Small callgrind-informed optimisation.

diff --git a/src/sound_asset_writer.cc b/src/sound_asset_writer.cc
index 35c3dab0d4b6222bb17ca89d91bc00bc1f68a0df..fccdda2814a2e8fc64df514b5eafe0d525a0dd86 100644 (file)
--- a/src/sound_asset_writer.cc
+++ b/src/sound_asset_writer.cc
@@ -91,19 +91,27 @@ SoundAssetWriter::write (float const * const * data, int frames)
                _started = true;
        }
 
+       int const ch = _sound_asset->channels ();
+
        for (int i = 0; i < frames; ++i) {
 
                byte_t* out = _state->frame_buffer.Data() + _frame_buffer_offset;
 
                /* Write one sample per channel */
-               for (int j = 0; j < _sound_asset->channels(); ++j) {
+               for (int j = 0; j < ch; ++j) {
                        /* Convert sample to 24-bit int, clipping if necessary. */
-                       int32_t const s = min (clip, max (-clip, data[j][i])) * (1 << 23);
+                       float x = data[j][i];
+                       if (x > clip) {
+                               x = clip;
+                       } else if (x < -clip) {
+                               x = -clip;
+                       }
+                       int32_t const s = x * (1 << 23);
                        *out++ = (s & 0xff);
                        *out++ = (s & 0xff00) >> 8;
                        *out++ = (s & 0xff0000) >> 16;
                }
-               _frame_buffer_offset += 3 * _sound_asset->channels();
+               _frame_buffer_offset += 3 * ch;
 
                DCP_ASSERT (_frame_buffer_offset <= int (_state->frame_buffer.Capacity()));