From 6690f9e7e8b46b96aea222d3e85315cd63545953 Mon Sep 17 00:00:00 2001
From: Ben Schreiber <bjschre2@illinois.edu>
Date: Sat, 18 Jan 2020 16:07:11 -0600
Subject: [PATCH] Port spmv and lbm
---
hpvm/test/parboil/benchmarks/lbm/Makefile | 2 +-
.../benchmarks/lbm/src/visc/lbm_macros.h | 2 +-
.../parboil/benchmarks/lbm/src/visc/main.cpp | 69 +++++++-
hpvm/test/parboil/benchmarks/spmv/Makefile | 2 +-
.../parboil/benchmarks/spmv/src/visc/main.cpp | 165 +++++++++++++++---
5 files changed, 207 insertions(+), 33 deletions(-)
diff --git a/hpvm/test/parboil/benchmarks/lbm/Makefile b/hpvm/test/parboil/benchmarks/lbm/Makefile
index 5644a7d42c..ee7bcb33b4 100644
--- a/hpvm/test/parboil/benchmarks/lbm/Makefile
+++ b/hpvm/test/parboil/benchmarks/lbm/Makefile
@@ -1,4 +1,4 @@
-PARBOIL_ROOT = $(LLVM_SRC_ROOT)/test/VISC/parboil
+PARBOIL_ROOT = $(LLVM_SRC_ROOT)/../test/parboil
APP = lbm
ifeq ($(NUM_CORES),)
diff --git a/hpvm/test/parboil/benchmarks/lbm/src/visc/lbm_macros.h b/hpvm/test/parboil/benchmarks/lbm/src/visc/lbm_macros.h
index 2f8ba8a09c..d8ceb373df 100644
--- a/hpvm/test/parboil/benchmarks/lbm/src/visc/lbm_macros.h
+++ b/hpvm/test/parboil/benchmarks/lbm/src/visc/lbm_macros.h
@@ -6,7 +6,7 @@
*cr
***************************************************************************/
-#ifndef _LBM_MARCOS_H
+#ifndef _LBM_MACROS_H_
#define _LBM_MACROS_H_
#define OMEGA (1.95f)
diff --git a/hpvm/test/parboil/benchmarks/lbm/src/visc/main.cpp b/hpvm/test/parboil/benchmarks/lbm/src/visc/main.cpp
index 255036539c..605513847a 100644
--- a/hpvm/test/parboil/benchmarks/lbm/src/visc/main.cpp
+++ b/hpvm/test/parboil/benchmarks/lbm/src/visc/main.cpp
@@ -82,21 +82,35 @@ void MAIN_printInfo( const MAIN_Param* param ) {
}
/*############################################################################*/
-void performStreamCollide_kernel( float* srcG, float* dstG )
+
+typedef struct __attribute__((__packed__)) {
+ float* srcG; size_t bytes_srcG;
+ float* dstG; size_t bytes_dstG;
+ size_t dim_X1, dim_X2, dim_Y2;
+} RootIn;
+
+void performStreamCollide_kernel( float* srcG, size_t bytes_srcG, float* dstG, size_t bytes_dstG )
{
__visc__hint(visc::DEVICE);
__visc__attributes(2, srcG, dstG, 1, dstG);
+
+ void* thisNode = __visc__getNode();
+ void* parentNode = __visc__getParentNode(thisNode);
+
srcG += MARGIN;
dstG += MARGIN;
+ int lx = __visc__getNodeInstanceID_x(thisNode);
+ int gx = __visc__getNodeInstanceID_x(parentNode);
+ int gy = __visc__getNodeInstanceID_y(parentNode);
//Using some predefined macros here. Consider this the declaration
// and initialization of the variables SWEEP_X, SWEEP_Y and SWEEP_Z
SWEEP_VAR
- SWEEP_X = get_local_id(0);
- SWEEP_Y = get_group_id(0);
- SWEEP_Z = get_group_id(1);
+ SWEEP_X = lx; // get_local_id(0)
+ SWEEP_Y = gx; // get_group_id(0)
+ SWEEP_Z = gy; // get_group_id(1)
float temp_swp, tempC, tempN, tempS, tempE, tempW, tempT, tempB;
float tempNE, tempNW, tempSE, tempSW, tempNT, tempNB, tempST ;
@@ -262,13 +276,58 @@ void performStreamCollide_kernel( float* srcG, float* dstG )
DST_WB( dstG ) = tempWB;
}
+void lbmLvl1(float* srcG, size_t bytes_srcG, float* dstG, size_t bytes_dstG, size_t dim_X1)
+{
+ __visc__hint(visc::DEVICE);
+ __visc__attributes(2, srcG, dstG, 1, dstG);
+ void* lbm_node = __visc__createNodeND(1, performStreamCollide_kernel, dim_X1);
+ __visc__bindIn(lbm_node, 0, 0, 0);
+ __visc__bindIn(lbm_node, 1, 1, 0);
+ __visc__bindIn(lbm_node, 2, 2, 0);
+ __visc__bindIn(lbm_node, 3, 3, 0);
+}
+
+void lbmLvl2(float* srcG, size_t bytes_srcG, float* dstG, size_t bytes_dstG, size_t dim_X1, size_t dim_X2, size_t dim_Y2)
+{
+ __visc__hint(visc::CPU_TARGET);
+ __visc__attributes(2, srcG, dstG, 1, dstG);
+ void* lbm_node = __visc__createNodeND(2, lbmLvl1, dim_X2, dim_Y2);
+ __visc__bindIn(lbm_node, 0, 0, 0);
+ __visc__bindIn(lbm_node, 1, 1, 0);
+ __visc__bindIn(lbm_node, 2, 2, 0);
+ __visc__bindIn(lbm_node, 3, 3, 0);
+ __visc__bindIn(lbm_node, 4, 4, 0);
+}
+
+void lbmLvl3(float* srcG, size_t bytes_srcG, float* dstG, size_t bytes_dstG, size_t dim_X1, size_t dim_X2, size_t dim_Y2)
+{
+ __visc__hint(visc::CPU_TARGET);
+ __visc__attributes(2, srcG, dstG, 1, dstG);
+ void* lbm_node = __visc__createNodeND(0, lbmLvl2);
+ __visc__bindIn(lbm_node, 0, 0, 0);
+ __visc__bindIn(lbm_node, 1, 1, 0);
+ __visc__bindIn(lbm_node, 2, 2, 0);
+ __visc__bindIn(lbm_node, 3, 3, 0);
+ __visc__bindIn(lbm_node, 4, 4, 0);
+ __visc__bindIn(lbm_node, 5, 5, 0);
+ __visc__bindIn(lbm_node, 6, 6, 0);
+}
__attribute__((noinline)) void MAIN_performStreamCollide( LBM_Grid src, LBM_Grid dst ) {
long dimBlock[3] = {SIZE_X,1,1};
long dimGrid[3] = {SIZE_X*SIZE_Y,SIZE_Z,1};
size_t size = TOTAL_PADDED_CELLS*N_CELL_ENTRIES*sizeof( float );
- void* lbmDFG = __visc__node(performStreamCollide_kernel, 2, 3, dimBlock[0], dimBlock[1], dimBlock[2], dimGrid[0]/dimBlock[0], dimGrid[1]/dimBlock[1], dimGrid[2]/dimBlock[2], 4, src-MARGIN, size, dst-MARGIN, size, 0);
+
+ void* root_in = malloc(sizeof(RootIn));
+ RootIn root_in_local = {
+ src - MARGIN, size,
+ dst - MARGIN, size,
+ SIZE_X, SIZE_Y, SIZE_Z
+ };
+ *(RootIn*)root_in = root_in_local;
+ void* lbmDFG = __visc__launch(0, lbmLvl3, root_in);
+
__visc__wait(lbmDFG);
}
diff --git a/hpvm/test/parboil/benchmarks/spmv/Makefile b/hpvm/test/parboil/benchmarks/spmv/Makefile
index e7ef8ca39c..71e2246343 100644
--- a/hpvm/test/parboil/benchmarks/spmv/Makefile
+++ b/hpvm/test/parboil/benchmarks/spmv/Makefile
@@ -1,4 +1,4 @@
-PARBOIL_ROOT = $(LLVM_SRC_ROOT)/test/VISC/parboil
+PARBOIL_ROOT = $(LLVM_SRC_ROOT)/../test/parboil
APP = spmv
# Default compile visc
diff --git a/hpvm/test/parboil/benchmarks/spmv/src/visc/main.cpp b/hpvm/test/parboil/benchmarks/spmv/src/visc/main.cpp
index bd457e6e13..eb4d4d7dbe 100644
--- a/hpvm/test/parboil/benchmarks/spmv/src/visc/main.cpp
+++ b/hpvm/test/parboil/benchmarks/spmv/src/visc/main.cpp
@@ -31,17 +31,38 @@ static int generate_vector(float *x_vector, int dim)
return 0;
}
-void spmv_jds(float *dst_vector, float *d_data,
- int *d_index, int *d_perm,
- float *x_vec, int dim,
- int *jds_ptr_int,
- int *sh_zcnt_int)
+typedef struct __attribute__((__packed__)) {
+ float* dst_vector; size_t bytes_dst_vector;
+ float* d_data; size_t bytes_d_data;
+ int* d_index; size_t bytes_d_index;
+ int* d_perm; size_t bytes_d_perm;
+ float* x_vec; size_t bytes_x_vec;
+ int dim;
+ int* jds_ptr_int; size_t bytes_jds_ptr_int;
+ int* sh_zcnt_int; size_t bytes_sh_zcnt_int;
+ size_t dim_X1, dim_X2;
+} RootIn;
+
+void spmv_jds(float* dst_vector, size_t bytes_dst_vector,
+ float* d_data, size_t bytes_d_data,
+ int* d_index, size_t bytes_d_index,
+ int* d_perm, size_t bytes_d_perm,
+ float* x_vec, size_t bytes_x_vec,
+ int dim,
+ int* jds_ptr_int, size_t bytes_jds_ptr_int,
+ int* sh_zcnt_int, size_t bytes_sh_zcnt_int)
{
__visc__hint(visc::DEVICE);
__visc__attributes(7, dst_vector, d_data, d_index, d_perm, x_vec, jds_ptr_int, sh_zcnt_int,
1, dst_vector);
- int ix = get_global_id(0);
+ void* thisNode = __visc__getNode();
+ void* parentNode = __visc__getParentNode(thisNode);
+ int lx = __visc__getNodeInstanceID_x(thisNode);
+ int gx = __visc__getNodeInstanceID_x(parentNode);
+ int gridx = __visc__getNumNodeInstances_x(thisNode);
+
+ int ix = gx * gridx + lx;
int warp_id=ix>>WARP_BITS;
if(ix<dim)
@@ -99,6 +120,102 @@ void spmv_jds(float *dst_vector, float *d_data,
}
}
+void spmvLvl1(float* dst_vector, size_t bytes_dst_vector,
+ float* d_data, size_t bytes_d_data,
+ int* d_index, size_t bytes_d_index,
+ int* d_perm, size_t bytes_d_perm,
+ float* x_vec, size_t bytes_x_vec,
+ int dim,
+ int* jds_ptr_int, size_t bytes_jds_ptr_int,
+ int* sh_zcnt_int, size_t bytes_sh_zcnt_int,
+ size_t dim_X1)
+{
+ __visc__hint(visc::DEVICE);
+ __visc__attributes(7, dst_vector, d_data, d_index, d_perm, x_vec, jds_ptr_int, sh_zcnt_int,
+ 1, dst_vector);
+ void* spmv_node = __visc__createNodeND(1, spmv_jds, dim_X1);
+ __visc__bindIn(spmv_node, 0, 0, 0);
+ __visc__bindIn(spmv_node, 1, 1, 0);
+ __visc__bindIn(spmv_node, 2, 2, 0);
+ __visc__bindIn(spmv_node, 3, 3, 0);
+ __visc__bindIn(spmv_node, 4, 4, 0);
+ __visc__bindIn(spmv_node, 5, 5, 0);
+ __visc__bindIn(spmv_node, 6, 6, 0);
+ __visc__bindIn(spmv_node, 7, 7, 0);
+ __visc__bindIn(spmv_node, 8, 8, 0);
+ __visc__bindIn(spmv_node, 9, 9, 0);
+ __visc__bindIn(spmv_node, 10, 10, 0);
+ __visc__bindIn(spmv_node, 11, 11, 0);
+ __visc__bindIn(spmv_node, 12, 12, 0);
+ __visc__bindIn(spmv_node, 13, 13, 0);
+ __visc__bindIn(spmv_node, 14, 14, 0);
+}
+
+void spmvLvl2(float* dst_vector, size_t bytes_dst_vector,
+ float* d_data, size_t bytes_d_data,
+ int* d_index, size_t bytes_d_index,
+ int* d_perm, size_t bytes_d_perm,
+ float* x_vec, size_t bytes_x_vec,
+ int dim,
+ int* jds_ptr_int, size_t bytes_jds_ptr_int,
+ int* sh_zcnt_int, size_t bytes_sh_zcnt_int,
+ size_t dim_X1, size_t dim_X2)
+{
+ __visc__hint(visc::CPU_TARGET);
+ __visc__attributes(7, dst_vector, d_data, d_index, d_perm, x_vec, jds_ptr_int, sh_zcnt_int,
+ 1, dst_vector);
+ void* spmv_node = __visc__createNodeND(1, spmvLvl1, dim_X2);
+ __visc__bindIn(spmv_node, 0, 0, 0);
+ __visc__bindIn(spmv_node, 1, 1, 0);
+ __visc__bindIn(spmv_node, 2, 2, 0);
+ __visc__bindIn(spmv_node, 3, 3, 0);
+ __visc__bindIn(spmv_node, 4, 4, 0);
+ __visc__bindIn(spmv_node, 5, 5, 0);
+ __visc__bindIn(spmv_node, 6, 6, 0);
+ __visc__bindIn(spmv_node, 7, 7, 0);
+ __visc__bindIn(spmv_node, 8, 8, 0);
+ __visc__bindIn(spmv_node, 9, 9, 0);
+ __visc__bindIn(spmv_node, 10, 10, 0);
+ __visc__bindIn(spmv_node, 11, 11, 0);
+ __visc__bindIn(spmv_node, 12, 12, 0);
+ __visc__bindIn(spmv_node, 13, 13, 0);
+ __visc__bindIn(spmv_node, 14, 14, 0);
+ __visc__bindIn(spmv_node, 15, 15, 0);
+}
+
+void spmvLvl3(float* dst_vector, size_t bytes_dst_vector,
+ float* d_data, size_t bytes_d_data,
+ int* d_index, size_t bytes_d_index,
+ int* d_perm, size_t bytes_d_perm,
+ float* x_vec, size_t bytes_x_vec,
+ int dim,
+ int* jds_ptr_int, size_t bytes_jds_ptr_int,
+ int* sh_zcnt_int, size_t bytes_sh_zcnt_int,
+ size_t dim_X1, size_t dim_X2)
+{
+ __visc__hint(visc::CPU_TARGET);
+ __visc__attributes(7, dst_vector, d_data, d_index, d_perm, x_vec, jds_ptr_int, sh_zcnt_int,
+ 1, dst_vector);
+ void* spmv_node = __visc__createNodeND(1, spmvLvl2, dim_X2);
+ __visc__bindIn(spmv_node, 0, 0, 0);
+ __visc__bindIn(spmv_node, 1, 1, 0);
+ __visc__bindIn(spmv_node, 2, 2, 0);
+ __visc__bindIn(spmv_node, 3, 3, 0);
+ __visc__bindIn(spmv_node, 4, 4, 0);
+ __visc__bindIn(spmv_node, 5, 5, 0);
+ __visc__bindIn(spmv_node, 6, 6, 0);
+ __visc__bindIn(spmv_node, 7, 7, 0);
+ __visc__bindIn(spmv_node, 8, 8, 0);
+ __visc__bindIn(spmv_node, 9, 9, 0);
+ __visc__bindIn(spmv_node, 10, 10, 0);
+ __visc__bindIn(spmv_node, 11, 11, 0);
+ __visc__bindIn(spmv_node, 12, 12, 0);
+ __visc__bindIn(spmv_node, 13, 13, 0);
+ __visc__bindIn(spmv_node, 14, 14, 0);
+ __visc__bindIn(spmv_node, 15, 15, 0);
+ __visc__bindIn(spmv_node, 16, 16, 0);
+}
+
int main(int argc, char** argv) {
struct pb_TimerSet timers;
struct pb_Parameters *parameters;
@@ -185,24 +302,22 @@ int main(int argc, char** argv) {
for(i=0; i<50; i++)
{
pb_SwitchToTimer(&timers, pb_TimerID_NONE);
- void* spmvDFG = __visc__node(spmv_jds, 2, 1, block, (grid/block),
- 15,
- h_Ax_vector,
- dim*sizeof(float),
- h_data,
- len*sizeof(float),
- h_indices,
- len*sizeof(int),
- h_perm,
- dim*sizeof(int),
- h_x_vector,
- dim*sizeof(float),
- dim,
- h_ptr,
- depth*sizeof(int),
- h_nzcnt,
- nzcnt_len*sizeof(int),
- 0);
+
+ void* root_in = malloc(sizeof(RootIn));
+ RootIn root_in_local = {
+ h_Ax_vector, dim * sizeof(float),
+ h_data, len * sizeof(float),
+ h_indices, len * sizeof(int),
+ h_perm, dim * sizeof(int),
+ h_x_vector, dim * sizeof(float),
+ dim,
+ h_ptr, depth * sizeof(int),
+ h_nzcnt, nzcnt_len * sizeof(int),
+ block, (grid/block)
+ };
+ *(RootIn*)root_in = root_in_local;
+ void* spmvDFG = __visc__launch(0, spmvLvl3, root_in);
+
__visc__wait(spmvDFG);
pb_SwitchToTimer(&timers, pb_TimerID_COMPUTE);
@@ -224,7 +339,7 @@ int main(int argc, char** argv) {
llvm_visc_request_mem(h_Ax_vector, dim*sizeof(float));
- pb_SwitchToTimer(&timers, visc_TimerID_MEM_UNTRACK);
+ pb_SwitchToTimer(&timers, visc_TimerID_MEM_UNTRACK);
llvm_visc_untrack_mem(h_Ax_vector);
llvm_visc_untrack_mem(h_data);
--
GitLab