_outfile_remove (void *p)
 
 static void pw_add (hc_device_param_t *device_param, const u8 *pw_buf, const int pw_len)
 {
-  if (device_param->pws_cnt < device_param->kernel_power)
-  {
+  //if (device_param->pws_cnt < device_param->kernel_power)
+  //{
     pw_t *pw = (pw_t *) device_param->pws_buf + device_param->pws_cnt;
 
     u8 *ptr = (u8 *) pw->i;
@@ -4269,16 +4417,55 @@ static void pw_add (hc_device_param_t *device_param, const u8 *pw_buf, const int
     pw->pw_len = pw_len;
 
     device_param->pws_cnt++;
+  //}
+  //else
+  //{
+  //  fprintf (stderr, "BUG pw_add()!!\n");
+  //
+  //  return;
+  //}
+}
+
+static void set_kernel_power_final (const u64 kernel_power_final)
+{
+  if (data.quiet == 0)
+  {
+    clear_prompt ();
+
+    //log_info ("");
+
+    log_info ("INFO: approaching final keyspace, workload adjusted");
+    log_info ("");
+
+    fprintf (stdout, "%s", PROMPT);
+
+    fflush (stdout);
   }
-  else
+
+  data.kernel_power_final = kernel_power_final;
+}
+
+static u32 get_power (hc_device_param_t *device_param)
+{
+  const u64 kernel_power_final = data.kernel_power_final;
+
+  if (kernel_power_final)
   {
-    fprintf (stderr, "BUG pw_add()!!\n");
+    const double device_factor = (double) device_param->hardware_power / data.hardware_power_all;
 
-    return;
+    const u64 words_left_device = CEIL ((double) kernel_power_final * device_factor);
+
+    // work should be at least the hardware power available without any accelerator
+
+    const u64 work = MAX (words_left_device, device_param->hardware_power);
+
+    return work;
   }
+
+  return device_param->kernel_power;
 }
 
-static uint get_work (hc_device_param_t *device_param, const u64 max, const bool allow_div)
+static uint get_work (hc_device_param_t *device_param, const u64 max)
 {
   hc_thread_mutex_lock (mux_dispatcher);
 
@@ -4287,33 +4474,19 @@ static uint get_work (hc_device_param_t *device_param, const u64 max, const bool
 
   device_param->words_off = words_cur;
 
+  const u64 kernel_power_all = data.kernel_power_all;
+
   const u64 words_left = words_base - words_cur;
 
-  if (allow_div)
+  if (words_left < kernel_power_all)
   {
-    if (data.kernel_power_all > words_left)
-    {
-      if (data.kernel_power_div == 0)
-      {
-        data.kernel_power_div = find_kernel_power_div (words_left, data.kernel_power_all);
-      }
-    }
-
-    if (data.kernel_power_div)
+    if (data.kernel_power_final == 0)
     {
-      if (device_param->kernel_power == device_param->kernel_power_user)
-      {
-        const u32 kernel_power_new = (float) device_param->kernel_power * data.kernel_power_div;
-
-        if (kernel_power_new < device_param->kernel_power)
-        {
-          device_param->kernel_power = kernel_power_new;
-        }
-      }
+      set_kernel_power_final (words_left);
     }
   }
 
-  const uint kernel_power = device_param->kernel_power;
+  const u32 kernel_power = get_power (device_param);
 
   uint work = MIN (words_left, kernel_power);
 
@@ -4326,7 +4499,7 @@ static uint get_work (hc_device_param_t *device_param, const u64 max, const bool
   return work;
 }
 
-static void *thread_calc_stdin (void *p)
+static void *thread_autotune (void *p)
 {
   hc_device_param_t *device_param = (hc_device_param_t *) p;
 
@@ -4334,12 +4507,19 @@ static void *thread_calc_stdin (void *p)
 
   autotune (device_param);
 
+  return NULL;
+}
+
+static void *thread_calc_stdin (void *p)
+{
+  hc_device_param_t *device_param = (hc_device_param_t *) p;
+
+  if (device_param->skipped) return NULL;
+
   char *buf = (char *) mymalloc (HCBUFSIZ);
 
   const uint attack_kern = data.attack_kern;
 
-  const uint kernel_power = device_param->kernel_power;
-
   while ((data.devices_status != STATUS_EXHAUSTED) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
   {
     hc_thread_mutex_lock (mux_dispatcher);
@@ -4353,7 +4533,7 @@ static void *thread_calc_stdin (void *p)
 
     uint words_cur = 0;
 
-    while (words_cur < kernel_power)
+    while (words_cur < device_param->kernel_power)
     {
       char *line_buf = fgets (buf, HCBUFSIZ - 1, stdin);
 
@@ -4387,6 +4567,8 @@ static void *thread_calc_stdin (void *p)
         continue;
       }
 
+      // hmm that's always the case, or?
+
       if (attack_kern == ATTACK_KERN_STRAIGHT)
       {
         if ((line_len < data.pw_min) || (line_len > data.pw_max))
@@ -4403,25 +4585,6 @@ static void *thread_calc_stdin (void *p)
           continue;
         }
       }
-      else if (attack_kern == ATTACK_KERN_COMBI)
-      {
-        // do not check if minimum restriction is satisfied (line_len >= data.pw_min) here
-        // since we still need to combine the plains
-
-        if (line_len > data.pw_max)
-        {
-          hc_thread_mutex_lock (mux_counter);
-
-          for (uint salt_pos = 0; salt_pos < data.salts_cnt; salt_pos++)
-          {
-            data.words_progress_rejected[salt_pos] += data.combs_cnt;
-          }
-
-          hc_thread_mutex_unlock (mux_counter);
-
-          continue;
-        }
-      }
 
       pw_add (device_param, (u8 *) line_buf, line_len);
 
@@ -4480,8 +4643,6 @@ static void *thread_calc (void *p)
 
   if (device_param->skipped) return NULL;
 
-  autotune (device_param);
-
   const uint attack_mode = data.attack_mode;
   const uint attack_kern = data.attack_kern;
 
@@ -4489,7 +4650,7 @@ static void *thread_calc (void *p)
   {
     while ((data.devices_status != STATUS_EXHAUSTED) && (data.devices_status != STATUS_CRACKED) && (data.devices_status != STATUS_ABORTED) && (data.devices_status != STATUS_QUIT))
     {
-      const uint work = get_work (device_param, -1, true);
+      const uint work = get_work (device_param, -1);
 
       if (work == 0) break;
 
@@ -4604,18 +4765,16 @@ static void *thread_calc (void *p)
       u64 words_off = 0;
       u64 words_fin = 0;
 
-      bool allow_div = true;
-
       u64 max = -1;
 
       while (max)
       {
-        const uint work = get_work (device_param, max, allow_div);
-
-        allow_div = false;
+        const uint work = get_work (device_param, max);
 
         if (work == 0) break;
 
+        max = 0;
+
         words_off = device_param->words_off;
         words_fin = words_off + work;
 
@@ -4624,8 +4783,6 @@ static void *thread_calc (void *p)
 
         for ( ; words_cur < words_off; words_cur++) get_next_word (wl_data, fd, &line_buf, &line_len);
 
-        max = 0;
-
         for ( ; words_cur < words_fin; words_cur++)
         {
           get_next_word (wl_data, fd, &line_buf, &line_len);
@@ -5211,8 +5368,8 @@ static uint hlfmt_detect (FILE *fp, uint max_check)
 
 // wrapper around mymalloc for ADL
 
-#if defined(HAVE_HWMON) && defined(HAVE_ADL)
-void *__stdcall ADL_Main_Memory_Alloc (const int iSize)
+#if defined(HAVE_HWMON)
+void *HC_API_CALL ADL_Main_Memory_Alloc (const int iSize)
 {
   return mymalloc (iSize);
 }
@@ -5234,8 +5391,6 @@ static uint generate_bitmaps (const uint digests_cnt, const uint dgst_size, cons
 
   for (uint i = 0; i < digests_cnt; i++)
   {
-    if (data.digests_shown[i] == 1) continue; // can happen with potfile
-
     uint *digest_ptr = (uint *) digests_buf_ptr;
 
     digests_buf_ptr += dgst_size;
@@ -5270,35 +5425,82 @@ static uint generate_bitmaps (const uint digests_cnt, const uint dgst_size, cons
  * main
  */
 
-int main (int argc, char **argv)
+#ifdef LINUX
+int (*clock_gettime_orig) (clockid_t clk_id, struct timespec *tp);
+
+int clock_gettime (clockid_t clk_id, struct timespec *tp)
 {
-  /**
-   * To help users a bit
-   */
+  int r = clock_gettime_orig (clk_id, tp);
 
-  char *compute = getenv ("COMPUTE");
+  usleep (NVIDIA_100PERCENTCPU_WORKAROUND);
 
-  if (compute)
-  {
-    static char display[100];
+  return r;
+}
+#endif
 
-    snprintf (display, sizeof (display) - 1, "DISPLAY=%s", compute);
+#ifdef WIN
+void SetConsoleWindowSize (const int x)
+{
+  HANDLE h = GetStdHandle (STD_OUTPUT_HANDLE);
 
-    putenv (display);
-  }
-  else
-  {
-    if (getenv ("DISPLAY") == NULL)
-      putenv ((char *) "DISPLAY=:0");
-  }
+  if (h == INVALID_HANDLE_VALUE) return;
 
-  if (getenv ("GPU_MAX_ALLOC_PERCENT") == NULL)
-    putenv ((char *) "GPU_MAX_ALLOC_PERCENT=100");
+  CONSOLE_SCREEN_BUFFER_INFO bufferInfo;
 
-  if (getenv ("CPU_MAX_ALLOC_PERCENT") == NULL)
-    putenv ((char *) "CPU_MAX_ALLOC_PERCENT=100");
+  if (!GetConsoleScreenBufferInfo (h, &bufferInfo)) return;
 
-  if (getenv ("GPU_USE_SYNC_OBJECTS") == NULL)
+  SMALL_RECT *sr = &bufferInfo.srWindow;
+
+  sr->Right = MAX (sr->Right, x - 1);
+
+  COORD co;
+
+  co.X = sr->Right + 1;
+  co.Y = 9999;
+
+  if (!SetConsoleScreenBufferSize (h, co)) return;
+
+  if (!SetConsoleWindowInfo (h, TRUE, sr)) return;
+}
+#endif
+
+int main (int argc, char **argv)
+{
+  #ifdef LINUX
+  clock_gettime_orig = dlsym (RTLD_NEXT, "clock_gettime");
+  #endif
+
+  #ifdef WIN
+  SetConsoleWindowSize (132);
+  #endif
+
+  /**
+   * To help users a bit
+   */
+
+  char *compute = getenv ("COMPUTE");
+
+  if (compute)
+  {
+    static char display[100];
+
+    snprintf (display, sizeof (display) - 1, "DISPLAY=%s", compute);
+
+    putenv (display);
+  }
+  else
+  {
+    if (getenv ("DISPLAY") == NULL)
+      putenv ((char *) "DISPLAY=:0");
+  }
+
+  if (getenv ("GPU_MAX_ALLOC_PERCENT") == NULL)
+    putenv ((char *) "GPU_MAX_ALLOC_PERCENT=100");
+
+  if (getenv ("CPU_MAX_ALLOC_PERCENT") == NULL)
+    putenv ((char *) "CPU_MAX_ALLOC_PERCENT=100");
+
+  if (getenv ("GPU_USE_SYNC_OBJECTS") == NULL)
     putenv ((char *) "GPU_USE_SYNC_OBJECTS=1");
 
   if (getenv ("CUDA_CACHE_DISABLE") == NULL)
@@ -5333,87 +5535,85 @@ int main (int argc, char **argv)
    * commandline parameters
    */
 
-  uint  usage                 = USAGE;
-  uint  version               = VERSION;
-  uint  quiet                 = QUIET;
-  uint  benchmark             = BENCHMARK;
-  uint  show                  = SHOW;
-  uint  left                  = LEFT;
-  uint  username              = USERNAME;
-  uint  remove                = REMOVE;
-  uint  remove_timer          = REMOVE_TIMER;
-  u64   skip                  = SKIP;
-  u64   limit                 = LIMIT;
-  uint  keyspace              = KEYSPACE;
-  uint  potfile_disable       = POTFILE_DISABLE;
-  char *potfile_path          = NULL;
-  uint  debug_mode            = DEBUG_MODE;
-  char *debug_file            = NULL;
-  char *induction_dir         = NULL;
-  char *outfile_check_dir     = NULL;
-  uint  force                 = FORCE;
-  uint  runtime               = RUNTIME;
-  uint  hash_mode             = HASH_MODE;
-  uint  attack_mode           = ATTACK_MODE;
-  uint  markov_disable        = MARKOV_DISABLE;
-  uint  markov_classic        = MARKOV_CLASSIC;
-  uint  markov_threshold      = MARKOV_THRESHOLD;
-  char *markov_hcstat         = NULL;
-  char *outfile               = NULL;
-  uint  outfile_format        = OUTFILE_FORMAT;
-  uint  outfile_autohex       = OUTFILE_AUTOHEX;
-  uint  outfile_check_timer   = OUTFILE_CHECK_TIMER;
-  uint  restore               = RESTORE;
-  uint  restore_timer         = RESTORE_TIMER;
-  uint  restore_disable       = RESTORE_DISABLE;
-  uint  status                = STATUS;
-  uint  status_timer          = STATUS_TIMER;
-  uint  status_automat        = STATUS_AUTOMAT;
-  uint  loopback              = LOOPBACK;
-  uint  weak_hash_threshold   = WEAK_HASH_THRESHOLD;
-  char *session               = NULL;
-  uint  hex_charset           = HEX_CHARSET;
-  uint  hex_salt              = HEX_SALT;
-  uint  hex_wordlist          = HEX_WORDLIST;
-  uint  rp_gen                = RP_GEN;
-  uint  rp_gen_func_min       = RP_GEN_FUNC_MIN;
-  uint  rp_gen_func_max       = RP_GEN_FUNC_MAX;
-  uint  rp_gen_seed           = RP_GEN_SEED;
-  char *rule_buf_l            = (char *) RULE_BUF_L;
-  char *rule_buf_r            = (char *) RULE_BUF_R;
-  uint  increment             = INCREMENT;
-  uint  increment_min         = INCREMENT_MIN;
-  uint  increment_max         = INCREMENT_MAX;
-  char *cpu_affinity          = NULL;
-  OCL_PTR *ocl                = NULL;
-  char *opencl_devices        = NULL;
-  char *opencl_platforms      = NULL;
-  char *opencl_device_types   = NULL;
-  uint  opencl_vector_width   = OPENCL_VECTOR_WIDTH;
-  char *truecrypt_keyfiles    = NULL;
-  char *veracrypt_keyfiles    = NULL;
-  uint  veracrypt_pim         = 0;
-  uint  workload_profile      = WORKLOAD_PROFILE;
-  uint  kernel_accel          = KERNEL_ACCEL;
-  uint  kernel_loops          = KERNEL_LOOPS;
-  uint  gpu_temp_disable      = GPU_TEMP_DISABLE;
+  uint  usage                     = USAGE;
+  uint  version                   = VERSION;
+  uint  quiet                     = QUIET;
+  uint  benchmark                 = BENCHMARK;
+  uint  show                      = SHOW;
+  uint  left                      = LEFT;
+  uint  username                  = USERNAME;
+  uint  remove                    = REMOVE;
+  uint  remove_timer              = REMOVE_TIMER;
+  u64   skip                      = SKIP;
+  u64   limit                     = LIMIT;
+  uint  keyspace                  = KEYSPACE;
+  uint  potfile_disable           = POTFILE_DISABLE;
+  char *potfile_path              = NULL;
+  uint  debug_mode                = DEBUG_MODE;
+  char *debug_file                = NULL;
+  char *induction_dir             = NULL;
+  char *outfile_check_dir         = NULL;
+  uint  force                     = FORCE;
+  uint  runtime                   = RUNTIME;
+  uint  hash_mode                 = HASH_MODE;
+  uint  attack_mode               = ATTACK_MODE;
+  uint  markov_disable            = MARKOV_DISABLE;
+  uint  markov_classic            = MARKOV_CLASSIC;
+  uint  markov_threshold          = MARKOV_THRESHOLD;
+  char *markov_hcstat             = NULL;
+  char *outfile                   = NULL;
+  uint  outfile_format            = OUTFILE_FORMAT;
+  uint  outfile_autohex           = OUTFILE_AUTOHEX;
+  uint  outfile_check_timer       = OUTFILE_CHECK_TIMER;
+  uint  restore                   = RESTORE;
+  uint  restore_timer             = RESTORE_TIMER;
+  uint  restore_disable           = RESTORE_DISABLE;
+  uint  status                    = STATUS;
+  uint  status_timer              = STATUS_TIMER;
+  uint  machine_readable          = MACHINE_READABLE;
+  uint  loopback                  = LOOPBACK;
+  uint  weak_hash_threshold       = WEAK_HASH_THRESHOLD;
+  char *session                   = NULL;
+  uint  hex_charset               = HEX_CHARSET;
+  uint  hex_salt                  = HEX_SALT;
+  uint  hex_wordlist              = HEX_WORDLIST;
+  uint  rp_gen                    = RP_GEN;
+  uint  rp_gen_func_min           = RP_GEN_FUNC_MIN;
+  uint  rp_gen_func_max           = RP_GEN_FUNC_MAX;
+  uint  rp_gen_seed               = RP_GEN_SEED;
+  char *rule_buf_l                = (char *) RULE_BUF_L;
+  char *rule_buf_r                = (char *) RULE_BUF_R;
+  uint  increment                 = INCREMENT;
+  uint  increment_min             = INCREMENT_MIN;
+  uint  increment_max             = INCREMENT_MAX;
+  char *cpu_affinity              = NULL;
+  OCL_PTR *ocl                    = NULL;
+  char *opencl_devices            = NULL;
+  char *opencl_platforms          = NULL;
+  char *opencl_device_types       = NULL;
+  uint  opencl_vector_width       = OPENCL_VECTOR_WIDTH;
+  char *truecrypt_keyfiles        = NULL;
+  char *veracrypt_keyfiles        = NULL;
+  uint  veracrypt_pim             = 0;
+  uint  workload_profile          = WORKLOAD_PROFILE;
+  uint  kernel_accel              = KERNEL_ACCEL;
+  uint  kernel_loops              = KERNEL_LOOPS;
+  uint  gpu_temp_disable          = GPU_TEMP_DISABLE;
   #ifdef HAVE_HWMON
-  uint  gpu_temp_abort        = GPU_TEMP_ABORT;
-  uint  gpu_temp_retain       = GPU_TEMP_RETAIN;
-  #ifdef HAVE_ADL
-  uint  powertune_enable      = POWERTUNE_ENABLE;
-  #endif
+  uint  gpu_temp_abort            = GPU_TEMP_ABORT;
+  uint  gpu_temp_retain           = GPU_TEMP_RETAIN;
+  uint  powertune_enable          = POWERTUNE_ENABLE;
   #endif
-  uint  logfile_disable       = LOGFILE_DISABLE;
-  uint  segment_size          = SEGMENT_SIZE;
-  uint  scrypt_tmto           = SCRYPT_TMTO;
-  char  separator             = SEPARATOR;
-  uint  bitmap_min            = BITMAP_MIN;
-  uint  bitmap_max            = BITMAP_MAX;
-  char *custom_charset_1      = NULL;
-  char *custom_charset_2      = NULL;
-  char *custom_charset_3      = NULL;
-  char *custom_charset_4      = NULL;
+  uint  logfile_disable           = LOGFILE_DISABLE;
+  uint  segment_size              = SEGMENT_SIZE;
+  uint  scrypt_tmto               = SCRYPT_TMTO;
+  char  separator                 = SEPARATOR;
+  uint  bitmap_min                = BITMAP_MIN;
+  uint  bitmap_max                = BITMAP_MAX;
+  char *custom_charset_1          = NULL;
+  char *custom_charset_2          = NULL;
+  char *custom_charset_3          = NULL;
+  char *custom_charset_4          = NULL;
 
   #define IDX_HELP                      'h'
   #define IDX_VERSION                   'V'
@@ -5456,7 +5656,7 @@ int main (int argc, char **argv)
   #define IDX_RESTORE_DISABLE           0xff27
   #define IDX_STATUS                    0xff17
   #define IDX_STATUS_TIMER              0xff18
-  #define IDX_STATUS_AUTOMAT            0xff50
+  #define IDX_MACHINE_READABLE          0xff50
   #define IDX_LOOPBACK                  0xff38
   #define IDX_WEAK_HASH_THRESHOLD       0xff42
   #define IDX_SESSION                   0xff19
@@ -5520,7 +5720,7 @@ int main (int argc, char **argv)
     {"restore-disable",           no_argument,       0, IDX_RESTORE_DISABLE},
     {"status",                    no_argument,       0, IDX_STATUS},
     {"status-timer",              required_argument, 0, IDX_STATUS_TIMER},
-    {"status-automat",            no_argument,       0, IDX_STATUS_AUTOMAT},
+    {"machine-readable",          no_argument,       0, IDX_MACHINE_READABLE},
     {"loopback",                  no_argument,       0, IDX_LOOPBACK},
     {"weak-hash-threshold",       required_argument, 0, IDX_WEAK_HASH_THRESHOLD},
     {"session",                   required_argument, 0, IDX_SESSION},
@@ -5557,9 +5757,7 @@ int main (int argc, char **argv)
     #ifdef HAVE_HWMON
     {"gpu-temp-abort",            required_argument, 0, IDX_GPU_TEMP_ABORT},
     {"gpu-temp-retain",           required_argument, 0, IDX_GPU_TEMP_RETAIN},
-    #ifdef HAVE_ADL
     {"powertune-enable",          no_argument,       0, IDX_POWERTUNE_ENABLE},
-    #endif
     #endif // HAVE_HWMON
     {"logfile-disable",           no_argument,       0, IDX_LOGFILE_DISABLE},
     {"truecrypt-keyfiles",        required_argument, 0, IDX_TRUECRYPT_KEYFILES},
@@ -5783,11 +5981,6 @@ int main (int argc, char **argv)
   uint workload_profile_chgd    = 0;
   uint opencl_vector_width_chgd = 0;
 
-  #if defined(HAVE_HWMON) && defined(HAVE_ADL)
-  uint gpu_temp_retain_chgd   = 0;
-  uint gpu_temp_abort_chgd    = 0;
-  #endif
-
   optind = 1;
   optopt = 0;
   option_index = 0;
@@ -5823,7 +6016,7 @@ int main (int argc, char **argv)
       case IDX_RESTORE_DISABLE:           restore_disable           = 1;              break;
       case IDX_STATUS:                    status                    = 1;              break;
       case IDX_STATUS_TIMER:              status_timer              = atoi (optarg);  break;
-      case IDX_STATUS_AUTOMAT:            status_automat            = 1;              break;
+      case IDX_MACHINE_READABLE:          machine_readable          = 1;              break;
       case IDX_LOOPBACK:                  loopback                  = 1;              break;
       case IDX_WEAK_HASH_THRESHOLD:       weak_hash_threshold       = atoi (optarg);  break;
     //case IDX_SESSION:                   session                   = optarg;         break;
@@ -5868,19 +6061,9 @@ int main (int argc, char **argv)
                                           kernel_loops_chgd         = 1;              break;
       case IDX_GPU_TEMP_DISABLE:          gpu_temp_disable          = 1;              break;
       #ifdef HAVE_HWMON
-      case IDX_GPU_TEMP_ABORT:            gpu_temp_abort            = atoi (optarg);
-      #ifdef HAVE_ADL
-                                          gpu_temp_abort_chgd       = 1;
-      #endif
-                                                                                      break;
-      case IDX_GPU_TEMP_RETAIN:           gpu_temp_retain           = atoi (optarg);
-      #ifdef HAVE_ADL
-                                          gpu_temp_retain_chgd      = 1;
-      #endif
-                                                                                      break;
-      #ifdef HAVE_ADL
+      case IDX_GPU_TEMP_ABORT:            gpu_temp_abort            = atoi (optarg);  break;
+      case IDX_GPU_TEMP_RETAIN:           gpu_temp_retain           = atoi (optarg);  break;
       case IDX_POWERTUNE_ENABLE:          powertune_enable          = 1;              break;
-      #endif
       #endif // HAVE_HWMON
       case IDX_LOGFILE_DISABLE:           logfile_disable           = 1;              break;
       case IDX_TRUECRYPT_KEYFILES:        truecrypt_keyfiles        = optarg;         break;
@@ -5924,10 +6107,13 @@ int main (int argc, char **argv)
   {
     if (benchmark == 1)
     {
-      if (status_automat == 0)
+      if (machine_readable == 0)
       {
         log_info ("%s (%s) starting in benchmark-mode...", PROGNAME, VERSION_TAG);
         log_info ("");
+        log_info ("Note: Reported benchmark cracking speed = real cracking speed");
+        log_info ("To verify, run hashcat like this: only_one_hash.txt -a 3 -w 3 ?b?b?b?b?b?b?b");
+        log_info ("");
       }
       else
       {
@@ -5964,7 +6150,7 @@ int main (int argc, char **argv)
     return (-1);
   }
 
-  if (hash_mode_chgd && hash_mode > 13799) // just added to remove compiler warnings for hash_mode_chgd
+  if (hash_mode_chgd && hash_mode > 13800) // just added to remove compiler warnings for hash_mode_chgd
   {
     log_error ("ERROR: Invalid hash-type specified");
 
@@ -6554,47 +6740,45 @@ int main (int argc, char **argv)
    * store stuff
    */
 
-  data.hash_mode          = hash_mode;
-  data.restore            = restore;
-  data.restore_timer      = restore_timer;
-  data.restore_disable    = restore_disable;
-  data.status             = status;
-  data.status_timer       = status_timer;
-  data.status_automat     = status_automat;
-  data.loopback           = loopback;
-  data.runtime            = runtime;
-  data.remove             = remove;
-  data.remove_timer       = remove_timer;
-  data.debug_mode         = debug_mode;
-  data.debug_file         = debug_file;
-  data.username           = username;
-  data.quiet              = quiet;
-  data.outfile            = outfile;
-  data.outfile_format     = outfile_format;
-  data.outfile_autohex    = outfile_autohex;
-  data.hex_charset        = hex_charset;
-  data.hex_salt           = hex_salt;
-  data.hex_wordlist       = hex_wordlist;
-  data.separator          = separator;
-  data.rp_files           = rp_files;
-  data.rp_files_cnt       = rp_files_cnt;
-  data.rp_gen             = rp_gen;
-  data.rp_gen_seed        = rp_gen_seed;
-  data.force              = force;
-  data.benchmark          = benchmark;
-  data.skip               = skip;
-  data.limit              = limit;
+  data.hash_mode               = hash_mode;
+  data.restore                 = restore;
+  data.restore_timer           = restore_timer;
+  data.restore_disable         = restore_disable;
+  data.status                  = status;
+  data.status_timer            = status_timer;
+  data.machine_readable        = machine_readable;
+  data.loopback                = loopback;
+  data.runtime                 = runtime;
+  data.remove                  = remove;
+  data.remove_timer            = remove_timer;
+  data.debug_mode              = debug_mode;
+  data.debug_file              = debug_file;
+  data.username                = username;
+  data.quiet                   = quiet;
+  data.outfile                 = outfile;
+  data.outfile_format          = outfile_format;
+  data.outfile_autohex         = outfile_autohex;
+  data.hex_charset             = hex_charset;
+  data.hex_salt                = hex_salt;
+  data.hex_wordlist            = hex_wordlist;
+  data.separator               = separator;
+  data.rp_files                = rp_files;
+  data.rp_files_cnt            = rp_files_cnt;
+  data.rp_gen                  = rp_gen;
+  data.rp_gen_seed             = rp_gen_seed;
+  data.force                   = force;
+  data.benchmark               = benchmark;
+  data.skip                    = skip;
+  data.limit                   = limit;
   #ifdef HAVE_HWMON
-  #ifdef HAVE_ADL
-  data.powertune_enable   = powertune_enable;
-  #endif
+  data.powertune_enable        = powertune_enable;
   #endif
-  data.logfile_disable    = logfile_disable;
-  data.truecrypt_keyfiles = truecrypt_keyfiles;
-  data.veracrypt_keyfiles = veracrypt_keyfiles;
-  data.veracrypt_pim      = veracrypt_pim;
-  data.scrypt_tmto        = scrypt_tmto;
-  data.workload_profile   = workload_profile;
+  data.logfile_disable         = logfile_disable;
+  data.truecrypt_keyfiles      = truecrypt_keyfiles;
+  data.veracrypt_keyfiles      = veracrypt_keyfiles;
+  data.veracrypt_pim           = veracrypt_pim;
+  data.scrypt_tmto             = scrypt_tmto;
+  data.workload_profile        = workload_profile;
 
   /**
    * cpu affinity
@@ -6690,7 +6874,7 @@ int main (int argc, char **argv)
   logfile_top_uint   (outfile_format);
   logfile_top_uint   (potfile_disable);
   logfile_top_string (potfile_path);
-  #if defined(HAVE_HWMON) && defined(HAVE_ADL)
+  #if defined(HAVE_HWMON)
   logfile_top_uint   (powertune_enable);
   #endif
   logfile_top_uint   (scrypt_tmto);
@@ -6708,7 +6892,7 @@ int main (int argc, char **argv)
   logfile_top_uint   (segment_size);
   logfile_top_uint   (show);
   logfile_top_uint   (status);
-  logfile_top_uint   (status_automat);
+  logfile_top_uint   (machine_readable);
   logfile_top_uint   (status_timer);
   logfile_top_uint   (usage);
   logfile_top_uint   (username);
@@ -6787,6 +6971,13 @@ int main (int argc, char **argv)
     weak_hash_threshold   = 0;
     gpu_temp_disable      = 1;
 
+    #ifdef HAVE_HWMON
+    if (powertune_enable == 1)
+    {
+      gpu_temp_disable = 0;
+    }
+    #endif
+
     data.status_timer     = status_timer;
     data.restore_timer    = restore_timer;
     data.restore_disable  = restore_disable;
@@ -10635,6 +10826,26 @@ int main (int argc, char **argv)
                    dgst_pos3   = 3;
                    break;
 
+      case 13800:  hash_type   = HASH_TYPE_SHA256;
+                   salt_type   = SALT_TYPE_EMBEDDED;
+                   attack_exec = ATTACK_EXEC_INSIDE_KERNEL;
+                   opts_type   = OPTS_TYPE_PT_GENERATE_BE
+                               | OPTS_TYPE_PT_UNICODE;
+                   kern_type   = KERN_TYPE_WIN8PHONE;
+                   dgst_size   = DGST_SIZE_4_8;
+                   parse_func  = win8phone_parse_hash;
+                   sort_by_digest = sort_by_digest_4_8;
+                   opti_type   = OPTI_TYPE_ZERO_BYTE
+                               | OPTI_TYPE_PRECOMPUTE_INIT
+                               | OPTI_TYPE_EARLY_SKIP
+                               | OPTI_TYPE_NOT_ITERATED
+                               | OPTI_TYPE_RAW_HASH;
+                   dgst_pos0   = 3;
+                   dgst_pos1   = 7;
+                   dgst_pos2   = 2;
+                   dgst_pos3   = 6;
+                   break;
+
 
       default:     usage_mini_print (PROGNAME); return (-1);
     }
@@ -10761,6 +10972,7 @@ int main (int argc, char **argv)
       case 13761:  esalt_size = sizeof (tc_t);            break;
       case 13762:  esalt_size = sizeof (tc_t);            break;
       case 13763:  esalt_size = sizeof (tc_t);            break;
+      case 13800:  esalt_size = sizeof (win8phone_t);     break;
     }
 
     data.esalt_size = esalt_size;
@@ -11094,6 +11306,8 @@ int main (int argc, char **argv)
                   break;
       case  7400: if (pw_max > 16) pw_max = 16;
                   break;
+      case  7700: if (pw_max >  8) pw_max =  8;
+                  break;
       case  7900: if (pw_max > 48) pw_max = 48;
                   break;
       case  8500: if (pw_max >  8) pw_max =  8;
@@ -13115,32 +13329,6 @@ int main (int argc, char **argv)
       }
     }
 
-    /**
-     * OpenCL platforms: For each platform check if we need to unset features that we can not use, eg: temp_retain
-     */
-
-    for (uint platform_id = 0; platform_id < platforms_cnt; platform_id++)
-    {
-      cl_platform_id platform = platforms[platform_id];
-
-      char platform_vendor[INFOSZ] = { 0 };
-
-      hc_clGetPlatformInfo (data.ocl, platform, CL_PLATFORM_VENDOR, sizeof (platform_vendor), platform_vendor, NULL);
-
-      #ifdef HAVE_HWMON
-      #if defined(HAVE_NVML) || defined(HAVE_NVAPI)
-      if (strcmp (platform_vendor, CL_VENDOR_NV) == 0)
-      {
-        // make sure that we do not directly control the fan for NVidia
-
-        gpu_temp_retain = 0;
-
-        data.gpu_temp_retain = gpu_temp_retain;
-      }
-      #endif // HAVE_NVML || HAVE_NVAPI
-      #endif
-    }
-
     /**
      * OpenCL device types:
      *   In case the user did not specify --opencl-device-types and the user runs hashcat in a system with only a CPU only he probably want to use that CPU.
@@ -13181,6 +13369,10 @@ int main (int argc, char **argv)
      * OpenCL devices: simply push all devices from all platforms into the same device array
      */
 
+    int need_adl   = 0;
+    int need_nvapi = 0;
+    int need_nvml  = 0;
+
     hc_device_param_t *devices_param = (hc_device_param_t *) mycalloc (DEVICES_MAX, sizeof (hc_device_param_t));
 
     data.devices_param = devices_param;
@@ -13205,39 +13397,43 @@ int main (int argc, char **argv)
       // this causes trouble with vendor id based macros
       // we'll assign generic to those without special optimization available
 
-      cl_uint vendor_id = 0;
+      cl_uint platform_vendor_id = 0;
 
       if (strcmp (platform_vendor, CL_VENDOR_AMD) == 0)
       {
-        vendor_id = VENDOR_ID_AMD;
+        platform_vendor_id = VENDOR_ID_AMD;
+      }
+      else if (strcmp (platform_vendor, CL_VENDOR_AMD_USE_INTEL) == 0)
+      {
+        platform_vendor_id = VENDOR_ID_AMD_USE_INTEL;
       }
       else if (strcmp (platform_vendor, CL_VENDOR_APPLE) == 0)
       {
-        vendor_id = VENDOR_ID_APPLE;
+        platform_vendor_id = VENDOR_ID_APPLE;
       }
       else if (strcmp (platform_vendor, CL_VENDOR_INTEL_BEIGNET) == 0)
       {
-        vendor_id = VENDOR_ID_INTEL_BEIGNET;
+        platform_vendor_id = VENDOR_ID_INTEL_BEIGNET;
       }
       else if (strcmp (platform_vendor, CL_VENDOR_INTEL_SDK) == 0)
       {
-        vendor_id = VENDOR_ID_INTEL_SDK;
+        platform_vendor_id = VENDOR_ID_INTEL_SDK;
       }
       else if (strcmp (platform_vendor, CL_VENDOR_MESA) == 0)
       {
-        vendor_id = VENDOR_ID_MESA;
+        platform_vendor_id = VENDOR_ID_MESA;
       }
       else if (strcmp (platform_vendor, CL_VENDOR_NV) == 0)
       {
-        vendor_id = VENDOR_ID_NV;
+        platform_vendor_id = VENDOR_ID_NV;
       }
       else if (strcmp (platform_vendor, CL_VENDOR_POCL) == 0)
       {
-        vendor_id = VENDOR_ID_POCL;
+        platform_vendor_id = VENDOR_ID_POCL;
       }
       else
       {
-        vendor_id = VENDOR_ID_GENERIC;
+        platform_vendor_id = VENDOR_ID_GENERIC;
       }
 
       for (uint platform_devices_id = 0; platform_devices_id < platform_devices_cnt; platform_devices_id++)
@@ -13248,7 +13444,7 @@ int main (int argc, char **argv)
 
         hc_device_param_t *device_param = &data.devices_param[device_id];
 
-        device_param->vendor_id = vendor_id;
+        device_param->platform_vendor_id = platform_vendor_id;
 
         device_param->device = platform_devices[platform_devices_id];
 
@@ -13276,6 +13472,57 @@ int main (int argc, char **argv)
 
         device_param->device_name = device_name;
 
+        // device_vendor
+
+        hc_clGetDeviceInfo (data.ocl, device_param->device, CL_DEVICE_VENDOR, 0, NULL, &param_value_size);
+
+        char *device_vendor = (char *) mymalloc (param_value_size);
+
+        hc_clGetDeviceInfo (data.ocl, device_param->device, CL_DEVICE_VENDOR, param_value_size, device_vendor, NULL);
+
+        device_param->device_vendor = device_vendor;
+
+        cl_uint device_vendor_id = 0;
+
+        if (strcmp (device_vendor, CL_VENDOR_AMD) == 0)
+        {
+          device_vendor_id = VENDOR_ID_AMD;
+        }
+        else if (strcmp (device_vendor, CL_VENDOR_AMD_USE_INTEL) == 0)
+        {
+          device_vendor_id = VENDOR_ID_AMD_USE_INTEL;
+        }
+        else if (strcmp (device_vendor, CL_VENDOR_APPLE) == 0)
+        {
+          device_vendor_id = VENDOR_ID_APPLE;
+        }
+        else if (strcmp (device_vendor, CL_VENDOR_INTEL_BEIGNET) == 0)
+        {
+          device_vendor_id = VENDOR_ID_INTEL_BEIGNET;
+        }
+        else if (strcmp (device_vendor, CL_VENDOR_INTEL_SDK) == 0)
+        {
+          device_vendor_id = VENDOR_ID_INTEL_SDK;
+        }
+        else if (strcmp (device_vendor, CL_VENDOR_MESA) == 0)
+        {
+          device_vendor_id = VENDOR_ID_MESA;
+        }
+        else if (strcmp (device_vendor, CL_VENDOR_NV) == 0)
+        {
+          device_vendor_id = VENDOR_ID_NV;
+        }
+        else if (strcmp (device_vendor, CL_VENDOR_POCL) == 0)
+        {
+          device_vendor_id = VENDOR_ID_POCL;
+        }
+        else
+        {
+          device_vendor_id = VENDOR_ID_GENERIC;
+        }
+
+        device_param->device_vendor_id = device_vendor_id;
+
         // tuning db
 
         tuning_db_entry_t *tuningdb_entry = tuning_db_search (tuning_db, device_param, attack_mode, hash_mode);
@@ -13465,6 +13712,28 @@ int main (int argc, char **argv)
           device_param->skipped = 1;
         }
 
+        // If there's both an Intel CPU and an AMD OpenCL runtime it's a tricky situation
+        // Both platforms support CPU device types and therefore both will try to use 100% of the physical resources
+        // This results in both utilizing it for 50%
+        // However, Intel has much better SIMD control over their own hardware
+        // It makes sense to give them full control over their own hardware
+
+        if (device_type & CL_DEVICE_TYPE_CPU)
+        {
+          if (device_param->device_vendor_id == VENDOR_ID_AMD_USE_INTEL)
+          {
+            if (data.force == 0)
+            {
+              if (algorithm_pos == 0)
+              {
+                log_info ("Device #%u: WARNING: not native intel opencl runtime, expect massive speed loss", device_id + 1);
+                log_info ("           You can use --force to override this but do not post error reports if you do so");
+              }
+
+              device_param->skipped = 1;
+            }
+          }
+        }
 
         // skipped
 
@@ -13486,9 +13755,9 @@ int main (int argc, char **argv)
         char *device_name_chksum = (char *) mymalloc (INFOSZ);
 
         #if __x86_64__
-        snprintf (device_name_chksum, INFOSZ - 1, "%u-%u-%u-%s-%s-%s-%u", 64, device_param->vendor_id, device_param->vector_width, device_param->device_name, device_param->device_version, device_param->driver_version, COMPTIME);
+        snprintf (device_name_chksum, INFOSZ - 1, "%u-%u-%u-%s-%s-%s-%u", 64, device_param->platform_vendor_id, device_param->vector_width, device_param->device_name, device_param->device_version, device_param->driver_version, COMPTIME);
         #else
-        snprintf (device_name_chksum, INFOSZ - 1, "%u-%u-%u-%s-%s-%s-%u", 32, device_param->vendor_id, device_param->vector_width, device_param->device_name, device_param->device_version, device_param->driver_version, COMPTIME);
+        snprintf (device_name_chksum, INFOSZ - 1, "%u-%u-%u-%s-%s-%s-%u", 32, device_param->platform_vendor_id, device_param->vector_width, device_param->device_name, device_param->device_version, device_param->driver_version, COMPTIME);
         #endif
 
         uint device_name_digest[4] = { 0 };
@@ -13501,6 +13770,25 @@ int main (int argc, char **argv)
 
         // device_processor_cores
 
+        if (device_param->device_type & CL_DEVICE_TYPE_GPU)
+        {
+          if ((device_param->platform_vendor_id == VENDOR_ID_AMD) && (device_param->device_vendor_id == VENDOR_ID_AMD))
+          {
+            need_adl = 1;
+          }
+
+          if ((device_param->platform_vendor_id == VENDOR_ID_NV) && (device_param->device_vendor_id == VENDOR_ID_NV))
+          {
+            need_nvml = 1;
+
+            #ifdef _WIN
+            need_nvapi = 1;
+            #endif
+          }
+        }
+
+        // device_processor_cores
+
         if (device_type & CL_DEVICE_TYPE_CPU)
         {
           cl_uint device_processor_cores = 1;
@@ -13510,7 +13798,7 @@ int main (int argc, char **argv)
 
         if (device_type & CL_DEVICE_TYPE_GPU)
         {
-          if (vendor_id == VENDOR_ID_AMD)
+          if (device_vendor_id == VENDOR_ID_AMD)
           {
             cl_uint device_processor_cores = 0;
 
@@ -13520,7 +13808,7 @@ int main (int argc, char **argv)
 
             device_param->device_processor_cores = device_processor_cores;
           }
-          else if (vendor_id == VENDOR_ID_NV)
+          else if (device_vendor_id == VENDOR_ID_NV)
           {
             cl_uint kernel_exec_timeout = 0;
 
@@ -13562,16 +13850,15 @@ int main (int argc, char **argv)
 
         if ((benchmark == 1 || quiet == 0) && (algorithm_pos == 0))
         {
-          if (status_automat == 0)
+          if (machine_readable == 0)
           {
             if (device_param->skipped == 0)
             {
-              log_info ("Device #%u: %s, %lu/%lu MB allocatable, %dMhz, %uMCU",
+              log_info ("Device #%u: %s, %lu/%lu MB allocatable, %uMCU",
                         device_id + 1,
                         device_name,
                         (unsigned int) (device_maxmem_alloc / 1024 / 1024),
                         (unsigned int) (device_global_mem   / 1024 / 1024),
-                        (unsigned int) (device_maxclock_frequency),
                         (unsigned int)  device_processors);
             }
             else
@@ -13589,7 +13876,7 @@ int main (int argc, char **argv)
         {
           if (device_type & CL_DEVICE_TYPE_GPU)
           {
-            if (vendor_id == VENDOR_ID_AMD)
+            if (platform_vendor_id == VENDOR_ID_AMD)
             {
               int catalyst_check = (force == 1) ? 0 : 1;
 
@@ -13637,7 +13924,7 @@ int main (int argc, char **argv)
                 return (-1);
               }
             }
-            else if (vendor_id == VENDOR_ID_NV)
+            else if (platform_vendor_id == VENDOR_ID_NV)
             {
               if (device_param->kernel_exec_timeout != 0)
               {
@@ -13647,9 +13934,10 @@ int main (int argc, char **argv)
             }
           }
 
+          /* turns out pocl still creates segfaults (because of llvm)
           if (device_type & CL_DEVICE_TYPE_CPU)
           {
-            if (vendor_id == VENDOR_ID_AMD)
+            if (platform_vendor_id == VENDOR_ID_AMD)
             {
               if (force == 0)
               {
@@ -13664,6 +13952,7 @@ int main (int argc, char **argv)
               }
             }
           }
+          */
 
           /**
            * kernel accel and loops tuning db adjustment
@@ -13758,7 +14047,7 @@ int main (int argc, char **argv)
 
     if ((benchmark == 1 || quiet == 0) && (algorithm_pos == 0))
     {
-      if (status_automat == 0)
+      if (machine_readable == 0)
       {
         log_info ("");
       }
@@ -13769,101 +14058,93 @@ int main (int argc, char **argv)
      */
 
     #ifdef HAVE_HWMON
-    #if defined(HAVE_NVML) || defined(HAVE_NVAPI)
-    hm_attrs_t hm_adapters_nv[DEVICES_MAX]  = { { { 0 }, 0, 0 } };
-    #endif
-
-    #ifdef HAVE_ADL
-    hm_attrs_t hm_adapters_amd[DEVICES_MAX] = { { { 0 }, 0, 0 } };
-    #endif
+    hm_attrs_t hm_adapters_adl[DEVICES_MAX]   = { { 0 } };
+    hm_attrs_t hm_adapters_nvapi[DEVICES_MAX] = { { 0 } };
+    hm_attrs_t hm_adapters_nvml[DEVICES_MAX]  = { { 0 } };
 
     if (gpu_temp_disable == 0)
     {
-      #if defined(WIN) && defined(HAVE_NVAPI)
+      ADL_PTR   *adl   = (ADL_PTR *)   mymalloc (sizeof (ADL_PTR));
       NVAPI_PTR *nvapi = (NVAPI_PTR *) mymalloc (sizeof (NVAPI_PTR));
+      NVML_PTR  *nvml  = (NVML_PTR *)  mymalloc (sizeof (NVML_PTR));
+
+      data.hm_adl   = NULL;
+      data.hm_nvapi = NULL;
+      data.hm_nvml  = NULL;
 
-      if (nvapi_init (nvapi) == 0)
-        data.hm_nv = nvapi;
+      if ((need_nvml == 1) && (nvml_init (nvml) == 0))
+      {
+        data.hm_nvml = nvml;
+      }
 
-      if (data.hm_nv)
+      if (data.hm_nvml)
       {
-        if (hm_NvAPI_Initialize (data.hm_nv) == NVAPI_OK)
+        if (hm_NVML_nvmlInit (data.hm_nvml) == NVML_SUCCESS)
         {
-          HM_ADAPTER_NV nvGPUHandle[DEVICES_MAX] = { 0 };
+          HM_ADAPTER_NVML nvmlGPUHandle[DEVICES_MAX] = { 0 };
 
-          int tmp_in = hm_get_adapter_index_nv (nvGPUHandle);
+          int tmp_in = hm_get_adapter_index_nvml (nvmlGPUHandle);
 
           int tmp_out = 0;
 
           for (int i = 0; i < tmp_in; i++)
           {
-            hm_adapters_nv[tmp_out++].adapter_index.nv = nvGPUHandle[i];
+            hm_adapters_nvml[tmp_out++].nvml = nvmlGPUHandle[i];
           }
 
           for (int i = 0; i < tmp_out; i++)
           {
-            NV_GPU_COOLER_SETTINGS pCoolerSettings;
+            unsigned int speed;
+
+            if (hm_NVML_nvmlDeviceGetFanSpeed (data.hm_nvml, 0, hm_adapters_nvml[i].nvml, &speed) == NVML_SUCCESS) hm_adapters_nvml[i].fan_get_supported = 1;
 
-            pCoolerSettings.Version = GPU_COOLER_SETTINGS_VER | sizeof (NV_GPU_COOLER_SETTINGS);
+            hm_NVML_nvmlDeviceSetComputeMode (data.hm_nvml, 1, hm_adapters_nvml[i].nvml, NVML_COMPUTEMODE_EXCLUSIVE_PROCESS);
 
-            if (hm_NvAPI_GPU_GetCoolerSettings (data.hm_nv, hm_adapters_nv[i].adapter_index.nv, 0, &pCoolerSettings) != NVAPI_NOT_SUPPORTED) hm_adapters_nv[i].fan_supported = 1;
+            hm_NVML_nvmlDeviceSetGpuOperationMode (data.hm_nvml, 1, hm_adapters_nvml[i].nvml, NVML_GOM_ALL_ON);
           }
         }
       }
-      #endif // WIN && HAVE_NVAPI
 
-      #if defined(LINUX) && defined(HAVE_NVML)
-      NVML_PTR *nvml = (NVML_PTR *) mymalloc (sizeof (NVML_PTR));
-
-      if (nvml_init (nvml) == 0)
-        data.hm_nv = nvml;
+      if ((need_nvapi == 1) && (nvapi_init (nvapi) == 0))
+      {
+        data.hm_nvapi = nvapi;
+      }
 
-      if (data.hm_nv)
+      if (data.hm_nvapi)
       {
-        if (hm_NVML_nvmlInit (data.hm_nv) == NVML_SUCCESS)
+        if (hm_NvAPI_Initialize (data.hm_nvapi) == NVAPI_OK)
         {
-          HM_ADAPTER_NV nvGPUHandle[DEVICES_MAX] = { 0 };
+          HM_ADAPTER_NVAPI nvGPUHandle[DEVICES_MAX] = { 0 };
 
-          int tmp_in = hm_get_adapter_index_nv (nvGPUHandle);
+          int tmp_in = hm_get_adapter_index_nvapi (nvGPUHandle);
 
           int tmp_out = 0;
 
           for (int i = 0; i < tmp_in; i++)
           {
-            hm_adapters_nv[tmp_out++].adapter_index.nv = nvGPUHandle[i];
-          }
-
-          for (int i = 0; i < tmp_out; i++)
-          {
-            unsigned int speed;
-
-            if (hm_NVML_nvmlDeviceGetFanSpeed (data.hm_nv, 1, hm_adapters_nv[i].adapter_index.nv, &speed) != NVML_ERROR_NOT_SUPPORTED) hm_adapters_nv[i].fan_supported = 1;
+            hm_adapters_nvapi[tmp_out++].nvapi = nvGPUHandle[i];
           }
         }
       }
-      #endif // LINUX && HAVE_NVML
-
-      data.hm_amd = NULL;
 
-      #ifdef HAVE_ADL
-      ADL_PTR *adl = (ADL_PTR *) mymalloc (sizeof (ADL_PTR));
-
-      if (adl_init (adl) == 0)
-        data.hm_amd = adl;
+      if ((need_adl == 1) && (adl_init (adl) == 0))
+      {
+        data.hm_adl = adl;
+      }
 
-      if (data.hm_amd)
+      if (data.hm_adl)
       {
-        if (hm_ADL_Main_Control_Create (data.hm_amd, ADL_Main_Memory_Alloc, 0) == ADL_OK)
+        if (hm_ADL_Main_Control_Create (data.hm_adl, ADL_Main_Memory_Alloc, 0) == ADL_OK)
         {
           // total number of adapters
 
           int hm_adapters_num;
 
-          if (get_adapters_num_amd (data.hm_amd, &hm_adapters_num) != 0) return (-1);
+          if (get_adapters_num_adl (data.hm_adl, &hm_adapters_num) != 0) return (-1);
 
           // adapter info
 
-          LPAdapterInfo lpAdapterInfo = hm_get_adapter_info_amd (data.hm_amd, hm_adapters_num);
+          LPAdapterInfo lpAdapterInfo = hm_get_adapter_info_adl (data.hm_adl, hm_adapters_num);
 
           if (lpAdapterInfo == NULL) return (-1);
 
@@ -13877,12 +14158,12 @@ int main (int argc, char **argv)
           {
             hc_thread_mutex_lock (mux_adl);
 
-            // hm_get_opencl_busid_devid (hm_adapters_amd, devices_all_cnt, devices_all);
+            // hm_get_opencl_busid_devid (hm_adapters_adl, devices_all_cnt, devices_all);
 
-            hm_get_adapter_index_amd (hm_adapters_amd, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
+            hm_get_adapter_index_adl (hm_adapters_adl, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
 
-            hm_get_overdrive_version  (data.hm_amd, hm_adapters_amd, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
-            hm_check_fanspeed_control (data.hm_amd, hm_adapters_amd, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
+            hm_get_overdrive_version  (data.hm_adl, hm_adapters_adl, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
+            hm_check_fanspeed_control (data.hm_adl, hm_adapters_adl, valid_adl_device_list, num_adl_adapters, lpAdapterInfo);
 
             hc_thread_mutex_unlock (mux_adl);
           }
@@ -13891,9 +14172,8 @@ int main (int argc, char **argv)
           myfree (lpAdapterInfo);
         }
       }
-      #endif // HAVE_ADL
 
-      if (data.hm_amd == NULL && data.hm_nv == NULL)
+      if (data.hm_adl == NULL && data.hm_nvml == NULL)
       {
         gpu_temp_disable = 1;
       }
@@ -13903,34 +14183,19 @@ int main (int argc, char **argv)
      * OpenCL devices: allocate buffer for device specific information
      */
 
-    #ifdef HAVE_HWMON
-    int *temp_retain_fanspeed_value = (int *) mycalloc (data.devices_cnt, sizeof (int));
+    int *temp_retain_fanspeed_value  = (int *) mycalloc (data.devices_cnt, sizeof (int));
+    int *temp_retain_fanpolicy_value = (int *) mycalloc (data.devices_cnt, sizeof (int));
 
-    #ifdef HAVE_ADL
     ADLOD6MemClockState *od_clock_mem_status = (ADLOD6MemClockState *) mycalloc (data.devices_cnt, sizeof (ADLOD6MemClockState));
 
     int *od_power_control_status = (int *) mycalloc (data.devices_cnt, sizeof (int));
-    #endif // ADL
-    #endif
-
-    /**
-     * enable custom signal handler(s)
-     */
 
-    if (benchmark == 0)
-    {
-      hc_signal (sigHandler_default);
-    }
-    else
-    {
-      hc_signal (sigHandler_benchmark);
-    }
+    unsigned int *nvml_power_limit = (unsigned int *) mycalloc (data.devices_cnt, sizeof (unsigned int));
 
     /**
      * User-defined GPU temp handling
      */
 
-    #ifdef HAVE_HWMON
     if (gpu_temp_disable == 1)
     {
       gpu_temp_abort  = 0;
@@ -13952,6 +14217,19 @@ int main (int argc, char **argv)
     data.gpu_temp_retain  = gpu_temp_retain;
     #endif
 
+    /**
+     * enable custom signal handler(s)
+     */
+
+    if (benchmark == 0)
+    {
+      hc_signal (sigHandler_default);
+    }
+    else
+    {
+      hc_signal (sigHandler_benchmark);
+    }
+
     /**
      * inform the user
      */
@@ -13984,7 +14262,7 @@ int main (int argc, char **argv)
        */
 
       #ifdef HAVE_HWMON
-      if (gpu_temp_disable == 0 && data.hm_amd == NULL && data.hm_nv == NULL)
+      if (gpu_temp_disable == 0 && data.hm_adl == NULL && data.hm_nvml == NULL)
       {
         log_info ("Watchdog: Hardware Monitoring Interface not found on your system");
       }
@@ -14011,6 +14289,8 @@ int main (int argc, char **argv)
       #endif
     }
 
+    #ifdef HAVE_HWMON
+
     /**
      * HM devices: copy
      */
@@ -14027,31 +14307,32 @@ int main (int argc, char **argv)
 
         const uint platform_devices_id = device_param->platform_devices_id;
 
-        #if defined(HAVE_NVML) || defined(HAVE_NVAPI)
-        if (device_param->vendor_id == VENDOR_ID_NV)
+        if (device_param->device_vendor_id == VENDOR_ID_AMD)
         {
-          memcpy (&data.hm_device[device_id], &hm_adapters_nv[platform_devices_id], sizeof (hm_attrs_t));
+          data.hm_device[device_id].adl               = hm_adapters_adl[platform_devices_id].adl;
+          data.hm_device[device_id].nvapi             = 0;
+          data.hm_device[device_id].nvml              = 0;
+          data.hm_device[device_id].od_version        = hm_adapters_adl[platform_devices_id].od_version;
+          data.hm_device[device_id].fan_get_supported = hm_adapters_adl[platform_devices_id].fan_get_supported;
+          data.hm_device[device_id].fan_set_supported = hm_adapters_adl[platform_devices_id].fan_set_supported;
         }
-        #endif
 
-        #ifdef HAVE_ADL
-        if (device_param->vendor_id == VENDOR_ID_AMD)
+        if (device_param->device_vendor_id == VENDOR_ID_NV)
         {
-          memcpy (&data.hm_device[device_id], &hm_adapters_amd[platform_devices_id], sizeof (hm_attrs_t));
+          data.hm_device[device_id].adl               = 0;
+          data.hm_device[device_id].nvapi             = hm_adapters_nvapi[platform_devices_id].nvapi;
+          data.hm_device[device_id].nvml              = hm_adapters_nvml[platform_devices_id].nvml;
+          data.hm_device[device_id].od_version        = 0;
+          data.hm_device[device_id].fan_get_supported = hm_adapters_nvml[platform_devices_id].fan_get_supported;
+          data.hm_device[device_id].fan_set_supported = 0;
         }
-        #endif
       }
     }
 
-   /*
-    * Temporary fix:
-    * with AMD r9 295x cards it seems that we need to set the powertune value just AFTER the ocl init stuff
-    * otherwise after hc_clCreateContext () etc, powertune value was set back to "normal" and cards unfortunately
-    * were not working @ full speed (setting hm_ADL_Overdrive_PowerControl_Set () here seems to fix the problem)
-    * Driver / ADL bug?
-    */
+    /**
+     * powertune on user request
+     */
 
-    #ifdef HAVE_ADL
     if (powertune_enable == 1)
     {
       hc_thread_mutex_lock (mux_adl);
@@ -14062,83 +14343,220 @@ int main (int argc, char **argv)
 
         if (device_param->skipped) continue;
 
-        if (data.hm_device[device_id].od_version == 6)
+        if (data.devices_param[device_id].device_vendor_id == VENDOR_ID_AMD)
         {
-          // set powertune value only
-
-          int powertune_supported = 0;
-
-          int ADL_rc = 0;
+          /**
+           * Temporary fix:
+           * with AMD r9 295x cards it seems that we need to set the powertune value just AFTER the ocl init stuff
+           * otherwise after hc_clCreateContext () etc, powertune value was set back to "normal" and cards unfortunately
+           * were not working @ full speed (setting hm_ADL_Overdrive_PowerControl_Set () here seems to fix the problem)
+           * Driver / ADL bug?
+           */
 
-          if ((ADL_rc = hm_ADL_Overdrive6_PowerControl_Caps (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &powertune_supported)) != ADL_OK)
+          if (data.hm_device[device_id].od_version == 6)
           {
-            log_error ("ERROR: Failed to get ADL PowerControl Capabilities");
+            int ADL_rc;
 
-            return (-1);
-          }
+            // check powertune capabilities first, if not available then skip device
 
-          if (powertune_supported != 0)
-          {
-            // powertune set
-            ADLOD6PowerControlInfo powertune = {0, 0, 0, 0, 0};
+            int powertune_supported = 0;
 
-            if ((ADL_rc = hm_ADL_Overdrive_PowerControlInfo_Get (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &powertune)) != ADL_OK)
+            if ((ADL_rc = hm_ADL_Overdrive6_PowerControl_Caps (data.hm_adl, data.hm_device[device_id].adl, &powertune_supported)) != ADL_OK)
             {
-              log_error ("ERROR: Failed to get current ADL PowerControl settings");
+              log_error ("ERROR: Failed to get ADL PowerControl Capabilities");
 
               return (-1);
             }
 
-            if ((ADL_rc = hm_ADL_Overdrive_PowerControl_Set (data.hm_amd, data.hm_device[device_id].adapter_index.amd, powertune.iMaxValue)) != ADL_OK)
+            // first backup current value, we will restore it later
+
+            if (powertune_supported != 0)
             {
-              log_error ("ERROR: Failed to set new ADL PowerControl values");
+              // powercontrol settings
 
-              return (-1);
-            }
-          }
-        }
-      }
+              ADLOD6PowerControlInfo powertune = {0, 0, 0, 0, 0};
 
-      hc_thread_mutex_unlock (mux_adl);
-    }
-    #endif // HAVE_ADK
-    #endif // HAVE_HWMON
+              if ((ADL_rc = hm_ADL_Overdrive_PowerControlInfo_Get (data.hm_adl, data.hm_device[device_id].adl, &powertune)) == ADL_OK)
+              {
+                ADL_rc = hm_ADL_Overdrive_PowerControl_Get (data.hm_adl, data.hm_device[device_id].adl, &od_power_control_status[device_id]);
+              }
 
-    #ifdef DEBUG
-    if (benchmark == 1) log_info ("Hashmode: %d", data.hash_mode);
-    #endif
+              if (ADL_rc != ADL_OK)
+              {
+                log_error ("ERROR: Failed to get current ADL PowerControl settings");
 
-    if (data.quiet == 0) log_info_nn ("Initializing device kernels and memory...");
+                return (-1);
+              }
 
-    uint kernel_power_all = 0;
+              if ((ADL_rc = hm_ADL_Overdrive_PowerControl_Set (data.hm_adl, data.hm_device[device_id].adl, powertune.iMaxValue)) != ADL_OK)
+              {
+                log_error ("ERROR: Failed to set new ADL PowerControl values");
 
-    for (uint device_id = 0; device_id < data.devices_cnt; device_id++)
-    {
-      /**
-       * host buffer
-       */
+                return (-1);
+              }
 
-      hc_device_param_t *device_param = &data.devices_param[device_id];
+              // clocks
 
-      if (device_param->skipped) continue;
+              memset (&od_clock_mem_status[device_id], 0, sizeof (ADLOD6MemClockState));
 
-      /**
-       * device properties
-       */
+              od_clock_mem_status[device_id].state.iNumberOfPerformanceLevels = 2;
 
-      const char *device_name_chksum      = device_param->device_name_chksum;
-      const u32   device_processors       = device_param->device_processors;
-      const u32   device_processor_cores  = device_param->device_processor_cores;
+              if ((ADL_rc = hm_ADL_Overdrive_StateInfo_Get (data.hm_adl, data.hm_device[device_id].adl, ADL_OD6_GETSTATEINFO_CUSTOM_PERFORMANCE, &od_clock_mem_status[device_id])) != ADL_OK)
+              {
+                log_error ("ERROR: Failed to get ADL memory and engine clock frequency");
 
-      /**
-       * create context for each device
-       */
+                return (-1);
+              }
 
-      device_param->context = hc_clCreateContext (data.ocl, NULL, 1, &device_param->device, NULL, NULL);
+              // Query capabilities only to see if profiles were not "damaged", if so output a warning but do accept the users profile settings
 
-      /**
-       * create command-queue
-       */
+              ADLOD6Capabilities caps = {0, 0, 0, {0, 0, 0}, {0, 0, 0}, 0, 0};
+
+              if ((ADL_rc = hm_ADL_Overdrive_Capabilities_Get (data.hm_adl, data.hm_device[device_id].adl, &caps)) != ADL_OK)
+              {
+                log_error ("ERROR: Failed to get ADL device capabilities");
+
+                return (-1);
+              }
+
+              int engine_clock_max = caps.sEngineClockRange.iMax * 0.6666;
+              int memory_clock_max = caps.sMemoryClockRange.iMax * 0.6250;
+
+              int warning_trigger_engine = (int) (0.25 * (float) engine_clock_max);
+              int warning_trigger_memory = (int) (0.25 * (float) memory_clock_max);
+
+              int engine_clock_profile_max = od_clock_mem_status[device_id].state.aLevels[1].iEngineClock;
+              int memory_clock_profile_max = od_clock_mem_status[device_id].state.aLevels[1].iMemoryClock;
+
+              // warning if profile has too low max values
+
+              if ((engine_clock_max - engine_clock_profile_max) > warning_trigger_engine)
+              {
+                log_info ("WARN: the custom profile seems to have too low maximum engine clock values. You therefore may not reach full performance");
+              }
+
+              if ((memory_clock_max - memory_clock_profile_max) > warning_trigger_memory)
+              {
+                log_info ("WARN: the custom profile seems to have too low maximum memory clock values. You therefore may not reach full performance");
+              }
+
+              ADLOD6StateInfo *performance_state = (ADLOD6StateInfo*) mycalloc (1, sizeof (ADLOD6StateInfo) + sizeof (ADLOD6PerformanceLevel));
+
+              performance_state->iNumberOfPerformanceLevels = 2;
+
+              performance_state->aLevels[0].iEngineClock = engine_clock_profile_max;
+              performance_state->aLevels[1].iEngineClock = engine_clock_profile_max;
+              performance_state->aLevels[0].iMemoryClock = memory_clock_profile_max;
+              performance_state->aLevels[1].iMemoryClock = memory_clock_profile_max;
+
+              if ((ADL_rc = hm_ADL_Overdrive_State_Set (data.hm_adl, data.hm_device[device_id].adl, ADL_OD6_SETSTATE_PERFORMANCE, performance_state)) != ADL_OK)
+              {
+                log_info ("ERROR: Failed to set ADL performance state");
+
+                return (-1);
+              }
+
+              local_free (performance_state);
+            }
+
+            // set powertune value only
+
+            if (powertune_supported != 0)
+            {
+              // powertune set
+              ADLOD6PowerControlInfo powertune = {0, 0, 0, 0, 0};
+
+              if ((ADL_rc = hm_ADL_Overdrive_PowerControlInfo_Get (data.hm_adl, data.hm_device[device_id].adl, &powertune)) != ADL_OK)
+              {
+                log_error ("ERROR: Failed to get current ADL PowerControl settings");
+
+                return (-1);
+              }
+
+              if ((ADL_rc = hm_ADL_Overdrive_PowerControl_Set (data.hm_adl, data.hm_device[device_id].adl, powertune.iMaxValue)) != ADL_OK)
+              {
+                log_error ("ERROR: Failed to set new ADL PowerControl values");
+
+                return (-1);
+              }
+            }
+          }
+        }
+
+        if (data.devices_param[device_id].device_vendor_id == VENDOR_ID_NV)
+        {
+          // first backup current value, we will restore it later
+
+          unsigned int limit;
+
+          int powertune_supported = 0;
+
+          if (hm_NVML_nvmlDeviceGetPowerManagementLimit (data.hm_nvml, 0, data.hm_device[device_id].nvml, &limit) == NVML_SUCCESS)
+          {
+            powertune_supported = 1;
+          }
+
+          // if backup worked, activate the maximum allowed
+
+          if (powertune_supported != 0)
+          {
+            unsigned int minLimit;
+            unsigned int maxLimit;
+
+            if (hm_NVML_nvmlDeviceGetPowerManagementLimitConstraints (data.hm_nvml, 0, data.hm_device[device_id].nvml, &minLimit, &maxLimit) == NVML_SUCCESS)
+            {
+              if (maxLimit > 0)
+              {
+                if (hm_NVML_nvmlDeviceSetPowerManagementLimit (data.hm_nvml, 0, data.hm_device[device_id].nvml, maxLimit) == NVML_SUCCESS)
+                {
+                  // now we can be sure we need to reset later
+
+                  nvml_power_limit[device_id] = limit;
+                }
+              }
+            }
+          }
+        }
+      }
+
+      hc_thread_mutex_unlock (mux_adl);
+    }
+
+    #endif // HAVE_HWMON
+
+    #ifdef DEBUG
+    if (benchmark == 1) log_info ("Hashmode: %d", data.hash_mode);
+    #endif
+
+    if (data.quiet == 0) log_info_nn ("Initializing device kernels and memory...");
+
+    for (uint device_id = 0; device_id < data.devices_cnt; device_id++)
+    {
+      /**
+       * host buffer
+       */
+
+      hc_device_param_t *device_param = &data.devices_param[device_id];
+
+      if (device_param->skipped) continue;
+
+      /**
+       * device properties
+       */
+
+      const char *device_name_chksum      = device_param->device_name_chksum;
+      const u32   device_processors       = device_param->device_processors;
+      const u32   device_processor_cores  = device_param->device_processor_cores;
+
+      /**
+       * create context for each device
+       */
+
+      device_param->context = hc_clCreateContext (data.ocl, NULL, 1, &device_param->device, NULL, NULL);
+
+      /**
+       * create command-queue
+       */
 
       // not supported with NV
       // device_param->command_queue = hc_clCreateCommandQueueWithProperties (device_param->context, device_param->device, NULL);
@@ -14172,6 +14590,10 @@ int main (int argc, char **argv)
       if (hash_mode == 10500) kernel_threads = 64; // RC4
       if (hash_mode == 13100) kernel_threads = 64; // RC4
 
+      device_param->kernel_threads = kernel_threads;
+
+      device_param->hardware_power = device_processors * kernel_threads;
+
       /**
        * create input buffers on device : calculate size of fixed memory buffers
        */
@@ -14223,22 +14645,22 @@ int main (int argc, char **argv)
 
           if (hash_mode == 8900)
           {
-            if (device_param->vendor_id == VENDOR_ID_AMD)
+            if (device_param->device_vendor_id == VENDOR_ID_AMD)
             {
               tmto_start = 1;
             }
-            else if (device_param->vendor_id == VENDOR_ID_NV)
+            else if (device_param->device_vendor_id == VENDOR_ID_NV)
             {
               tmto_start = 2;
             }
           }
           else if (hash_mode == 9300)
           {
-            if (device_param->vendor_id == VENDOR_ID_AMD)
+            if (device_param->device_vendor_id == VENDOR_ID_AMD)
             {
               tmto_start = 2;
             }
-            else if (device_param->vendor_id == VENDOR_ID_NV)
+            else if (device_param->device_vendor_id == VENDOR_ID_NV)
             {
               tmto_start = 2;
             }
@@ -14286,7 +14708,7 @@ int main (int argc, char **argv)
        * some algorithms need a fixed kernel-loops count
        */
 
-      if (hash_mode == 1500)
+      if (hash_mode == 1500 && attack_mode == ATTACK_MODE_BF)
       {
         const u32 kernel_loops_fixed = 1024;
 
@@ -14294,7 +14716,7 @@ int main (int argc, char **argv)
         device_param->kernel_loops_max = kernel_loops_fixed;
       }
 
-      if (hash_mode == 3000)
+      if (hash_mode == 3000 && attack_mode == ATTACK_MODE_BF)
       {
         const u32 kernel_loops_fixed = 1024;
 
@@ -14469,36 +14891,38 @@ int main (int argc, char **argv)
 
         int skip = 0;
 
-        if (size_pws   > device_param->device_maxmem_alloc) skip = 1;
-        if (size_tmps  > device_param->device_maxmem_alloc) skip = 1;
-        if (size_hooks > device_param->device_maxmem_alloc) skip = 1;
-
-        if (( bitmap_size
-            + bitmap_size
-            + bitmap_size
-            + bitmap_size
-            + bitmap_size
-            + bitmap_size
-            + bitmap_size
-            + bitmap_size
-            + size_bfs
-            + size_combs
-            + size_digests
-            + size_esalts
-            + size_hooks
-            + size_markov_css
-            + size_plains
-            + size_pws
-            + size_pws // not a bug
-            + size_results
-            + size_root_css
-            + size_rules
-            + size_rules_c
-            + size_salts
-            + size_scryptV
-            + size_shown
-            + size_tm
-            + size_tmps) > device_param->device_global_mem) skip = 1;
+        const u64 size_total
+          = bitmap_size
+          + bitmap_size
+          + bitmap_size
+          + bitmap_size
+          + bitmap_size
+          + bitmap_size
+          + bitmap_size
+          + bitmap_size
+          + size_bfs
+          + size_combs
+          + size_digests
+          + size_esalts
+          + size_hooks
+          + size_markov_css
+          + size_plains
+          + size_pws
+          + size_pws // not a bug
+          + size_results
+          + size_root_css
+          + size_rules
+          + size_rules_c
+          + size_salts
+          + size_scryptV
+          + size_shown
+          + size_tm
+          + size_tmps;
+
+        // Don't ask me, ask AMD!
+
+        if (size_total > device_param->device_maxmem_alloc) skip = 1;
+        if (size_total > device_param->device_global_mem)   skip = 1;
 
         if (skip == 1)
         {
@@ -14539,15 +14963,6 @@ int main (int argc, char **argv)
       device_param->size_tmps    = size_tmps;
       device_param->size_hooks   = size_hooks;
 
-      // do not confuse kernel_accel_max with kernel_accel here
-
-      const u32 kernel_power = device_processors * kernel_threads * kernel_accel_max;
-
-      device_param->kernel_threads    = kernel_threads;
-      device_param->kernel_power_user = kernel_power;
-
-      kernel_power_all += kernel_power;
-
       /**
        * default building options
        */
@@ -14557,19 +14972,19 @@ int main (int argc, char **argv)
       // we don't have sm_* on vendors not NV but it doesn't matter
 
       #if _WIN
-      snprintf (build_opts, sizeof (build_opts) - 1, "-I \"%s\\OpenCL\\\" -I '%s\\OpenCL\\' -I %s\\OpenCL\\", shared_dir, shared_dir, shared_dir);
+      snprintf (build_opts, sizeof (build_opts) - 1, "-I \"%s\\OpenCL\\\" -I '%s\\OpenCL\\' -I %s\\OpenCL\\ -I\"%s\\OpenCL\\\" -I'%s\\OpenCL\\' -I%s\\OpenCL\\", shared_dir, shared_dir, shared_dir, shared_dir, shared_dir, shared_dir);
       #else
-      snprintf (build_opts, sizeof (build_opts) - 1, "-I \"%s/OpenCL/\" -I '%s/OpenCL/' -I %s/OpenCL/", shared_dir, shared_dir, shared_dir);
+      snprintf (build_opts, sizeof (build_opts) - 1, "-I \"%s/OpenCL/\" -I '%s/OpenCL/' -I %s/OpenCL/ -I\"%s/OpenCL/\" -I'%s/OpenCL/' -I%s/OpenCL/", shared_dir, shared_dir, shared_dir, shared_dir, shared_dir, shared_dir);
       #endif
 
       char build_opts_new[1024] = { 0 };
 
-      snprintf (build_opts_new, sizeof (build_opts_new) - 1, "%s -DVENDOR_ID=%u -DCUDA_ARCH=%d -DVECT_SIZE=%u -DDEVICE_TYPE=%u -DKERN_TYPE=%u -D_unroll -cl-std=CL1.1", build_opts, device_param->vendor_id, (device_param->sm_major * 100) + device_param->sm_minor, device_param->vector_width, (u32) device_param->device_type, kern_type);
+      snprintf (build_opts_new, sizeof (build_opts_new) - 1, "%s -DVENDOR_ID=%u -DCUDA_ARCH=%d -DVECT_SIZE=%u -DDEVICE_TYPE=%u -DKERN_TYPE=%u -D_unroll -cl-std=CL1.1", build_opts, device_param->device_vendor_id, (device_param->sm_major * 100) + device_param->sm_minor, device_param->vector_width, (u32) device_param->device_type, kern_type);
 
       strncpy (build_opts, build_opts_new, sizeof (build_opts) - 1);
 
-
-      if (device_param->vendor_id == VENDOR_ID_INTEL_SDK)
+      /*
+      if (device_param->device_vendor_id == VENDOR_ID_INTEL_SDK)
       {
         // we do vectorizing much better than the auto-vectorizer
 
@@ -14577,6 +14992,7 @@ int main (int argc, char **argv)
 
         strncpy (build_opts, build_opts_new, sizeof (build_opts) - 1);
       }
+      */
 
       #ifdef DEBUG
       log_info ("Device #%u: build_opts '%s'\n", device_id + 1, build_opts);
@@ -14998,13 +15414,6 @@ int main (int argc, char **argv)
       hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_digests_shown,  CL_TRUE, 0, size_shown,   data.digests_shown, 0, NULL, NULL);
       hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_salt_bufs,      CL_TRUE, 0, size_salts,   data.salts_buf,     0, NULL, NULL);
 
-      run_kernel_bzero (device_param, device_param->d_pws_buf,        size_pws);
-      run_kernel_bzero (device_param, device_param->d_pws_amp_buf,    size_pws);
-      run_kernel_bzero (device_param, device_param->d_tmps,           size_tmps);
-      run_kernel_bzero (device_param, device_param->d_hooks,          size_hooks);
-      run_kernel_bzero (device_param, device_param->d_plain_bufs,     size_plains);
-      run_kernel_bzero (device_param, device_param->d_result,         size_results);
-
       /**
        * special buffers
        */
@@ -15015,8 +15424,6 @@ int main (int argc, char **argv)
         device_param->d_rules_c = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_ONLY, size_rules_c, NULL);
 
         hc_clEnqueueWriteBuffer (data.ocl, device_param->command_queue, device_param->d_rules, CL_TRUE, 0, size_rules, kernel_rules_buf, 0, NULL, NULL);
-
-        run_kernel_bzero (device_param, device_param->d_rules_c, size_rules_c);
       }
       else if (attack_kern == ATTACK_KERN_COMBI)
       {
@@ -15024,11 +15431,6 @@ int main (int argc, char **argv)
         device_param->d_combs_c         = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_ONLY, size_combs,      NULL);
         device_param->d_root_css_buf    = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_ONLY, size_root_css,   NULL);
         device_param->d_markov_css_buf  = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_ONLY, size_markov_css, NULL);
-
-        run_kernel_bzero (device_param, device_param->d_combs,          size_combs);
-        run_kernel_bzero (device_param, device_param->d_combs_c,        size_combs);
-        run_kernel_bzero (device_param, device_param->d_root_css_buf,   size_root_css);
-        run_kernel_bzero (device_param, device_param->d_markov_css_buf, size_markov_css);
       }
       else if (attack_kern == ATTACK_KERN_BF)
       {
@@ -15037,12 +15439,6 @@ int main (int argc, char **argv)
         device_param->d_tm_c            = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_ONLY, size_tm,         NULL);
         device_param->d_root_css_buf    = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_ONLY, size_root_css,   NULL);
         device_param->d_markov_css_buf  = hc_clCreateBuffer (data.ocl, device_param->context, CL_MEM_READ_ONLY, size_markov_css, NULL);
-
-        run_kernel_bzero (device_param, device_param->d_bfs,            size_bfs);
-        run_kernel_bzero (device_param, device_param->d_bfs_c,          size_bfs);
-        run_kernel_bzero (device_param, device_param->d_tm_c,           size_tm);
-        run_kernel_bzero (device_param, device_param->d_root_css_buf,   size_root_css);
-        run_kernel_bzero (device_param, device_param->d_markov_css_buf, size_markov_css);
       }
 
       if (size_esalts)
@@ -15186,6 +15582,13 @@ int main (int argc, char **argv)
       device_param->kernel_params_tm[0] = &device_param->d_bfs_c;
       device_param->kernel_params_tm[1] = &device_param->d_tm_c;
 
+      device_param->kernel_params_memset_buf32[1] = 0; // value
+      device_param->kernel_params_memset_buf32[2] = 0; // gid_max
+
+      device_param->kernel_params_memset[0] = NULL;
+      device_param->kernel_params_memset[1] = &device_param->kernel_params_memset_buf32[1];
+      device_param->kernel_params_memset[2] = &device_param->kernel_params_memset_buf32[2];
+
       /**
        * kernel name
        */
@@ -15294,6 +15697,18 @@ int main (int argc, char **argv)
         if (opts_type & OPTS_TYPE_HOOK23) hc_clSetKernelArg (data.ocl, device_param->kernel23, i, sizeof (cl_uint), device_param->kernel_params[i]);
       }
 
+      // GPU memset
+
+      device_param->kernel_memset = hc_clCreateKernel (data.ocl, device_param->program, "gpu_memset");
+
+      hc_clGetKernelWorkGroupInfo (data.ocl, device_param->kernel_memset, device_param->device, CL_KERNEL_WORK_GROUP_SIZE, sizeof (size_t), &kernel_wgs_tmp, NULL); kernel_threads = MIN (kernel_threads, kernel_wgs_tmp);
+
+      hc_clSetKernelArg (data.ocl, device_param->kernel_memset, 0, sizeof (cl_mem),  device_param->kernel_params_memset[0]);
+      hc_clSetKernelArg (data.ocl, device_param->kernel_memset, 1, sizeof (cl_uint), device_param->kernel_params_memset[1]);
+      hc_clSetKernelArg (data.ocl, device_param->kernel_memset, 2, sizeof (cl_uint), device_param->kernel_params_memset[2]);
+
+      // MP start
+
       if (attack_mode == ATTACK_MODE_BF)
       {
         device_param->kernel_mp_l = hc_clCreateKernel (data.ocl, device_param->program_mp, "l_markov");
@@ -15354,185 +15769,100 @@ int main (int argc, char **argv)
 
       device_param->kernel_threads = kernel_threads;
 
+      // zero some data buffers
+
+      run_kernel_bzero (device_param, device_param->d_pws_buf,        size_pws);
+      run_kernel_bzero (device_param, device_param->d_pws_amp_buf,    size_pws);
+      run_kernel_bzero (device_param, device_param->d_tmps,           size_tmps);
+      run_kernel_bzero (device_param, device_param->d_hooks,          size_hooks);
+      run_kernel_bzero (device_param, device_param->d_plain_bufs,     size_plains);
+      run_kernel_bzero (device_param, device_param->d_result,         size_results);
+
       /**
-       * Store initial fanspeed if gpu_temp_retain is enabled
+       * special buffers
        */
 
-      #if defined(HAVE_HWMON) && defined(HAVE_ADL)
-      int gpu_temp_retain_set = 0;
-
-      if (gpu_temp_disable == 0)
+      if (attack_kern == ATTACK_KERN_STRAIGHT)
       {
-        if (gpu_temp_retain != 0) // VENDOR_ID_AMD implied
-        {
-          hc_thread_mutex_lock (mux_adl);
-
-          if (data.hm_device[device_id].fan_supported == 1)
-          {
-            if (gpu_temp_retain_chgd == 0)
-            {
-              uint cur_temp = 0;
-              uint default_temp = 0;
-
-              int ADL_rc = hm_ADL_Overdrive6_TargetTemperatureData_Get (data.hm_amd, data.hm_device[device_id].adapter_index.amd, (int *) &cur_temp, (int *) &default_temp);
-
-              if (ADL_rc == ADL_OK)
-              {
-                #define GPU_TEMP_RETAIN_ABORT_DIFF 15
-
-                const uint gpu_temp_retain_target = default_temp - GPU_TEMP_RETAIN_ABORT_DIFF;
-
-                // special case with multi gpu setups: always use minimum retain
-
-                if (gpu_temp_retain_set == 0)
-                {
-                  gpu_temp_retain = gpu_temp_retain_target;
-                  gpu_temp_retain_set = 1;
-                }
-                else
-                {
-                  gpu_temp_retain = MIN (gpu_temp_retain, gpu_temp_retain_target);
-                }
-
-                if (gpu_temp_abort_chgd == 0) gpu_temp_abort = gpu_temp_retain + GPU_TEMP_RETAIN_ABORT_DIFF;
-              }
-            }
-
-            const int fan_speed = hm_get_fanspeed_with_device_id (device_id);
-
-            temp_retain_fanspeed_value[device_id] = fan_speed;
-
-            if (fan_speed == -1)
-            {
-              log_info ("WARNING: Failed to get current fan speed settings for gpu number: %i:", device_id + 1);
-
-              temp_retain_fanspeed_value[device_id] = 0;
-            }
-          }
-
-          hc_thread_mutex_unlock (mux_adl);
-        }
+        run_kernel_bzero (device_param, device_param->d_rules_c, size_rules_c);
+      }
+      else if (attack_kern == ATTACK_KERN_COMBI)
+      {
+        run_kernel_bzero (device_param, device_param->d_combs,          size_combs);
+        run_kernel_bzero (device_param, device_param->d_combs_c,        size_combs);
+        run_kernel_bzero (device_param, device_param->d_root_css_buf,   size_root_css);
+        run_kernel_bzero (device_param, device_param->d_markov_css_buf, size_markov_css);
+      }
+      else if (attack_kern == ATTACK_KERN_BF)
+      {
+        run_kernel_bzero (device_param, device_param->d_bfs,            size_bfs);
+        run_kernel_bzero (device_param, device_param->d_bfs_c,          size_bfs);
+        run_kernel_bzero (device_param, device_param->d_tm_c,           size_tm);
+        run_kernel_bzero (device_param, device_param->d_root_css_buf,   size_root_css);
+        run_kernel_bzero (device_param, device_param->d_markov_css_buf, size_markov_css);
       }
 
+      #if defined(HAVE_HWMON)
+
       /**
-       * Store original powercontrol/clocks settings, set overdrive 6 performance tuning settings
+       * Store initial fanspeed if gpu_temp_retain is enabled
        */
 
-      if (powertune_enable == 1) // VENDOR_ID_AMD implied
+      if (gpu_temp_disable == 0)
       {
-        hc_thread_mutex_lock (mux_adl);
-
-        if (data.hm_device[device_id].od_version == 6)
+        if (gpu_temp_retain != 0)
         {
-          int ADL_rc;
-
-          // check powertune capabilities first, if not available then skip device
-
-          int powertune_supported = 0;
-
-          if ((ADL_rc = hm_ADL_Overdrive6_PowerControl_Caps (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &powertune_supported)) != ADL_OK)
-          {
-            log_error ("ERROR: Failed to get ADL PowerControl Capabilities");
-
-            return (-1);
-          }
+          hc_thread_mutex_lock (mux_adl);
 
-          if (powertune_supported != 0)
+          if (data.hm_device[device_id].fan_get_supported == 1)
           {
-            // powercontrol settings
+            const int fanspeed  = hm_get_fanspeed_with_device_id  (device_id);
+            const int fanpolicy = hm_get_fanpolicy_with_device_id (device_id);
 
-            ADLOD6PowerControlInfo powertune = {0, 0, 0, 0, 0};
+            temp_retain_fanspeed_value[device_id]  = fanspeed;
+            temp_retain_fanpolicy_value[device_id] = fanpolicy;
 
-            if ((ADL_rc = hm_ADL_Overdrive_PowerControlInfo_Get (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &powertune)) == ADL_OK)
-            {
-              ADL_rc = hm_ADL_Overdrive_PowerControl_Get (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &od_power_control_status[device_id]);
-            }
+            // we also set it to tell the OS we take control over the fan and it's automatic controller
+            // if it was set to automatic. we do not control user-defined fanspeeds.
 
-            if (ADL_rc != ADL_OK)
+            if (fanpolicy == 1)
             {
-              log_error ("ERROR: Failed to get current ADL PowerControl settings");
-
-              return (-1);
-            }
+              data.hm_device[device_id].fan_set_supported = 1;
 
-            if ((ADL_rc = hm_ADL_Overdrive_PowerControl_Set (data.hm_amd, data.hm_device[device_id].adapter_index.amd, powertune.iMaxValue)) != ADL_OK)
-            {
-              log_error ("ERROR: Failed to set new ADL PowerControl values");
+              int rc = -1;
 
-              return (-1);
-            }
-
-            // clocks
-
-            memset (&od_clock_mem_status[device_id], 0, sizeof (ADLOD6MemClockState));
-
-            od_clock_mem_status[device_id].state.iNumberOfPerformanceLevels = 2;
-
-            if ((ADL_rc = hm_ADL_Overdrive_StateInfo_Get (data.hm_amd, data.hm_device[device_id].adapter_index.amd, ADL_OD6_GETSTATEINFO_CUSTOM_PERFORMANCE, &od_clock_mem_status[device_id])) != ADL_OK)
-            {
-              log_error ("ERROR: Failed to get ADL memory and engine clock frequency");
-
-              return (-1);
-            }
-
-            // Query capabilities only to see if profiles were not "damaged", if so output a warning but do accept the users profile settings
-
-            ADLOD6Capabilities caps = {0, 0, 0, {0, 0, 0}, {0, 0, 0}, 0, 0};
-
-            if ((ADL_rc = hm_ADL_Overdrive_Capabilities_Get (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &caps)) != ADL_OK)
-            {
-              log_error ("ERROR: Failed to get ADL device capabilities");
-
-              return (-1);
-            }
-
-            int engine_clock_max = caps.sEngineClockRange.iMax * 0.6666;
-            int memory_clock_max = caps.sMemoryClockRange.iMax * 0.6250;
-
-            int warning_trigger_engine = (int) (0.25 * (float) engine_clock_max);
-            int warning_trigger_memory = (int) (0.25 * (float) memory_clock_max);
-
-            int engine_clock_profile_max = od_clock_mem_status[device_id].state.aLevels[1].iEngineClock;
-            int memory_clock_profile_max = od_clock_mem_status[device_id].state.aLevels[1].iMemoryClock;
+              if (device_param->device_vendor_id == VENDOR_ID_AMD)
+              {
+                rc = hm_set_fanspeed_with_device_id_adl (device_id, fanspeed, 1);
+              }
+              else if (device_param->device_vendor_id == VENDOR_ID_NV)
+              {
 
-            // warning if profile has too low max values
+              }
 
-            if ((engine_clock_max - engine_clock_profile_max) > warning_trigger_engine)
-            {
-              log_info ("WARN: the custom profile seems to have too low maximum engine clock values. You therefore may not reach full performance");
-            }
+              if (rc == 0)
+              {
+                data.hm_device[device_id].fan_set_supported = 1;
+              }
+              else
+              {
+                log_info ("WARNING: Failed to set initial fan speed for device #%u", device_id + 1);
 
-            if ((memory_clock_max - memory_clock_profile_max) > warning_trigger_memory)
-            {
-              log_info ("WARN: the custom profile seems to have too low maximum memory clock values. You therefore may not reach full performance");
+                data.hm_device[device_id].fan_set_supported = 0;
+              }
             }
-
-            ADLOD6StateInfo *performance_state = (ADLOD6StateInfo*) mycalloc (1, sizeof (ADLOD6StateInfo) + sizeof (ADLOD6PerformanceLevel));
-
-            performance_state->iNumberOfPerformanceLevels = 2;
-
-            performance_state->aLevels[0].iEngineClock = engine_clock_profile_max;
-            performance_state->aLevels[1].iEngineClock = engine_clock_profile_max;
-            performance_state->aLevels[0].iMemoryClock = memory_clock_profile_max;
-            performance_state->aLevels[1].iMemoryClock = memory_clock_profile_max;
-
-            if ((ADL_rc = hm_ADL_Overdrive_State_Set (data.hm_amd, data.hm_device[device_id].adapter_index.amd, ADL_OD6_SETSTATE_PERFORMANCE, performance_state)) != ADL_OK)
+            else
             {
-              log_info ("ERROR: Failed to set ADL performance state");
-
-              return (-1);
+              data.hm_device[device_id].fan_set_supported = 0;
             }
-
-            local_free (performance_state);
           }
-        }
 
-        hc_thread_mutex_unlock (mux_adl);
+          hc_thread_mutex_unlock (mux_adl);
+        }
       }
-      #endif // HAVE_HWMON && HAVE_ADL
-    }
 
-    data.kernel_power_all = kernel_power_all;
+      #endif // HAVE_HWMON
+    }
 
     if (data.quiet == 0) log_info_nn ("");
 
@@ -15542,7 +15872,7 @@ int main (int argc, char **argv)
 
     if (benchmark == 1)
     {
-      if (status_automat == 0)
+      if (machine_readable == 0)
       {
         quiet = 0;
 
@@ -16758,6 +17088,8 @@ int main (int argc, char **argv)
 
         data.ms_paused = 0;
 
+        data.kernel_power_final = 0;
+
         data.words_cur = rd->words_cur;
 
         for (uint device_id = 0; device_id < data.devices_cnt; device_id++)
@@ -16775,8 +17107,6 @@ int main (int argc, char **argv)
 
           memset (device_param->exec_ms, 0, EXEC_CACHE * sizeof (double));
 
-          device_param->kernel_power = device_param->kernel_power_user;
-
           device_param->outerloop_pos  = 0;
           device_param->outerloop_left = 0;
           device_param->innerloop_pos  = 0;
@@ -16792,8 +17122,6 @@ int main (int argc, char **argv)
           device_param->words_done = 0;
         }
 
-        data.kernel_power_div = 0;
-
         // figure out some workload
 
         if (attack_mode == ATTACK_MODE_STRAIGHT)
@@ -17235,26 +17563,6 @@ int main (int argc, char **argv)
           }
         }
 
-        /*
-         * Inform user about possible slow speeds
-         */
-
-        if ((wordlist_mode == WL_MODE_FILE) || (wordlist_mode == WL_MODE_MASK))
-        {
-          if (data.words_base < kernel_power_all)
-          {
-            if (quiet == 0)
-            {
-              log_info ("ATTENTION!");
-              log_info ("  The wordlist or mask you are using is too small.");
-              log_info ("  Therefore, hashcat is unable to utilize the full parallelization power of your device(s).");
-              log_info ("  The cracking speed will drop.");
-              log_info ("  Workaround: https://hashcat.net/wiki/doku.php?id=frequently_asked_questions#how_to_create_more_work_for_full_speed");
-              log_info ("");
-            }
-          }
-        }
-
         /*
          * Update loopback file
          */
@@ -17290,6 +17598,64 @@ int main (int argc, char **argv)
           }
         }
 
+        /**
+         * create autotune threads
+         */
+
+        data.devices_status = STATUS_AUTOTUNE;
+
+        hc_thread_t *c_threads = (hc_thread_t *) mycalloc (data.devices_cnt, sizeof (hc_thread_t));
+
+        for (uint device_id = 0; device_id < data.devices_cnt; device_id++)
+        {
+          hc_device_param_t *device_param = &devices_param[device_id];
+
+          hc_thread_create (c_threads[device_id], thread_autotune, device_param);
+        }
+
+        hc_thread_wait (data.devices_cnt, c_threads);
+
+        /*
+         * Inform user about possible slow speeds
+         */
+
+        uint hardware_power_all = 0;
+
+        uint kernel_power_all = 0;
+
+        for (uint device_id = 0; device_id < data.devices_cnt; device_id++)
+        {
+          hc_device_param_t *device_param = &devices_param[device_id];
+
+          hardware_power_all += device_param->hardware_power;
+
+          kernel_power_all += device_param->kernel_power;
+        }
+
+        data.hardware_power_all = hardware_power_all; // hardware_power_all is the same as kernel_power_all but without the influence of kernel_accel on the devices
+
+        data.kernel_power_all = kernel_power_all;
+
+        if ((wordlist_mode == WL_MODE_FILE) || (wordlist_mode == WL_MODE_MASK))
+        {
+          if (data.words_base < kernel_power_all)
+          {
+            if (quiet == 0)
+            {
+              log_info ("ATTENTION!");
+              log_info ("  The wordlist or mask you are using is too small.");
+              log_info ("  Therefore, hashcat is unable to utilize the full parallelization power of your device(s).");
+              log_info ("  The cracking speed will drop.");
+              log_info ("  Workaround: https://hashcat.net/wiki/doku.php?id=frequently_asked_questions#how_to_create_more_work_for_full_speed");
+              log_info ("");
+            }
+          }
+        }
+
+        /**
+         * create cracker threads
+         */
+
         data.devices_status = STATUS_RUNNING;
 
         if (initial_restore_done == 0)
@@ -17321,12 +17687,6 @@ int main (int argc, char **argv)
 
         data.runtime_start = runtime_start;
 
-        /**
-         * create cracker threads
-         */
-
-        hc_thread_t *c_threads = (hc_thread_t *) mycalloc (data.devices_cnt, sizeof (hc_thread_t));
-
         for (uint device_id = 0; device_id < data.devices_cnt; device_id++)
         {
           hc_device_param_t *device_param = &devices_param[device_id];
@@ -17341,8 +17701,6 @@ int main (int argc, char **argv)
           }
         }
 
-        // wait for crack threads to exit
-
         hc_thread_wait (data.devices_cnt, c_threads);
 
         local_free (c_threads);
@@ -17546,7 +17904,7 @@ int main (int argc, char **argv)
     {
       status_benchmark ();
 
-      if (status_automat == 0)
+      if (machine_readable == 0)
       {
         log_info ("");
       }
@@ -17627,6 +17985,7 @@ int main (int argc, char **argv)
       if (device_param->kernel_mp_r)        hc_clReleaseKernel        (data.ocl, device_param->kernel_mp_r);
       if (device_param->kernel_tm)          hc_clReleaseKernel        (data.ocl, device_param->kernel_tm);
       if (device_param->kernel_amp)         hc_clReleaseKernel        (data.ocl, device_param->kernel_amp);
+      if (device_param->kernel_memset)      hc_clReleaseKernel        (data.ocl, device_param->kernel_memset);
 
       if (device_param->program)            hc_clReleaseProgram       (data.ocl, device_param->program);
       if (device_param->program_mp)         hc_clReleaseProgram       (data.ocl, device_param->program_mp);
@@ -17641,7 +18000,6 @@ int main (int argc, char **argv)
     #ifdef HAVE_HWMON
     if (gpu_temp_disable == 0)
     {
-      #ifdef HAVE_ADL
       if (gpu_temp_retain != 0) // VENDOR_ID_AMD is implied here
       {
         hc_thread_mutex_lock (mux_adl);
@@ -17652,24 +18010,33 @@ int main (int argc, char **argv)
 
           if (device_param->skipped) continue;
 
-          if (data.hm_device[device_id].fan_supported == 1)
+          if (data.hm_device[device_id].fan_set_supported == 1)
           {
-            int fanspeed = temp_retain_fanspeed_value[device_id];
+            int fanspeed  = temp_retain_fanspeed_value[device_id];
+            int fanpolicy = temp_retain_fanpolicy_value[device_id];
 
-            if (fanspeed == -1) continue;
+            if (fanpolicy == 1)
+            {
+              int rc = -1;
+
+              if (device_param->device_vendor_id == VENDOR_ID_AMD)
+              {
+                rc = hm_set_fanspeed_with_device_id_adl (device_id, fanspeed, 0);
+              }
+              else if (device_param->device_vendor_id == VENDOR_ID_NV)
+              {
 
-            int rc = hm_set_fanspeed_with_device_id_amd (device_id, fanspeed);
+              }
 
-            if (rc == -1) log_info ("WARNING: Failed to restore default fan speed for gpu number: %i:", device_id);
+              if (rc == -1) log_info ("WARNING: Failed to restore default fan speed and policy for device #%", device_id + 1);
+            }
           }
         }
 
         hc_thread_mutex_unlock (mux_adl);
       }
-      #endif // HAVE_ADL
     }
 
-    #ifdef HAVE_ADL
     // reset power tuning
 
     if (powertune_enable == 1) // VENDOR_ID_AMD is implied here
@@ -17682,89 +18049,88 @@ int main (int argc, char **argv)
 
         if (device_param->skipped) continue;
 
-        if (data.hm_device[device_id].od_version == 6)
+        if (data.devices_param[device_id].device_vendor_id == VENDOR_ID_AMD)
         {
-          // check powertune capabilities first, if not available then skip device
-
-          int powertune_supported = 0;
-
-          if ((hm_ADL_Overdrive6_PowerControl_Caps (data.hm_amd, data.hm_device[device_id].adapter_index.amd, &powertune_supported)) != ADL_OK)
+          if (data.hm_device[device_id].od_version == 6)
           {
-            log_error ("ERROR: Failed to get ADL PowerControl Capabilities");
-
-            return (-1);
-          }
+            // check powertune capabilities first, if not available then skip device
 
-          if (powertune_supported != 0)
-          {
-            // powercontrol settings
+            int powertune_supported = 0;
 
-            if ((hm_ADL_Overdrive_PowerControl_Set (data.hm_amd, data.hm_device[device_id].adapter_index.amd, od_power_control_status[device_id])) != ADL_OK)
+            if ((hm_ADL_Overdrive6_PowerControl_Caps (data.hm_adl, data.hm_device[device_id].adl, &powertune_supported)) != ADL_OK)
             {
-              log_info ("ERROR: Failed to restore the ADL PowerControl values");
+              log_error ("ERROR: Failed to get ADL PowerControl Capabilities");
 
               return (-1);
             }
 
-            // clocks
+            if (powertune_supported != 0)
+            {
+              // powercontrol settings
+
+              if ((hm_ADL_Overdrive_PowerControl_Set (data.hm_adl, data.hm_device[device_id].adl, od_power_control_status[device_id])) != ADL_OK)
+              {
+                log_info ("ERROR: Failed to restore the ADL PowerControl values");
+
+                return (-1);
+              }
 
-            ADLOD6StateInfo *performance_state = (ADLOD6StateInfo*) mycalloc (1, sizeof (ADLOD6StateInfo) + sizeof (ADLOD6PerformanceLevel));
+              // clocks
 
-            performance_state->iNumberOfPerformanceLevels = 2;
+              ADLOD6StateInfo *performance_state = (ADLOD6StateInfo*) mycalloc (1, sizeof (ADLOD6StateInfo) + sizeof (ADLOD6PerformanceLevel));
 
-            performance_state->aLevels[0].iEngineClock = od_clock_mem_status[device_id].state.aLevels[0].iEngineClock;
-            performance_state->aLevels[1].iEngineClock = od_clock_mem_status[device_id].state.aLevels[1].iEngineClock;
-            performance_state->aLevels[0].iMemoryClock = od_clock_mem_status[device_id].state.aLevels[0].iMemoryClock;
-            performance_state->aLevels[1].iMemoryClock = od_clock_mem_status[device_id].state.aLevels[1].iMemoryClock;
+              performance_state->iNumberOfPerformanceLevels = 2;
 
-            if ((hm_ADL_Overdrive_State_Set (data.hm_amd, data.hm_device[device_id].adapter_index.amd, ADL_OD6_SETSTATE_PERFORMANCE, performance_state)) != ADL_OK)
-            {
-              log_info ("ERROR: Failed to restore ADL performance state");
+              performance_state->aLevels[0].iEngineClock = od_clock_mem_status[device_id].state.aLevels[0].iEngineClock;
+              performance_state->aLevels[1].iEngineClock = od_clock_mem_status[device_id].state.aLevels[1].iEngineClock;
+              performance_state->aLevels[0].iMemoryClock = od_clock_mem_status[device_id].state.aLevels[0].iMemoryClock;
+              performance_state->aLevels[1].iMemoryClock = od_clock_mem_status[device_id].state.aLevels[1].iMemoryClock;
 
-              return (-1);
+              if ((hm_ADL_Overdrive_State_Set (data.hm_adl, data.hm_device[device_id].adl, ADL_OD6_SETSTATE_PERFORMANCE, performance_state)) != ADL_OK)
+              {
+                log_info ("ERROR: Failed to restore ADL performance state");
+
+                return (-1);
+              }
+
+              local_free (performance_state);
             }
+          }
+        }
 
-            local_free (performance_state);
+        if (data.devices_param[device_id].device_vendor_id == VENDOR_ID_NV)
+        {
+          unsigned int limit = nvml_power_limit[device_id];
+
+          if (limit > 0)
+          {
+            hm_NVML_nvmlDeviceSetPowerManagementLimit (data.hm_nvml, 0, data.hm_device[device_id].nvml, limit);
           }
         }
       }
 
       hc_thread_mutex_unlock (mux_adl);
     }
-    #endif // HAVE_ADL
 
     if (gpu_temp_disable == 0)
     {
-      #if defined(HAVE_NVML) || defined(HAVE_NVAPI)
-      if (data.hm_nv)
+      if (data.hm_nvml)
       {
-        #if defined(LINUX) && defined(HAVE_NVML)
-
-        hm_NVML_nvmlShutdown (data.hm_nv);
-
-        nvml_close (data.hm_nv);
-
-        #elif defined(WIN) && (HAVE_NVAPI)
+        hm_NVML_nvmlShutdown (data.hm_nvml);
 
-        hm_NvAPI_Unload (data.hm_nv);
+        nvml_close (data.hm_nvml);
 
-        nvapi_close (data.hm_nv);
-
-        #endif
-
-        data.hm_nv = NULL;
+        data.hm_nvml = NULL;
       }
-      #endif
 
-      #ifdef HAVE_ADL
-      if (data.hm_amd)
+      if (data.hm_adl)
       {
-        hm_ADL_Main_Control_Destroy (data.hm_amd);
+        hm_ADL_Main_Control_Destroy (data.hm_adl);
+
+        adl_close (data.hm_adl);
 
-        adl_close (data.hm_amd);
-        data.hm_amd = NULL;
+        data.hm_adl = NULL;
       }
-      #endif
     }
     #endif // HAVE_HWMON
 
@@ -17807,10 +18173,9 @@ int main (int argc, char **argv)
 
     #ifdef HAVE_HWMON
     local_free (temp_retain_fanspeed_value);
-    #ifdef HAVE_ADL
     local_free (od_clock_mem_status);
     local_free (od_power_control_status);
-    #endif // ADL
+    local_free (nvml_power_limit);
     #endif
 
     global_free (devices_param);