Skip to content

Testing #333

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
shubhammalhotra28 merged 4 commits into from
Feb 6, 2026
Merged

Testing #333

Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
8caec34
Optimise resampling and heap thrashing
VyasGuru Feb 5, 2026
2bea5e4
Merge branch 'testing' of https://github.com/VyasGuru/runanywhere-sdk…
VyasGuru Feb 5, 2026
f901cf0
improved even more for integer ratios
VyasGuru Feb 5, 2026
3367adb
Update whispercpp_backend.cpp
VyasGuru Feb 5, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
73 changes: 58 additions & 15 deletions sdk/runanywhere-commons/src/backends/whispercpp/whispercpp_backend.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -274,35 +274,41 @@ STTResult WhisperCppSTT::transcribe_internal(const std::vector<float>& audio,

const int n_segments = whisper_full_n_segments(ctx_);
std::string full_text;
full_text.reserve(n_segments * 64);
Copy link
Copy Markdown
Contributor

@greptile-apps greptile-apps bot Feb 5, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Trailing whitespace after reserve call

Suggested change
full_text.reserve(n_segments * 64);
full_text.reserve(n_segments * 64);

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

Prompt To Fix With AI 
This is a comment left during a code review.
Path: sdk/runanywhere-commons/src/backends/whispercpp/whispercpp_backend.cpp
Line: 277:277

Comment:
Trailing whitespace after reserve call

```suggestion
    full_text.reserve(n_segments * 64);
```

<sub>Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!</sub>

How can I resolve this? If you propose a fix, please make it concise.


result.segments.reserve(n_segments);

if (word_timestamps) {
result.word_timings.reserve(n_segments * 15);
}

for (int i = 0; i < n_segments; ++i) {
const char* text = whisper_full_get_segment_text(ctx_, i);
if (text) {
full_text += text;

AudioSegment segment;
result.segments.emplace_back();
AudioSegment& segment = result.segments.back();
segment.text = text;
segment.start_time_ms = whisper_full_get_segment_t0(ctx_, i) * 10.0;
segment.end_time_ms = whisper_full_get_segment_t1(ctx_, i) * 10.0;

float no_speech_prob = whisper_full_get_segment_no_speech_prob(ctx_, i);
segment.confidence = 1.0f - no_speech_prob;

result.segments.push_back(segment);

if (word_timestamps) {
const int n_tokens = whisper_full_n_tokens(ctx_, i);
for (int j = 0; j < n_tokens; ++j) {
whisper_token_data token_data = whisper_full_get_token_data(ctx_, i, j);
const char* token_text = whisper_full_get_token_text(ctx_, i, j);

if (token_text && token_text[0] != '\0' && token_text[0] != '<') {
WordTiming word;
result.word_timings.emplace_back();
WordTiming& word = result.word_timings.back();
word.word = token_text;
word.start_time_ms = token_data.t0 * 10.0;
word.end_time_ms = token_data.t1 * 10.0;
word.confidence = token_data.p;
result.word_timings.push_back(word);
}
}
}
Expand Down Expand Up @@ -547,22 +553,59 @@ std::vector<std::string> WhisperCppSTT::get_supported_languages() const {

std::vector<float> WhisperCppSTT::resample_to_16khz(const std::vector<float>& samples,
int source_rate) {
if (source_rate == WHISPER_SAMPLE_RATE) {
if (source_rate == WHISPER_SAMPLE_RATE || samples.empty()) {
return samples;
}

const double ratio = static_cast<double>(WHISPER_SAMPLE_RATE) / source_rate;
const size_t output_size = static_cast<size_t>(samples.size() * ratio);
const double step = static_cast<double>(source_rate) / WHISPER_SAMPLE_RATE;
const size_t output_size = static_cast<size_t>(samples.size() / step);

std::vector<float> output;

Copy link
Copy Markdown
Contributor

@greptile-apps greptile-apps bot Feb 5, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Blank line with trailing whitespace after declaration

Suggested change
std::vector<float> output;

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

Prompt To Fix With AI 
This is a comment left during a code review.
Path: sdk/runanywhere-commons/src/backends/whispercpp/whispercpp_backend.cpp
Line: 564:564

Comment:
Blank line with trailing whitespace after declaration

```suggestion
    std::vector<float> output;
```

<sub>Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!</sub>

How can I resolve this? If you propose a fix, please make it concise.


if (source_rate % WHISPER_SAMPLE_RATE == 0) {
int stride = source_rate / WHISPER_SAMPLE_RATE;
output.resize(samples.size() / stride);
for (size_t i = 0; i < output.size(); ++i) {
output[i] = samples[i * stride];
}
return output;
}
Copy link
Copy Markdown
Contributor

@greptile-apps greptile-apps bot Feb 5, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Indentation inconsistent with surrounding code - missing 4 spaces

Suggested change
if (source_rate % WHISPER_SAMPLE_RATE == 0) {
int stride = source_rate / WHISPER_SAMPLE_RATE;
output.resize(samples.size() / stride);
for (size_t i = 0; i < output.size(); ++i) {
output[i] = samples[i * stride];
}
return output;
}
if (source_rate % WHISPER_SAMPLE_RATE == 0) {
int stride = source_rate / WHISPER_SAMPLE_RATE;
output.resize(samples.size() / stride);
for (size_t i = 0; i < output.size(); ++i) {
output[i] = samples[i * stride];
}
return output;
}

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

Prompt To Fix With AI 
This is a comment left during a code review.
Path: sdk/runanywhere-commons/src/backends/whispercpp/whispercpp_backend.cpp
Line: 566:573

Comment:
Indentation inconsistent with surrounding code - missing 4 spaces

```suggestion
    if (source_rate % WHISPER_SAMPLE_RATE == 0) {
        int stride = source_rate / WHISPER_SAMPLE_RATE;
        output.resize(samples.size() / stride);
        for (size_t i = 0; i < output.size(); ++i) {
            output[i] = samples[i * stride];
        }
        return output;
    }
```

<sub>Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!</sub>

How can I resolve this? If you propose a fix, please make it concise.


output.reserve(output_size);

const float* __restrict src_ptr = samples.data();
const size_t src_size = samples.size();

const size_t safe_output_limit = (output_size > 0) ? output_size - 1 : 0;

double pos = 0.0;

for (size_t i = 0; i < safe_output_limit; ++i) {
size_t idx0 = static_cast<size_t>(pos);
if (idx0 >= src_size - 1) break;

double frac = pos - idx0;

float val0 = src_ptr[idx0];
float val1 = src_ptr[idx0 + 1];

output.push_back(val0 + static_cast<float>(frac) * (val1 - val0));
pos += step;
}

while (output.size() < output_size) {
size_t idx0 = static_cast<size_t>(pos);
if (idx0 >= src_size) idx0 = src_size - 1;

std::vector<float> output(output_size);
size_t idx1 = (idx0 + 1 < src_size) ? idx0 + 1 : src_size - 1;

for (size_t i = 0; i < output_size; ++i) {
const double src_idx = i / ratio;
const size_t idx0 = static_cast<size_t>(src_idx);
const size_t idx1 = std::min(idx0 + 1, samples.size() - 1);
const double frac = src_idx - idx0;
double frac = pos - static_cast<double>(idx0);
float val0 = src_ptr[idx0];
float val1 = src_ptr[idx1];

output[i] = static_cast<float>(samples[idx0] * (1.0 - frac) + samples[idx1] * frac);
output.push_back(val0 + static_cast<float>(frac) * (val1 - val0));
pos += step;
}
Copy link
Copy Markdown
Contributor

@greptile-apps greptile-apps bot Feb 5, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looping with push_back after exact size calculation reduces performance benefit of vectorization optimization. Consider using indexed writes with pre-sized vector since output_size is known upfront

Prompt To Fix With AI 
This is a comment left during a code review.
Path: sdk/runanywhere-commons/src/backends/whispercpp/whispercpp_backend.cpp
Line: 597:609

Comment:
Looping with `push_back` after exact size calculation reduces performance benefit of vectorization optimization. Consider using indexed writes with pre-sized vector since `output_size` is known upfront

How can I resolve this? If you propose a fix, please make it concise.


LOGI("Resampled audio from %d Hz to %d Hz (%zu -> %zu samples)", source_rate,
Expand Down
Loading