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 5.4 kB

7 years ago
5 years ago
7 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195
  1. ifdef TARGET_CORE
  2. TARGET = $(TARGET_CORE)
  3. endif
  4. TOPDIR = ..
  5. include $(TOPDIR)/Makefile.system
  6. ifdef HAVE_SSE3
  7. CFLAGS += -msse3
  8. endif
  9. ifeq ($(C_COMPILER), GCC)
  10. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  11. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  12. endif
  13. ifeq ($(ARCH), power)
  14. ifeq ($(C_COMPILER), CLANG)
  15. override CFLAGS += -fno-integrated-as
  16. endif
  17. endif
  18. AVX2OPT =
  19. ifeq ($(C_COMPILER), GCC)
  20. # AVX2 support was added in 4.7.0
  21. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  22. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 7)
  23. ifeq ($(GCCVERSIONGTEQ4)$(GCCMINORVERSIONGTEQ7), 11)
  24. AVX2OPT = -mavx2
  25. endif
  26. endif
  27. ifeq ($(C_COMPILER), CLANG)
  28. # Any clang posing as gcc 4.2 should be new enough (3.4 or later)
  29. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  30. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 2)
  31. ifeq ($(GCCVERSIONGTEQ4)$(GCCMINORVERSIONGTEQ2), 11)
  32. AVX2OPT = -mavx2
  33. endif
  34. endif
  35. ifdef NO_AVX2
  36. AVX2OPT=
  37. endif
  38. ifdef TARGET_CORE
  39. ifeq ($(TARGET_CORE), COOPERLAKE)
  40. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
  41. ifeq ($(GCCVERSIONGTEQ10), 1)
  42. override CFLAGS += -march=cooperlake
  43. else
  44. override CFLAGS += -march=skylake-avx512
  45. endif
  46. ifeq ($(OSNAME), CYGWIN_NT)
  47. override CFLAGS += -fno-asynchronous-unwind-tables
  48. endif
  49. ifeq ($(OSNAME), WINNT)
  50. ifeq ($(C_COMPILER), GCC)
  51. override CFLAGS += -fno-asynchronous-unwind-tables
  52. endif
  53. endif
  54. else ifeq ($(TARGET_CORE), SKYLAKEX)
  55. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) -march=skylake-avx512
  56. ifeq ($(OSNAME), CYGWIN_NT)
  57. override CFLAGS += -fno-asynchronous-unwind-tables
  58. endif
  59. ifeq ($(OSNAME), WINNT)
  60. ifeq ($(C_COMPILER), GCC)
  61. override CFLAGS += -fno-asynchronous-unwind-tables
  62. endif
  63. endif
  64. else ifeq ($(TARGET_CORE), HASWELL)
  65. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) $(AVX2OPT)
  66. else
  67. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
  68. endif
  69. BUILD_KERNEL = 1
  70. KDIR =
  71. TSUFFIX = _$(TARGET_CORE)
  72. else
  73. TARGET_CORE = $(CORE)
  74. KDIR =
  75. TSUFFIX =
  76. endif
  77. -include $(KERNELDIR)/KERNEL.$(TARGET_CORE)
  78. include $(KERNELDIR)/KERNEL
  79. include Makefile.L1
  80. include Makefile.L2
  81. include Makefile.L3
  82. include Makefile.LA
  83. HPLOBJS = \
  84. dgemm_kernel.$(SUFFIX) \
  85. $(DGEMMINCOPYOBJ) $(DGEMMITCOPYOBJ) \
  86. $(DGEMMONCOPYOBJ) $(DGEMMOTCOPYOBJ) \
  87. dtrsm_kernel_LN.$(SUFFIX) dtrsm_kernel_LT.$(SUFFIX) \
  88. dtrsm_kernel_RN.$(SUFFIX) dtrsm_kernel_RT.$(SUFFIX) \
  89. daxpy_k.$(SUFFIX) dcopy_k.$(SUFFIX) ddot_k.$(SUFFIX) \
  90. dger_k.$(SUFFIX) dscal_k.$(SUFFIX) idamax_k.$(SUFFIX) \
  91. dgemv_n.$(SUFFIX) dgemv_t.$(SUFFIX) dgemm_beta.$(SUFFIX) \
  92. dtrsm_iunucopy.$(SUFFIX) dtrsm_iunncopy.$(SUFFIX) \
  93. dtrsm_ilnucopy.$(SUFFIX) dtrsm_ilnncopy.$(SUFFIX) \
  94. dtrsm_iutucopy.$(SUFFIX) dtrsm_iutncopy.$(SUFFIX) \
  95. dtrsm_iltucopy.$(SUFFIX) dtrsm_iltncopy.$(SUFFIX) \
  96. dtrsm_ounucopy.$(SUFFIX) dtrsm_ounncopy.$(SUFFIX) \
  97. dtrsm_olnucopy.$(SUFFIX) dtrsm_olnncopy.$(SUFFIX) \
  98. dtrsm_outucopy.$(SUFFIX) dtrsm_outncopy.$(SUFFIX) \
  99. dtrsm_oltucopy.$(SUFFIX) dtrsm_oltncopy.$(SUFFIX)
  100. COMMONOBJS += lsame.$(SUFFIX) scabs1.$(SUFFIX) dcabs1.$(SUFFIX)
  101. ifeq ($(DYNAMIC_ARCH), 1)
  102. SBLASOBJS += setparam$(TSUFFIX).$(SUFFIX)
  103. CCOMMON_OPT += -DTS=$(TSUFFIX)
  104. endif
  105. KERNEL_INTERFACE = ../common_level1.h ../common_level2.h ../common_level3.h
  106. ifneq ($(NO_LAPACK), 1)
  107. KERNEL_INTERFACE += ../common_lapack.h
  108. endif
  109. ifeq ($(ARCH), x86)
  110. COMMONOBJS += cpuid.$(SUFFIX)
  111. endif
  112. ifdef EXPRECISION
  113. COMMONOBJS += qconjg.$(SUFFIX) qcabs1.$(SUFFIX)
  114. endif
  115. ifdef QUAD_PRECISION
  116. COMMONOBJS += qconjg.$(SUFFIX) qcabs1.$(SUFFIX)
  117. endif
  118. all : libs
  119. scabs1.$(SUFFIX): $(KERNELDIR)/$(SCABS_KERNEL)
  120. $(CC) -c $(CFLAGS) -DCOMPLEX -UDOUBLE -DF_INTERFACE $< -o $(@F)
  121. dcabs1.$(SUFFIX): $(KERNELDIR)/$(DCABS_KERNEL)
  122. $(CC) -c $(CFLAGS) -DCOMPLEX -DDOUBLE -DF_INTERFACE $< -o $(@F)
  123. qcabs1.$(SUFFIX): $(KERNELDIR)/$(QCABS_KERNEL)
  124. $(CC) -c $(CFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  125. qconjg.$(SUFFIX): $(KERNELDIR)/qconjg.S
  126. $(CC) -c $(CFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  127. lsame.$(SUFFIX): $(KERNELDIR)/$(LSAME_KERNEL)
  128. $(CC) -c $(CFLAGS) -DF_INTERFACE $< -o $(@F)
  129. setparam$(TSUFFIX).$(SUFFIX): setparam$(TSUFFIX).c kernel$(TSUFFIX).h
  130. ifeq ($(USE_GEMM3M), 1)
  131. $(CC) -c $(CFLAGS) -DUSE_GEMM3M $< -o $@
  132. else
  133. $(CC) -c $(CFLAGS) $< -o $@
  134. endif
  135. setparam$(TSUFFIX).c : setparam-ref.c
  136. sed 's/TS/$(TSUFFIX)/g' $< > $(@F)
  137. kernel$(TSUFFIX).h : $(KERNEL_INTERFACE)
  138. sed 's/\ *(/$(TSUFFIX)(/g' $^ > $(@F)
  139. cpuid.$(SUFFIX): $(KERNELDIR)/cpuid.S
  140. $(CC) -c $(CFLAGS) $< -o $(@F)
  141. scabs1.$(PSUFFIX): $(KERNELDIR)/$(SCABS_KERNEL)
  142. $(CC) -c $(PFLAGS) -DCOMPLEX -UDOUBLE -DF_INTERFACE $< -o $(@F)
  143. dcabs1.$(PSUFFIX): $(KERNELDIR)/$(DCABS_KERNEL)
  144. $(CC) -c $(PFLAGS) -DCOMPLEX -DDOUBLE -DF_INTERFACE $< -o $(@F)
  145. qcabs1.$(PSUFFIX): $(KERNELDIR)/$(QCABS_KERNEL)
  146. $(CC) -c $(PFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  147. qconjg.$(PSUFFIX): $(KERNELDIR)/qconjg.S
  148. $(CC) -c $(PFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  149. lsame.$(PSUFFIX): $(KERNELDIR)/$(LSAME_KERNEL)
  150. $(CC) -c $(PFLAGS) -DF_INTERFACE $< -o $(@F)
  151. cpuid.$(PSUFFIX): $(KERNELDIR)/cpuid.S
  152. $(CC) -c $(PFLAGS) $< -o $(@F)
  153. #ifdef DYNAMIC_ARCH
  154. clean ::
  155. @rm -f setparam_*.c kernel_*.h setparam.h kernel.h
  156. #endif
  157. include $(TOPDIR)/Makefile.tail