Pd-cvs July 2004

pd-cvs@lists.iem.at

7 participants
73 discussions

externals/sc4pd/source main.cpp,1.1.1.1,1.2 support.hpp,1.1.1.1,1.2 Dust.cpp,1.1.1.1,1.2 Dust2.cpp,1.1.1.1,1.2
by Tim Blechmann 14 Jul '04

14 Jul '04

Update of /cvsroot/pure-data/externals/sc4pd/source In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv7437 Modified Files: main.cpp support.hpp Dust.cpp Dust2.cpp Log Message: update & bugfix Index: support.hpp =================================================================== RCS file: /cvsroot/pure-data/externals/sc4pd/source/support.hpp,v retrieving revision 1.1.1.1 retrieving revision 1.2 diff -C2 -d -r1.1.1.1 -r1.2 *** support.hpp 14 Jul 2004 16:21:44 -0000 1.1.1.1 --- support.hpp 14 Jul 2004 18:29:52 -0000 1.2 *************** *** 61,65 **** inline float sc_getfloatarg (flext::AtomList a,int i) { ! if (a.Count() >0 || a.Count() <i) return flext::GetAFloat(a[i]); else --- 61,65 ---- inline float sc_getfloatarg (flext::AtomList a,int i) { ! if (a.Count() > 0 && a.Count() > i) return flext::GetAFloat(a[i]); else *************** *** 67,70 **** --- 67,84 ---- } + inline bool sc_ar(flext::AtomList a) + { + for (int i = 0; i!=a.Count();++i) + { + if ( flext::IsSymbol(a[i]) ) + { + const char * teststring; + teststring = flext::GetString(a[i]); + if((strcmp(teststring,"ar"))==0) + return true; + } + } + return false; + } // macros to put rgen state in registers Index: main.cpp =================================================================== RCS file: /cvsroot/pure-data/externals/sc4pd/source/main.cpp,v retrieving revision 1.1.1.1 retrieving revision 1.2 diff -C2 -d -r1.1.1.1 -r1.2 *** main.cpp 14 Jul 2004 16:21:44 -0000 1.1.1.1 --- main.cpp 14 Jul 2004 18:29:52 -0000 1.2 *************** *** 50,56 **** post("compiled on "__DATE__); post("contains: Dust(~), MantissaMask(~), Hasher(~), Median(~), " ! "BrownNoise(~),"); ! post(" ClipNoise(~), GrayNoise(~), Dust2(~), WhiteNoise(~), " ! "PinkNoise(~)\n"); //initialize objects --- 50,56 ---- post("compiled on "__DATE__); post("contains: Dust(~), MantissaMask(~), Hasher(~), Median(~), " ! "BrownNoise(~),\n" ! " ClipNoise(~), GrayNoise(~), Dust2(~), WhiteNoise(~), " ! "PinkNoise(~), \n Crackle(~), Rand(~)\n"); //initialize objects *************** *** 87,90 **** --- 87,94 ---- FLEXT_DSP_SETUP(Crackle_ar); FLEXT_SETUP(Crackle_kr); + + FLEXT_DSP_SETUP(Rand_ar); + FLEXT_SETUP(Rand_kr); + } Index: Dust.cpp =================================================================== RCS file: /cvsroot/pure-data/externals/sc4pd/source/Dust.cpp,v retrieving revision 1.1.1.1 retrieving revision 1.2 diff -C2 -d -r1.1.1.1 -r1.2 *** Dust.cpp 14 Jul 2004 16:21:44 -0000 1.1.1.1 --- Dust.cpp 14 Jul 2004 18:29:52 -0000 1.2 *************** *** 129,133 **** void m_set(float f); Timer Dust_timer; ! void m_doit(); private: --- 129,133 ---- void m_set(float f); Timer Dust_timer; ! void m_doit(void*); private: *************** *** 172,176 **** } ! void Dust_kr::m_doit() { ToOutFloat(0,rgen.frand()); --- 172,176 ---- } ! void Dust_kr::m_doit(void*) { ToOutFloat(0,rgen.frand()); Index: Dust2.cpp =================================================================== RCS file: /cvsroot/pure-data/externals/sc4pd/source/Dust2.cpp,v retrieving revision 1.1.1.1 retrieving revision 1.2 diff -C2 -d -r1.1.1.1 -r1.2 *** Dust2.cpp 14 Jul 2004 16:21:44 -0000 1.1.1.1 --- Dust2.cpp 14 Jul 2004 18:29:52 -0000 1.2 *************** *** 130,134 **** void m_set(float f); Timer Dust2_timer; ! void m_doit(); private: --- 130,134 ---- void m_set(float f); Timer Dust2_timer; ! void m_doit(void*); private: *************** *** 173,177 **** } ! void Dust2_kr::m_doit() { ToOutFloat(0,2*rgen.frand() - 1 ); --- 173,177 ---- } ! void Dust2_kr::m_doit(void*) { ToOutFloat(0,2*rgen.frand() - 1 );

1 0

externals/tbext/source rfftw~.cpp,1.1,NONE rifftw~.cpp,1.1,NONE
by Tim Blechmann 13 Jul '04

13 Jul '04

Update of /cvsroot/pure-data/externals/tbext/source In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv32390 Removed Files: rfftw~.cpp rifftw~.cpp Log Message: obsolete ... part of pd now... --- rifftw~.cpp DELETED --- --- rfftw~.cpp DELETED ---

1 0

externals/tbext/source tbfft2.cpp,1.2,1.3
by Tim Blechmann 13 Jul '04

13 Jul '04

Update of /cvsroot/pure-data/externals/tbext/source In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv31075 Modified Files: tbfft2.cpp Log Message: cleanup Index: tbfft2.cpp =================================================================== RCS file: /cvsroot/pure-data/externals/tbext/source/tbfft2.cpp,v retrieving revision 1.2 retrieving revision 1.3 diff -C2 -d -r1.2 -r1.3 *** tbfft2.cpp 28 Jan 2004 11:09:13 -0000 1.2 --- tbfft2.cpp 13 Jul 2004 20:06:10 -0000 1.3 *************** *** 50,82 **** class tbfft2: public flext_dsp { ! FLEXT_HEADER(tbfft2,flext_dsp); ! public: // constructor ! tbfft2(); protected: ! virtual void m_signal (int n, float *const *in, float *const *out); ! void set_freq(t_float); ! void set_width(t_float); ! private: ! FLEXT_CALLBACK_1(set_freq,t_float) ! FLEXT_CALLBACK_1(set_width,t_float) ! ! t_int center; ! t_int width; ! ! t_float pos; ! t_int posi; ! ! float *ins; ! float *outs; ! float *tmps; ! t_float tmp[17000]; ! ! t_float s; ! t_float b; ! t_int n0; }; --- 50,75 ---- class tbfft2: public flext_dsp { ! FLEXT_HEADER(tbfft2,flext_dsp); ! public: // constructor ! tbfft2(); ! ~tbfft2(); protected: ! virtual void m_signal (int n, float *const *in, float *const *out); ! virtual void m_dsp (int n, float *const *in, float *const *out); ! void set_freq(t_float); ! void set_width(t_float); ! private: ! FLEXT_CALLBACK_1(set_freq,t_float); ! FLEXT_CALLBACK_1(set_width,t_float); ! t_int center; ! t_int width; ! ! t_float * tmp; ! ! t_int n0; }; *************** *** 92,122 **** } ! void tbfft2::m_signal(int n, t_float *const *in, t_float *const *out) { ! ins = in[0]; ! outs = out[0]; - CopySamples(tmp,ins,n); - - n0=n/2; ! if (center-width>0) { ! n=center-width; } ! else ! n=0; ! ! while (n<center+width) { ! tmp[n]=*(ins+2*center-n); ! ++n; } ! ! ! //memcp ! CopySamples(outs,tmp,n0*2); } --- 85,127 ---- } + tbfft2::~tbfft2() + { + free(tmp); + } ! void tbfft2::m_dsp(int n, t_float *const *in, t_float *const *out) { ! free(tmp); ! tmp=(t_float*)malloc(n*sizeof(t_float)); ! } ! ! void tbfft2::m_signal(int n, t_float *const *in, t_float *const *out) ! { ! t_float * ins = in[0]; ! t_float * outs = out[0]; ! ! CopySamples(tmp,ins,n); ! ! n0=n/2; ! ! if (center-width>0) { ! n=center-width; } ! else ! n=0; ! ! while (n<center+width) { ! tmp[n]=*(ins+2*center-n); ! ++n; } ! ! ! //memcpy ! CopySamples(outs,tmp,n0*2); } *************** *** 124,129 **** void tbfft2::set_freq(t_float freq) { ! center=freq; ! set_width(width); } --- 129,134 ---- void tbfft2::set_freq(t_float freq) { ! center=freq; ! set_width(width); }

1 0

externals/tbext/source him.cpp,1.3,1.4
by Tim Blechmann 13 Jul '04

13 Jul '04

Update of /cvsroot/pure-data/externals/tbext/source In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv30617 Modified Files: him.cpp Log Message: hopefully a speedup Index: him.cpp =================================================================== RCS file: /cvsroot/pure-data/externals/tbext/source/him.cpp,v retrieving revision 1.3 retrieving revision 1.4 diff -C2 -d -r1.3 -r1.4 *** him.cpp 9 Apr 2004 13:42:01 -0000 1.3 --- him.cpp 13 Jul 2004 20:04:52 -0000 1.4 *************** *** 63,67 **** protected: virtual void m_signal (int n, float *const *in, float *const *out); - t_float *outs; void set_mu(t_float); --- 63,66 ---- *************** *** 78,88 **** // contains DGL-System t_float deriv(t_float x[],int eq); - t_float result; // 4th order Runge Kutta update of the dynamical variables void runge_kutta_4(t_float dt); - int i; - t_float k1[NUMB_EQ],k2[NUMB_EQ],k3[NUMB_EQ],k4[NUMB_EQ]; - t_float temp1[NUMB_EQ], temp2[NUMB_EQ], temp3[NUMB_EQ]; //these are our data --- 77,83 ---- *************** *** 113,118 **** (data[0]*data[0]*data[0]*data[0]*data[1]*data[1]) + 16); } ! void reset_muv() { --- 108,115 ---- (data[0]*data[0]*data[0]*data[0]*data[1]*data[1]) + 16); + if (fabs((data[3]))<1e-5) + data[3]=0; } ! void reset_muv() { *************** *** 122,125 **** --- 119,124 ---- (data[0]*data[0]*data[0]*data[0]*data[1]*data[1]) + 16); + if (fabs((data[1]))<1e-5) + data[1]=0; } *************** *** 164,170 **** } ! t_float him::deriv(t_float x[], int eq) { ! // set DGL-System here if (eq == 0) result = x[1]; if (eq == 1) result = 2*E*x[0]-0.25*x[0]*x[2]*x[2]*(2*x[0]*x[0]+x[2]*x[2]); --- 163,170 ---- } ! inline t_float him::deriv(t_float * x, int eq) { ! t_float result; ! // set DGL-System here if (eq == 0) result = x[1]; if (eq == 1) result = 2*E*x[0]-0.25*x[0]*x[2]*x[2]*(2*x[0]*x[0]+x[2]*x[2]); *************** *** 175,216 **** } ! void him::runge_kutta_4(t_float dt) { ! for(i=0;i<=NUMB_EQ-1;i++) // iterate over equations ! { ! k1[i] = dt * deriv(data,i); ! temp1[i] = data[i] + 0.5*k1[i]; ! } ! for(i=0;i<=NUMB_EQ-1;i++) ! { ! k2[i] = dt * deriv(temp1,i); ! temp2[i] = data[i] + 0.5*k2[i]; ! } ! for(i=0;i<=NUMB_EQ-1;i++) ! { ! k3[i] = dt * deriv(temp2,i); ! temp3[i] = data[i] + k3[i]; ! } ! for(i=0;i<=NUMB_EQ-1;i++) ! { ! k4[i] = dt * deriv(temp3,i); ! data[i] = data[i] + (k1[i] + (2.*(k2[i]+k3[i])) + k4[i])/6.; ! } ! /* the system might become unstable ... in this case, we'll reset the system */ ! for(i=0;i<=NUMB_EQ-1;i++) if(data[i]>2) reset(); ! else ! if(PD_BADFLOAT(data[i])) //not that we get some troubles with denormals ! data[i]=0; ! } --- 175,221 ---- } ! inline void him::runge_kutta_4(t_float dt) { ! t_float k1[NUMB_EQ],k2[NUMB_EQ],k3[NUMB_EQ],k4[NUMB_EQ]; ! t_float temp1[NUMB_EQ], temp2[NUMB_EQ], temp3[NUMB_EQ]; ! for(int i=0;i<=NUMB_EQ-1;i++) // iterate over equations ! { ! k1[i] = dt * deriv(data,i); ! temp1[i] = data[i] + 0.5*k1[i]; ! } ! for(int i=0;i<=NUMB_EQ-1;i++) ! { ! k2[i] = dt * deriv(temp1,i); ! temp2[i] = data[i] + 0.5*k2[i]; ! } ! for(int i=0;i<=NUMB_EQ-1;i++) ! { ! k3[i] = dt * deriv(temp2,i); ! temp3[i] = data[i] + k3[i]; ! } ! ! for(int i=0;i<=NUMB_EQ-1;i++) ! { ! k4[i] = dt * deriv(temp3,i); ! data[i] = data[i] + (k1[i] + (2.*(k2[i]+k3[i])) + k4[i])/6.; ! ! // we don't want to experience denormals in the next step */ ! if(fabs((data[i]))<1e-5) ! data[i]=0; ! } ! /* the system might become unstable ... in this case, we'll reset the system */ ! for(int i=0;i<=NUMB_EQ-1;i++) ! { if(data[i]>2) reset(); ! } } *************** *** 219,249 **** void him::m_signal(int n, t_float *const *in, t_float *const *out) { if (regtime) { ! for (int j=0;j!=n;++j) ! { ! runge_kutta_4(dt); ! *(out[0]+j)=data[0]; ! *(out[1]+j)=data[1]; ! *(out[2]+j)=data[2]; ! *(out[3]+j)=data[3]; ! *(out[4]+j)=data[0]*data[2]; ! *(out[5]+j)=(data[0]*data[0]-data[2]*data[2])*0.5; ! } } else { ! for (int j=0;j!=n;++j) ! { ! runge_kutta_4(dt/ ! (2*sqrt(data[0]*data[0]+data[2]*data[2]))); ! *(out[0]+j)=data[0]; ! *(out[1]+j)=data[1]; ! *(out[2]+j)=data[2]; ! *(out[3]+j)=data[3]; ! *(out[4]+j)=data[0]*data[2]; ! *(out[5]+j)=(data[0]*data[0]-data[2]*data[2])*0.5; ! } } } --- 224,261 ---- void him::m_signal(int n, t_float *const *in, t_float *const *out) { + t_float * out0 = out[0]; + t_float * out1 = out[1]; + t_float * out2 = out[2]; + t_float * out3 = out[3]; + t_float * out4 = out[4]; + t_float * out5 = out[5]; + + if (regtime) + { + for (int j=0;j!=n;++j) { ! runge_kutta_4(dt); ! (*(out0)++)=data[0]; ! (*(out1)++)=data[1]; ! (*(out2)++)=data[2]; ! (*(out3)++)=data[3]; ! (*(out4)++)=data[0]*data[2]; ! (*(out5)++)=(data[0]*data[0]-data[2]*data[2])*0.5; } + } else + { + for (int j=0;j!=n;++j) { ! runge_kutta_4(dt/(2*sqrt(data[0]*data[0]+data[2]*data[2]))); ! (*(out0)++)=data[0]; ! (*(out1)++)=data[1]; ! (*(out2)++)=data[2]; ! (*(out3)++)=data[3]; ! (*(out4)++)=data[0]*data[2]; ! (*(out5)++)=(data[0]*data[0]-data[2]*data[2])*0.5; } + } }

1 0

externals/tbext build-max-msvc.bat,NONE,1.1 build-pd-bcc.bat,NONE,1.1 build-pd-cygwin.sh,NONE,1.1 build-pd-darwin.sh,NONE,1.1 build-pd-linux.sh,NONE,1.1 build-pd-mingw.bat,NONE,1.1 build-pd-msvc.bat,NONE,1.1 config-max-msvc.txt,NONE,1.1 config-pd-bcc.txt,NONE,1.1 config-pd-cygwin.txt,NONE,1.1 config-pd-darwin.txt,NONE,1.1 config-pd-mingw.txt,NONE,1.1 config-pd-msvc.txt,NONE,1.1 makefile.pd-cygwin,NONE,1.1 makefile.pd-darwin,NONE,1.1 makefile.pd-mingw,NONE,1.1 makefile.pd-msvc,NONE,1.1
by Tim Blechmann 13 Jul '04

13 Jul '04

Update of /cvsroot/pure-data/externals/tbext In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv29750 Added Files: build-max-msvc.bat build-pd-bcc.bat build-pd-cygwin.sh build-pd-darwin.sh build-pd-linux.sh build-pd-mingw.bat build-pd-msvc.bat config-max-msvc.txt config-pd-bcc.txt config-pd-cygwin.txt config-pd-darwin.txt config-pd-mingw.txt config-pd-msvc.txt makefile.pd-cygwin makefile.pd-darwin makefile.pd-mingw makefile.pd-msvc Log Message: adapted thomas' build system --- NEW FILE: config-pd-cygwin.txt --- # config file for tbext, adapted from Thomas Grill's xsample makefile # # your c++ compiler (define only if it's different than g++) # CXX=g++ # where does the PD installation reside? PD=/cygdrive/c/programme/audio/pd # where are the PD header files? # leave it blank if it is a system directory (like /usr/local/include), # since gcc 3.2 complains about it PDINC=${PD}/src # where do the flext libraries reside? FLEXTPATH=${PD}/flext # where should tbext objects be built? TARGDIR=./pd-cygwin # where should tbext be installed? # (leave blank to omit installation) INSTDIR=${PD}/extra # where should the tbext help be installed? # (leave blank to omit installation) HELPDIR=${PD}/doc/5.reference # additional compiler flags # (check whether they fit to your system!) UFLAGS=-mcpu=pentium4 -msse -mfpmath=sse # gcc 3.2 # UFLAGS=-mcpu=pentiumpro # gcc 2.95 # define to build with shared flext library #FLEXT_SHARED=1 --- NEW FILE: build-pd-linux.sh --- #!/bin/sh SYS=pd-linux . config-${SYS}.txt make -f makefile.${SYS} && { if [ $INSTDIR != "" ]; then echo Now install as root su -c "make -f makefile.${SYS} install" fi if [ $HELPDIR != "" ]; then echo Now install help as root su -c "make -f makefile.${SYS} install-help" fi } --- NEW FILE: build-pd-cygwin.sh --- #!/bin/sh SYS=pd-cygwin . config-${SYS}.txt make -f makefile.${SYS} && { if [ $INSTDIR != "" ]; then make -f makefile.${SYS} install fi if [ $HELPDIR != "" ]; then make -f makefile.${SYS} install-help fi } --- NEW FILE: makefile.pd-cygwin --- # makefile adapted from thomas grill's xsample makefile # # Makefile for gcc @ cygwin # # usage: # to build run "make -f makefile.pd-cygwin" # to install (as root), do "make -f makefile.pd-cygwin install" # CONFIG=config-pd-cygwin.txt MAKEFILE=makefile.pd-cygwin include ${CONFIG} # compiler stuff # /usr/include holds the cygwin pthread headers and must be first! INCLUDES=/usr/include $(PDINC) ./headers/plugin_interface ./headers/common ./headers/server ./headers/app ./headers/lang FLAGS=-DFLEXT_SYS=2 CFLAGS=-O2 -funroll-loops -fmove-all-movables -frerun-loop-opt -finline-functions -fno-rtti -fno-exceptions ${UFLAGS} PDLIBS=$(PD)/bin/pd.dll $(PD)/bin/pthreadVC.dll ifdef FLEXT_SHARED CFLAGS+=-DFLEXT_SHARED FLEXTLIB=$(FLEXTPATH)/flext.dll else FLEXTLIB=$(FLEXTPATH)/flext-pdwin.lib endif # ---------------------------------------------- # the rest can stay untouched # ---------------------------------------------- NAME=tbext include make-files.txt TARGET=$(TARGDIR)/$(NAME).dll # default target all: $(TARGDIR) $(TARGET) $(patsubst %,$(SRCDIR)/%,$(SRCS)): $(patsubst %,$(SRCDIR)/%,$(HDRS)) $(MAKEFILE) $(CONFIG) touch $(patsubst %,$(SRCDIR)/%,$(SRCS)) $(TARGDIR): -mkdir $(TARGDIR) $(TARGDIR)/%.o : $(SRCDIR)/%.cpp $(CXX) -c $(CFLAGS) $(FLAGS) $(patsubst %,-I%,$(INCLUDES) $(FLEXTPATH)) $< -o $@ $(TARGET) : $(patsubst %.cpp,$(TARGDIR)/%.o,$(SRCS)) $(FLEXTLIB) $(CXX) -shared $(LDFLAGS) $^ ${PDLIBS} $(patsubst %,-l%,$(LIBS)) -o $@ strip --strip-unneeded $@ chmod 755 $@ $(INSTDIR): -mkdir $(INSTDIR) install:: $(INSTDIR) install:: $(TARGET) cp $^ $(INSTDIR) # chown root.root $(patsubst %,$(INSTDIR)/%,$(notdir $^)) $(HELPDIR): -mkdir $(HELPDIR) install-help:: $(HELPDIR) install-help:: ./pd chmod 644 $^/*.* cp $^/*.* $(HELPDIR) .PHONY: clean clean: rm -f $(TARGDIR)/*.o $(TARGET) --- NEW FILE: config-pd-msvc.txt --- # config file for tbext, adapted from Thomas Grill's xsample makefile # # where is PD? PDPATH=c:\programme\audio\pd # where do the flext libraries reside? FLEXTPATH=$(PDPATH)\flext # where is MS VC++? # (not necessary if the build started with the compiler environment) # MSVCPATH=c:\programme\prog\microsoft visual studio\VC98 # where should the external be built? OUTPATH=pd-msvc # where should the external be installed? # (leave blank to omit installation) INSTDIR=$(PDPATH)\extra --- NEW FILE: build-pd-mingw.bat --- mingw32-make -f makefile.pd-mingw mingw32-make -f makefile.pd-mingw install mingw32-make -f makefile.pd-mingw install-help --- NEW FILE: makefile.pd-darwin --- # makefile adapted from thomas grill's xsample makefile # # Makefile for gcc @ OSX (darwin) # # usage: # to build run "make -f makefile.pd-darwin" # to install (as root), do "make -f makefile.pd-darwin install" # to install help, do "make -f makefile.pd-darwin install-help" # CONFIG=config-pd-darwin.txt include ${CONFIG} FLEXTLIB=$(FLEXTPATH)/libflext.a # compiler stuff INCLUDES=$(PDINC) ./headers/plugin_interface ./headers/common ./headers/server ./headers/app ./headers/lang FLAGS=-DFLEXT_SYS=2 CFLAGS=${UFLAGS} -dynamic -O2 -Wno-unused -Wno-parentheses -Wno-switch -Wstrict-prototypes -funroll-loops -fmove-all-movables -frerun-loop-opt -fno-rtti -fno-exceptions LIBS=m gcc LDFLAGS=$(FLEXTLIB) -bundle -bundle_loader $(PDBIN) FRAMEWORKS=Carbon veclib # ---------------------------------------------- # the rest can stay untouched # ---------------------------------------------- NAME=tbext include make-files.txt MAKEFILE=makefile.pd-darwin TARGET=$(TARGDIR)/$(NAME).pd_darwin # default target all: $(TARGDIR) $(TARGET) $(patsubst %,$(SRCDIR)/%,$(SRCS)): $(patsubst %,$(SRCDIR)/%,$(HDRS)) $(MAKEFILE) $(CONFIG) touch $(patsubst %,$(SRCDIR)/%,$(SRCS)) $(TARGDIR): -mkdir $(TARGDIR) $(TARGDIR)/%.o : $(SRCDIR)/%.cpp $(CXX) -c $(CFLAGS) $(FLAGS) $(patsubst %,-I%,$(INCLUDES) $(FLEXTPATH)) $< -o $@ $(TARGET) : $(patsubst %.cpp,$(TARGDIR)/%.o,$(SRCS)) $(CXX) $(LDFLAGS) $^ $(patsubst %,-framework %,$(FRAMEWORKS)) $(patsubst %,-L%,$(LIBPATH)) $(patsubst %,-l%,$(LIBS)) -o $@ chmod 755 $@ $(INSTDIR): -mkdir $(INSTDIR) install:: $(INSTDIR) install:: $(TARGET) cp -R $^ $(INSTDIR) # chown -R root.root $(INSTDIR) $(HELPDIR): -mkdir $(HELPDIR) install-help:: $(HELPDIR) install-help:: ./pd cp $^/*.* $(HELPDIR) # chown -R root.root $(HELPDIR) .PHONY: clean clean: rm -f $(TARGDIR)/*.o $(TARGET) --- NEW FILE: makefile.pd-msvc --- # makefile adapted from thomas grill's xsample makefile # # Makefile for MSVC++ 6 # # usage: # to build run "make -f makefile.pd-msvc" # !include config-pd-msvc.txt # includes INCPATH=/I"$(MSVCPATH)\include" /I"$(PDPATH)\src" /I"$(FLEXTPATH)" /I"./headers/plugin_interface" /I"./headers/common" /I"/headers/server" /I"./headers/app" /I"./headers/lang LIBPATH=/LIBPATH:"$(MSVCPATH)\lib" /LIBPATH:"$(PDPATH)\bin" /LIBPATH:"$(FLEXTPATH)" LIBS=pd.lib pthreadVC.lib flext-pdwin.lib # compiler definitions and flags DEFS=/DFLEXT_SYS=2 CFLAGS=/GR- /GX- /GD /G6 /Ox /EHsc /ML # the rest can stay untouched # ---------------------------------------------- NAME=tbext !include make-files.txt all: $(OUTPATH) $(OUTPATH)\$(NAME).dll # remove build clean: -del /q $(OUTPATH) > nul -rmdir $(OUTPATH) > nul OBJS= $(SRCS:.cpp=.obj) $(OUTPATH): -mkdir $(OUTPATH) > nul {$(SRCDIR)}.cpp{}.obj: cl /c /LD $(CFLAGS) $(DEFS) $(INCPATH) $** /Fo$(OUTPATH)\$@ $(OUTPATH)\$(NAME).dll: $(OBJS) cd $(OUTPATH) link /DLL /out:$(NAME).dll /INCREMENTAL:NO $** $(LIBS) $(LIBPATH) @-del *.exp @-del *.lib cd .. !ifdef INSTPATH @-if not exist $(INSTPATH) mkdir $(INSTPATH) copy $@ $(INSTPATH) > nul !endif --- NEW FILE: config-pd-darwin.txt --- # config file for tbext, adapted from Thomas Grill's xsample makefile # # your c++ compiler (if not g++) #CXX=g++-3.3 # where does the PD installation reside? PD=/usr/local/lib/pd # where are the PD header files? # leave it blank if it is a system directory (like /usr/local/include), # since gcc 3.2 complains about it PDINC=${PD}/src # where is the PD executable? PDBIN=/usr/local/bin/pd # where do the flext libraries reside? FLEXTPATH=${PD}/flext # where should the tbext objects be built? TARGDIR=./pd-darwin # where should tbext be installed? # (leave blank to omit installation) INSTDIR=${PD}/extra # where should the tbext help be installed? # (leave blank to omit installation) HELPDIR=${PD}/doc/5.reference # additional compiler flags # (check whether they fit your system!) UFLAGS=-malign-power -maltivec -faltivec --- NEW FILE: build-pd-darwin.sh --- #!/bin/sh SYS=pd-darwin . config-${SYS}.txt make -f makefile.${SYS} && { if [ $INSTDIR != "" ]; then echo Now install as root sudo make -f makefile.${SYS} install fi if [ $HELPDIR != "" ]; then echo Now install help as root sudo make -f makefile.${SYS} install-help fi } --- NEW FILE: build-pd-msvc.bat --- @echo --- Building with MS Visual C++ --- nmake -f makefile.pd-msvc clean nmake -f makefile.pd-msvc --- NEW FILE: config-pd-bcc.txt --- # config file for tbext, adapted from Thomas Grill's xsample makefile # # where is PD? PDPATH=c:\programme\audio\pd # where do the flext libraries reside? FLEXTPATH=$(PDPATH)\flext # where is BorlandC++? BCCPATH=c:\programme\prog\bcc55 # where should the external(s) be built? OUTPATH=.\pd-bcc # should the external be installed? (yes/no) INSTALL=yes # where should the external be installed? INSTDIR=$(PDPATH)\extra # additional compiler flags UFLAGS=-6 -O2 -OS -ff --- NEW FILE: config-max-msvc.txt --- # config file for tbext, adapted from Thomas Grill's xsample makefile # # where is the Max SDK? # you should have the latest version! MAXSDKPATH="F:\prog\audio\MaxWinSDK\c74support" # where do the flext libraries reside? FLEXTPATH="%CommonProgramFiles%\Cycling '74\flext" # where is MS VC++? # (need not be defined if the build is started with the compiler environment set) # MSVCPATH="c:\programme\prog\microsoft visual studio\VC98" # where should the external be built? # (path for temporary files) OUTPATH=max-msvc # where should the external be installed? # (leave blank to omit installation) INSTPATH="%CommonProgramFiles%\Cycling '74\externals\flext" # some user-definable flags # (check if they match your system!) UFLAGS=/G6 /arch:SSE --- NEW FILE: build-max-msvc.bat --- @echo --- Building with MS Visual C++ --- nmake -f makefile.max-msvc clean nmake -f makefile.max-msvc --- NEW FILE: build-pd-bcc.bat --- @echo --- Building with BorlandC++ --- make -f makefile.pd-bcc --- NEW FILE: makefile.pd-mingw --- # makefile adapted from thomas grill's xsample makefile # # Makefile for gcc @ minGW # # usage: # to build run "make -f makefile.pd-mingw" # to install (as root), do "make -f makefile.pd-mingw install" # CONFIG=config-pd-mingw.txt MAKEFILE=makefile.pd-mingw include ${CONFIG} # compiler stuff INCLUDES=$(PDINC) ./headers/plugin_interface ./headers/common ./headers/server ./headers/app ./headers/lang FLAGS=-DFLEXT_SYS=2 CFLAGS=-O2 -funroll-loops -fmove-all-movables -frerun-loop-opt -finline-functions -fno-rtti -fno-exceptions ${UFLAGS} PDLIBS=$(PD)/bin/pd.dll $(PD)/bin/pthreadVC.dll ifdef FLEXT_SHARED CFLAGS+=-DFLEXT_SHARED FLEXTLIB=$(FLEXTPATH)/flext.dll else FLEXTLIB=$(FLEXTPATH)/flext-pdwin.lib endif # ---------------------------------------------- # the rest can stay untouched # ---------------------------------------------- NAME=tbext include make-files.txt TARGET=$(TARGDIR)/$(NAME).dll # default target all: $(TARGDIR) $(TARGET) $(patsubst %,$(SRCDIR)/%,$(SRCS)): $(patsubst %,$(SRCDIR)/%,$(HDRS)) $(MAKEFILE) $(CONFIG) # echo touch $(patsubst %,$(SRCDIR)/%,$(SRCS)) # minGW has no touch $(TARGDIR): -mkdir $(TARGDIR) $(TARGDIR)/%.o : $(SRCDIR)/%.cpp $(CXX) -c $(CFLAGS) $(FLAGS) $(patsubst %,-I%,$(INCLUDES) $(FLEXTPATH)) $< -o $@ $(TARGET) : $(patsubst %.cpp,$(TARGDIR)/%.o,$(SRCS)) $(FLEXTLIB) $(CXX) -shared $(LDFLAGS) $^ ${PDLIBS} $(patsubst %,-l%,$(LIBS)) -o $@ strip --strip-unneeded $@ chmod 755 $@ $(INSTDIR): -mkdir $(INSTDIR) install:: $(INSTDIR) install:: $(TARGET) cp $^ $(INSTDIR) # chown root.root $(patsubst %,$(INSTDIR)/%,$(notdir $^)) $(HELPDIR): -mkdir $(HELPDIR) install-help:: $(HELPDIR) install-help:: ./pd chmod 644 $^/*.* cp $^/*.* $(HELPDIR) .PHONY: clean clean: rm -f $(TARGDIR)/*.o $(TARGET) --- NEW FILE: config-pd-mingw.txt --- # config file for tbext, adapted from Thomas Grill's xsample makefile # # your c++ compiler (define only if it's different than g++) # CXX=g++ # where does the PD installation reside? PD=c:/programme/audio/pd # where are the PD header files? # leave it blank if it is a system directory (like /usr/local/include), # since gcc 3.2 complains about it PDINC=${PD}/src # where do the flext libraries reside? FLEXTPATH=${PD}/flext # where should tbext objects be built? TARGDIR=./pd-mingw # where should tbext be installed? # (leave blank to omit installation) INSTDIR=${PD}/extra # where should the tbext help be installed? # (leave blank to omit installation) HELPDIR=${PD}/doc/5.reference # additional compiler flags # (check whether they fit to your system!) UFLAGS=-mcpu=pentium4 -msse -mfpmath=sse # gcc 3.2 # UFLAGS=-mcpu=pentiumpro # gcc 2.95 # define to build with shared flext library # FLEXT_SHARED=1

1 0

externals/tbext config-pd-linux.txt,1.6,1.7
by Tim Blechmann 13 Jul '04

13 Jul '04

Update of /cvsroot/pure-data/externals/tbext In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv29437 Modified Files: config-pd-linux.txt Log Message: minor changes Index: config-pd-linux.txt =================================================================== RCS file: /cvsroot/pure-data/externals/tbext/config-pd-linux.txt,v retrieving revision 1.6 retrieving revision 1.7 diff -C2 -d -r1.6 -r1.7 *** config-pd-linux.txt 30 Mar 2004 18:41:12 -0000 1.6 --- config-pd-linux.txt 13 Jul 2004 20:00:39 -0000 1.7 *************** *** 17,28 **** FLEXTPATH=/usr/lib/flext ! # where should xsample objects be built? TARGDIR=./pd-linux ! # where should xsample be installed? # (leave blank to omit installation) INSTDIR=${PD}/externs ! # where should the xsample help be installed? # (leave blank to omit installation) HELPDIR=${PD}/doc/5.reference --- 17,28 ---- FLEXTPATH=/usr/lib/flext ! # where should tbext objects be built? TARGDIR=./pd-linux ! # where should tbext be installed? # (leave blank to omit installation) INSTDIR=${PD}/externs ! # where should the tbext help be installed? # (leave blank to omit installation) HELPDIR=${PD}/doc/5.reference

1 0

externals/hardware/memPIO README.txt,1.1.1.1,1.2
by IOhannes m zmölnig 13 Jul '04

13 Jul '04

Update of /cvsroot/pure-data/externals/hardware/memPIO In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv16469 Modified Files: README.txt Log Message: forgot to mention that memPIO is windoze only (what a shame) Index: README.txt =================================================================== RCS file: /cvsroot/pure-data/externals/hardware/memPIO/README.txt,v retrieving revision 1.1.1.1 retrieving revision 1.2 diff -C2 -d -r1.1.1.1 -r1.2 *** README.txt 13 Jul 2004 16:33:26 -0000 1.1.1.1 --- README.txt 13 Jul 2004 16:49:31 -0000 1.2 *************** *** 1,8 **** ! memPIO: an external to read from/write to the "mem-PIO"-device by bmcm ( httP://www.bmcm.de ) with pure-data. the "mem-PIO" is a USB-device that offers digital (this is: 1 & 0) I/O. there are 3 ports, each 8 channels. each port can be set to either input or output ! (not quite true: port3 is split into 2 subports (4 channels each) that can be set to input resp. output separately; however this is not supported by this external) --- 1,10 ---- ! memPIO: an external to read from/write to the "mem-PIO"-device ! by bmcm ( httP://www.bmcm.de ) with pure-data. the "mem-PIO" is a USB-device that offers digital (this is: 1 & 0) I/O. there are 3 ports, each 8 channels. each port can be set to either input or output ! (not quite true: port3 is split into 2 subports (4 channels each) ! that can be set to input resp. output separately; however this is not supported by this external) *************** *** 10,22 **** general:: ! memPIO is a plugin for miller.s.puckette's realtime-computermusic-environment "puredata" (or abbreviated "pd") this software will be of no use, if you don't have a running version of pd on your system. check out for http://pd.iem.at to learn more about pd and how to get it. note: [memPIO] should be published under the Gnu General Public License. ! However, it is heavily based on the example-code provided by bmcm, which does not come with a copyright notice ! (at least i didn't find one) ! unfortunately [memPIO] is windows only (there are no drivers for other OS! shame on bmcm) win32 : --- 12,27 ---- general:: ! memPIO is a plugin for miller.s.puckette's realtime-computermusic-environment ! "puredata" (or abbreviated "pd") this software will be of no use, if you don't have a running version of pd on your system. check out for http://pd.iem.at to learn more about pd and how to get it. note: [memPIO] should be published under the Gnu General Public License. ! However, it is heavily based on the example-code provided by bmcm, ! which does not come with a copyright notice (at least i didn't find one) ! ! IMPORTANT NOTE: ! unfortunately [memPIO] is WINDOZE only (there are no drivers for other OS! shame on bmcm) win32 : *************** *** 27,34 **** compiling: ! if you want to compile it for yourself, use the workspace memPIO.dsw. i am not sure, whether it works with anything but m$ visual-c++ 6.0 authors:: ! this software is copyleft by iohannes m zmoelnig <zmoelnig(a)iem.kug.ac.at> --- 32,40 ---- compiling: ! if you want to compile it for yourself, use the workspace memPIO.dsw. ! i am not sure, whether it works with anything but m$ visual-c++ 6.0 authors:: ! this software is copyleft by iohannes m zmoelnig <zmoelnig(a)iem.at>

1 0

pd/src d_soundfile.c,1.1.1.2.2.16,1.1.1.2.2.17
by Tim Blechmann 12 Jul '04

12 Jul '04

Update of /cvsroot/pure-data/pd/src In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv25747 Modified Files: Tag: devel_0_37 d_soundfile.c Log Message: main thread waits for helper thread Index: d_soundfile.c =================================================================== RCS file: /cvsroot/pure-data/pd/src/d_soundfile.c,v retrieving revision 1.1.1.2.2.16 retrieving revision 1.1.1.2.2.17 diff -C2 -d -r1.1.1.2.2.16 -r1.1.1.2.2.17 *** d_soundfile.c 29 Jun 2004 20:48:18 -0000 1.1.1.2.2.16 --- d_soundfile.c 12 Jul 2004 14:15:07 -0000 1.1.1.2.2.17 *************** *** 1004,1007 **** --- 1004,1009 ---- void soundfiler_thread(void) { + /* thread has been started, the main thread can continue */ + pthread_cond_signal(&(soundfiler_queue->cond)); while (1) *************** *** 1180,1183 **** --- 1182,1189 ---- post("global soundfiler thread launched, priority: %d", sf_param.sched_priority); + + /* we wait until the helper thread is ready */ + pthread_cond_wait(&soundfiler_queue->cond, &soundfiler_queue->mutex); + }

1 0

pd/src cvs_changes.txt,1.1.2.2,1.1.2.3
by Tim Blechmann 12 Jul '04

12 Jul '04

Update of /cvsroot/pure-data/pd/src In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv25433 Modified Files: Tag: devel_0_37 cvs_changes.txt Log Message: updates Index: cvs_changes.txt =================================================================== RCS file: /cvsroot/pure-data/pd/src/Attic/cvs_changes.txt,v retrieving revision 1.1.2.2 retrieving revision 1.1.2.3 diff -C2 -d -r1.1.2.2 -r1.1.2.3 *** cvs_changes.txt 6 May 2004 08:50:07 -0000 1.1.2.2 --- cvs_changes.txt 12 Jul 2004 14:13:00 -0000 1.1.2.3 *************** *** 1,13 **** Changes between Miller Puckette's branch and the current devel_0_37 branch: ! - Support for the fftw library (--enable-fftw) ! - Comments in .pdrc files. Lines beginning with '#' will be dropped ! - Changes to the autoconf script: icc compiler (linux) and pentium3/4 optimizations without editing the makefiles (--enable-icc, --enable-optimize=pentium3/pentium4) ! - Added Yves Degoyon's patch for the threaded gui (--enable-threadedgui) - GG: inlets take tooltips (first parameters) --- 1,13 ---- Changes between Miller Puckette's branch and the current devel_0_37 branch: ! - TB: Support for the fftw library (--enable-fftw) ! - TB: Comments in .pdrc files. Lines beginning with '#' will be dropped ! - TB: Changes to the autoconf script: icc compiler (linux) and pentium3/4 optimizations without editing the makefiles (--enable-icc, --enable-optimize=pentium3/pentium4) ! - TB: Added Yves Degoyon's patch for the threaded gui (--enable-threadedgui) - GG: inlets take tooltips (first parameters) *************** *** 15,16 **** --- 15,22 ---- - GG: -rt mode does not check for uid() and gui() (on linux with realtime-lsm -rt works without being user root). + + - TB: thread lock for t_garrays (--enable-threadedsf) + + - TB: threaded soundfiler (--enable-threadedsf) + + - TB: port of thomas grill's vectorized simd funcions for vc to gcc \ No newline at end of file

1 0

pd/src m_simd_sse_gcc.c,1.1.2.1,1.1.2.2 m_simd_sse_gcc.h,1.1.2.3,1.1.2.4
by Tim Blechmann 10 Jul '04

10 Jul '04

Update of /cvsroot/pure-data/pd/src In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv20643 Modified Files: Tag: devel_0_37 m_simd_sse_gcc.c m_simd_sse_gcc.h Log Message: port of thomas grill's vectorized simd functions for vc to gcc Index: m_simd_sse_gcc.c =================================================================== RCS file: /cvsroot/pure-data/pd/src/Attic/m_simd_sse_gcc.c,v retrieving revision 1.1.2.1 retrieving revision 1.1.2.2 diff -C2 -d -r1.1.2.1 -r1.1.2.2 *** m_simd_sse_gcc.c 23 Dec 2003 01:15:39 -0000 1.1.2.1 --- m_simd_sse_gcc.c 10 Jul 2004 19:57:20 -0000 1.1.2.2 *************** *** 10,12 **** --- 10,997 ---- + /* TB: adapted from the xxx_sse_vc routines + since gcc can't access the c code as vc can, the functions itself have + been adapted to assembler */ + + /* zero_perf_sse_gcc (t_int * w)*/ + asm( + ".set T_FLOAT,4 \n" /* sizeof(t_float) */ + ".set T_INT,4 \n" /* sizeof(t_int) */ + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl zero_perf_sse_gcc \n" + "zero_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "pushl %esi \n" + "movl %esi, -4(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument pointer */ + "movl T_INT(%esi), %edx \n" /* out */ + "movl 2*T_INT(%esi), %ecx \n" /* n */ + + "xorps %xmm0, %xmm0 \n" /* load zero */ + "shrl $4, %ecx \n" /* divide by 16 */ + + /* loop: *out=0 */ + "zpsg_loop: \n" + "movaps %xmm0, (%edx) \n" + "movaps %xmm0, 4*T_FLOAT(%edx) \n" + "movaps %xmm0, 8*T_FLOAT(%edx) \n" + "movaps %xmm0, 12*T_FLOAT(%edx) \n" + + "addl $64, %edx \n" /* out+=16 */ + "loop zpsg_loop \n" + + + /* return w+3; */ + "movl -4(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $12, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + + ".align 4 \n" /* alignment */ + ".type zero_perf_sse_gcc, @function \n" + + ); + + /* copy_perf_sse_gcc (t_int * w)*/ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl copy_perf_sse_gcc \n" + "copy_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument pointer */ + "movl 1*T_INT(%esi), %ebx \n" /* in1 */ + "movl 2*T_INT(%esi), %edx \n" /* out */ + "movl 3*T_INT(%esi), %ecx \n" /* n */ + "shrl $4, %ecx \n" /* divide by 16 */ + + /* loop: *out = *in */ + "cpsg_loop: \n" + "movaps (%ebx), %xmm0 \n" + "movaps %xmm0, (%edx) \n" + "movaps 4*T_FLOAT(%ebx), %xmm1 \n" + "movaps %xmm1, 4*T_FLOAT(%edx) \n" + + "movaps 8*T_FLOAT(%ebx), %xmm2 \n" + "movaps %xmm2, 8*T_FLOAT(%edx) \n" + "movaps 12*T_FLOAT(%ebx), %xmm3 \n" + "movaps %xmm3, 12*T_FLOAT(%edx) \n" + + "addl $64, %ebx \n" /* in1 +=16 */ + "addl $64, %edx \n" /* out +=16 */ + "loop cpsg_loop \n" + + /* return w+4; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $16, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + + ".align 4 \n" /* alignment */ + ".type copy_perf_sse_gcc, @function \n" + ); + + + /* sig_tilde_perf_sse_gcc(t_int * w) */ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl sig_tilde_perf_sse_gcc \n" + "sig_tilde_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "pushl %esi \n" + "movl %esi, -4(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument pointer */ + "movl 2*T_INT(%esi), %edx \n" /* out */ + "movl 1*T_INT(%esi), %eax \n" /* f */ + "movl 3*T_INT(%esi), %ecx \n" /* n */ + + /* set registers to f */ + "movss (%eax), %xmm0 \n" + "shufps $0, %xmm0, %xmm0 \n" + "movl 12(%esi), %ecx \n" + "shrl $4, %ecx \n" /* divide by 16 */ + + /* loop: *out = f */ + "sigtpsg_loop: \n" + "movaps %xmm0, (%edx) \n" + "movaps %xmm0, 16(%edx) \n" + "movaps %xmm0, 32(%edx) \n" + "movaps %xmm0, 48(%edx) \n" + "addl $64, %edx \n" /* out+=16 */ + "loop sigtpsg_loop \n" + + /* return w+4; */ + "movl -4(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $16, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + ".align 4 \n" /* alignment */ + ".type sig_tilde_perf_sse_gcc, @function \n" + + ); + + + + /* plus_perf_sse_gcc (t_int * w)*/ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl plus_perf_sse_gcc \n" + "plus_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument vector */ + + "movl 1*T_INT(%esi), %eax \n" /* in1 */ + "movl 2*T_INT(%esi), %ebx \n" /* in2 */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + "shrl $4, %ecx \n" /* divide by 16 */ + "xorl %esi, %esi \n" /* reset index */ + + /* loop: *out = *in1 + *in2 */ + "ppsg_loop: \n" + "movaps (%eax,%esi), %xmm0 \n" + "movaps (%ebx,%esi), %xmm1 \n" + "addps %xmm1, %xmm0 \n" + "movaps %xmm0, (%edx,%esi) \n" + + "movaps 4*T_FLOAT(%eax,%esi), %xmm2 \n" + "movaps 4*T_FLOAT(%ebx,%esi), %xmm3 \n" + "addps %xmm3, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx,%esi) \n" + + "movaps 8*T_FLOAT(%eax,%esi), %xmm4 \n" + "movaps 8*T_FLOAT(%ebx,%esi), %xmm5 \n" + "addps %xmm5, %xmm4 \n" + "movaps %xmm4, 8*T_FLOAT(%edx,%esi) \n" + + "movaps 12*T_FLOAT(%eax,%esi), %xmm6 \n" + "movaps 12*T_FLOAT(%ebx,%esi), %xmm7 \n" + "addps %xmm7, %xmm6 \n" + "movaps %xmm6, 12*T_FLOAT(%edx,%esi) \n" + + "addl $64, %esi \n" /* out+=16; */ + "loop ppsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + + ".align 4 \n" /* alignment */ + ".type plus_perf_sse_gcc, @function \n" + + ); + + /* scalarplus_perf_sse_gcc(t_int *w) */ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl scalarplus_perf_sse_gcc \n" + "scalarplus_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument pointer */ + "movl T_INT(%esi), %ebx \n" /* in */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 2*T_INT(%esi), %eax \n" /* value */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + + "movss (%eax), %xmm0 \n" + "shufps $0, %xmm0, %xmm0 \n" + "shrl $4, %ecx \n" /* divide by 16 */ + + /* loop: *out = *in + value */ + "sppsg_loop: \n" + + "movaps (%ebx), %xmm1 \n" + "addps %xmm0, %xmm1 \n" + "movaps %xmm1, (%edx) \n" + + "movaps 4*T_FLOAT(%ebx), %xmm2 \n" + "addps %xmm0, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx) \n" + + "movaps 8*T_FLOAT(%ebx), %xmm3 \n" + "addps %xmm0, %xmm3 \n" + "movaps %xmm3, 8*T_FLOAT(%edx) \n" + + "movaps 12*T_FLOAT(%ebx), %xmm4 \n" + "addps %xmm0, %xmm4 \n" + "movaps %xmm4, 12*T_FLOAT(%edx) \n" + + "addl $64, %ebx \n" /* in += 16 */ + "addl $64, %edx \n" /* out += 16 */ + "loop sppsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + ".align 4 \n" /* alignment */ + ".type scalarplus_perf_sse_gcc, @function \n" + + ); + + + /* minus_perf_sse_gcc(t_int *w) */ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl minus_perf_sse_gcc \n" + "minus_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument pointer */ + "movl 1*T_INT(%esi), %eax \n" /* in1 */ + "movl 2*T_INT(%esi), %ebx \n" /* in2 */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + "shrl $4, %ecx \n" /* divide by 16 */ + "xorl %esi, %esi \n" /* reset register */ + + + /* loop: *out = *in1 - *in2 */ + "mpsg_loop: \n" + + "movaps (%eax,%esi), %xmm0 \n" + "movaps (%ebx,%esi), %xmm1 \n" + "subps %xmm1, %xmm0 \n" + "movaps %xmm0, (%edx,%esi) \n" + + "movaps 4*T_FLOAT(%eax,%esi), %xmm2 \n" + "movaps 4*T_FLOAT(%ebx,%esi), %xmm3 \n" + "subps %xmm3, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx,%esi) \n" + + "movaps 8*T_FLOAT(%eax,%esi), %xmm4 \n" + "movaps 8*T_FLOAT(%ebx,%esi), %xmm5 \n" + "subps %xmm5, %xmm4 \n" + "movaps %xmm4, 8*T_FLOAT(%edx,%esi) \n" + + "movaps 12*T_FLOAT(%eax,%esi), %xmm6 \n" + "movaps 12*T_FLOAT(%ebx,%esi), %xmm7 \n" + "subps %xmm7, %xmm6 \n" + "movaps %xmm6, 12*T_FLOAT(%edx,%esi) \n" + "addl $64, %esi \n" /* out +=16 */ + + "loop mpsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + ".align 4 \n" /* alignment */ + ".type minus_perf_sse_gcc, @function \n" + + ); + + /* scalarminus_perf_sse_gcc(t_int *w) */ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl scalarminus_perf_sse_gcc \n" + "scalarminus_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument pointer */ + "movl T_INT(%esi), %ebx \n" /* in */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 2*T_INT(%esi), %eax \n" /* value */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + + "movss (%eax), %xmm0 \n" + "shufps $0, %xmm0, %xmm0 \n" + "shrl $4, %ecx \n" /* divide by 16 */ + + /* loop: *out = *in - value */ + "smpsg_loop: \n" + + "movaps (%ebx), %xmm1 \n" + "subps %xmm0, %xmm1 \n" + "movaps %xmm1, (%edx) \n" + + "movaps 4*T_FLOAT(%ebx), %xmm2 \n" + "subps %xmm0, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx) \n" + + "movaps 8*T_FLOAT(%ebx), %xmm3 \n" + "subps %xmm0, %xmm3 \n" + "movaps %xmm3, 8*T_FLOAT(%edx) \n" + + "movaps 12*T_FLOAT(%ebx), %xmm4 \n" + "subps %xmm0, %xmm4 \n" + "movaps %xmm4, 12*T_FLOAT(%edx) \n" + + "addl $64, %ebx \n" /* in += 16 */ + "addl $64, %edx \n" /* out += 16 */ + "loop smpsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + ".align 4 \n" /* alignment */ + ".type scalarminus_perf_sse_gcc, @function\n" + + ); + + /* times_perf_sse_gcc (t_int * w)*/ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl times_perf_sse_gcc \n" + "times_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument vector */ + + "movl 1*T_INT(%esi), %eax \n" /* in1 */ + "movl 2*T_INT(%esi), %ebx \n" /* in2 */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + "shrl $4, %ecx \n" /* divide by 16 */ + "xorl %esi, %esi \n" /* reset index */ + + /* loop: *out = *in1 * *in2 */ + "tpsg_loop: \n" + "movaps (%eax,%esi), %xmm0 \n" + "movaps (%ebx,%esi), %xmm1 \n" + "mulps %xmm1, %xmm0 \n" + "movaps %xmm0, (%edx,%esi) \n" + + "movaps 4*T_FLOAT(%eax,%esi), %xmm2 \n" + "movaps 4*T_FLOAT(%ebx,%esi), %xmm3 \n" + "mulps %xmm3, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx,%esi) \n" + + "movaps 8*T_FLOAT(%eax,%esi), %xmm4 \n" + "movaps 8*T_FLOAT(%ebx,%esi), %xmm5 \n" + "mulps %xmm5, %xmm4 \n" + "movaps %xmm4, 8*T_FLOAT(%edx,%esi) \n" + + "movaps 12*T_FLOAT(%eax,%esi), %xmm6 \n" + "movaps 12*T_FLOAT(%ebx,%esi), %xmm7 \n" + "mulps %xmm7, %xmm6 \n" + "movaps %xmm6, 12*T_FLOAT(%edx,%esi) \n" + + "addl $64, %esi \n" /* out+=16; */ + "loop tpsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + + ".align 4 \n" /* alignment */ + ".type times_perf_sse_gcc, @function \n" + + ); + + /* scalartimes_perf_sse_gcc(t_int *w) */ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl scalartimes_perf_sse_gcc \n" + "scalartimes_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument pointer */ + "movl T_INT(%esi), %ebx \n" /* in */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 2*T_INT(%esi), %eax \n" /* value */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + + "movss (%eax), %xmm0 \n" + "shufps $0, %xmm0, %xmm0 \n" + "shrl $4, %ecx \n" /* divide by 16 */ + + /* loop: *out = *in * n */ + "stpsg_loop: \n" + + "movaps (%ebx), %xmm1 \n" + "mulps %xmm0, %xmm1 \n" + "movaps %xmm1, (%edx) \n" + + "movaps 4*T_FLOAT(%ebx), %xmm2 \n" + "mulps %xmm0, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx) \n" + + "movaps 8*T_FLOAT(%ebx), %xmm3 \n" + "mulps %xmm0, %xmm3 \n" + "movaps %xmm3, 8*T_FLOAT(%edx) \n" + + "movaps 12*T_FLOAT(%ebx), %xmm4 \n" + "mulps %xmm0, %xmm4 \n" + "movaps %xmm4, 12*T_FLOAT(%edx) \n" + + "addl $64, %ebx \n" /* in += 16 */ + "addl $64, %edx \n" /* out += 16 */ + "loop stpsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + ".align 4 \n" /* alignment */ + ".type scalartimes_perf_sse_gcc, @function\n" + + ); + + /* over_perf_sse_gcc (t_int * w)*/ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl over_perf_sse_gcc \n" + "over_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument vector */ + + "movl 1*T_INT(%esi), %eax \n" /* in1 */ + "movl 2*T_INT(%esi), %ebx \n" /* in2 */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + "shrl $4, %ecx \n" /* divide by 16 */ + "xorl %esi, %esi \n" /* reset index */ + + /* loop: *out = *in1 / *in2 */ + "opsg_loop: \n" + "movaps (%eax,%esi), %xmm0 \n" + "movaps (%ebx,%esi), %xmm1 \n" + "divps %xmm1, %xmm0 \n" + "movaps %xmm0, (%edx,%esi) \n" + + "movaps 4*T_FLOAT(%eax,%esi), %xmm2 \n" + "movaps 4*T_FLOAT(%ebx,%esi), %xmm3 \n" + "divps %xmm3, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx,%esi) \n" + + "movaps 8*T_FLOAT(%eax,%esi), %xmm4 \n" + "movaps 8*T_FLOAT(%ebx,%esi), %xmm5 \n" + "divps %xmm5, %xmm4 \n" + "movaps %xmm4, 8*T_FLOAT(%edx,%esi) \n" + + "movaps 12*T_FLOAT(%eax,%esi), %xmm6 \n" + "movaps 12*T_FLOAT(%ebx,%esi), %xmm7 \n" + "divps %xmm7, %xmm6 \n" + "movaps %xmm6, 12*T_FLOAT(%edx,%esi) \n" + + "addl $64, %esi \n" /* out+=16; */ + "loop opsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + + ".align 4 \n" /* alignment */ + ".type over_perf_sse_gcc, @function \n" + + ); + + /* scalarover_perf_sse_gcc(t_int *w) */ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl scalarover_perf_sse_gcc \n" + "scalarover_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument pointer */ + "movl T_INT(%esi), %ebx \n" /* in */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 2*T_INT(%esi), %eax \n" /* value */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + + "movss (%eax), %xmm0 \n" + "shufps $0, %xmm0, %xmm0 \n" + "shrl $4, %ecx \n" /* divide by 16 */ + + /* loop: *out = *in / n */ + "sopsg_loop: \n" + + "movaps (%ebx), %xmm1 \n" + "divps %xmm0, %xmm1 \n" + "movaps %xmm1, (%edx) \n" + + "movaps 4*T_FLOAT(%ebx), %xmm2 \n" + "divps %xmm0, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx) \n" + + "movaps 8*T_FLOAT(%ebx), %xmm3 \n" + "divps %xmm0, %xmm3 \n" + "movaps %xmm3, 8*T_FLOAT(%edx) \n" + + "movaps 12*T_FLOAT(%ebx), %xmm4 \n" + "divps %xmm0, %xmm4 \n" + "movaps %xmm4, 12*T_FLOAT(%edx) \n" + + "addl $64, %ebx \n" /* in += 16 */ + "addl $64, %edx \n" /* out += 16 */ + "loop sopsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + ".align 4 \n" /* alignment */ + ".type scalarover_perf_sse_gcc, @function \n" + + ); + + /* min_perf_sse_gcc (t_int * w)*/ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl min_perf_sse_gcc \n" + "min_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument vector */ + + "movl 1*T_INT(%esi), %eax \n" /* in1 */ + "movl 2*T_INT(%esi), %ebx \n" /* in2 */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + "shrl $4, %ecx \n" /* divide by 16 */ + "xorl %esi, %esi \n" /* reset index */ + + /* loop: *out = min(*in1, *in2) */ + "minpsg_loop: \n" + "movaps (%eax,%esi), %xmm0 \n" + "movaps (%ebx,%esi), %xmm1 \n" + "minps %xmm1, %xmm0 \n" + "movaps %xmm0, (%edx,%esi) \n" + + "movaps 4*T_FLOAT(%eax,%esi), %xmm2 \n" + "movaps 4*T_FLOAT(%ebx,%esi), %xmm3 \n" + "minps %xmm3, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx,%esi) \n" + + "movaps 8*T_FLOAT(%eax,%esi), %xmm4 \n" + "movaps 8*T_FLOAT(%ebx,%esi), %xmm5 \n" + "minps %xmm5, %xmm4 \n" + "movaps %xmm4, 8*T_FLOAT(%edx,%esi) \n" + + "movaps 12*T_FLOAT(%eax,%esi), %xmm6 \n" + "movaps 12*T_FLOAT(%ebx,%esi), %xmm7 \n" + "minps %xmm7, %xmm6 \n" + "movaps %xmm6, 12*T_FLOAT(%edx,%esi) \n" + + "addl $64, %esi \n" /* out+=16; */ + "loop minpsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + + ".align 4 \n" /* alignment */ + ".type min_perf_sse_gcc, @function \n" + + ); + + /* scalarmin_perf_sse_gcc(t_int *w) */ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl scalarmin_perf_sse_gcc \n" + "scalarmin_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument pointer */ + "movl T_INT(%esi), %ebx \n" /* in */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 2*T_INT(%esi), %eax \n" /* value */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + + "movss (%eax), %xmm0 \n" + "shufps $0, %xmm0, %xmm0 \n" + "shrl $4, %ecx \n" /* divide by 16 */ + + /* loop: *out = min(*in,value) */ + "sminopsg_loop: \n" + + "movaps (%ebx), %xmm1 \n" + "minps %xmm0, %xmm1 \n" + "movaps %xmm1, (%edx) \n" + + "movaps 4*T_FLOAT(%ebx), %xmm2 \n" + "minps %xmm0, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx) \n" + + "movaps 8*T_FLOAT(%ebx), %xmm3 \n" + "minps %xmm0, %xmm3 \n" + "movaps %xmm3, 8*T_FLOAT(%edx) \n" + + "movaps 12*T_FLOAT(%ebx), %xmm4 \n" + "minps %xmm0, %xmm4 \n" + "movaps %xmm4, 12*T_FLOAT(%edx) \n" + + "addl $64, %ebx \n" /* in += 16 */ + "addl $64, %edx \n" /* out += 16 */ + "loop sminopsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + ".align 4 \n" /* alignment */ + ".type scalarmin_perf_sse_gcc, @function \n" + + ); + + /* max_perf_sse_gcc (t_int * w)*/ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl max_perf_sse_gcc \n" + "max_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument vector */ + + "movl 1*T_INT(%esi), %eax \n" /* in1 */ + "movl 2*T_INT(%esi), %ebx \n" /* in2 */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + "shrl $4, %ecx \n" /* divide by 16 */ + "xorl %esi, %esi \n" /* reset index */ + + /* loop: *out = max(*in1, *in2) */ + "maxpsg_loop: \n" + "movaps (%eax,%esi), %xmm0 \n" + "movaps (%ebx,%esi), %xmm1 \n" + "maxps %xmm1, %xmm0 \n" + "movaps %xmm0, (%edx,%esi) \n" + + "movaps 4*T_FLOAT(%eax,%esi), %xmm2 \n" + "movaps 4*T_FLOAT(%ebx,%esi), %xmm3 \n" + "maxps %xmm3, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx,%esi) \n" + + "movaps 8*T_FLOAT(%eax,%esi), %xmm4 \n" + "movaps 8*T_FLOAT(%ebx,%esi), %xmm5 \n" + "maxps %xmm5, %xmm4 \n" + "movaps %xmm4, 8*T_FLOAT(%edx,%esi) \n" + + "movaps 12*T_FLOAT(%eax,%esi), %xmm6 \n" + "movaps 12*T_FLOAT(%ebx,%esi), %xmm7 \n" + "maxps %xmm7, %xmm6 \n" + "movaps %xmm6, 12*T_FLOAT(%edx,%esi) \n" + + "addl $64, %esi \n" /* out+=16; */ + "loop maxpsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + + ".align 4 \n" /* alignment */ + ".type max_perf_sse_gcc, @function \n" + + ); + + /* scalarmax_perf_sse_gcc(t_int *w) */ + asm( + ".set T_FLOAT,4 \n" + ".set T_INT,4 \n" + + /* header */ + ".text \n" + ".align 4 \n" /* alignment */ + + ".globl scalarmax_perf_sse_gcc \n" + "scalarmax_perf_sse_gcc: \n" + + /* head of function */ + "pushl %ebp \n" + "movl %esp, %ebp \n" + "subl $8, %esp \n" + "movl %ebx, -4(%ebp) \n" + "movl %esi, -8(%ebp) \n" + + /* get arguments */ + "movl 8(%ebp), %esi \n" /* argument pointer */ + "movl T_INT(%esi), %ebx \n" /* in */ + "movl 3*T_INT(%esi), %edx \n" /* out */ + "movl 2*T_INT(%esi), %eax \n" /* value */ + "movl 4*T_INT(%esi), %ecx \n" /* n */ + + "movss (%eax), %xmm0 \n" + "shufps $0, %xmm0, %xmm0 \n" + "shrl $4, %ecx \n" /* divide by 16 */ + + /* loop: *out = max(*in,value) */ + "smaxopsg_loop: \n" + + "movaps (%ebx), %xmm1 \n" + "maxps %xmm0, %xmm1 \n" + "movaps %xmm1, (%edx) \n" + + "movaps 4*T_FLOAT(%ebx), %xmm2 \n" + "maxps %xmm0, %xmm2 \n" + "movaps %xmm2, 4*T_FLOAT(%edx) \n" + + "movaps 8*T_FLOAT(%ebx), %xmm3 \n" + "maxps %xmm0, %xmm3 \n" + "movaps %xmm3, 8*T_FLOAT(%edx) \n" + + "movaps 12*T_FLOAT(%ebx), %xmm4 \n" + "maxps %xmm0, %xmm4 \n" + "movaps %xmm4, 12*T_FLOAT(%edx) \n" + + "addl $64, %ebx \n" /* in += 16 */ + "addl $64, %edx \n" /* out += 16 */ + "loop smaxopsg_loop \n" + + /* return w+5; */ + "movl -4(%ebp), %ebx \n" + "movl -8(%ebp), %esi \n" + "movl 8(%ebp), %eax \n" + "addl $20, %eax \n" + "movl %ebp, %esp \n" + "popl %ebp \n" + "ret \n" + + ".align 4 \n" /* alignment */ + ".type scalarmax_perf_sse_gcc, @function \n" + + ); + + #endif + Index: m_simd_sse_gcc.h =================================================================== RCS file: /cvsroot/pure-data/pd/src/Attic/m_simd_sse_gcc.h,v retrieving revision 1.1.2.3 retrieving revision 1.1.2.4 diff -C2 -d -r1.1.2.3 -r1.1.2.4 *** m_simd_sse_gcc.h 29 Dec 2003 02:01:57 -0000 1.1.2.3 --- m_simd_sse_gcc.h 10 Jul 2004 19:57:21 -0000 1.1.2.4 *************** *** 33,58 **** /* functions in d_ugen.c */ ! #define zero_perf_simd zero_perf8 /* SIMD not implemented */ /* functions in d_dac.c */ ! #define copy_perf_simd copy_perf8 /* SIMD not implemented */ /* functions in d_ctl.c */ ! #define sig_tilde_perf_simd sig_tilde_perf8 /* SIMD not implemented */ /* functions in d_arithmetic.c */ ! #define plus_perf_simd plus_perf8 /* SIMD not implemented */ ! #define scalarplus_perf_simd scalarplus_perf8 /* SIMD not implemented */ ! #define minus_perf_simd minus_perf8 /* SIMD not implemented */ ! #define scalarminus_perf_simd scalarminus_perf8 /* SIMD not implemented */ ! #define times_perf_simd times_perf8 /* SIMD not implemented */ ! #define scalartimes_perf_simd scalartimes_perf8 /* SIMD not implemented */ #define sqr_perf_simd sqr_perf8 /* SIMD not implemented */ ! #define over_perf_simd over_perf8 /* SIMD not implemented */ ! #define scalarover_perf_simd scalarover_perf8 /* SIMD not implemented */ ! #define min_perf_simd min_perf8 /* SIMD not implemented */ ! #define scalarmin_perf_simd scalarmin_perf8 /* SIMD not implemented */ ! #define max_perf_simd max_perf8 /* SIMD not implemented */ ! #define scalarmax_perf_simd scalarmax_perf8 /* SIMD not implemented */ /* functions in d_math.c */ --- 33,58 ---- /* functions in d_ugen.c */ ! #define zero_perf_simd zero_perf_sse_gcc /* functions in d_dac.c */ ! #define copy_perf_simd copy_perf_sse_gcc /* functions in d_ctl.c */ ! #define sig_tilde_perf_simd sig_tilde_perf_sse_gcc /* functions in d_arithmetic.c */ ! #define plus_perf_simd plus_perf_sse_gcc ! #define scalarplus_perf_simd scalarplus_perf_sse_gcc ! #define minus_perf_simd minus_perf_sse_gcc ! #define scalarminus_perf_simd scalarminus_perf_sse_gcc ! #define times_perf_simd times_perf_sse_gcc ! #define scalartimes_perf_simd scalartimes_perf_sse_gcc #define sqr_perf_simd sqr_perf8 /* SIMD not implemented */ ! #define over_perf_simd over_perf_sse_gcc ! #define scalarover_perf_simd scalarover_perf_sse_gcc ! #define min_perf_simd min_perf_sse_gcc ! #define scalarmin_perf_simd scalarmin_perf_sse_gcc ! #define max_perf_simd max_perf_sse_gcc ! #define scalarmax_perf_simd scalarmax_perf_sse_gcc /* functions in d_math.c */

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

Pd-cvs July 2004