AuroraMiddleware/OpenSubdiv
1
0
Fork 0
mirror of https://github.com/PixarAnimationStudios/OpenSubdiv synced 2025-01-01 12:50:13 +00:00
Code Issues Releases Wiki Activity

gcd: convert most of loops to submit work in 32 item size batches.

Browse Source 
MBP6,2 (2 core 4 thread), catmark_car Lv3: 6.9ms CPU, 5.2ms GCD; Lv4: 31.2ms CPU, 22.ms GCD
This commit is contained in:
Aras Pranckevicius 2013-02-03 09:05:22 +02:00
parent b556e9d6cd
commit c4b463668a
1 changed files with 37 additions and 73 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

110
opensubdiv/osd/gcdKernel.cpp
View File

@ -56,6 +56,7 @@
//
#include "../osd/gcdKernel.h"
#include "../osd/cpuKernel.h"
#include "../osd/vertexDescriptor.h"
#include <math.h>
@ -63,6 +64,9 @@
namespace OpenSubdiv {
namespace OPENSUBDIV_VERSION {
const int GCD_WORK_STRIDE = 32;
void OsdGcdComputeFace(
const OsdVertexDescriptor *vdesc, float * vertex, float * varying,
const int *F_IT, const int *F_ITa, int offset, int start, int end,
@ -93,31 +97,18 @@ void OsdGcdComputeEdge(
const int *E_IT, const float *E_W, int offset, int start, int end,
dispatch_queue_t gcdq) {
dispatch_apply(end-start, gcdq, ^(size_t blockIdx){
int i = start+blockIdx;
int eidx0 = E_IT[4*i+0];
int eidx1 = E_IT[4*i+1];
int eidx2 = E_IT[4*i+2];
int eidx3 = E_IT[4*i+3];
float vertWeight = E_W[i*2+0];
int dstIndex = offset + i;
vdesc->Clear(vertex, varying, dstIndex);
vdesc->AddWithWeight(vertex, dstIndex, eidx0, vertWeight);
vdesc->AddWithWeight(vertex, dstIndex, eidx1, vertWeight);
if (eidx2 != -1) {
float faceWeight = E_W[i*2+1];
vdesc->AddWithWeight(vertex, dstIndex, eidx2, faceWeight);
vdesc->AddWithWeight(vertex, dstIndex, eidx3, faceWeight);
}
vdesc->AddVaryingWithWeight(varying, dstIndex, eidx0, 0.5f);
vdesc->AddVaryingWithWeight(varying, dstIndex, eidx1, 0.5f);
const int workSize = end-start;
dispatch_apply(workSize/GCD_WORK_STRIDE, gcdq, ^(size_t blockIdx){
const int start_i = start + blockIdx*GCD_WORK_STRIDE;
const int end_i = start_i + GCD_WORK_STRIDE;
OsdCpuComputeEdge(vdesc, vertex, varying, E_IT, E_W, offset,
start_i, end_i);
});
const int start_e = end - workSize%GCD_WORK_STRIDE;
const int end_e = end;
if (start_e < end_e)
OsdCpuComputeEdge(vdesc, vertex, varying, E_IT, E_W, offset,
start_e, end_e);
}
void OsdGcdComputeVertexA(
@ -126,36 +117,18 @@ void OsdGcdComputeVertexA(
int offset, int start, int end, int pass,
dispatch_queue_t gcdq) {
dispatch_apply(end-start, gcdq, ^(size_t blockIdx){
int i = start+blockIdx;
int n = V_ITa[5*i+1];
int p = V_ITa[5*i+2];
int eidx0 = V_ITa[5*i+3];
int eidx1 = V_ITa[5*i+4];
float weight = (pass == 1) ? V_W[i] : 1.0f - V_W[i];
// In the case of fractional weight, the weight must be inverted since
// the value is shared with the k_Smooth kernel (statistically the
// k_Smooth kernel runs much more often than this one)
if (weight > 0.0f && weight < 1.0f && n > 0)
weight = 1.0f - weight;
int dstIndex = offset + i;
if (not pass)
vdesc->Clear(vertex, varying, dstIndex);
if (eidx0 == -1 || (pass == 0 && (n == -1))) {
vdesc->AddWithWeight(vertex, dstIndex, p, weight);
} else {
vdesc->AddWithWeight(vertex, dstIndex, p, weight * 0.75f);
vdesc->AddWithWeight(vertex, dstIndex, eidx0, weight * 0.125f);
vdesc->AddWithWeight(vertex, dstIndex, eidx1, weight * 0.125f);
}
if (not pass)
vdesc->AddVaryingWithWeight(varying, dstIndex, p, 1.0f);
const int workSize = end-start;
dispatch_apply(workSize/GCD_WORK_STRIDE, gcdq, ^(size_t blockIdx){
const int start_i = start + blockIdx*GCD_WORK_STRIDE;
const int end_i = start_i + GCD_WORK_STRIDE;
OsdCpuComputeVertexA(vdesc, vertex, varying, V_ITa, V_W, offset,
start_i, end_i, pass);
});
const int start_e = end - workSize%GCD_WORK_STRIDE;
const int end_e = end;
if (start_e < end_e)
OsdCpuComputeVertexA(vdesc, vertex, varying, V_ITa, V_W, offset,
start_e, end_e, pass);
}
void OsdGcdComputeVertexB(
@ -164,27 +137,18 @@ void OsdGcdComputeVertexB(
int offset, int start, int end,
dispatch_queue_t gcdq) {
dispatch_apply(end-start, gcdq, ^(size_t blockIdx){
int i = start+blockIdx;
int h = V_ITa[5*i];
int n = V_ITa[5*i+1];
int p = V_ITa[5*i+2];
float weight = V_W[i];
float wp = 1.0f/static_cast<float>(n*n);
float wv = (n-2.0f) * n * wp;
int dstIndex = offset + i;
vdesc->Clear(vertex, varying, dstIndex);
vdesc->AddWithWeight(vertex, dstIndex, p, weight * wv);
for (int j = 0; j < n; ++j) {
vdesc->AddWithWeight(vertex, dstIndex, V_IT[h+j*2], weight * wp);
vdesc->AddWithWeight(vertex, dstIndex, V_IT[h+j*2+1], weight * wp);
}
vdesc->AddVaryingWithWeight(varying, dstIndex, p, 1.0f);
const int workSize = end-start;
dispatch_apply(workSize/GCD_WORK_STRIDE, gcdq, ^(size_t blockIdx){
const int start_i = start + blockIdx*GCD_WORK_STRIDE;
const int end_i = start_i + GCD_WORK_STRIDE;
OsdCpuComputeVertexB(vdesc, vertex, varying, V_ITa, V_IT, V_W, offset,
start_i, end_i);
});
const int start_e = end - workSize%GCD_WORK_STRIDE;
const int end_e = end;
if (start_e < end_e)
OsdCpuComputeVertexB(vdesc, vertex, varying, V_ITa, V_IT, V_W, offset,
start_e, end_e);
}
void OsdGcdComputeLoopVertexB(