Merge pull request #4597 from martin-frbg/issue4581

Add support for the ARM Cortex-A76 cpu

Author: martin-frbg

Makefile.arm64

@@ -58,6 +58,13 @@ FCOMMON_OPT += -march=armv8-a -mtune=cortex-a73
 endif
 endif
 
+ifeq ($(CORE), CORTEXA76)
+CCOMMON_OPT += -march=armv8.2-a -mtune=cortex-a76
+ifneq ($(F_COMPILER), NAG)
+FCOMMON_OPT += -march=armv8.2-a -mtune=cortex-a76
+endif
+endif
+
 ifeq ($(CORE), FT2000)
 CCOMMON_OPT += -march=armv8-a -mtune=cortex-a72
 ifneq ($(F_COMPILER), NAG)

README.md

@@ -167,6 +167,7 @@ Please read `GotoBLAS_01Readme.txt` for older CPU models already supported by th
 - **Cortex A57**: Optimized Level-3 and Level-2 functions
 - **Cortex A72**: same as A57 ( different cpu specifications)
 - **Cortex A73**: same as A57 (different cpu specifications)
+- **Cortex A76**: same as A57 (different cpu specifications)
 - **Falkor**: same as A57 (different cpu specifications)
 - **ThunderX**: Optimized some Level-1 functions
 - **ThunderX2T99**: Optimized Level-3 BLAS and parts of Levels 1 and 2

TargetList.txt

@@ -93,6 +93,7 @@ CORTEXA53
 CORTEXA57
 CORTEXA72
 CORTEXA73
+CORTEXA76
 CORTEXA510
 CORTEXA710
 CORTEXX1

cmake/prebuild.cmake

@@ -932,7 +932,7 @@ endif ()
     set(ZGEMM_UNROLL_M 4)
     set(ZGEMM_UNROLL_N 4)
     set(SYMV_P 16)
-  elseif ("${TCORE}" STREQUAL "CORTEXA72" OR "${TCORE}" STREQUAL "CORTEXA73")
+  elseif ("${TCORE}" STREQUAL "CORTEXA72" OR "${TCORE}" STREQUAL "CORTEXA73" OR "${TCORE}" STREQUAL "CORTEXA76")
     file(APPEND ${TARGET_CONF_TEMP}
       "#define L1_CODE_SIZE\t49152\n"
       "#define L1_CODE_LINESIZE\t64\n"

cpuid_arm64.c

@@ -42,6 +42,7 @@ size_t length64=sizeof(value64);
 #define CPU_CORTEXA57     3
 #define CPU_CORTEXA72     4
 #define CPU_CORTEXA73     5
+#define CPU_CORTEXA76	  23
 #define CPU_NEOVERSEN1    11
 #define CPU_NEOVERSEV1    16
 #define CPU_NEOVERSEN2    17
@@ -89,7 +90,8 @@ static char *cpuname[] = {
   "CORTEXX2",
   "CORTEXA510",
   "CORTEXA710",
-  "FT2000"
+  "FT2000",
+  "CORTEXA76"
 };
 
 static char *cpuname_lower[] = {
@@ -115,7 +117,8 @@ static char *cpuname_lower[] = {
   "cortexx2",
   "cortexa510",
   "cortexa710",
-  "ft2000"
+  "ft2000",
+  "cortexa76"
 };
 
 int get_feature(char *search)
@@ -210,6 +213,8 @@ int detect(void)
 	return CPU_CORTEXX2;
       else if (strstr(cpu_part, "0xd4e")) //X3
 	return CPU_CORTEXX2;
+      else if (strstr(cpu_part, "0xd0b")) 
+	return CPU_CORTEXA76;
     }
     // Qualcomm
     else if (strstr(cpu_implementer, "0x51") && strstr(cpu_part, "0xc00"))
@@ -391,6 +396,7 @@ void get_cpuconfig(void)
 		break;
 
 	    case CPU_NEOVERSEV1:
+	    case CPU_CORTEXA76:
                 printf("#define %s\n", cpuname[d]);
                 printf("#define L1_CODE_SIZE 65536\n");
                 printf("#define L1_CODE_LINESIZE 64\n");

getarch.c

@@ -1331,6 +1331,21 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define CORENAME  "CORTEXA73"
 #endif
 
+#ifdef FORCE_CORTEXA76
+#define FORCE
+#define ARCHITECTURE    "ARM64"
+#define SUBARCHITECTURE "CORTEXA76"
+#define SUBDIRNAME      "arm64"
+#define ARCHCONFIG   "-DCORTEXA76 " \
+       "-DL1_CODE_SIZE=49152 -DL1_CODE_LINESIZE=64 -DL1_CODE_ASSOCIATIVE=3 " \
+       "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 -DL1_DATA_ASSOCIATIVE=2 " \
+       "-DL2_SIZE=2097152 -DL2_LINESIZE=64 -DL2_ASSOCIATIVE=16 " \
+       "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
+       "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
+#define LIBNAME   "cortexa76"
+#define CORENAME  "CORTEXA76"
+#endif
+
 #ifdef FORCE_CORTEXX1
 #define FORCE
 #define ARCHITECTURE    "ARM64"

kernel/arm64/KERNEL.CORTEXA76

@@ -0,0 +1,3 @@
+include $(KERNELDIR)/KERNEL.CORTEXA57
+
+

param.h

@@ -3351,6 +3351,41 @@ is a big desktop or server with abundant cache rather than a phone or embedded d
 #define CGEMM_DEFAULT_R 4096
 #define ZGEMM_DEFAULT_R 2048
 
+#elif defined(CORTEXA76)
+
+#define SGEMM_DEFAULT_UNROLL_M  16
+#define SGEMM_DEFAULT_UNROLL_N  4
+
+#define DGEMM_DEFAULT_UNROLL_M  8
+#define DGEMM_DEFAULT_UNROLL_N  4
+
+#define CGEMM_DEFAULT_UNROLL_M  8
+#define CGEMM_DEFAULT_UNROLL_N  4
+
+#define ZGEMM_DEFAULT_UNROLL_M  4
+#define ZGEMM_DEFAULT_UNROLL_N  4
+
+#if defined(XDOUBLE) || defined(DOUBLE)
+#define SWITCH_RATIO            8
+#else
+#define SWITCH_RATIO            16
+#endif
+
+  #define SGEMM_DEFAULT_P 256
+  #define DGEMM_DEFAULT_P 128
+  #define CGEMM_DEFAULT_P 128
+  #define ZGEMM_DEFAULT_P 64
+
+  #define SGEMM_DEFAULT_Q 512
+  #define DGEMM_DEFAULT_Q 256
+  #define CGEMM_DEFAULT_Q 256
+  #define ZGEMM_DEFAULT_Q 256
+
+#define SGEMM_DEFAULT_R 4096
+#define DGEMM_DEFAULT_R 4096
+#define CGEMM_DEFAULT_R 4096
+#define ZGEMM_DEFAULT_R 4096
+
 #elif defined(CORTEXA53) || defined(CORTEXA55)
 
 #define SGEMM_DEFAULT_UNROLL_M  8