Detect: Add NVidia decoders data

examples/get_detect_data.py

@@ -38,7 +38,11 @@ API_TO_HWACCEL = {
 NVIDIA_GPU_MATRIX_URL = (
     'https://developer.nvidia.com/video-encode-decode-gpu-support-matrix')
 NVIDIA_LINE_SUFFIXES = {'geforce': ['gtx titan', 'gtx', 'gt', 'rtx']}
-NVIDIA_CODEC_COLUMN_PREFIXES = {'h.264': 'h264', 'h.265': 'hevc'}
+NVIDIA_CODEC_COLUMN_PREFIXES = {
+    'mpeg-1': 'mpeg1video', 'mpeg-2': 'mpeg2video',
+    'vc-1': 'vc1',
+    'vp8': 'vp8', 'vp9': 'vp9',
+    'h.264': 'h264', 'h.265': 'hevc'}
 
 
 def get_hwaccel_data():
@@ -75,14 +79,18 @@ def get_nvidia_data():
     (
         nvenc_recent, nvenc_consumer, nvenc_workstation, nvenc_virt,
         nvdec_recent, nvdec_consumer, nvdec_workstation, nvdec_virt) = tables
-    nvidia = collections.OrderedDict(
-        lines=[], model_lines=collections.OrderedDict(),
-        boards=collections.OrderedDict())
+    nv_coders = dict(
+        encoders=(
+            nvenc_recent, nvenc_consumer, nvenc_workstation, nvenc_virt),
+        decoders=(
+            nvdec_recent, nvdec_consumer, nvdec_workstation, nvdec_virt))
+    nvidia = collections.OrderedDict(lines=[])
 
     # Compile aggregate data needed to parse individual rows
-    for nvenc_table in (
-            nvenc_recent, nvenc_consumer, nvenc_workstation, nvenc_virt):
-        for board in nvenc_table['BOARD']:
+    for nv_coder_table in tables:
+        for board in nv_coder_table['BOARD']:
+            if board == 'BOARD':
+                continue
             line = board.replace('\xa0', ' ').split(None, 1)[0].lower()
             if line not in nvidia['lines']:
                 nvidia['lines'].append(line)
@@ -90,62 +98,79 @@ def get_nvidia_data():
         for line_suffix in reversed(line_suffixes):
             nvidia['lines'].insert(0, ' '.join((line, line_suffix)))
 
-    for nvenc_table in (
-            nvenc_recent, nvenc_consumer, nvenc_workstation, nvenc_virt):
-        for nvenc_row_idx, nvenc_row in nvenc_table.iterrows():
-            nvenc_row_values = {
-                idx: cell for idx, cell in enumerate(nvenc_row[1:]) if (
+    for coder_type, nv_coder_tables in nv_coders.items():
+        coder_data = nvidia[coder_type] = collections.OrderedDict(
+            model_lines=collections.OrderedDict(),
+            boards=collections.OrderedDict())
+        for nv_coder_table in nv_coder_tables:
+            for nv_coder_row_idx, nv_coder_row in nv_coder_table.iterrows():
+                nv_coder_row_values = {
+                    idx: cell for idx, cell in enumerate(nv_coder_row[1:]) if (
                         cell and
                         not (isinstance(cell, float) and math.isnan(cell)))}
-            if not nvenc_row_values:
+                if not nv_coder_row_values:
                     # Divider row
                     continue
 
                 # Assemble the data for this row to use for each model or range
                 model_data = collections.OrderedDict()
-            for key, value in nvenc_row.items():
+                for key, value in nv_coder_row.items():
+                    if isinstance(key, tuple):
+                        if key[0] == key[1]:
+                            key = key[0]
+                        else:
+                            key = ' '.join(key)
                     if value in {'YES', 'NO'}:
                         model_data[key] = value == 'YES'
                     else:
                         model_data[key] = value
                 model_data['BOARD'] = model_data['BOARD'].replace('\xa0', ' ')
-            # Add keys for the data for the ffmpeg codec names for fast lookup
-            for codec_prefix, codec in NVIDIA_CODEC_COLUMN_PREFIXES.items():
-                for column_idx, column in enumerate(nvenc_row.keys()):
+                # Add keys for the ffmpeg codec names for fast lookup
+                for codec_prefix, codec in (
+                        NVIDIA_CODEC_COLUMN_PREFIXES.items()):
+                    for column_idx, column in enumerate(nv_coder_row.keys()):
+                        if isinstance(column, tuple):
+                            if column[0] == column[1]:
+                                column = column[0]
+                            else:
+                                column = ' '.join(column)
                         if column.lower().startswith(codec_prefix):
-                        model_data[codec] = nvenc_row[column_idx] == 'YES'
+                            model_data[codec] = nv_coder_row[
+                                column_idx] == 'YES'
                             break
-            nvidia['boards'][model_data['BOARD']] = model_data
+                coder_data['boards'][model_data['BOARD']] = model_data
 
                 _detect._parse_models(
                     model_lines=nvidia['lines'],
                     boards=model_data['BOARD'].lower(),
                     model_data=model_data['BOARD'],
-                model_lines_data=nvidia['model_lines'])
+                    model_lines_data=coder_data['model_lines'])
 
-    # Clean up some annoying clashes between the titan model line and GeForce
-    # GTX model numbers
-    for model_line, model_line_suffixes in NVIDIA_LINE_SUFFIXES.items():
-        models_data = nvidia['model_lines'][model_line]['models']
-        for model_num in list(models_data):
-            for model_line_suffix in model_line_suffixes:
-                if model_num.startswith(model_line_suffix + ' '):
-                    models_data[model_num[
-                        len(model_line_suffix + ' '):]] = models_data.pop(
-                            model_num)
-    for titan_model_num in {'black', 'xp'}:
-        nvidia['model_lines']['geforce gtx']['models'][
-            'titan ' + titan_model_num] = nvidia['model_lines'][
-                'titan']['models'].pop(titan_model_num)
-    for titan_model_num in list(nvidia['model_lines'][
-            'geforce gtx titan']['models'].keys()):
-        nvidia['model_lines']['geforce gtx']['models'][
-            'titan ' + titan_model_num] = nvidia['model_lines'][
-                'geforce gtx titan']['models'].pop(titan_model_num)
-    nvidia['model_lines']['geforce gtx']['models']['titan'] = nvidia[
-        'model_lines']['geforce gtx']['models']['titan black']
-    del nvidia['model_lines']['geforce gtx']['models']['titan ']
-    del nvidia['model_lines']['geforce gtx titan']
+        # Cleanup any deviations from the convention where models from
+        # multiple lines are in the same BOARD cell
+        for model_line, model_line_data in coder_data['model_lines'].items():
+            for line, line_suffixes in NVIDIA_LINE_SUFFIXES.items():
+                if not model_line.startswith(line):
+                    continue
+                for model_num, boards in list(
+                        model_line_data['models'].items()):
+                    for line_suffix in line_suffixes:
+                        if not model_num.startswith(line_suffix + ' '):
+                            continue
+                        coder_data['model_lines'][
+                            ' '.join((line, line_suffix))]['models'][
+                                model_num[len(line_suffix + ' '):]
+                            ] = model_line_data['models'].pop(model_num)
+        # Clean up some annoying clashes between the titan model line and
+        # GeForce GTX model numbers
+        del coder_data['model_lines']['geforce gtx titan']['models']['']
+        coder_data['model_lines']['geforce gtx titan']['models'][
+            'xp'] = coder_data['model_lines']['titan']['models'].pop('xp')
+        coder_data['model_lines']['geforce gtx titan']['models'][
+            'black'] = titan_black = coder_data['model_lines'][
+                'titan']['models'].pop('black')
+        coder_data['model_lines']['geforce gtx']['models'][
+            'titan'] = titan_black
 
     return nvidia