diff --git a/hysop/backend/device/codegen/base/codegen.py b/hysop/backend/device/codegen/base/codegen.py
index 0d313e3e82a795fbdabe2f52122bfaa259b6c9a2..d93e69142ac3e4cd0ee6a10fefb8b4fe340ce5f4 100644
--- a/hysop/backend/device/codegen/base/codegen.py
+++ b/hysop/backend/device/codegen/base/codegen.py
@@ -249,24 +249,34 @@ class CodeGenerator(object):
self.append(code)
def decl_vars(self, *variables, **kargs):
+ codegen = kargs.pop('codegen', None)
+ align = kargs.pop('align', False)
assert len(set(var.base_ctype() for var in variables))==1
base= variables[0].base_ctype()
svars=[]
for var in variables:
check_instance(var, CodegenVariable)
svars.append(var.declare(multidecl=True, **kargs))
- decl = '{} {};'.format(base, ', '.join(svars))
- self.append(decl)
+ decl = '{} ${};'.format(base, ', '.join(svars))
+ if not align:
+ decl = decl.replace('$','')
+ if (codegen is None):
+ self.append(decl)
+ else:
+ codegen.append(decl)
def decl_aligned_vars(self, *variables, **kargs):
assert len(variables)>0
jmp = kargs.pop('jmp', True)
- with self._align_() as al:
- for var in variables:
- check_instance(var, CodegenVariable)
- var.declare(codegen=al, align=True, **kargs)
- if jmp:
- al.jumpline()
+ try:
+ with self._align_() as al:
+ for var in variables:
+ check_instance(var, CodegenVariable)
+ var.declare(codegen=al, align=True, **kargs)
+ if jmp:
+ al.jumpline()
+ except:
+ raise
class VarBlock(object):
diff --git a/hysop/backend/device/codegen/base/utils.py b/hysop/backend/device/codegen/base/utils.py
index ee926da994be988509a1a5e8f51f590ad9082194..067786d87e1ef686e9a7876bbccdb868d029c76e 100644
--- a/hysop/backend/device/codegen/base/utils.py
+++ b/hysop/backend/device/codegen/base/utils.py
@@ -1,6 +1,5 @@
import hashlib
-
class WriteOnceDict(dict):
def __init__(self,**kargs):
super(WriteOnceDict,self).__init__(**kargs)
diff --git a/hysop/backend/device/codegen/base/variables.py b/hysop/backend/device/codegen/base/variables.py
index 3d454ea62c9eaa21bccdf9487c5e9c83f66f89b2..0eb544376ba7fd5470272084b7101889770d71e5 100644
--- a/hysop/backend/device/codegen/base/variables.py
+++ b/hysop/backend/device/codegen/base/variables.py
@@ -351,8 +351,13 @@ class CodegenVariable(object):
# static array ctype needs to be split
name = self.decl_name()
-
- init = init if (init is not None) else self.init
+
+ if init==False:
+ msg='Const variable should be initialized at declaration.'
+ assert (not self.const) or (not self.add_impl_const), msg
+ init=None
+ else:
+ init = init if (init is not None) else self.init
if (len(ctype)>0) and ctype[-1]=='*':
code = '{}${}'.format(ctype, name)
else:
@@ -382,6 +387,22 @@ class CodegenVariable(object):
codegen.append(code)
return code.strip()
+ def affect(self, codegen=None, align=False, init=None,
+ compact=False):
+ msg='Cannot affect a const variable.'
+ assert (not self.const) or (not self.add_impl_const), msg
+ init = self.init if (init is None) else init
+ if compact:
+ code = '{}$={};'
+ else:
+ code = '{} $= {};'
+ code = code.format(self, init)
+ if not align:
+ code = code.replace('$','')
+ if codegen is not None:
+ codegen.append(code)
+ return code
+
def __getitem__(self,ss):
if self.is_ptr:
return '{}[{}]'.format(self.name,ss)
@@ -736,14 +757,14 @@ class CodegenVectorClBuiltin(CodegenVector):
else:
raise TypeError, 'Invalid key type!'
- def declare(self, init=None, **kargs):
+ def declare(self, codegen=None, init=None, **kargs):
if isinstance(init,int):
init = ','.join([self.typegen.dump(init) for _ in xrange(self.dim)])
init = '({})({})'.format(self.ctype,init)
elif init.__class__ in [list,tuple,np.ndarray]:
init = ','.join([self.typegen.dump(init[i]) for i in xrange(self.dim)])
init = '({})({})'.format(self.ctype,init)
- return super(CodegenVectorClBuiltin,self).declare(init=init, **kargs)
+ return super(CodegenVectorClBuiltin,self).declare(init=init, codegen=codegen, **kargs)
class CodegenVectorClBuiltinFunc(CodegenVectorClBuiltin):
def __init__(self,fname,name,btype,dim,typegen,
@@ -762,13 +783,12 @@ class CodegenVectorClBuiltinFunc(CodegenVectorClBuiltin):
assert i<self.dim
return 'get_{}({})'.format(self.fname,i)
- def declare(self,codegen=None,align=False,const=None,init=None):
+ def declare(self, codegen=None, init=None, **kargs):
if init is False:
init = None
else:
init = self.fval() if not self.known() else init
- return super(CodegenVectorClBuiltinFunc,self).declare(codegen=codegen,align=align,
- const=const, init=init)
+ return super(CodegenVectorClBuiltinFunc,self).declare(init=init, codegen=codegen, **kargs)
class CodegenStruct(CodegenVariable):
def __init__(self, name, struct,
diff --git a/hysop/backend/device/codegen/kernels/directional_remesh.py b/hysop/backend/device/codegen/kernels/directional_remesh.py
index 2199743e92c9df8a3f48b77aef871999744de610..adea049a7fe1e018d3b2b739056e6b305aba7e7a 100644
--- a/hysop/backend/device/codegen/kernels/directional_remesh.py
+++ b/hysop/backend/device/codegen/kernels/directional_remesh.py
@@ -27,14 +27,13 @@ from hysop.backend.device.codegen.base.variables import CodegenStruct
from hysop.backend.device.codegen.structs.mesh_info import MeshBaseStruct, MeshInfoStruct
from hysop.backend.device.codegen.functions.compute_index import ComputeIndexFunction
-from hysop.backend.device.codegen.functions.cache_load import CacheLoadFunction
-from hysop.backend.device.codegen.functions.runge_kutta import RungeKuttaFunction
from hysop.backend.device.opencl import cl, clCharacterize
from hysop.backend.device.opencl.opencl_env import OpenClEnvironment
from hysop.backend.device.opencl.opencl_kernel import OpenClKernelLauncher
from hysop.backend.device.kernel_autotuner import KernelAutotuner, AutotunerFlags, \
KernelGenerationError
+
from hysop.fields.discrete import DiscreteField
from hysop.core.arrays.all import OpenClArrayBackend
from hysop.numerics.remesh.remesh import RemeshKernel
@@ -142,6 +141,10 @@ class DirectionalRemeshKernel(KernelCodeGenerator):
mesh_info_struct = MeshInfoStruct(typegen=typegen, typedef='MeshInfo_s')
reqs['MeshInfoStruct'] = mesh_info_struct
+ compute_id = ComputeIndexFunction(typegen=typegen, dim=work_dim, itype=itype,
+ wrap=is_periodic)
+ reqs['compute_id'] = compute_id
+
return reqs
def gen_kernel_arguments(self, typegen, work_dim, itype, ftype,
@@ -209,6 +212,8 @@ class DirectionalRemeshKernel(KernelCodeGenerator):
is_periodic = s.is_periodic
local_size_known = s.local_size_known
+ compute_id = s.reqs['compute_id']
+
global_id = s.vars['global_id']
local_id = s.vars['local_id']
group_id = s.vars['group_id']
@@ -228,15 +233,36 @@ class DirectionalRemeshKernel(KernelCodeGenerator):
position = s.position
scalars_in = s.scalars_in
scalars_out = s.scalars_out
-
+
+
+ #common to all scalars and position
compute_grid_size = position_mesh_info['local_mesh']['compute_resolution'].view(
'compute_grid_size',slice(None,work_dim),const=True)
-
+
+ # may be topology specific
+ position_grid_size = position_mesh_info['local_mesh']['resolution'].view(
+ 'pos_grid_size', slice(0,work_dim), const=True)
+ scalars_grid_size = tuple(smi['local_mesh']['resolution'].view('S{}_grid_size'.format(i),
+ slice(0,work_dim), const=True) for (i,smi) in enumerate(scalars_mesh_info))
+
position_grid_ghosts = position_mesh_info['ghosts'].view(
'pos_grid_ghosts', slice(0,work_dim), const=True)
scalars_grid_ghosts = tuple(smi['ghosts'].view('S{}_grid_ghosts'.format(i),
slice(0,work_dim), const=True) for (i,smi) in enumerate(scalars_mesh_info))
+ position_GID = CodegenVariable('pos_global_id', itype, typegen=tg)
+ scalars_GID = tuple(CodegenVariable('S{}_global_id'.format(i), itype, typegen=tg) for i in xrange(nscalars))
+
+ position_global_id = CodegenVectorClBuiltin('pos_gid', itype, work_dim, typegen=tg)
+ scalars_global_id = tuple(CodegenVectorClBuiltin('S{}_gid'.format(i),
+ itype, work_dim, typegen=tg) for i in xrange(nscalars))
+
+ grid_ghosts = (position_grid_ghosts,) + scalars_grid_ghosts
+ grid_sizes = (position_grid_size,) + scalars_grid_size
+ global_ids = (position_global_id,) + scalars_global_id
+ GIDs = (position_GID,) + scalars_GID
+
+
dx = position_mesh_info['dx'].view('dx', slice(0,1), const=True)
inv_dx = position_mesh_info['inv_dx'].view('inv_dx', slice(0,1), const=True)
xmin = position_mesh_info['local_mesh']['xmin'].view('xmin', slice(0,1), const=True)
@@ -254,10 +280,6 @@ class DirectionalRemeshKernel(KernelCodeGenerator):
line_velocity = CodegenVariable(name='line_velocity', ctype=ftype, ptr=True,
storage='__global', ptr_restrict=True, ptr_const=True, const=True, typegen=tg)
- position_global_id = CodegenVectorClBuiltin('pos_gid', itype, work_dim, typegen=tg)
- scalars_global_id = tuple(CodegenVectorClBuiltin('S{}_gid'.format(i),
- itype, work_dim, typegen=tg) for i in xrange(nscalars))
-
npart = CodegenVariable(name='nparticles', ctype=itype, typegen=tg, value=nparticles)
cache_ghosts = CodegenVariable('cache_ghosts', itype, typegen=tg,
@@ -280,15 +302,18 @@ class DirectionalRemeshKernel(KernelCodeGenerator):
for i in xrange(nscalars):
Si = CodegenVariable(name='S{}'.format(i),ctype=ftype,typegen=tg,
ptr_restrict=True, ptr=True, storage=self._local,
- const=True,
+ ptr_const=True,
init='{} + {}*{}'.format(buf,i,cache_width))
cached_scalars.append(Si)
- first = CodegenVariable('first','bool',tg,init='true')
- active = CodegenVariable('active','bool',tg)
+ first = CodegenVariable('first','bool',tg,init='true')
+ active = CodegenVariable('active','bool',tg)
particle_offset = CodegenVariable('particle_offset',itype,tg)
+ pos = CodegenVectorClBuiltin('p', ftype, nparticles, tg)
+ scalars = [CodegenVectorClBuiltin('s{}'.format(i), ftype, nparticles, tg) for i in xrange(nscalars)]
+
@contextlib.contextmanager
def _work_iterate_(i):
try:
@@ -315,7 +340,8 @@ class DirectionalRemeshKernel(KernelCodeGenerator):
active.declare(s, init='({} < {})'.format(particle_offset, compute_grid_size[0]))
elif i==1:
first.declare(s)
-
+ s.jumpline()
+
with s._align_() as al:
for field_gid, field_ghosts in zip((position_global_id,)+scalars_global_id,
(position_grid_ghosts,)+scalars_grid_ghosts):
@@ -325,13 +351,11 @@ class DirectionalRemeshKernel(KernelCodeGenerator):
else:
al.append('{} $= {} $+ {};'.format(field_gid[i],
loopvars[i], field_ghosts[i]))
-
-
+ al.jumpline()
yield ctx
except:
raise
-
with s._kernel_():
s.jumpline()
@@ -346,10 +370,10 @@ class DirectionalRemeshKernel(KernelCodeGenerator):
s.decl_aligned_vars(npart, cache_ghosts, cache_width, local_work, const=True)
- s.decl_aligned_vars(compute_grid_size, position_grid_ghosts, *scalars_grid_ghosts)
-
- s.decl_vars(position_global_id, *scalars_global_id)
+ s.decl_vars(compute_grid_size)
s.jumpline()
+ s.decl_aligned_vars(*grid_sizes)
+ s.decl_aligned_vars(*grid_ghosts)
if local_size_known:
s.decl_vars(*cached_scalars)
@@ -357,13 +381,26 @@ class DirectionalRemeshKernel(KernelCodeGenerator):
else:
s.decl_aligned_vars(*cached_scalars)
-
+ with s._align_() as al:
+ s.decl_vars(*global_ids, align=True, codegen=al)
+ s.decl_vars(*GIDs, align=True, codegen=al)
+ s.decl_vars(pos, *scalars, align=True, codegen=al)
+ s.jumpline()
nested_loops = [_work_iterate_(i) for i in xrange(dim-1,-1,-1)]
with contextlib.nested(*nested_loops):
- # s.barrier(_local=True)
- # s.jumpline()
- pass
+ with s._align_() as al:
+ for (_GID, _global_id, _grid_size) in zip(GIDs, global_ids, grid_sizes):
+ al.append('{} $= {};'.format(_GID,
+ compute_id(idx=_global_id, size=_grid_size)))
+ s.jumpline()
+
+ with s._align_() as al:
+ pos.affect(al, align=True, init=position[position_GID()])
+ for si, sin, sgid in zip(scalars, scalars_in, scalars_GID):
+ si.affect(al, align=True, init=sin[sgid])
+ s.jumpline()
+
if __name__ == '__main__':