From 61bbae3ac1a9e8c4399c9889af6b5533636c03c6 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sat, 18 Apr 2020 21:09:32 +0200
Subject: [PATCH 01/14] Handle  MIPS24K like P5600

and allow enforcing TARGET=1004K as well (omission from earlier 1004K merge and later introduction of TARGET check)
---
 cpuid_mips.c | 20 +++++++++++++++++---
 getarch.c    | 28 ++++++++++++++++++++++++++++
 param.h      |  8 +++++++-
 3 files changed, 52 insertions(+), 4 deletions(-)

diff --git a/cpuid_mips.c b/cpuid_mips.c
index 6f2932c94..df3541536 100644
--- a/cpuid_mips.c
+++ b/cpuid_mips.c
@@ -73,11 +73,13 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define CPU_UNKNOWN     0
 #define CPU_P5600       1
 #define CPU_1004K	2
+#define CPU_24K		3
 
 static char *cpuname[] = {
   "UNKNOWN",
   "P5600",
-  "1004K"
+  "1004K",
+  "24K"
 };
 
 int detect(void){
@@ -105,6 +107,8 @@ int detect(void){
     return CPU_P5600;
   } else if (strstr(p, "1004K")) {
     return CPU_1004K;
+  } else if (strstr(p, " 24K")) {
+    return CPU_24K;
   } else  
     return CPU_UNKNOWN;
   }
@@ -121,7 +125,7 @@ void get_architecture(void){
 }
 
 void get_subarchitecture(void){
-  if(detect()==CPU_P5600|| detect()==CPU_1004K){
+  if(detect()==CPU_P5600|| detect()==CPU_1004K|| detect()==CPU_24K){
     printf("P5600");
   }else{
     printf("UNKNOWN");
@@ -146,7 +150,15 @@ void get_cpuconfig(void){
     printf("#define MIPS1004K\n");
     printf("#define L1_DATA_SIZE 32768\n");
     printf("#define L1_DATA_LINESIZE 32\n");
-    printf("#define L2_SIZE 26144\n");
+    printf("#define L2_SIZE 262144\n");
+    printf("#define DTB_DEFAULT_ENTRIES 8\n");
+    printf("#define DTB_SIZE 4096\n");
+    printf("#define L2_ASSOCIATIVE 4\n");
+  } else if (detect()==CPU_24K) {
+    printf("#define MIPS24K\n");
+    printf("#define L1_DATA_SIZE 32768\n");
+    printf("#define L1_DATA_LINESIZE 32\n");
+    printf("#define L2_SIZE 32768\n");
     printf("#define DTB_DEFAULT_ENTRIES 8\n");
     printf("#define DTB_SIZE 4096\n");
     printf("#define L2_ASSOCIATIVE 4\n");
@@ -160,6 +172,8 @@ void get_libname(void){
     printf("p5600\n");
   } else if (detect()==CPU_1004K) {
     printf("1004K\n");
+  } else if (detect()==CPU_24K) {
+    printf("24K\n");
   }else{
     printf("mips\n");
   }
diff --git a/getarch.c b/getarch.c
index 145753bcc..8a6684975 100644
--- a/getarch.c
+++ b/getarch.c
@@ -812,6 +812,34 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #else
 #endif
 
+#ifdef FORCE_1004K
+#define FORCE
+#define ARCHITECTURE    "MIPS"
+#define SUBARCHITECTURE "1004K"
+#define SUBDIRNAME      "mips"
+#define ARCHCONFIG   "-D1004K " \
+       "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
+       "-DL2_SIZE=262144 -DL2_LINESIZE=32 " \
+       "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
+#define LIBNAME   "1004K"
+#define CORENAME  "1004K"
+#else
+#endif
+
+#ifdef FORCE_24K
+#define FORCE
+#define ARCHITECTURE    "MIPS"
+#define SUBARCHITECTURE "24K"
+#define SUBDIRNAME      "mips"
+#define ARCHCONFIG   "-D24K " \
+       "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
+       "-DL2_SIZE=32768 -DL2_LINESIZE=32 " \
+       "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
+#define LIBNAME   "24K"
+#define CORENAME  "24K"
+#else
+#endif
+
 #ifdef FORCE_I6500
 #define FORCE
 #define ARCHITECTURE    "MIPS"
diff --git a/param.h b/param.h
index d6cbe544a..2795947c5 100644
--- a/param.h
+++ b/param.h
@@ -72,6 +72,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifndef PARAM_H
 #define PARAM_H
 
+#define SHGEMM_DEFAULT_UNROLL_N 4
+#define SHGEMM_DEFAULT_UNROLL_M 8
+#define SHGEMM_DEFAULT_UNROLL_MN 32
+#define SHGEMM_DEFAULT_P 256
+#define SHGEMM_DEFAULT_R 256
+#define SHGEMM_DEFAULT_Q 256
 #ifdef OPTERON
 
 #define SNUMOPT		4
@@ -2468,7 +2474,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define SYMV_P	16
 #endif
 
-#if defined(P5600) || defined(MIPS1004K) || defined(I6400) || defined(P6600) || defined(I6500)
+#if defined(P5600) || defined(MIPS1004K) defined(MIPS24K) || defined(I6400) || defined(P6600) || defined(I6500)
 #define SNUMOPT  2
 #define DNUMOPT  2
 

From d712ea724cbf517fb8a40607cf5381dd453dbf92 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sat, 18 Apr 2020 21:10:18 +0200
Subject: [PATCH 02/14] Add MIPS24K support

---
 kernel/mips/KERNEL.24K | 1 +
 1 file changed, 1 insertion(+)
 create mode 100644 kernel/mips/KERNEL.24K

diff --git a/kernel/mips/KERNEL.24K b/kernel/mips/KERNEL.24K
new file mode 100644
index 000000000..67135356e
--- /dev/null
+++ b/kernel/mips/KERNEL.24K
@@ -0,0 +1 @@
+include $(KERNELDIR)/KERNEL.P5600

From 00172d440bfc7dedc8523a4cdad58b685801bb76 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sat, 18 Apr 2020 21:16:49 +0200
Subject: [PATCH 03/14] Typo fix in MIPS24K addition

---
 param.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/param.h b/param.h
index 2795947c5..9fdf40fe2 100644
--- a/param.h
+++ b/param.h
@@ -2474,7 +2474,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define SYMV_P	16
 #endif
 
-#if defined(P5600) || defined(MIPS1004K) defined(MIPS24K) || defined(I6400) || defined(P6600) || defined(I6500)
+#if defined(P5600) || defined(MIPS1004K) || defined(MIPS24K) || defined(I6400) || defined(P6600) || defined(I6500)
 #define SNUMOPT  2
 #define DNUMOPT  2
 

From a1fc98dc57f896450c3a807814ad36f541eb112f Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sat, 18 Apr 2020 23:50:23 +0200
Subject: [PATCH 04/14] rename 1004K, 24K to MIPS1004K, MIPS24K to avoid
 identifier naming problem

---
 cpuid_mips.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/cpuid_mips.c b/cpuid_mips.c
index df3541536..3a2e12393 100644
--- a/cpuid_mips.c
+++ b/cpuid_mips.c
@@ -78,8 +78,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 static char *cpuname[] = {
   "UNKNOWN",
   "P5600",
-  "1004K",
-  "24K"
+  "MIPS1004K",
+  "MIPS24K"
 };
 
 int detect(void){
@@ -171,9 +171,9 @@ void get_libname(void){
   if(detect()==CPU_P5600) {
     printf("p5600\n");
   } else if (detect()==CPU_1004K) {
-    printf("1004K\n");
+    printf("mips1004K\n");
   } else if (detect()==CPU_24K) {
-    printf("24K\n");
+    printf("mips24K\n");
   }else{
     printf("mips\n");
   }

From b0b02a080d06f41d4132c75876c073fade5feb8b Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sun, 19 Apr 2020 06:50:51 +0200
Subject: [PATCH 05/14] Add compiler options for MIPS32 24K/1004K

---
 Makefile.prebuild | 6 +++++-
 Makefile.system   | 9 ++++++++-
 2 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/Makefile.prebuild b/Makefile.prebuild
index b00f13368..48fb5e991 100644
--- a/Makefile.prebuild
+++ b/Makefile.prebuild
@@ -17,7 +17,11 @@ ifdef CPUIDEMU
 EXFLAGS = -DCPUIDEMU -DVENDOR=99
 endif
 
-ifeq ($(TARGET), 1004K)
+ifeq ($(TARGET), MIPS24K)
+TARGET_FLAGS = -mips32r2
+endif
+
+ifeq ($(TARGET), MIPS1004K)
 TARGET_FLAGS = -mips32r2
 endif
 
diff --git a/Makefile.system b/Makefile.system
index 2998c0e6a..51bd1c4bd 100644
--- a/Makefile.system
+++ b/Makefile.system
@@ -690,7 +690,12 @@ CCOMMON_OPT += -march=mips64
 FCOMMON_OPT += -march=mips64
 endif
 
-ifeq ($(CORE), 1004K)
+ifeq ($(CORE), MIPS24K)
+CCOMMON_OPT += -mips32r2 -mtune=24kc  $(MSA_FLAGS)
+FCOMMON_OPT += -mips32r2 -mtune=24kc  $(MSA_FLAGS)
+endif
+
+ifeq ($(CORE), MIPS1004K)
 CCOMMON_OPT += -mips32r2   $(MSA_FLAGS)
 FCOMMON_OPT += -mips32r2   $(MSA_FLAGS)
 endif
@@ -1390,6 +1395,8 @@ export FUNCTION_PROFILE
 export TARGET_CORE
 export NO_AVX512
 
+export SHGEMM_UNROLL_M
+export SHGEMM_UNROLL_N
 export SGEMM_UNROLL_M
 export SGEMM_UNROLL_N
 export DGEMM_UNROLL_M

From 6721f2750eabc6b9839e1b4d1aef39fa46810bc7 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sun, 19 Apr 2020 06:51:57 +0200
Subject: [PATCH 06/14] Update TargetList.txt

---
 TargetList.txt | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/TargetList.txt b/TargetList.txt
index f4a40ed02..e2d2f4026 100644
--- a/TargetList.txt
+++ b/TargetList.txt
@@ -58,7 +58,8 @@ CELL
 
 3.MIPS CPU:
 P5600
-1004K
+MIPS1004K
+MIPS24K
 
 4.MIPS64 CPU:
 SICORTEX

From 577c5d9f8fc3b1776bbaf5d6f15eabc3c0b8d170 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sun, 19 Apr 2020 06:54:52 +0200
Subject: [PATCH 07/14] Update README.md

---
 README.md | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/README.md b/README.md
index 61393bd8f..6dc3c7b42 100644
--- a/README.md
+++ b/README.md
@@ -122,6 +122,11 @@ Please read `GotoBLAS_01Readme.txt` for older CPU models already supported by th
 - **AMD STEAMROLLER**: Uses Bulldozer codes with some optimizations.
 - **AMD ZEN**: Uses Haswell codes with some optimizations.
 
+#### MIPS32
+
+- **MIPS 1004K**: uses P5600 codes
+- **MIPS 24K**: uses P5600 codes
+
 #### MIPS64
 
 - **ICT Loongson 3A**: Optimized Level-3 BLAS and the part of Level-1,2.

From 8792fc4d5f0dd69de1024963611304dd291e9792 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sun, 19 Apr 2020 07:21:48 +0200
Subject: [PATCH 08/14] Disable RPCC macro on MIPS24K

---
 common_mips.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/common_mips.h b/common_mips.h
index 2cc923043..dd2f8d558 100644
--- a/common_mips.h
+++ b/common_mips.h
@@ -43,6 +43,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #ifndef ASSEMBLER
 
+#if !defined(MIPS24K)
 static inline unsigned int rpcc(void){
   unsigned long ret;
 
@@ -53,6 +54,7 @@ static inline unsigned int rpcc(void){
   return ret;
 }
 #define RPCC_DEFINED
+#endif
 
 static inline int blas_quickdivide(blasint x, blasint y){
   return x / y;

From 2f4a8e5bc4504c0ba9faca82e0ebfb4d50120a48 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sun, 19 Apr 2020 13:22:19 +0200
Subject: [PATCH 09/14] Rename the FORCE entries for 24K and 1004K to include
 the MIPS prefix

---
 getarch.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/getarch.c b/getarch.c
index 8a6684975..6be006aee 100644
--- a/getarch.c
+++ b/getarch.c
@@ -812,7 +812,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #else
 #endif
 
-#ifdef FORCE_1004K
+#ifdef FORCE_MIPS1004K
 #define FORCE
 #define ARCHITECTURE    "MIPS"
 #define SUBARCHITECTURE "1004K"
@@ -826,7 +826,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #else
 #endif
 
-#ifdef FORCE_24K
+#ifdef FORCE_MIPS24K
 #define FORCE
 #define ARCHITECTURE    "MIPS"
 #define SUBARCHITECTURE "24K"

From 0d18f231fc4d5591a30e995c695c3b729b4178a5 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sun, 19 Apr 2020 13:52:58 +0200
Subject: [PATCH 10/14] Update getarch.c

---
 getarch.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/getarch.c b/getarch.c
index 6be006aee..0bd5bc601 100644
--- a/getarch.c
+++ b/getarch.c
@@ -815,28 +815,28 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifdef FORCE_MIPS1004K
 #define FORCE
 #define ARCHITECTURE    "MIPS"
-#define SUBARCHITECTURE "1004K"
+#define SUBARCHITECTURE "MIPS1004K"
 #define SUBDIRNAME      "mips"
 #define ARCHCONFIG   "-D1004K " \
        "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
        "-DL2_SIZE=262144 -DL2_LINESIZE=32 " \
        "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
-#define LIBNAME   "1004K"
-#define CORENAME  "1004K"
+#define LIBNAME   "mips1004K"
+#define CORENAME  "MIPS1004K"
 #else
 #endif
 
 #ifdef FORCE_MIPS24K
 #define FORCE
 #define ARCHITECTURE    "MIPS"
-#define SUBARCHITECTURE "24K"
+#define SUBARCHITECTURE "MIPS24K"
 #define SUBDIRNAME      "mips"
 #define ARCHCONFIG   "-D24K " \
        "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
        "-DL2_SIZE=32768 -DL2_LINESIZE=32 " \
        "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
-#define LIBNAME   "24K"
-#define CORENAME  "24K"
+#define LIBNAME   "mips24K"
+#define CORENAME  "MIPS24K"
 #else
 #endif
 

From 5afb66812f21406b290878388bc3a74dde455910 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sun, 19 Apr 2020 14:55:31 +0200
Subject: [PATCH 11/14] Update getarch.c

---
 getarch.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/getarch.c b/getarch.c
index 0bd5bc601..e739d2de9 100644
--- a/getarch.c
+++ b/getarch.c
@@ -817,7 +817,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define ARCHITECTURE    "MIPS"
 #define SUBARCHITECTURE "MIPS1004K"
 #define SUBDIRNAME      "mips"
-#define ARCHCONFIG   "-D1004K " \
+#define ARCHCONFIG   "-DMIPS1004K " \
        "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
        "-DL2_SIZE=262144 -DL2_LINESIZE=32 " \
        "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
@@ -831,7 +831,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define ARCHITECTURE    "MIPS"
 #define SUBARCHITECTURE "MIPS24K"
 #define SUBDIRNAME      "mips"
-#define ARCHCONFIG   "-D24K " \
+#define ARCHCONFIG   "-DMIPS24K " \
        "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
        "-DL2_SIZE=32768 -DL2_LINESIZE=32 " \
        "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "

From 6a04efb1227fc6afd2ba72987b1780344da9eae6 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sun, 19 Apr 2020 15:43:54 +0200
Subject: [PATCH 12/14] Rename KERNEL files to include MIPS prefix

---
 kernel/mips/KERNEL.MIPS1004K | 1 +
 kernel/mips/KERNEL.MIPS24K   | 1 +
 2 files changed, 2 insertions(+)
 create mode 100644 kernel/mips/KERNEL.MIPS1004K
 create mode 100644 kernel/mips/KERNEL.MIPS24K

diff --git a/kernel/mips/KERNEL.MIPS1004K b/kernel/mips/KERNEL.MIPS1004K
new file mode 100644
index 000000000..67135356e
--- /dev/null
+++ b/kernel/mips/KERNEL.MIPS1004K
@@ -0,0 +1 @@
+include $(KERNELDIR)/KERNEL.P5600
diff --git a/kernel/mips/KERNEL.MIPS24K b/kernel/mips/KERNEL.MIPS24K
new file mode 100644
index 000000000..67135356e
--- /dev/null
+++ b/kernel/mips/KERNEL.MIPS24K
@@ -0,0 +1 @@
+include $(KERNELDIR)/KERNEL.P5600

From 7353ea5afc682d12944cfd97ca8a24daa83304a3 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sun, 19 Apr 2020 15:44:19 +0200
Subject: [PATCH 13/14] Delete KERNEL.24K

---
 kernel/mips/KERNEL.24K | 1 -
 1 file changed, 1 deletion(-)
 delete mode 100644 kernel/mips/KERNEL.24K

diff --git a/kernel/mips/KERNEL.24K b/kernel/mips/KERNEL.24K
deleted file mode 100644
index 67135356e..000000000
--- a/kernel/mips/KERNEL.24K
+++ /dev/null
@@ -1 +0,0 @@
-include $(KERNELDIR)/KERNEL.P5600

From e55ec82bb92338d09ecd77357da3fcdfac0a7902 Mon Sep 17 00:00:00 2001
From: Martin Kroeker <martin@ruby.chemie.uni-freiburg.de>
Date: Sun, 19 Apr 2020 15:44:30 +0200
Subject: [PATCH 14/14] Delete KERNEL.1004K

---
 kernel/mips/KERNEL.1004K | 1 -
 1 file changed, 1 deletion(-)
 delete mode 100644 kernel/mips/KERNEL.1004K

diff --git a/kernel/mips/KERNEL.1004K b/kernel/mips/KERNEL.1004K
deleted file mode 100644
index 67135356e..000000000
--- a/kernel/mips/KERNEL.1004K
+++ /dev/null
@@ -1 +0,0 @@
-include $(KERNELDIR)/KERNEL.P5600