Audio: Keep fixed buffer style consistent
This commit is contained in:
parent
67131caef3
commit
496f01253e
1 changed files with 5 additions and 5 deletions
|
@ -309,7 +309,7 @@ namespace Ryujinx.HLE.HOS.Services.Aud.AudioRenderer
|
|||
{
|
||||
int[] MixBuffer = new int[MixBufferSamplesCount * AudioConsts.HostChannelsCount];
|
||||
|
||||
fixed (int* mixptr = &MixBuffer[0])
|
||||
fixed (int* mixptr = MixBuffer)
|
||||
{
|
||||
foreach (VoiceContext Voice in Voices)
|
||||
{
|
||||
|
@ -339,7 +339,7 @@ namespace Ryujinx.HLE.HOS.Services.Aud.AudioRenderer
|
|||
|
||||
if(Avx2.IsSupported && Avx.IsSupported)
|
||||
{
|
||||
fixed (int* samptr = &Samples[0])
|
||||
fixed (int* samptr = Samples)
|
||||
{
|
||||
// Load our scale factor as a scalar
|
||||
Vector256<float> volume = Avx.SetAllVector256(Volume);
|
||||
|
@ -369,7 +369,7 @@ namespace Ryujinx.HLE.HOS.Services.Aud.AudioRenderer
|
|||
}
|
||||
else if (Sse2.IsSupported && Sse.IsSupported)
|
||||
{
|
||||
fixed (int* samptr = &Samples[0])
|
||||
fixed (int* samptr = Samples)
|
||||
{
|
||||
// Load our scale factor as a scalar
|
||||
Vector128<float> volume = Sse.SetAllVector128(Volume);
|
||||
|
@ -419,8 +419,8 @@ namespace Ryujinx.HLE.HOS.Services.Aud.AudioRenderer
|
|||
// Perform Saturation using SSE2 if supported
|
||||
if (Sse2.IsSupported)
|
||||
{
|
||||
fixed (int* inptr = &Buffer[0])
|
||||
fixed (short* outptr = &Output[0])
|
||||
fixed (int* inptr = Buffer)
|
||||
fixed (short* outptr = Output)
|
||||
{
|
||||
for (; Offset + 32 <= Buffer.Length; Offset += 32)
|
||||
{
|
||||
|
|
Loading…
Add table
Reference in a new issue