Reset ctx->t_start_us when calling whisper_reset_timings()

[bjnortier]

Currently ctx->t_start_us is not reset when calling whisper_reset_timings(). This leads to an incorrect "total time" when doing consecutive transcription calls. In the SwiftUI example app it also outputs the total time from app startup (which loads the model) to when the transcription finishes.

See below for example log outputs

[bjnortier]

Before. "total time" is time from initial app startup to finish of first transcription. Subsequent transcriptions show total time from app startup.

Selecting 8 threads
About to run whisper_full
[00:00:00.000 --> 00:00:07.300]   And so my fellow Americans ask not what your country can do for you
[00:00:07.300 --> 00:00:12.400]   ask what you can do for your country.

whisper_print_timings:     load time =    47.99 ms
whisper_print_timings:     fallbacks =   0 p /   0 h
whisper_print_timings:      mel time =     6.67 ms
whisper_print_timings:   sample time =    11.16 ms /    27 runs (    0.41 ms per run)
whisper_print_timings:   encode time =   383.20 ms /     1 runs (  383.20 ms per run)
whisper_print_timings:   decode time =    29.17 ms /    27 runs (    1.08 ms per run)
whisper_print_timings:   prompt time =     0.00 ms /     1 runs (    0.00 ms per run)
whisper_print_timings:    total time =  1797.31 ms

Selecting 8 threads
About to run whisper_full
[00:00:00.000 --> 00:00:07.300]   And so my fellow Americans ask not what your country can do for you
[00:00:07.300 --> 00:00:12.400]   ask what you can do for your country.

whisper_print_timings:     load time =    47.99 ms
whisper_print_timings:     fallbacks =   0 p /   0 h
whisper_print_timings:      mel time =    16.72 ms
whisper_print_timings:   sample time =    11.08 ms /    27 runs (    0.41 ms per run)
whisper_print_timings:   encode time =   385.28 ms /     1 runs (  385.28 ms per run)
whisper_print_timings:   decode time =    40.26 ms /    27 runs (    1.49 ms per run)
whisper_print_timings:   prompt time =     0.00 ms /     1 runs (    0.00 ms per run)
whisper_print_timings:    total time =  7639.08 ms

Selecting 8 threads
About to run whisper_full
[00:00:00.000 --> 00:00:07.300]   And so my fellow Americans ask not what your country can do for you
[00:00:07.300 --> 00:00:12.400]   ask what you can do for your country.

whisper_print_timings:     load time =    47.99 ms
whisper_print_timings:     fallbacks =   0 p /   0 h
whisper_print_timings:      mel time =    27.72 ms
whisper_print_timings:   sample time =    11.13 ms /    27 runs (    0.41 ms per run)
whisper_print_timings:   encode time =   385.56 ms /     1 runs (  385.56 ms per run)
whisper_print_timings:   decode time =    36.23 ms /    27 runs (    1.34 ms per run)
whisper_print_timings:   prompt time =     0.00 ms /     1 runs (    0.00 ms per run)
whisper_print_timings:    total time = 11275.45 ms

[bjnortier]

After. "total time" shows transcription time

Selecting 8 threads
About to run whisper_full
[00:00:00.000 --> 00:00:07.300]   And so my fellow Americans ask not what your country can do for you
[00:00:07.300 --> 00:00:12.400]   ask what you can do for your country.

whisper_print_timings:     load time =    55.53 ms
whisper_print_timings:     fallbacks =   0 p /   0 h
whisper_print_timings:      mel time =     7.42 ms
whisper_print_timings:   sample time =    11.24 ms /    27 runs (    0.42 ms per run)
whisper_print_timings:   encode time =   415.34 ms /     1 runs (  415.34 ms per run)
whisper_print_timings:   decode time =    50.04 ms /    27 runs (    1.85 ms per run)
whisper_print_timings:   prompt time =     0.00 ms /     1 runs (    0.00 ms per run)
whisper_print_timings:    total time =   484.90 ms

Selecting 8 threads
About to run whisper_full
[00:00:00.000 --> 00:00:07.300]   And so my fellow Americans ask not what your country can do for you
[00:00:07.300 --> 00:00:12.400]   ask what you can do for your country.

whisper_print_timings:     load time =    55.53 ms
whisper_print_timings:     fallbacks =   0 p /   0 h
whisper_print_timings:      mel time =    18.01 ms
whisper_print_timings:   sample time =    11.34 ms /    27 runs (    0.42 ms per run)
whisper_print_timings:   encode time =   396.28 ms /     1 runs (  396.28 ms per run)
whisper_print_timings:   decode time =    35.22 ms /    27 runs (    1.30 ms per run)
whisper_print_timings:   prompt time =     0.00 ms /     1 runs (    0.00 ms per run)
whisper_print_timings:    total time =   453.65 ms

Selecting 8 threads
About to run whisper_full
[00:00:00.000 --> 00:00:07.300]   And so my fellow Americans ask not what your country can do for you
[00:00:07.300 --> 00:00:12.400]   ask what you can do for your country.

whisper_print_timings:     load time =    55.53 ms
whisper_print_timings:     fallbacks =   0 p /   0 h
whisper_print_timings:      mel time =    28.65 ms
whisper_print_timings:   sample time =    11.17 ms /    27 runs (    0.41 ms per run)
whisper_print_timings:   encode time =   400.93 ms /     1 runs (  400.93 ms per run)
whisper_print_timings:   decode time =    38.94 ms /    27 runs (    1.44 ms per run)
whisper_print_timings:   prompt time =     0.00 ms /     1 runs (    0.00 ms per run)
whisper_print_timings:    total time =   461.91 ms