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 6.2 kB

6 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213
  1. ifdef TARGET_CORE
  2. TARGET = $(TARGET_CORE)
  3. endif
  4. TOPDIR = ..
  5. include $(TOPDIR)/Makefile.system
  6. ifeq ($(ARCH), power)
  7. ifeq ($(C_COMPILER), CLANG)
  8. override CFLAGS += -fno-integrated-as
  9. endif
  10. endif
  11. AVX2OPT =
  12. ifeq ($(C_COMPILER), GCC)
  13. # AVX2 support was added in 4.7.0
  14. GCCVERSIONCHECK := $(GCCVERSIONGT4)$(GCCVERSIONGTEQ4)$(GCCMINORVERSIONGTEQ7)
  15. ifeq ($(GCCVERSIONCHECK), $(filter $(GCCVERSIONCHECK), 011 110 111))
  16. AVX2OPT = -mavx2
  17. endif
  18. endif
  19. ifeq ($(C_COMPILER), CLANG)
  20. # Any clang posing as gcc 4.2 should be new enough (3.4 or later)
  21. GCCVERSIONCHECK := $(GCCVERSIONGT4)$(GCCVERSIONGTEQ4)$(GCCMINORVERSIONGTEQ2)
  22. ifeq ($(GCCVERSIONCHECK), $(filter $(GCCVERSIONCHECK), 011 110 111))
  23. AVX2OPT = -mavx2 -mfma
  24. endif
  25. endif
  26. ifdef NO_AVX2
  27. AVX2OPT=
  28. endif
  29. ifdef TARGET_CORE
  30. ifeq ($(TARGET_CORE), SAPPHIRERAPIDS)
  31. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
  32. ifeq (1, $(filter 1,$(GCCVERSIONGTEQ11) $(CLANGVERSIONGTEQ12)))
  33. override CFLAGS += -march=sapphirerapids
  34. else
  35. override CFLAGS += -march=skylake-avx512 -mavx512f
  36. endif
  37. ifeq ($(OSNAME), CYGWIN_NT)
  38. override CFLAGS += -fno-asynchronous-unwind-tables
  39. endif
  40. ifeq ($(OSNAME), WINNT)
  41. ifeq ($(C_COMPILER), GCC)
  42. override CFLAGS += -fno-asynchronous-unwind-tables
  43. endif
  44. endif
  45. else ifeq ($(TARGET_CORE), COOPERLAKE)
  46. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
  47. ifeq (1, $(filter 1,$(GCCVERSIONGTEQ10) $(CLANGVERSIONGTEQ9)))
  48. override CFLAGS += -march=cooperlake
  49. else
  50. override CFLAGS += -march=skylake-avx512 -mavx512f
  51. endif
  52. ifeq ($(OSNAME), CYGWIN_NT)
  53. override CFLAGS += -fno-asynchronous-unwind-tables
  54. endif
  55. ifeq ($(OSNAME), WINNT)
  56. ifeq ($(C_COMPILER), GCC)
  57. override CFLAGS += -fno-asynchronous-unwind-tables
  58. endif
  59. endif
  60. else ifeq ($(TARGET_CORE), SKYLAKEX)
  61. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) -march=skylake-avx512 -mavx512f
  62. ifeq ($(OSNAME), CYGWIN_NT)
  63. override CFLAGS += -fno-asynchronous-unwind-tables
  64. endif
  65. ifeq ($(OSNAME), WINNT)
  66. ifeq ($(C_COMPILER), GCC)
  67. override CFLAGS += -fno-asynchronous-unwind-tables
  68. endif
  69. endif
  70. else ifeq ($(TARGET_CORE), HASWELL)
  71. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) $(AVX2OPT)
  72. else ifeq ($(TARGET_CORE), ZEN)
  73. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) $(AVX2OPT)
  74. else ifeq ($(TARGET_CORE), LOONGSON3R4)
  75. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) $(MSA_FLAGS)
  76. else ifneq ($(filter NEOVERSEN2 NEOVERSEV1, $(TARGET_CORE)),)
  77. ifeq ($(C_COMPILER), PGI)
  78. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE) -Msve_intrinsics
  79. else
  80. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
  81. endif
  82. else
  83. override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
  84. endif
  85. BUILD_KERNEL = 1
  86. KDIR =
  87. TSUFFIX = _$(TARGET_CORE)
  88. else
  89. TARGET_CORE = $(CORE)
  90. KDIR =
  91. TSUFFIX =
  92. ifeq ($(TARGET_CORE), LOONGSON3R4)
  93. override CFLAGS += $(MSA_FLAGS)
  94. endif
  95. endif
  96. -include $(KERNELDIR)/KERNEL.$(TARGET_CORE)
  97. include $(KERNELDIR)/KERNEL
  98. include Makefile.L1
  99. include Makefile.L2
  100. include Makefile.L3
  101. include Makefile.LA
  102. HPLOBJS = \
  103. dgemm_kernel.$(SUFFIX) \
  104. $(DGEMMINCOPYOBJ) $(DGEMMITCOPYOBJ) \
  105. $(DGEMMONCOPYOBJ) $(DGEMMOTCOPYOBJ) \
  106. dtrsm_kernel_LN.$(SUFFIX) dtrsm_kernel_LT.$(SUFFIX) \
  107. dtrsm_kernel_RN.$(SUFFIX) dtrsm_kernel_RT.$(SUFFIX) \
  108. daxpy_k.$(SUFFIX) dcopy_k.$(SUFFIX) ddot_k.$(SUFFIX) \
  109. dger_k.$(SUFFIX) dscal_k.$(SUFFIX) idamax_k.$(SUFFIX) \
  110. dgemv_n.$(SUFFIX) dgemv_t.$(SUFFIX) dgemm_beta.$(SUFFIX) \
  111. dtrsm_iunucopy.$(SUFFIX) dtrsm_iunncopy.$(SUFFIX) \
  112. dtrsm_ilnucopy.$(SUFFIX) dtrsm_ilnncopy.$(SUFFIX) \
  113. dtrsm_iutucopy.$(SUFFIX) dtrsm_iutncopy.$(SUFFIX) \
  114. dtrsm_iltucopy.$(SUFFIX) dtrsm_iltncopy.$(SUFFIX) \
  115. dtrsm_ounucopy.$(SUFFIX) dtrsm_ounncopy.$(SUFFIX) \
  116. dtrsm_olnucopy.$(SUFFIX) dtrsm_olnncopy.$(SUFFIX) \
  117. dtrsm_outucopy.$(SUFFIX) dtrsm_outncopy.$(SUFFIX) \
  118. dtrsm_oltucopy.$(SUFFIX) dtrsm_oltncopy.$(SUFFIX)
  119. COMMONOBJS += lsame.$(SUFFIX) scabs1.$(SUFFIX) dcabs1.$(SUFFIX)
  120. ifeq ($(DYNAMIC_ARCH), 1)
  121. SBLASOBJS += setparam$(TSUFFIX).$(SUFFIX)
  122. CCOMMON_OPT += -DTS=$(TSUFFIX)
  123. endif
  124. KERNEL_INTERFACE = ../common_level1.h ../common_level2.h ../common_level3.h
  125. ifneq ($(NO_LAPACK), 1)
  126. KERNEL_INTERFACE += ../common_lapack.h
  127. endif
  128. ifeq ($(ARCH), x86)
  129. COMMONOBJS += cpuid.$(SUFFIX)
  130. endif
  131. ifdef EXPRECISION
  132. COMMONOBJS += qconjg.$(SUFFIX) qcabs1.$(SUFFIX)
  133. endif
  134. ifdef QUAD_PRECISION
  135. COMMONOBJS += qconjg.$(SUFFIX) qcabs1.$(SUFFIX)
  136. endif
  137. all : libs
  138. scabs1.$(SUFFIX): $(KERNELDIR)/$(SCABS_KERNEL)
  139. $(CC) -c $(CFLAGS) -DCOMPLEX -UDOUBLE -DF_INTERFACE $< -o $(@F)
  140. dcabs1.$(SUFFIX): $(KERNELDIR)/$(DCABS_KERNEL)
  141. $(CC) -c $(CFLAGS) -DCOMPLEX -DDOUBLE -DF_INTERFACE $< -o $(@F)
  142. qcabs1.$(SUFFIX): $(KERNELDIR)/$(QCABS_KERNEL)
  143. $(CC) -c $(CFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  144. qconjg.$(SUFFIX): $(KERNELDIR)/qconjg.S
  145. $(CC) -c $(CFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  146. lsame.$(SUFFIX): $(KERNELDIR)/$(LSAME_KERNEL)
  147. $(CC) -c $(CFLAGS) -DF_INTERFACE $< -o $(@F)
  148. setparam$(TSUFFIX).$(SUFFIX): setparam$(TSUFFIX).c kernel$(TSUFFIX).h
  149. ifeq ($(USE_GEMM3M), 1)
  150. $(CC) -c $(CFLAGS) -DUSE_GEMM3M $< -o $@
  151. else
  152. $(CC) -c $(CFLAGS) $< -o $@
  153. endif
  154. setparam$(TSUFFIX).c : setparam-ref.c
  155. sed 's/TS/$(TSUFFIX)/g' $< > $(@F)
  156. kernel$(TSUFFIX).h : $(KERNEL_INTERFACE)
  157. sed 's/\ *(/$(TSUFFIX)(/g' $^ > $(@F)
  158. cpuid.$(SUFFIX): $(KERNELDIR)/cpuid.S
  159. $(CC) -c $(CFLAGS) $< -o $(@F)
  160. scabs1.$(PSUFFIX): $(KERNELDIR)/$(SCABS_KERNEL)
  161. $(CC) -c $(PFLAGS) -DCOMPLEX -UDOUBLE -DF_INTERFACE $< -o $(@F)
  162. dcabs1.$(PSUFFIX): $(KERNELDIR)/$(DCABS_KERNEL)
  163. $(CC) -c $(PFLAGS) -DCOMPLEX -DDOUBLE -DF_INTERFACE $< -o $(@F)
  164. qcabs1.$(PSUFFIX): $(KERNELDIR)/$(QCABS_KERNEL)
  165. $(CC) -c $(PFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  166. qconjg.$(PSUFFIX): $(KERNELDIR)/qconjg.S
  167. $(CC) -c $(PFLAGS) -DCOMPLEX -DXDOUBLE -DF_INTERFACE $< -o $(@F)
  168. lsame.$(PSUFFIX): $(KERNELDIR)/$(LSAME_KERNEL)
  169. $(CC) -c $(PFLAGS) -DF_INTERFACE $< -o $(@F)
  170. cpuid.$(PSUFFIX): $(KERNELDIR)/cpuid.S
  171. $(CC) -c $(PFLAGS) $< -o $(@F)
  172. #ifdef DYNAMIC_ARCH
  173. clean ::
  174. @rm -f setparam_*.c kernel_*.h setparam.h kernel.h
  175. #endif
  176. include $(TOPDIR)/Makefile.tail