claunia/cuetools.net
1
0
Fork 0
mirror of https://github.com/claunia/cuetools.net.git synced 2025-12-16 18:14:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

experiment with Burg LPC method

Browse Source
This commit is contained in:
chudov
2009-09-28 06:27:28 +00:00
parent 3c287e2ede
commit a02f79a3f8
4 changed files with 1078 additions and 1376 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
CUETools.FlaCuda/CUETools.Codecs.FlaCuda.csproj
View File

@@ -69,6 +69,6 @@
<PropertyGroup> <PropertyGroup>
<PostBuildEvent> <PostBuildEvent>
</PostBuildEvent> </PostBuildEvent>
<PreBuildEvent>nvcc $(ProjectDir)flacuda.cu -o $(ProjectDir)\flacuda.cubin --machine 32 --maxrregcount 10 --cubin --compiler-bindir "C:\Program Files (x86)\Microsoft Visual Studio 8\VC\bin" --system-include "C:\Program Files (x86)\Microsoft Visual Studio 8\VC\include"</PreBuildEvent> <PreBuildEvent>nvcc $(ProjectDir)flacuda.cu -o $(ProjectDir)\flacuda.cubin --machine 32 --cubin --compiler-bindir "C:\Program Files (x86)\Microsoft Visual Studio 8\VC\bin" --system-include "C:\Program Files (x86)\Microsoft Visual Studio 8\VC\include"</PreBuildEvent>
</PropertyGroup> </PropertyGroup>
</Project> </Project>

12
CUETools.FlaCuda/FlaCudaWriter.cs
View File

@@ -853,6 +853,7 @@ namespace CUETools.Codecs.FlaCuda
task.ResidualTasks[task.nResidualTasks].type = (int)SubframeType.LPC; task.ResidualTasks[task.nResidualTasks].type = (int)SubframeType.LPC;
task.ResidualTasks[task.nResidualTasks].channel = ch; task.ResidualTasks[task.nResidualTasks].channel = ch;
task.ResidualTasks[task.nResidualTasks].obits = (int)bits_per_sample + (channels == 2 && ch == 3 ? 1 : 0); task.ResidualTasks[task.nResidualTasks].obits = (int)bits_per_sample + (channels == 2 && ch == 3 ? 1 : 0);
task.ResidualTasks[task.nResidualTasks].abits = task.ResidualTasks[task.nResidualTasks].obits;
task.ResidualTasks[task.nResidualTasks].blocksize = blocksize; task.ResidualTasks[task.nResidualTasks].blocksize = blocksize;
task.ResidualTasks[task.nResidualTasks].residualOrder = order; task.ResidualTasks[task.nResidualTasks].residualOrder = order;
task.ResidualTasks[task.nResidualTasks].samplesOffs = ch * FlaCudaWriter.MAX_BLOCKSIZE + iFrame * blocksize; task.ResidualTasks[task.nResidualTasks].samplesOffs = ch * FlaCudaWriter.MAX_BLOCKSIZE + iFrame * blocksize;
@@ -866,6 +867,7 @@ namespace CUETools.Codecs.FlaCuda
task.ResidualTasks[task.nResidualTasks].type = (int)SubframeType.Constant; task.ResidualTasks[task.nResidualTasks].type = (int)SubframeType.Constant;
task.ResidualTasks[task.nResidualTasks].channel = ch; task.ResidualTasks[task.nResidualTasks].channel = ch;
task.ResidualTasks[task.nResidualTasks].obits = (int)bits_per_sample + (channels == 2 && ch == 3 ? 1 : 0); task.ResidualTasks[task.nResidualTasks].obits = (int)bits_per_sample + (channels == 2 && ch == 3 ? 1 : 0);
task.ResidualTasks[task.nResidualTasks].abits = task.ResidualTasks[task.nResidualTasks].obits;
task.ResidualTasks[task.nResidualTasks].blocksize = blocksize; task.ResidualTasks[task.nResidualTasks].blocksize = blocksize;
task.ResidualTasks[task.nResidualTasks].samplesOffs = ch * FlaCudaWriter.MAX_BLOCKSIZE + iFrame * blocksize; task.ResidualTasks[task.nResidualTasks].samplesOffs = ch * FlaCudaWriter.MAX_BLOCKSIZE + iFrame * blocksize;
task.ResidualTasks[task.nResidualTasks].residualOffs = task.ResidualTasks[task.nResidualTasks].samplesOffs; task.ResidualTasks[task.nResidualTasks].residualOffs = task.ResidualTasks[task.nResidualTasks].samplesOffs;
@@ -880,6 +882,7 @@ namespace CUETools.Codecs.FlaCuda
task.ResidualTasks[task.nResidualTasks].type = (int)SubframeType.Fixed; task.ResidualTasks[task.nResidualTasks].type = (int)SubframeType.Fixed;
task.ResidualTasks[task.nResidualTasks].channel = ch; task.ResidualTasks[task.nResidualTasks].channel = ch;
task.ResidualTasks[task.nResidualTasks].obits = (int)bits_per_sample + (channels == 2 && ch == 3 ? 1 : 0); task.ResidualTasks[task.nResidualTasks].obits = (int)bits_per_sample + (channels == 2 && ch == 3 ? 1 : 0);
task.ResidualTasks[task.nResidualTasks].abits = task.ResidualTasks[task.nResidualTasks].obits;
task.ResidualTasks[task.nResidualTasks].blocksize = blocksize; task.ResidualTasks[task.nResidualTasks].blocksize = blocksize;
task.ResidualTasks[task.nResidualTasks].residualOrder = order; task.ResidualTasks[task.nResidualTasks].residualOrder = order;
task.ResidualTasks[task.nResidualTasks].samplesOffs = ch * FlaCudaWriter.MAX_BLOCKSIZE + iFrame * blocksize; task.ResidualTasks[task.nResidualTasks].samplesOffs = ch * FlaCudaWriter.MAX_BLOCKSIZE + iFrame * blocksize;
@@ -916,6 +919,7 @@ namespace CUETools.Codecs.FlaCuda
task.ResidualTasks[task.nResidualTasks].type = (int)SubframeType.Verbatim; task.ResidualTasks[task.nResidualTasks].type = (int)SubframeType.Verbatim;
task.ResidualTasks[task.nResidualTasks].channel = ch; task.ResidualTasks[task.nResidualTasks].channel = ch;
task.ResidualTasks[task.nResidualTasks].obits = (int)bits_per_sample + (channels == 2 && ch == 3 ? 1 : 0); task.ResidualTasks[task.nResidualTasks].obits = (int)bits_per_sample + (channels == 2 && ch == 3 ? 1 : 0);
task.ResidualTasks[task.nResidualTasks].abits = task.ResidualTasks[task.nResidualTasks].obits;
task.ResidualTasks[task.nResidualTasks].blocksize = blocksize; task.ResidualTasks[task.nResidualTasks].blocksize = blocksize;
task.ResidualTasks[task.nResidualTasks].residualOrder = 0; task.ResidualTasks[task.nResidualTasks].residualOrder = 0;
task.ResidualTasks[task.nResidualTasks].samplesOffs = ch * FlaCudaWriter.MAX_BLOCKSIZE + iFrame * blocksize; task.ResidualTasks[task.nResidualTasks].samplesOffs = ch * FlaCudaWriter.MAX_BLOCKSIZE + iFrame * blocksize;
@@ -1101,7 +1105,8 @@ namespace CUETools.Codecs.FlaCuda
cuda.SetParameter(task.cudaComputeLPCLattice, 0, (uint)task.cudaResidualTasks.Pointer); cuda.SetParameter(task.cudaComputeLPCLattice, 0, (uint)task.cudaResidualTasks.Pointer);
cuda.SetParameter(task.cudaComputeLPCLattice, 1 * sizeof(uint), (uint)task.nResidualTasksPerChannel); cuda.SetParameter(task.cudaComputeLPCLattice, 1 * sizeof(uint), (uint)task.nResidualTasksPerChannel);
cuda.SetParameter(task.cudaComputeLPCLattice, 2 * sizeof(uint), (uint)task.cudaSamples.Pointer); cuda.SetParameter(task.cudaComputeLPCLattice, 2 * sizeof(uint), (uint)task.cudaSamples.Pointer);
cuda.SetParameter(task.cudaComputeLPCLattice, 3 * sizeof(uint), (uint)task.frameSize); cuda.SetParameter(task.cudaComputeLPCLattice, 3 * sizeof(uint), (uint)_windowcount);
//cuda.SetParameter(task.cudaComputeLPCLattice, 3 * sizeof(uint), (uint)task.frameSize);
cuda.SetParameter(task.cudaComputeLPCLattice, 4 * sizeof(uint), (uint)eparams.max_prediction_order); cuda.SetParameter(task.cudaComputeLPCLattice, 4 * sizeof(uint), (uint)eparams.max_prediction_order);
cuda.SetParameterSize(task.cudaComputeLPCLattice, 5U * sizeof(uint)); cuda.SetParameterSize(task.cudaComputeLPCLattice, 5U * sizeof(uint));
cuda.SetFunctionBlockShape(task.cudaComputeLPCLattice, 256, 1, 1); cuda.SetFunctionBlockShape(task.cudaComputeLPCLattice, 256, 1, 1);
@@ -1142,7 +1147,7 @@ namespace CUETools.Codecs.FlaCuda
// issue work to the GPU // issue work to the GPU
cuda.LaunchAsync(cudaChannelDecorr, (task.frameCount * task.frameSize + 255) / 256, channels == 2 ? 1 : channels, task.stream); cuda.LaunchAsync(cudaChannelDecorr, (task.frameCount * task.frameSize + 255) / 256, channels == 2 ? 1 : channels, task.stream);
if (task.frameSize <= 512 && _windowcount == 1) if (task.frameSize <= 512 && eparams.max_prediction_order <= 12)
cuda.LaunchAsync(task.cudaComputeLPCLattice, 1, channelsCount * task.frameCount, task.stream); cuda.LaunchAsync(task.cudaComputeLPCLattice, 1, channelsCount * task.frameCount, task.stream);
else else
{ {
@@ -1831,7 +1836,8 @@ namespace CUETools.Codecs.FlaCuda
public int channel; public int channel;
public int residualOffs; public int residualOffs;
public int wbits; public int wbits;
public fixed int reserved[4]; public int abits;
public fixed int reserved[3];
public fixed int coefs[32]; public fixed int coefs[32];
}; };

476
CUETools.FlaCuda/flacuda.cu
View File

@@ -51,10 +51,19 @@ typedef struct
int channel; int channel;
int residualOffs; int residualOffs;
int wbits; int wbits;
int reserved[4]; int abits;
int reserved[3];
int coefs[32]; int coefs[32];
} encodeResidualTaskStruct; } encodeResidualTaskStruct;
#define SUM32(buf,tid,op) buf[tid] op buf[tid + 16]; buf[tid] op buf[tid + 8]; buf[tid] op buf[tid + 4]; buf[tid] op buf[tid + 2]; buf[tid] op buf[tid + 1];
#define SUM64(buf,tid,op) if (tid < 32) buf[tid] op buf[tid + 32]; __syncthreads(); if (tid < 32) { SUM32(buf,tid,op) }
#define SUM128(buf,tid,op) if (tid < 64) buf[tid] op buf[tid + 64]; __syncthreads(); SUM64(buf,tid,op)
#define SUM256(buf,tid,op) if (tid < 128) buf[tid] op buf[tid + 128]; __syncthreads(); SUM128(buf,tid,op)
#define SUM512(buf,tid,op) if (tid < 256) buf[tid] op buf[tid + 256]; __syncthreads(); SUM256(buf,tid,op)
#define FSQR(s) ((s)*(s))
extern "C" __global__ void cudaStereoDecorr( extern "C" __global__ void cudaStereoDecorr(
int *samples, int *samples,
short2 *src, short2 *src,
@@ -107,28 +116,35 @@ extern "C" __global__ void cudaFindWastedBits(
{ {
__shared__ struct { __shared__ struct {
volatile int wbits[256]; volatile int wbits[256];
volatile int abits[256];
encodeResidualTaskStruct task; encodeResidualTaskStruct task;
} shared; } shared;
if (threadIdx.x < 16) if (threadIdx.x < 16)
((int*)&shared.task)[threadIdx.x] = ((int*)(&tasks[blockIdx.x * tasksPerChannel]))[threadIdx.x]; ((int*)&shared.task)[threadIdx.x] = ((int*)(&tasks[blockIdx.x * tasksPerChannel]))[threadIdx.x];
shared.wbits[threadIdx.x] = 0; shared.wbits[threadIdx.x] = 0;
shared.abits[threadIdx.x] = 0;
__syncthreads(); __syncthreads();
for (int pos = 0; pos < blocksize; pos += blockDim.x) for (int pos = 0; pos < blocksize; pos += blockDim.x)
shared.wbits[threadIdx.x] |= pos + threadIdx.x < blocksize ? samples[shared.task.samplesOffs + pos + threadIdx.x] : 0; {
int smp = pos + threadIdx.x < blocksize ? samples[shared.task.samplesOffs + pos + threadIdx.x] : 0;
shared.wbits[threadIdx.x] |= smp;
shared.abits[threadIdx.x] |= smp ^ (smp >> 31);
}
__syncthreads();
SUM256(shared.wbits, threadIdx.x, |=);
SUM256(shared.abits, threadIdx.x, |=);
if (threadIdx.x == 0)
shared.task.wbits = max(0,__ffs(shared.wbits[0]) - 1);
if (threadIdx.x == 0)
shared.task.abits = 32 - __clz(shared.abits[0]) - shared.task.wbits;
__syncthreads(); __syncthreads();
if (threadIdx.x < 128) shared.wbits[threadIdx.x] |= shared.wbits[threadIdx.x + 128]; __syncthreads();
if (threadIdx.x < 64) shared.wbits[threadIdx.x] |= shared.wbits[threadIdx.x + 64]; __syncthreads();
if (threadIdx.x < 32) shared.wbits[threadIdx.x] |= shared.wbits[threadIdx.x + 32]; __syncthreads();
shared.wbits[threadIdx.x] |= shared.wbits[threadIdx.x + 16];
shared.wbits[threadIdx.x] |= shared.wbits[threadIdx.x + 8];
shared.wbits[threadIdx.x] |= shared.wbits[threadIdx.x + 4];
shared.wbits[threadIdx.x] |= shared.wbits[threadIdx.x + 2];
shared.wbits[threadIdx.x] |= shared.wbits[threadIdx.x + 1];
if (threadIdx.x < tasksPerChannel) if (threadIdx.x < tasksPerChannel)
tasks[blockIdx.x * tasksPerChannel + threadIdx.x].wbits = max(0,__ffs(shared.wbits[0]) - 1); tasks[blockIdx.x * tasksPerChannel + threadIdx.x].wbits = shared.task.wbits;
if (threadIdx.x < tasksPerChannel)
tasks[blockIdx.x * tasksPerChannel + threadIdx.x].abits = shared.task.abits;
} }
extern "C" __global__ void cudaComputeAutocor( extern "C" __global__ void cudaComputeAutocor(
@@ -191,6 +207,7 @@ extern "C" __global__ void cudaComputeLPC(
{ {
__shared__ struct { __shared__ struct {
computeAutocorTaskStruct task; computeAutocorTaskStruct task;
encodeResidualTaskStruct task2;
volatile float ldr[32]; volatile float ldr[32];
volatile int bits[32]; volatile int bits[32];
volatile float autoc[33]; volatile float autoc[33];
@@ -205,6 +222,10 @@ extern "C" __global__ void cudaComputeLPC(
// fetch task data // fetch task data
if (tid < sizeof(shared.task) / sizeof(int)) if (tid < sizeof(shared.task) / sizeof(int))
((int*)&shared.task)[tid] = ((int*)(tasks + blockIdx.y))[tid]; ((int*)&shared.task)[tid] = ((int*)(tasks + blockIdx.y))[tid];
__syncthreads();
if (tid < sizeof(shared.task2) / sizeof(int))
((int*)&shared.task2)[tid] = ((int*)(output + shared.task.residualOffs))[tid];
__syncthreads();
// add up parts // add up parts
for (int order = 0; order <= max_order; order++) for (int order = 0; order <= max_order; order++)
@@ -251,7 +272,8 @@ extern "C" __global__ void cudaComputeLPC(
shared.ldr[tid] += (tid < order) * __fmul_rz(reff, shared.ldr[order - 1 - tid]) + (tid == order) * reff; shared.ldr[tid] += (tid < order) * __fmul_rz(reff, shared.ldr[order - 1 - tid]) + (tid == order) * reff;
// Quantization // Quantization
int precision = 13 - (order > 8) - (shared.task.blocksize <= 2304) - (shared.task.blocksize <= 1152) - (shared.task.blocksize <= 576); //int precision = 13 - (shared.task.blocksize <= 2304) - (shared.task.blocksize <= 1152) - (shared.task.blocksize <= 576);
int precision = max(3, min(13 - (shared.task.blocksize <= 2304) - (shared.task.blocksize <= 1152) - (shared.task.blocksize <= 576), shared.task2.abits));
int taskNo = shared.task.residualOffs + order; int taskNo = shared.task.residualOffs + order;
shared.bits[tid] = __mul24((33 - __clz(__float2int_rn(fabs(shared.ldr[tid]) * (1 << 15))) - precision), tid <= order); shared.bits[tid] = __mul24((33 - __clz(__float2int_rn(fabs(shared.ldr[tid]) * (1 << 15))) - precision), tid <= order);
shared.bits[tid] = max(shared.bits[tid], shared.bits[tid + 16]); shared.bits[tid] = max(shared.bits[tid], shared.bits[tid + 16]);
@@ -267,7 +289,7 @@ extern "C" __global__ void cudaComputeLPC(
output[taskNo].coefs[tid] = coef; output[taskNo].coefs[tid] = coef;
if (tid == 0) if (tid == 0)
output[taskNo].shift = sh; output[taskNo].shift = sh;
shared.bits[tid] = __mul24(33 - max(__clz(coef),__clz(-1 ^ coef)), tid <= order); shared.bits[tid] = __mul24(33 - __clz(coef ^ (coef >> 31)), tid <= order);
shared.bits[tid] = max(shared.bits[tid], shared.bits[tid + 16]); shared.bits[tid] = max(shared.bits[tid], shared.bits[tid + 16]);
shared.bits[tid] = max(shared.bits[tid], shared.bits[tid + 8]); shared.bits[tid] = max(shared.bits[tid], shared.bits[tid + 8]);
shared.bits[tid] = max(shared.bits[tid], shared.bits[tid + 4]); shared.bits[tid] = max(shared.bits[tid], shared.bits[tid + 4]);
@@ -280,39 +302,22 @@ extern "C" __global__ void cudaComputeLPC(
} }
} }
#define SUM32(buf,tid,op) buf[tid] op buf[tid + 16]; buf[tid] op buf[tid + 8]; buf[tid] op buf[tid + 4]; buf[tid] op buf[tid + 2]; buf[tid] op buf[tid + 1];
#define SUM64(buf,tid,op) if (tid < 32) buf[tid] op buf[tid + 32]; __syncthreads(); if (tid < 32) SUM32(buf,tid,op)
#define SUM128(buf,tid,op) if (tid < 64) buf[tid] op buf[tid + 64]; __syncthreads(); SUM64(buf,tid,op)
#define SUM256(buf,tid,op) if (tid < 128) buf[tid] op buf[tid + 128]; __syncthreads(); SUM128(buf,tid,op)
#define SUM512(buf,tid,op) if (tid < 256) buf[tid] op buf[tid + 256]; __syncthreads(); SUM256(buf,tid,op)
#define FSQR(s) ((s)*(s))
extern "C" __global__ void cudaComputeLPCLattice( extern "C" __global__ void cudaComputeLPCLattice(
encodeResidualTaskStruct *tasks, encodeResidualTaskStruct *tasks,
const int taskCount, // tasks per block const int taskCount, // tasks per block
const int *samples, const int *samples,
const int frameSize, // <= 512 const int precisions,
const int max_order // should be <= 32 const int max_order // should be <= 12
) )
{ {
__shared__ struct { __shared__ struct {
encodeResidualTaskStruct task; volatile encodeResidualTaskStruct task;
volatile float F[512];
volatile float lpc[12][32];
union { union {
volatile float F[512]; volatile float tmp[256];
volatile int tmpi[512]; volatile int tmpi[256];
}; };
union {
volatile float B[512];
volatile int smp[512];
};
volatile float tmp[256];
volatile float arp[32];
volatile float rc[32];
int bits[32];
volatile float PE[33];
volatile float DEN, reff;
int actual_bits;
} shared; } shared;
// fetch task data // fetch task data
@@ -321,247 +326,244 @@ extern "C" __global__ void cudaComputeLPCLattice(
__syncthreads(); __syncthreads();
// F = samples; B = samples // F = samples; B = samples
shared.tmpi[threadIdx.x] = shared.smp[threadIdx.x] = threadIdx.x < frameSize ? samples[shared.task.samplesOffs + threadIdx.x] : 0; //int frameSize = shared.task.blocksize;
shared.tmpi[threadIdx.x + 256] = shared.smp[threadIdx.x + 256] = threadIdx.x + 256 < frameSize ? samples[shared.task.samplesOffs + threadIdx.x + 256] : 0; int s1 = threadIdx.x < shared.task.blocksize ? samples[shared.task.samplesOffs + threadIdx.x] : 0;
int s2 = threadIdx.x + 256 < shared.task.blocksize ? samples[shared.task.samplesOffs + threadIdx.x + 256] : 0;
shared.tmpi[threadIdx.x] = s1|s2;
__syncthreads(); __syncthreads();
SUM512(shared.tmpi,threadIdx.x,|=); SUM256(shared.tmpi,threadIdx.x,|=);
if (threadIdx.x == 0) if (threadIdx.x == 0)
shared.task.wbits = max(0,__ffs(shared.tmpi[0]) - 1); shared.task.wbits = max(0,__ffs(shared.tmpi[0]) - 1);
__syncthreads(); __syncthreads();
shared.tmpi[threadIdx.x] = shared.smp[threadIdx.x] ^ (shared.smp[threadIdx.x] >> 31); if (threadIdx.x < taskCount)
shared.tmpi[threadIdx.x + 256] = shared.smp[threadIdx.x + 256] ^ (shared.smp[threadIdx.x + 256] >> 31); tasks[blockIdx.y * taskCount + threadIdx.x].wbits = shared.task.wbits;
SUM512(shared.tmpi,threadIdx.x,|=); shared.tmpi[threadIdx.x] = (s1 ^ (s1 >> 31)) | (s2 ^ (s2 >> 31));
if (threadIdx.x == 0)
shared.actual_bits = 32 - __clz(shared.tmpi[0]) - shared.task.wbits;
__syncthreads(); __syncthreads();
shared.F[threadIdx.x] = shared.smp[threadIdx.x] >> shared.task.wbits; SUM256(shared.tmpi,threadIdx.x,|=);
shared.F[threadIdx.x + 256] = shared.smp[threadIdx.x + 256] >> shared.task.wbits; if (threadIdx.x == 0)
shared.B[threadIdx.x] = shared.F[threadIdx.x]; shared.task.abits = 32 - __clz(shared.tmpi[0]) - shared.task.wbits;
shared.B[threadIdx.x + 256] = shared.F[threadIdx.x + 256]; __syncthreads();
s1 >>= shared.task.wbits;
s2 >>= shared.task.wbits;
shared.F[threadIdx.x] = s1;
shared.F[threadIdx.x + 256] = s2;
__syncthreads(); __syncthreads();
for (int order = 1; order <= max_order; order++) for (int order = 1; order <= max_order; order++)
{ {
// reff = F(order+1:frameSize) * B(1:frameSize-order)' / DEN // reff = F(order+1:frameSize) * B(1:frameSize-order)' / DEN
float f1 = (threadIdx.x + order < frameSize) * shared.F[threadIdx.x + order]; float f1 = (threadIdx.x + order < shared.task.blocksize) * shared.F[threadIdx.x + order];
float f2 = (threadIdx.x + 256 + order < frameSize) * shared.F[threadIdx.x + 256 + order]; float f2 = (threadIdx.x + 256 + order < shared.task.blocksize) * shared.F[threadIdx.x + 256 + order];
s1 *= (threadIdx.x + order < shared.task.blocksize);
s2 *= (threadIdx.x + 256 + order < shared.task.blocksize);
// DEN = F(order+1:frameSize) * F(order+1:frameSize)' + B(1:frameSize-order) * B(1:frameSize-order)' (BURG) // DEN = F(order+1:frameSize) * F(order+1:frameSize)' + B(1:frameSize-order) * B(1:frameSize-order)' (BURG)
shared.tmp[threadIdx.x] = FSQR(f1) + FSQR(f2); shared.tmp[threadIdx.x] = FSQR(f1) + FSQR(f2) + FSQR(s1) + FSQR(s2);
shared.tmp[threadIdx.x] += (threadIdx.x < frameSize - order) * FSQR(shared.B[threadIdx.x])
+ (threadIdx.x + 256 < frameSize - order) * FSQR(shared.B[threadIdx.x + 256]);
__syncthreads(); __syncthreads();
SUM256(shared.tmp, threadIdx.x, +=); SUM256(shared.tmp, threadIdx.x, +=);
if (threadIdx.x == 0) __syncthreads();
{ float DEN = shared.tmp[0] / 2;
shared.DEN = shared.tmp[0] / 2; //shared.PE[order-1] = shared.tmp[0] / 2 / (frameSize - order + 1);
shared.PE[order-1] = shared.tmp[0] / 2 / (frameSize - order + 1);
}
__syncthreads(); __syncthreads();
shared.tmp[threadIdx.x] = f1 * s1 + f2 * s2;
shared.tmp[threadIdx.x] = f1 * shared.B[threadIdx.x] + f2 * shared.B[threadIdx.x + 256];
__syncthreads(); __syncthreads();
SUM256(shared.tmp, threadIdx.x, +=); SUM256(shared.tmp, threadIdx.x, +=);
if (threadIdx.x == 0) __syncthreads();
shared.reff = shared.tmp[0] / shared.DEN; float reff = shared.tmp[0] / DEN;
__syncthreads(); __syncthreads();
// arp(order) = rc(order) = reff // arp(order) = rc(order) = reff
if (threadIdx.x == 0) if (threadIdx.x == 0)
shared.arp[order - 1] = shared.rc[order - 1] = shared.reff; shared.lpc[order - 1][order - 1] = reff;
//shared.rc[order - 1] = shared.lpc[order - 1][order - 1] = reff;
// Levinson-Durbin recursion // Levinson-Durbin recursion
// arp(1:order-1) = arp(1:order-1) - reff * arp(order-1:-1:1) // arp(1:order-1) = arp(1:order-1) - reff * arp(order-1:-1:1)
if (threadIdx.x < order - 1) if (threadIdx.x < order - 1)
shared.arp[threadIdx.x] -= shared.reff * shared.arp[order - 2 - threadIdx.x]; shared.lpc[order - 1][threadIdx.x] = shared.lpc[order - 2][threadIdx.x] - reff * shared.lpc[order - 2][order - 2 - threadIdx.x];
// F1 = F(order+1:frameSize) - reff * B(1:frameSize-order) // F1 = F(order+1:frameSize) - reff * B(1:frameSize-order)
// B(1:frameSize-order) = B(1:frameSize-order) - reff * F(order+1:frameSize) // B(1:frameSize-order) = B(1:frameSize-order) - reff * F(order+1:frameSize)
// F(order+1:frameSize) = F1 // F(order+1:frameSize) = F1
if (threadIdx.x < frameSize - order) if (threadIdx.x < shared.task.blocksize - order)
{ shared.F[order + threadIdx.x] -= reff * s1;
shared.F[order + threadIdx.x] -= shared.reff * shared.B[threadIdx.x]; if (threadIdx.x + 256 < shared.task.blocksize - order)
shared.B[threadIdx.x] -= shared.reff * f1; shared.F[order + threadIdx.x + 256] -= reff * s2;
} s1 -= reff * f1;
if (threadIdx.x + 256 < frameSize - order) s2 -= reff * f2;
{
shared.F[order + threadIdx.x + 256] -= shared.reff * shared.B[threadIdx.x + 256];
shared.B[threadIdx.x + 256] -= shared.reff * f2;
}
__syncthreads();
// Quantization
if (threadIdx.x < 32)
{
// get 15 bits of each coeff
shared.bits[threadIdx.x] = __mul24(__float2int_rn(shared.arp[threadIdx.x] * (1 << 15)), threadIdx.x < order);
// remove sign bits
shared.bits[threadIdx.x] = shared.bits[threadIdx.x] ^ (shared.bits[threadIdx.x] >> 31);
// OR reduction
SUM32(shared.bits,threadIdx.x,|=);
// choose precision
if (threadIdx.x == 0)
shared.task.cbits = max(3, min(10, shared.actual_bits)); // - __float2int_rn(shared.PE[order - 1])
// calculate shift based on precision and number of leading zeroes in coeffs
if (threadIdx.x == 0)
shared.task.shift = max(0,min(15, __clz(shared.bits[0]) - 18 + shared.task.cbits));
// quantize coeffs with given shift
shared.task.coefs[threadIdx.x] = max(-(1 << (shared.task.cbits - 1)), min((1 << (shared.task.cbits - 1))-1, __float2int_rn(shared.arp[threadIdx.x] * (1 << shared.task.shift))));
// error correction
//shared.tmp[threadIdx.x] = (threadIdx.x != 0) * (shared.arp[threadIdx.x - 1]*(1 << shared.task.shift) - shared.task.coefs[threadIdx.x - 1]);
//shared.task.coefs[threadIdx.x] = max(-(1 << (shared.task.cbits - 1)), min((1 << (shared.task.cbits - 1))-1, __float2int_rn((shared.arp[threadIdx.x]) * (1 << shared.task.shift) + shared.tmp[threadIdx.x])));
// remove sign bits
shared.bits[threadIdx.x] = __mul24(shared.task.coefs[threadIdx.x] ^ (shared.task.coefs[threadIdx.x] >> 31), threadIdx.x < order);
// OR reduction
SUM32(shared.bits,threadIdx.x,|=);
// calculate actual number of bits (+1 for sign)
if (threadIdx.x == 0)
shared.task.cbits = 1 + 32 - __clz(shared.bits[0]);
// output shift, cbits and output coeffs in reverse order
int taskNo = taskCount * blockIdx.y + order - 1;
if (threadIdx.x == 0)
tasks[taskNo].shift = shared.task.shift;
if (threadIdx.x == 0)
tasks[taskNo].cbits = shared.task.cbits;
if (threadIdx.x < order)
tasks[taskNo].coefs[threadIdx.x] = shared.task.coefs[order - 1 - threadIdx.x];
}
}
if (threadIdx.x < taskCount)
tasks[blockIdx.y * taskCount + threadIdx.x].wbits = shared.task.wbits;
}
extern "C" __global__ void cudaComputeLPCLattice512(
encodeResidualTaskStruct *tasks,
const int taskCount, // tasks per block
const int *samples,
const int frameSize, // <= 512
const int max_order // should be <= 32
)
{
__shared__ struct {
encodeResidualTaskStruct task;
float F[512];
float B[512];
float lpc[32][32];
volatile float tmp[512];
volatile float arp[32];
volatile float rc[32];
volatile int bits[512];
volatile float f, b;
} shared;
// fetch task data
if (threadIdx.x < sizeof(shared.task) / sizeof(int))
((int*)&shared.task)[threadIdx.x] = ((int*)(tasks + taskCount * blockIdx.y))[threadIdx.x];
__syncthreads();
// F = samples; B = samples
shared.F[threadIdx.x] = threadIdx.x < frameSize ? samples[shared.task.samplesOffs + threadIdx.x] >> shared.task.wbits : 0.0f;
shared.B[threadIdx.x] = shared.F[threadIdx.x];
__syncthreads();
// DEN = F*F'
shared.tmp[threadIdx.x] = FSQR(shared.F[threadIdx.x]);
__syncthreads();
SUM512(shared.tmp,threadIdx.x,+=);
__syncthreads();
if (threadIdx.x == 0)
shared.f = shared.b = shared.tmp[0];
// if (threadIdx.x == 0)
//shared.PE[0] = DEN / frameSize;
__syncthreads();
for (int order = 1; order <= max_order; order++)
{
// reff = F(order+1:frameSize) * B(1:frameSize-order)' / DEN
shared.tmp[threadIdx.x] = (threadIdx.x + order < frameSize) * shared.F[threadIdx.x + order] * shared.B[threadIdx.x];
__syncthreads();
SUM512(shared.tmp, threadIdx.x,+=);
__syncthreads();
//float reff = shared.tmp[0] * rsqrtf(shared.b * shared.f); // Geometric lattice
float reff = shared.tmp[0] * 2 / (shared.b + shared.f); // Burg method
__syncthreads();
// Levinson-Durbin recursion
// arp(order) = rc(order) = reff
// arp(1:order-1) = arp(1:order-1) - reff * arp(order-1:-1:1)
if (threadIdx.x == 32)
shared.arp[order - 1] = shared.rc[order - 1] = reff;
if (threadIdx.x < 32)
shared.arp[threadIdx.x] -= (threadIdx.x < order - 1) * __fmul_rz(reff, shared.arp[order - 2 - threadIdx.x]);
// F1 = F(order+1:frameSize) - reff * B(1:frameSize-order)
// B(1:frameSize-order) = B(1:frameSize-order) - reff * F(order+1:frameSize)
// F(order+1:frameSize) = F1
if (threadIdx.x < frameSize - order)
{
float f;// = shared.F[threadIdx.x + order];
shared.F[threadIdx.x + order] = (f = shared.F[threadIdx.x + order]) - reff * shared.B[threadIdx.x];
shared.B[threadIdx.x] -= reff * f;
}
__syncthreads();
// f = F(order+1:frameSize) * F(order+1:frameSize)'
// b = B(1:frameSize-order) * B(1:frameSize-order)'
shared.tmp[threadIdx.x] = (threadIdx.x < frameSize - order) * FSQR(shared.F[threadIdx.x + order]);
__syncthreads();
SUM512(shared.tmp, threadIdx.x,+=);
__syncthreads();
if (threadIdx.x == 0)
shared.f = shared.tmp[0];
__syncthreads();
shared.tmp[threadIdx.x] = (threadIdx.x < frameSize - order) * FSQR(shared.B[threadIdx.x]);
__syncthreads();
SUM512(shared.tmp, threadIdx.x,+=);
__syncthreads();
if (threadIdx.x == 0)
shared.b = shared.tmp[0];
__syncthreads();
if (threadIdx.x < 32)
shared.lpc[order - 1][threadIdx.x] = shared.arp[threadIdx.x];
//if (threadIdx.x == 0)
// shared.PE[order] = (shared.b + shared.f) / 2 / (frameSize - order);
__syncthreads(); __syncthreads();
} }
for (int order = 1 + (threadIdx.x >> 5); order <= max_order; order += 16) // Quantization
for (int order = (threadIdx.x >> 5); order < max_order; order += 8)
for (int precision = 0; precision < precisions; precision++)
{ {
// Quantization
int cn = threadIdx.x & 31; int cn = threadIdx.x & 31;
int precision = 10 - (order > 8) - min(2, shared.task.wbits); // get 15 bits of each coeff
int taskNo = taskCount * blockIdx.y + order - 1; int coef = cn <= order ? __float2int_rn(shared.lpc[order][cn] * (1 << 15)) : 0;
shared.bits[threadIdx.x] = __mul24((33 - __clz(__float2int_rn(fabs(shared.lpc[order - 1][cn]) * (1 << 15))) - precision), cn < order); // remove sign bits
shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 16]); shared.tmpi[threadIdx.x] = coef ^ (coef >> 31);
shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 8]); // OR reduction
shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 4]); SUM32(shared.tmpi,threadIdx.x,|=);
shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 2]); // choose precision
shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 1]); //int cbits = max(3, min(10, 5 + (shared.task.abits >> 1))); // - __float2int_rn(shared.PE[order - 1])
int sh = max(0,min(15, 15 - shared.bits[threadIdx.x - cn])); int cbits = max(3, min(10, shared.task.abits)) - precision;// + precision); // - __float2int_rn(shared.PE[order - 1])
// calculate shift based on precision and number of leading zeroes in coeffs
int shift = max(0,min(15, __clz(shared.tmpi[threadIdx.x & ~31]) - 18 + cbits));
//if (shared.task.abits + 32 - __clz(order) < shift
//int shift = max(0,min(15, (shared.task.abits >> 2) - 14 + __clz(shared.tmpi[threadIdx.x & ~31]) + ((32 - __clz(order))>>1)));
// quantize coeffs with given shift
coef = cn <= order ? max(-(1 << (cbits - 1)), min((1 << (cbits - 1)) -1, __float2int_rn(shared.lpc[order][order - cn] * (1 << shift)))) : 0;
// error correction
//shared.tmp[threadIdx.x] = (threadIdx.x != 0) * (shared.arp[threadIdx.x - 1]*(1 << shared.task.shift) - shared.task.coefs[threadIdx.x - 1]);
//shared.task.coefs[threadIdx.x] = max(-(1 << (shared.task.cbits - 1)), min((1 << (shared.task.cbits - 1))-1, __float2int_rn((shared.arp[threadIdx.x]) * (1 << shared.task.shift) + shared.tmp[threadIdx.x])));
// remove sign bits
shared.tmpi[threadIdx.x] = coef ^ (coef >> 31);
// OR reduction
SUM32(shared.tmpi,threadIdx.x,|=);
// calculate actual number of bits (+1 for sign)
cbits = 1 + 32 - __clz(shared.tmpi[threadIdx.x & ~31]);
// reverse coefs // output shift, cbits and output coeffs
int coef = max(-(1 << precision),min((1 << precision)-1,__float2int_rn(shared.lpc[order - 1][order - 1 - cn] * (1 << sh)))); int taskNo = taskCount * blockIdx.y + order + precision * max_order;
if (cn < order)
tasks[taskNo].coefs[cn] = coef;
if (cn == 0) if (cn == 0)
tasks[taskNo].shift = sh; tasks[taskNo].shift = shift;
shared.bits[threadIdx.x] = __mul24(33 - max(__clz(coef),__clz(-1 ^ coef)), cn < order);
shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 16]);
shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 8]);
shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 4]);
shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 2]);
shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 1]);
int cbits = shared.bits[threadIdx.x - cn];
if (cn == 0) if (cn == 0)
tasks[taskNo].cbits = cbits; tasks[taskNo].cbits = cbits;
if (cn <= order)
tasks[taskNo].coefs[cn] = coef;
} }
} }
//extern "C" __global__ void cudaComputeLPCLattice512(
// encodeResidualTaskStruct *tasks,
// const int taskCount, // tasks per block
// const int *samples,
// const int frameSize, // <= 512
// const int max_order // should be <= 32
//)
//{
// __shared__ struct {
// encodeResidualTaskStruct task;
// float F[512];
// float B[512];
// float lpc[32][32];
// volatile float tmp[512];
// volatile float arp[32];
// volatile float rc[32];
// volatile int bits[512];
// volatile float f, b;
// } shared;
//
// // fetch task data
// if (threadIdx.x < sizeof(shared.task) / sizeof(int))
// ((int*)&shared.task)[threadIdx.x] = ((int*)(tasks + taskCount * blockIdx.y))[threadIdx.x];
// __syncthreads();
//
// // F = samples; B = samples
// shared.F[threadIdx.x] = threadIdx.x < frameSize ? samples[shared.task.samplesOffs + threadIdx.x] >> shared.task.wbits : 0.0f;
// shared.B[threadIdx.x] = shared.F[threadIdx.x];
// __syncthreads();
//
// // DEN = F*F'
// shared.tmp[threadIdx.x] = FSQR(shared.F[threadIdx.x]);
// __syncthreads();
// SUM512(shared.tmp,threadIdx.x,+=);
// __syncthreads();
// if (threadIdx.x == 0)
// shared.f = shared.b = shared.tmp[0];
// // if (threadIdx.x == 0)
// //shared.PE[0] = DEN / frameSize;
// __syncthreads();
//
// for (int order = 1; order <= max_order; order++)
// {
// // reff = F(order+1:frameSize) * B(1:frameSize-order)' / DEN
// shared.tmp[threadIdx.x] = (threadIdx.x + order < frameSize) * shared.F[threadIdx.x + order] * shared.B[threadIdx.x];
// __syncthreads();
// SUM512(shared.tmp, threadIdx.x,+=);
// __syncthreads();
//
// //float reff = shared.tmp[0] * rsqrtf(shared.b * shared.f); // Geometric lattice
// float reff = shared.tmp[0] * 2 / (shared.b + shared.f); // Burg method
// __syncthreads();
//
// // Levinson-Durbin recursion
// // arp(order) = rc(order) = reff
// // arp(1:order-1) = arp(1:order-1) - reff * arp(order-1:-1:1)
// if (threadIdx.x == 32)
// shared.arp[order - 1] = shared.rc[order - 1] = reff;
// if (threadIdx.x < 32)
// shared.arp[threadIdx.x] -= (threadIdx.x < order - 1) * __fmul_rz(reff, shared.arp[order - 2 - threadIdx.x]);
//
// // F1 = F(order+1:frameSize) - reff * B(1:frameSize-order)
// // B(1:frameSize-order) = B(1:frameSize-order) - reff * F(order+1:frameSize)
// // F(order+1:frameSize) = F1
// if (threadIdx.x < frameSize - order)
// {
// float f;// = shared.F[threadIdx.x + order];
// shared.F[threadIdx.x + order] = (f = shared.F[threadIdx.x + order]) - reff * shared.B[threadIdx.x];
// shared.B[threadIdx.x] -= reff * f;
// }
// __syncthreads();
//
// // f = F(order+1:frameSize) * F(order+1:frameSize)'
// // b = B(1:frameSize-order) * B(1:frameSize-order)'
// shared.tmp[threadIdx.x] = (threadIdx.x < frameSize - order) * FSQR(shared.F[threadIdx.x + order]);
// __syncthreads();
// SUM512(shared.tmp, threadIdx.x,+=);
// __syncthreads();
// if (threadIdx.x == 0)
// shared.f = shared.tmp[0];
// __syncthreads();
//
// shared.tmp[threadIdx.x] = (threadIdx.x < frameSize - order) * FSQR(shared.B[threadIdx.x]);
// __syncthreads();
// SUM512(shared.tmp, threadIdx.x,+=);
// __syncthreads();
// if (threadIdx.x == 0)
// shared.b = shared.tmp[0];
// __syncthreads();
//
// if (threadIdx.x < 32)
// shared.lpc[order - 1][threadIdx.x] = shared.arp[threadIdx.x];
//
// //if (threadIdx.x == 0)
// // shared.PE[order] = (shared.b + shared.f) / 2 / (frameSize - order);
// __syncthreads();
// }
// for (int order = 1 + (threadIdx.x >> 5); order <= max_order; order += 16)
// {
// // Quantization
// int cn = threadIdx.x & 31;
// int precision = 10 - (order > 8) - min(2, shared.task.wbits);
// int taskNo = taskCount * blockIdx.y + order - 1;
// shared.bits[threadIdx.x] = __mul24((33 - __clz(__float2int_rn(fabs(shared.lpc[order - 1][cn]) * (1 << 15))) - precision), cn < order);
// shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 16]);
// shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 8]);
// shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 4]);
// shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 2]);
// shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 1]);
// int sh = max(0,min(15, 15 - shared.bits[threadIdx.x - cn]));
//
// // reverse coefs
// int coef = max(-(1 << precision),min((1 << precision)-1,__float2int_rn(shared.lpc[order - 1][order - 1 - cn] * (1 << sh))));
// if (cn < order)
// tasks[taskNo].coefs[cn] = coef;
// if (cn == 0)
// tasks[taskNo].shift = sh;
// shared.bits[threadIdx.x] = __mul24(33 - max(__clz(coef),__clz(-1 ^ coef)), cn < order);
// shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 16]);
// shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 8]);
// shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 4]);
// shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 2]);
// shared.bits[threadIdx.x] = max(shared.bits[threadIdx.x], shared.bits[threadIdx.x + 1]);
// int cbits = shared.bits[threadIdx.x - cn];
// if (cn == 0)
// tasks[taskNo].cbits = cbits;
// }
//}
// blockDim.x == 32 // blockDim.x == 32
// blockDim.y == 8 // blockDim.y == 8
extern "C" __global__ void cudaEstimateResidual( extern "C" __global__ void cudaEstimateResidual(

1960
CUETools.FlaCuda/flacuda.cubin
View File

File diff suppressed because it is too large Load Diff