source: trunk/src/testing/app/fluidsGL/common.mk @ 4

Revision 4, 10.6 KB checked in by ajaworski, 13 years ago (diff)

Added modified SAGE sources
Line 
1################################################################################
2#
3# Copyright 1993-2006 NVIDIA Corporation.  All rights reserved.
4#
5# NOTICE TO USER:   
6#
7# This source code is subject to NVIDIA ownership rights under U.S. and
8# international Copyright laws. 
9#
10# NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE
11# CODE FOR ANY PURPOSE.  IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR
12# IMPLIED WARRANTY OF ANY KIND.  NVIDIA DISCLAIMS ALL WARRANTIES WITH
13# REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF
14# MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.   
15# IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL,
16# OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS
17# OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
18# OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE
19# OR PERFORMANCE OF THIS SOURCE CODE. 
20#
21# U.S. Government End Users.  This source code is a "commercial item" as
22# that term is defined at 48 C.F.R. 2.101 (OCT 1995), consisting  of
23# "commercial computer software" and "commercial computer software
24# documentation" as such terms are used in 48 C.F.R. 12.212 (SEPT 1995)
25# and is provided to the U.S. Government only as a commercial end item. 
26# Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through
27# 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the
28# source code with only those rights set forth herein.
29#
30################################################################################
31#
32# Common build script
33#
34################################################################################
35
36.SUFFIXES : .cu .cu_dbg.o .c_dbg.o .cpp_dbg.o .cu_rel.o .c_rel.o .cpp_rel.o .cubin .ptx
37
38# Add new SM Versions here as devices with new Compute Capability are released
39SM_VERSIONS := sm_10 sm_11 sm_12 sm_13
40
41CUDA_INSTALL_PATH ?= /usr/local/cuda
42
43ifdef cuda-install
44        CUDA_INSTALL_PATH := $(cuda-install)
45endif
46
47# detect OS
48OSUPPER = $(shell uname -s 2>/dev/null | tr [:lower:] [:upper:])
49OSLOWER = $(shell uname -s 2>/dev/null | tr [:upper:] [:lower:])
50# 'linux' is output for Linux system, 'darwin' for OS X
51DARWIN = $(strip $(findstring DARWIN, $(OSUPPER)))
52
53# detect if 32 bit or 64 bit system
54HP_64 = $(shell uname -m | grep 64)
55
56# Basic directory setup for SDK
57# (override directories only if they are not already defined)
58SRCDIR     ?=
59ROOTDIR    ?= ${HOME}/NVIDIA_GPU_Computing_SDK/C
60ROOTBINDIR ?= ${SAGE_DIRECTORY}/bin
61BINDIR     ?= $(ROOTBINDIR)
62ROOTOBJDIR ?= obj
63LIBDIR     := $(ROOTDIR)/lib
64COMMONDIR  := $(ROOTDIR)/common
65
66# Compilers
67ifeq "$(strip $(HP_64))" ""
68   NVCC    := $(CUDA_INSTALL_PATH)/bin/nvcc
69else
70   NVCC    := $(CUDA_INSTALL_PATH)/bin/nvcc
71endif
72
73CXX        := g++
74CC         := gcc
75LINK       := g++ -fPIC
76
77# Includes
78INCLUDES  += -I. -I$(CUDA_INSTALL_PATH)/include -I$(COMMONDIR)/inc
79
80# architecture flag for cubin build
81CUBIN_ARCH_FLAG :=
82
83# Warning flags
84CXXWARN_FLAGS := \
85        -W -Wall \
86        -Wimplicit \
87        -Wswitch \
88        -Wformat \
89        -Wchar-subscripts \
90        -Wparentheses \
91        -Wmultichar \
92        -Wtrigraphs \
93        -Wpointer-arith \
94        -Wcast-align \
95        -Wreturn-type \
96        -Wno-unused-function \
97        $(SPACE)
98
99CWARN_FLAGS := $(CXXWARN_FLAGS) \
100        -Wstrict-prototypes \
101        -Wmissing-prototypes \
102        -Wmissing-declarations \
103        -Wnested-externs \
104        -Wmain \
105
106# Compiler-specific flags
107NVCCFLAGS :=
108CXXFLAGS  := $(CXXWARN_FLAGS)
109CFLAGS    := $(CWARN_FLAGS)
110
111# Common flags
112COMMONFLAGS += $(INCLUDES) -DUNIX
113
114# Debug/release configuration
115ifeq ($(dbg),1)
116        COMMONFLAGS += -g
117        NVCCFLAGS   += -D_DEBUG
118        CXXFLAGS    += -D_DEBUG
119        CFLAGS      += -D_DEBUG
120        BINSUBDIR   := debug
121        LIBSUFFIX   := D
122else
123        COMMONFLAGS += -O2
124        BINSUBDIR   := release
125        LIBSUFFIX   :=
126        NVCCFLAGS   += --compiler-options -fno-strict-aliasing
127        CXXFLAGS    += -fno-strict-aliasing
128        CFLAGS      += -fno-strict-aliasing
129endif
130
131# append optional arch/SM version flags (such as -arch sm_11)
132#NVCCFLAGS += $(SMVERSIONFLAGS)
133
134# architecture flag for cubin build
135CUBIN_ARCH_FLAG :=
136
137# OpenGL is used or not (if it is used, then it is necessary to include GLEW)
138ifeq ($(USEGLLIB),1)
139
140        ifneq ($(DARWIN),)
141                OPENGLLIB := -L/System/Library/Frameworks/OpenGL.framework/Libraries -lGL -lGLU $(COMMONDIR)/lib/$(OSLOWER)/libGLEW.a
142        else
143                OPENGLLIB := -lGL -lGLU -lX11 -lXi -lXmu
144
145                ifeq "$(strip $(HP_64))" ""
146                        OPENGLLIB += -lGLEW -L/usr/X11R6/lib
147                else
148                        OPENGLLIB += -lGLEW_x86_64 -L/usr/X11R6/lib64
149                endif
150        endif
151
152        CUBIN_ARCH_FLAG := -m64
153endif
154
155ifeq ($(USEGLUT),1)
156        ifneq ($(DARWIN),)
157                OPENGLLIB += -framework GLUT
158        else
159                OPENGLLIB += -lglut
160        endif
161endif
162
163ifeq ($(USEPARAMGL),1)
164        PARAMGLLIB := -lparamgl$(LIBSUFFIX)
165endif
166
167ifeq ($(USERENDERCHECKGL),1)
168        RENDERCHECKGLLIB := -lrendercheckgl$(LIBSUFFIX)
169endif
170
171ifeq ($(USECUDPP), 1)
172        ifeq "$(strip $(HP_64))" ""
173                CUDPPLIB := -lcudpp
174        else
175                CUDPPLIB := -lcudpp64
176        endif
177
178        CUDPPLIB := $(CUDPPLIB)$(LIBSUFFIX)
179
180        ifeq ($(emu), 1)
181                CUDPPLIB := $(CUDPPLIB)_emu
182        endif
183endif
184
185# Libs
186ifeq "$(strip $(HP_64))" ""
187   LIB       := -L$(CUDA_INSTALL_PATH)/lib -L$(LIBDIR) -L$(COMMONDIR)/lib/$(OSLOWER)
188else
189   LIB       := -L$(CUDA_INSTALL_PATH)/lib64 -L$(LIBDIR) -L$(COMMONDIR)/lib/$(OSLOWER)
190endif
191
192# If dynamically linking to CUDA and CUDART, we exclude the libraries from the LIB
193ifeq ($(USECUDADYNLIB),1)
194     LIB += ${OPENGLLIB} $(PARAMGLLIB) $(RENDERCHECKGLLIB) $(CUDPPLIB) ${LIB} -ldl -rdynamic
195else
196# static linking, we will statically link against CUDA and CUDART
197  ifeq ($(USEDRVAPI),1)
198     LIB += -lcuda   ${OPENGLLIB} $(PARAMGLLIB) $(RENDERCHECKGLLIB) $(CUDPPLIB) ${LIB}
199  else
200     LIB += -lcudart ${OPENGLLIB} $(PARAMGLLIB) $(RENDERCHECKGLLIB) $(CUDPPLIB) ${LIB}
201  endif
202endif
203
204ifeq ($(USECUFFT),1)
205  ifeq ($(emu),1)
206    LIB += -lcufftemu
207  else
208    LIB += -lcufft
209  endif
210endif
211
212ifeq ($(USECUBLAS),1)
213  ifeq ($(emu),1)
214    LIB += -lcublasemu
215  else
216    LIB += -lcublas
217  endif
218endif
219
220# Lib/exe configuration
221ifneq ($(STATIC_LIB),)
222        TARGETDIR := $(LIBDIR)
223        TARGET   := $(subst .a,$(LIBSUFFIX).a,$(LIBDIR)/$(STATIC_LIB))
224        LINKLINE  = ar rucv $(TARGET) $(OBJS)
225else
226        ifneq ($(OMIT_CUTIL_LIB),1)
227                LIB += -lcutil$(LIBSUFFIX)
228        endif
229        # Device emulation configuration
230        ifeq ($(emu), 1)
231                NVCCFLAGS   += -deviceemu
232                CUDACCFLAGS +=
233                BINSUBDIR   := emu$(BINSUBDIR)
234                # consistency, makes developing easier
235                CXXFLAGS                += -D__DEVICE_EMULATION__
236                CFLAGS                  += -D__DEVICE_EMULATION__
237        endif
238        TARGETDIR := $(BINDIR)
239        TARGET    := $(TARGETDIR)/$(EXECUTABLE)
240        LINKLINE  = $(LINK) -o $(TARGET) $(OBJS) $(LIB) -L${SAGE_DIRECTORY}/lib -lsail
241endif
242
243# check if verbose
244ifeq ($(verbose), 1)
245        VERBOSE :=
246else
247        VERBOSE := @
248endif
249
250################################################################################
251# Check for input flags and set compiler flags appropriately
252################################################################################
253ifeq ($(fastmath), 1)
254        NVCCFLAGS += -use_fast_math
255endif
256
257ifeq ($(keep), 1)
258        NVCCFLAGS += -keep
259        NVCC_KEEP_CLEAN := *.i* *.cubin *.cu.c *.cudafe* *.fatbin.c *.ptx
260endif
261
262ifdef maxregisters
263        NVCCFLAGS += -maxrregcount $(maxregisters)
264endif
265
266# Add cudacc flags
267NVCCFLAGS += $(CUDACCFLAGS)
268
269# Add common flags
270NVCCFLAGS += $(COMMONFLAGS)
271CXXFLAGS  += $(COMMONFLAGS)
272CFLAGS    += $(COMMONFLAGS)
273
274ifeq ($(nvcc_warn_verbose),1)
275        NVCCFLAGS += $(addprefix --compiler-options ,$(CXXWARN_FLAGS))
276        NVCCFLAGS += --compiler-options -fno-strict-aliasing
277endif
278
279################################################################################
280# Set up object files
281################################################################################
282OBJDIR := $(ROOTOBJDIR)/$(BINSUBDIR)
283OBJS +=  $(patsubst %.cpp,$(OBJDIR)/%.cpp.o,$(notdir $(CCFILES)))
284OBJS +=  $(patsubst %.c,$(OBJDIR)/%.c.o,$(notdir $(CFILES)))
285OBJS +=  $(patsubst %.cu,$(OBJDIR)/%.cu.o,$(notdir $(CUFILES)))
286
287################################################################################
288# Set up cubin output files
289################################################################################
290CUBINDIR := $(SRCDIR)data
291CUBINS +=  $(patsubst %.cu,$(CUBINDIR)/%.cubin,$(notdir $(CUBINFILES)))
292
293################################################################################
294# Set up PTX output files
295################################################################################
296PTXDIR := $(SRCDIR)data
297PTXBINS +=  $(patsubst %.cu,$(PTXDIR)/%.ptx,$(notdir $(PTXFILES)))
298
299################################################################################
300# Rules
301################################################################################
302$(OBJDIR)/%.c.o : $(SRCDIR)%.c $(C_DEPS)
303        $(VERBOSE)$(CC) $(CFLAGS) -o $@ -c $<
304
305$(OBJDIR)/%.cpp.o : $(SRCDIR)%.cpp $(C_DEPS)
306        $(VERBOSE)$(CXX) $(CXXFLAGS) -o $@ -c $<
307
308$(OBJDIR)/%.cu.o : $(SRCDIR)%.cu $(CU_DEPS)
309        $(VERBOSE)$(NVCC) $(NVCCFLAGS) $(SMVERSIONFLAGS) -o $@ -c $<
310
311$(CUBINDIR)/%.cubin : $(SRCDIR)%.cu cubindirectory
312        $(VERBOSE)$(NVCC) $(CUBIN_ARCH_FLAG) $(NVCCFLAGS) $(SMVERSIONFLAGS) -o $@ -cubin $<
313
314$(PTXDIR)/%.ptx : $(SRCDIR)%.cu ptxdirectory
315        $(VERBOSE)$(NVCC) $(CUBIN_ARCH_FLAG) $(NVCCFLAGS) $(SMVERSIONFLAGS) -o $@ -ptx $<
316
317#
318# The following definition is a template that gets instantiated for each SM
319# version (sm_10, sm_13, etc.) stored in SMVERSIONS.  It does 2 things:
320# 1. It adds to OBJS a .cu_sm_XX.o for each .cu file it finds in CUFILES_sm_XX.
321# 2. It generates a rule for building .cu_sm_XX.o files from the corresponding
322#    .cu file.
323#
324# The intended use for this is to allow Makefiles that use common.mk to compile
325# files to different Compute Capability targets (aka SM arch version).  To do
326# so, in the Makefile, list files for each SM arch separately, like so:
327#
328# CUFILES_sm_10 := mycudakernel_sm10.cu app.cu
329# CUFILES_sm_12 := anothercudakernel_sm12.cu
330#
331define SMVERSION_template
332OBJS += $(patsubst %.cu,$(OBJDIR)/%.cu_$(1).o,$(notdir $(CUFILES_$(1))))
333$(OBJDIR)/%.cu_$(1).o : $(SRCDIR)%.cu $(CU_DEPS)
334        $(VERBOSE)$(NVCC) -o $$@ -c $$< $(NVCCFLAGS) -arch $(1)
335endef
336
337# This line invokes the above template for each arch version stored in
338# SM_VERSIONS.  The call funtion invokes the template, and the eval
339# function interprets it as make commands.
340$(foreach smver,$(SM_VERSIONS),$(eval $(call SMVERSION_template,$(smver))))
341
342$(TARGET): makedirectories $(OBJS) $(CUBINS) $(PTXBINS) Makefile
343        $(VERBOSE)$(LINKLINE)
344
345install:$(TARGET)
346
347cubindirectory:
348        $(VERBOSE)mkdir -p $(CUBINDIR)
349
350ptxdirectory:
351        $(VERBOSE)mkdir -p $(PTXDIR)
352
353makedirectories:
354        $(VERBOSE)mkdir -p $(LIBDIR)
355        $(VERBOSE)mkdir -p $(OBJDIR)
356        $(VERBOSE)mkdir -p $(TARGETDIR)
357
358
359tidy :
360        $(VERBOSE)find . | egrep "#" | xargs rm -f
361        $(VERBOSE)find . | egrep "\~" | xargs rm -f
362
363clean : tidy
364        $(VERBOSE)rm -f $(OBJS)
365        $(VERBOSE)rm -f $(CUBINS)
366        $(VERBOSE)rm -f $(PTXBINS)
367        $(VERBOSE)rm -f $(TARGET)
368        $(VERBOSE)rm -f $(NVCC_KEEP_CLEAN)
369
370clobber : clean
371        $(VERBOSE)rm -rf $(ROOTOBJDIR)
372
Note: See TracBrowser for help on using the repository browser.