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

testing on Fermi

Browse Source
This commit is contained in:
chudov
2010-11-09 19:34:57 +00:00
parent f0a4de3f43
commit c3b0c1ae9c
1 changed files with 7 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
CUETools.Codecs.FLACCL/flac.cl
View File

@@ -799,7 +799,7 @@ void clEstimateResidual(
__global FLACCLSubframeTask *tasks __global FLACCLSubframeTask *tasks
) )
{ {
__local float data[GROUP_SIZE * 2]; __local float data[GROUP_SIZE * 2 + 32];
__local FLACCLSubframeTask task; __local FLACCLSubframeTask task;
__local int psum[64]; __local int psum[64];
__local float fcoef[32]; __local float fcoef[32];
@@ -818,11 +818,14 @@ void clEstimateResidual(
int bs = task.data.blocksize; int bs = task.data.blocksize;
if (tid < 32) if (tid < 32)
fcoef[tid] = select(0.0f, - ((float) task.coefs[tid]) / (1 << task.data.shift), tid < ro); //fcoef[tid] = select(0.0f, - ((float) task.coefs[tid]) / (1 << task.data.shift), tid < ro);
//fcoef[tid] = select(0.0f, - ((float) task.coefs[tid + ro - MAX_ORDER]) / (1 << task.data.shift), tid + ro >= MAX_ORDER && tid < MAX_ORDER); fcoef[tid] = tid < MAX_ORDER && tid + ro - MAX_ORDER >= 0 ? - ((float) task.coefs[tid + ro - MAX_ORDER]) / (1 << task.data.shift) : 0.0f;
if (tid < 64) if (tid < 64)
psum[tid] = 0; psum[tid] = 0;
data[tid] = 0.0f; data[tid] = 0.0f;
// need to initialize "extra" data, because NaNs can produce wierd results even when multipled by zero extra coefs
if (tid < 32)
data[GROUP_SIZE * 2 + tid] = 0.0f;
int partOrder = max(1, clz(64) - clz(bs - 1) + 1); int partOrder = max(1, clz(64) - clz(bs - 1) + 1);
@@ -844,7 +847,7 @@ void clEstimateResidual(
barrier(CLK_LOCAL_MEM_FENCE); barrier(CLK_LOCAL_MEM_FENCE);
// compute residual // compute residual
__local float* dptr = &data[tid + GROUP_SIZE - ro]; __local float* dptr = &data[tid + GROUP_SIZE - MAX_ORDER];
float4 sum float4 sum
#ifdef AMD #ifdef AMD
= fc0 * vload4(0, dptr) = fc0 * vload4(0, dptr)