You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

Makefile 5.9 kB

5 years ago
7 years ago
5 years ago
5 years ago
7 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204
  1. ifdef TARGET_CORE
  2. TARGET = $(TARGET_CORE)
  3. endif
  4. TOPDIR = ..
  5. include $(TOPDIR)/Makefile.system
  6. ifdef HAVE_SSE3
  7. CFLAGS += -msse3
  8. endif
  9. ifdef HAVE_SSSE3
  10. CFLAGS += -mssse3
  11. endif
  12. ifeq ($(C_COMPILER), GCC)
  13. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  14. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  15. endif
  16. ifeq ($(ARCH), power)
  17. ifeq ($(C_COMPILER), CLANG)
  18. override CFLAGS += -fno-integrated-as
  19. endif
  20. endif
  21. AVX2OPT =
  22. ifeq ($(C_COMPILER), GCC)
  23. # AVX2 support was added in 4.7.0
  24. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  25. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 7)
  26. ifeq ($(GCCVERSIONGTEQ4)$(GCCMINORVERSIONGTEQ7), 11)
  27. AVX2OPT = -mavx2
  28. endif
  29. endif
  30. ifeq ($(C_COMPILER), CLANG)
  31. # Any clang posing as gcc 4.2 should be new enough (3.4 or later)
  32. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  33. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 2)
  34. ifeq ($(GCCVERSIONGTEQ4)$(GCCMINORVERSIONGTEQ2), 11)
  35. AVX2OPT = -mavx2
  36. endif
  37. endif
  38. ifdef NO_AVX2
  39. AVX2OPT=
  40. endif
  41. ifdef TARGET_CORE
  42. ifeq ($(TARGET_CORE), $(filter $(TARGET_CORE),PRESCOTT CORE2 PENRYN DUNNINGTON ATOM NANO SANDYBRIDGE HASWELL NEHALEM ZEN BARCELONA BOBCAT BULLDOZER PILEDRIVER EXCAVATOR STEAMROLLER OPTERON_SSE3))
  43. override CFLAGS += -msse -msse2 -msse3 -mssse3 -msse4.1
  44. endif
  45. ifeq ($(TARGET_CORE), $(filter $(TARGET_CORE),KATMAI COPPERMINE BANIAS NORTHWOOD ATHLON OPTERON))
  46. override CFLAGS += -msse -msse2
  47. endif
  48. ifeq ($(TARGET_CORE), COOPERLAKE)
  49. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
  50. ifeq ($(GCCVERSIONGTEQ10), 1)
  51. override CFLAGS += -march=cooperlake
  52. else
  53. override CFLAGS += -march=skylake-avx512
  54. endif
  55. ifeq ($(OSNAME), CYGWIN_NT)
  56. override CFLAGS += -fno-asynchronous-unwind-tables
  57. endif
  58. ifeq ($(OSNAME), WINNT)
  59. ifeq ($(C_COMPILER), GCC)
  60. override CFLAGS += -fno-asynchronous-unwind-tables
  61. endif
  62. endif
  63. else ifeq ($(TARGET_CORE), SKYLAKEX)
  64. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) -march=skylake-avx512
  65. ifeq ($(OSNAME), CYGWIN_NT)
  66. override CFLAGS += -fno-asynchronous-unwind-tables
  67. endif
  68. ifeq ($(OSNAME), WINNT)
  69. ifeq ($(C_COMPILER), GCC)
  70. override CFLAGS += -fno-asynchronous-unwind-tables
  71. endif
  72. endif
  73. else ifeq ($(TARGET_CORE), HASWELL)
  74. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) $(AVX2OPT)
  75. else
  76. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
  77. endif
  78. BUILD_KERNEL = 1
  79. KDIR =
  80. TSUFFIX = _$(TARGET_CORE)
  81. else
  82. TARGET_CORE = $(CORE)
  83. KDIR =
  84. TSUFFIX =
  85. endif
  86. -include $(KERNELDIR)/KERNEL.$(TARGET_CORE)
  87. include $(KERNELDIR)/KERNEL
  88. include Makefile.L1
  89. include Makefile.L2
  90. include Makefile.L3
  91. include Makefile.LA
  92. HPLOBJS = \
  93. dgemm_kernel.$(SUFFIX) \
  94. $(DGEMMINCOPYOBJ) $(DGEMMITCOPYOBJ) \
  95. $(DGEMMONCOPYOBJ) $(DGEMMOTCOPYOBJ) \
  96. dtrsm_kernel_LN.$(SUFFIX) dtrsm_kernel_LT.$(SUFFIX) \
  97. dtrsm_kernel_RN.$(SUFFIX) dtrsm_kernel_RT.$(SUFFIX) \
  98. daxpy_k.$(SUFFIX) dcopy_k.$(SUFFIX) ddot_k.$(SUFFIX) \
  99. dger_k.$(SUFFIX) dscal_k.$(SUFFIX) idamax_k.$(SUFFIX) \
  100. dgemv_n.$(SUFFIX) dgemv_t.$(SUFFIX) dgemm_beta.$(SUFFIX) \
  101. dtrsm_iunucopy.$(SUFFIX) dtrsm_iunncopy.$(SUFFIX) \
  102. dtrsm_ilnucopy.$(SUFFIX) dtrsm_ilnncopy.$(SUFFIX) \
  103. dtrsm_iutucopy.$(SUFFIX) dtrsm_iutncopy.$(SUFFIX) \
  104. dtrsm_iltucopy.$(SUFFIX) dtrsm_iltncopy.$(SUFFIX) \
  105. dtrsm_ounucopy.$(SUFFIX) dtrsm_ounncopy.$(SUFFIX) \
  106. dtrsm_olnucopy.$(SUFFIX) dtrsm_olnncopy.$(SUFFIX) \
  107. dtrsm_outucopy.$(SUFFIX) dtrsm_outncopy.$(SUFFIX) \
  108. dtrsm_oltucopy.$(SUFFIX) dtrsm_oltncopy.$(SUFFIX)
  109. COMMONOBJS += lsame.$(SUFFIX) scabs1.$(SUFFIX) dcabs1.$(SUFFIX)
  110. ifeq ($(DYNAMIC_ARCH), 1)
  111. SBLASOBJS += setparam$(TSUFFIX).$(SUFFIX)
  112. CCOMMON_OPT += -DTS=$(TSUFFIX)
  113. endif
  114. KERNEL_INTERFACE = ../common_level1.h ../common_level2.h ../common_level3.h
  115. ifneq ($(NO_LAPACK), 1)
  116. KERNEL_INTERFACE += ../common_lapack.h
  117. endif
  118. ifeq ($(ARCH), x86)
  119. COMMONOBJS += cpuid.$(SUFFIX)
  120. endif
  121. ifdef EXPRECISION
  122. COMMONOBJS += qconjg.$(SUFFIX) qcabs1.$(SUFFIX)
  123. endif
  124. ifdef QUAD_PRECISION
  125. COMMONOBJS += qconjg.$(SUFFIX) qcabs1.$(SUFFIX)
  126. endif
  127. all : libs
  128. scabs1.$(SUFFIX): $(KERNELDIR)/$(SCABS_KERNEL)
  129. $(CC) -c $(CFLAGS) -DCOMPLEX -UDOUBLE -DF_INTERFACE $< -o $(@F)
  130. dcabs1.$(SUFFIX): $(KERNELDIR)/$(DCABS_KERNEL)
  131. $(CC) -c $(CFLAGS) -DCOMPLEX -DDOUBLE -DF_INTERFACE $< -o $(@F)
  132. qcabs1.$(SUFFIX): $(KERNELDIR)/$(QCABS_KERNEL)
  133. $(CC) -c $(CFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  134. qconjg.$(SUFFIX): $(KERNELDIR)/qconjg.S
  135. $(CC) -c $(CFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  136. lsame.$(SUFFIX): $(KERNELDIR)/$(LSAME_KERNEL)
  137. $(CC) -c $(CFLAGS) -DF_INTERFACE $< -o $(@F)
  138. setparam$(TSUFFIX).$(SUFFIX): setparam$(TSUFFIX).c kernel$(TSUFFIX).h
  139. ifeq ($(USE_GEMM3M), 1)
  140. $(CC) -c $(CFLAGS) -DUSE_GEMM3M $< -o $@
  141. else
  142. $(CC) -c $(CFLAGS) $< -o $@
  143. endif
  144. setparam$(TSUFFIX).c : setparam-ref.c
  145. sed 's/TS/$(TSUFFIX)/g' $< > $(@F)
  146. kernel$(TSUFFIX).h : $(KERNEL_INTERFACE)
  147. sed 's/\ *(/$(TSUFFIX)(/g' $^ > $(@F)
  148. cpuid.$(SUFFIX): $(KERNELDIR)/cpuid.S
  149. $(CC) -c $(CFLAGS) $< -o $(@F)
  150. scabs1.$(PSUFFIX): $(KERNELDIR)/$(SCABS_KERNEL)
  151. $(CC) -c $(PFLAGS) -DCOMPLEX -UDOUBLE -DF_INTERFACE $< -o $(@F)
  152. dcabs1.$(PSUFFIX): $(KERNELDIR)/$(DCABS_KERNEL)
  153. $(CC) -c $(PFLAGS) -DCOMPLEX -DDOUBLE -DF_INTERFACE $< -o $(@F)
  154. qcabs1.$(PSUFFIX): $(KERNELDIR)/$(QCABS_KERNEL)
  155. $(CC) -c $(PFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  156. qconjg.$(PSUFFIX): $(KERNELDIR)/qconjg.S
  157. $(CC) -c $(PFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  158. lsame.$(PSUFFIX): $(KERNELDIR)/$(LSAME_KERNEL)
  159. $(CC) -c $(PFLAGS) -DF_INTERFACE $< -o $(@F)
  160. cpuid.$(PSUFFIX): $(KERNELDIR)/cpuid.S
  161. $(CC) -c $(PFLAGS) $< -o $(@F)
  162. #ifdef DYNAMIC_ARCH
  163. clean ::
  164. @rm -f setparam_*.c kernel_*.h setparam.h kernel.h
  165. #endif
  166. include $(TOPDIR)/Makefile.tail