Resurrect and complete cblas_?gemm_batch

[martin-frbg]

fixes #3039 , xianyi's code has been languishing on the now heavily outdated small_matrix branch - the small matrix gemm support was already implemented in the meantime

[RajalakshmiSR]

Fails on Power Linux with gcc.

gemm_batch.c:117:6: note: in expansion of macro ‘CNAME’
  117 | void CNAME(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE *  transa_array, enum CBLAS_TRANSPOSE * transb_array,
      |      ^~~~~
In file included from ../common.h:835,
                 from gemm_batch.c:36:
../cblas.h:446:6: note: previous declaration of ‘cblas_sbgemm_batch’ with type ‘void(enum CBLAS_ORDER,  enum CBLAS_TRANSPOSE *, enum CBLAS_TRANSPOSE *, blasint *, blasint *, blasint *, float *, bfloat16 **, blasint *, bfloat16 **, blasint *, float *, float **, blasint *, blasint,  blasint *)’ {aka ‘void(enum CBLAS_ORDER,  enum CBLAS_TRANSPOSE *, enum CBLAS_TRANSPOSE *, int *, int *, int *, float *, short unsigned int **, int *, short unsigned int **, int *, float *, float **, int *, int,  int *)’}
  446 | void cblas_sbgemm_batch(OPENBLAS_CONST enum CBLAS_ORDER Order, OPENBLAS_CONST enum CBLAS_TRANSPOSE * TransA_array, OPENBLAS_CONST enum CBLAS_TRANSPOSE * TransB_array, OPENBLAS_CONST blasint * M_array, OPENBLAS_CONST blasint * N_array, OPENBLAS_CONST blasint * K_array,
      |      ^~~~~~~~~~~~~~~~~~
cc -m64 -O3 -O2 -Wno-error=incompatible-pointer-types -DMAX_STACK_ALLOC=2048 -DHAVE_P10_SUPPORT -Wall -m64 -DF_INTERFACE_GFORT -fPIC -DDYNAMIC_ARCH -DC_LAPACK -DSMP_SERVER -DNO_WARMUP -DMAX_CPU_NUMBER=32 -DMAX_PARALLEL_NUMBER=1 -DBUILD_SINGLE=1 -DBUILD_DOUBLE=1 -DBUILD_COMPLEX=1 -DBUILD_COMPLEX16=1 -DVERSION=\"0.3.27.dev\" -Ofast -mcpu=power10 -mtune=power10 -mvsx -fno-fast-math -DUSE_OPENMP -fopenmp -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME -DASMNAME= -DASMFNAME=_ -DNAME=_ -DCNAME= -DCHAR_NAME=\"_\" -DCHAR_CNAME=\"\" -DNO_AFFINITY -I. -O2 -DSMALL_MATRIX_OPT -Wno-error=incompatible-pointer-types -DMAX_STACK_ALLOC=2048 -fopenmp -DHAVE_P10_SUPPORT -Wall -m64 -DF_INTERFACE_GFORT -fPIC -DDYNAMIC_ARCH -DC_LAPACK -DSMP_SERVER -DUSE_OPENMP -DNO_WARMUP -DMAX_CPU_NUMBER=32 -DMAX_PARALLEL_NUMBER=1 -DBUILD_BFLOAT16 -DBUILD_SINGLE=1 -DBUILD_DOUBLE=1 -DBUILD_COMPLEX=1 -DBUILD_COMPLEX16=1 -DVERSION=\"0.3.27.dev\" -Ofast -mcpu=power10 -mtune=power10 -mvsx -fno-fast-math -DUSE_OPENMP -fopenmp -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME -DASMNAME=isamin -DASMFNAME=isamin_ -DNAME=isamin_ -DCNAME=isamin -DCHAR_NAME=\"isamin_\" -DCHAR_CNAME=\"isamin\" -DNO_AFFINITY -I.. -I. -UDOUBLE  -UCOMPLEX -c -DUSE_ABS -DUSE_MIN imax.c -o isamin.o
make[1]: *** [Makefile:2423: cblas_sbgemm_batch.o] Error 1
make[1]: *** Waiting for unfinished jobs..```

[RajalakshmiSR]

index 846d8e0f4..56ccc12ce 100644
--- a/interface/gemm_batch.c
+++ b/interface/gemm_batch.c
@@ -118,8 +118,8 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE *  transa_array, enum CB
           blasint * m_array, blasint * n_array, blasint * k_array,
 #ifndef COMPLEX
           FLOAT * alpha_array,
-          FLOAT ** a_array, blasint * lda_array,
-          FLOAT ** b_array, blasint * ldb_array,
+          IFLOAT ** a_array, blasint * lda_array,
+          IFLOAT ** b_array, blasint * ldb_array,
           FLOAT * beta_array,
           FLOAT ** c_array, blasint * ldc_array, blasint group_count, blasint * group_size) {
 #else

This should help to fix this issue.

[martin-frbg]

yes, that is what is in #4732