Hello, By default powerpc64 P8 system provides -maltivec gcc option preventing compilation - a patch is disabling that option with the -mno-altivec flag. In order to authorize that -maltivec flag to allow use of vector optimisation, two issues must be considered - vector name is a reserved keyword and can't be part of a struct. Commit a7e1d4c934e29d2861d95c2ee955707353b59db3 is renaming "vector" to "Vector". - in defbool.h, the bool definition must be replaced with a non vector one due to a a C++ issue. Commit ef1b194971d9e4ea1043e668940c83ee60eebc2d ensures the altivec definitions are correct. --
__________________________________________________________________________ thf - Thierry Fauck - tfa...@free.fr> /pubkey: 4096R/FCC181CE/ /fingerprint: 5CCF 6B82 DE4E E72A A40B B63E A153 BF4F FCC1 81CE/
From a7e1d4c934e29d2861d95c2ee955707353b59db3 Mon Sep 17 00:00:00 2001 From: Thierry Fauck <tfa...@free.fr> Date: Mon, 25 Apr 2016 09:04:37 -0400 Subject: [PATCH 1/2] vector is a gcc special word When using -maltivec option with gcc, vector is a special word and struct fields named vector are mis-understood. This patch just change vector name to Vector. modified: src/library/blas/functor/functor_xscal.cc modified: src/library/blas/generic/solution_seq_make.c modified: src/library/blas/gens/asum.cpp modified: src/library/blas/gens/axpy_reg.cpp modified: src/library/blas/gens/copy_reg.cpp modified: src/library/blas/gens/dot.cpp modified: src/library/blas/gens/gbmv.cpp modified: src/library/blas/gens/gemv.c modified: src/library/blas/gens/ger_lds.cpp modified: src/library/blas/gens/her2_lds.cpp modified: src/library/blas/gens/her_lds.cpp modified: src/library/blas/gens/iamax.cpp modified: src/library/blas/gens/nrm2.cpp modified: src/library/blas/gens/rotm_reg.cpp modified: src/library/blas/gens/scal_reg.cpp modified: src/library/blas/gens/swap_reg.cpp modified: src/library/blas/gens/symv.c modified: src/library/blas/gens/syr2_lds.cpp modified: src/library/blas/gens/syr_lds.cpp modified: src/library/blas/gens/trmv_reg.cpp modified: src/library/blas/gens/trsv_gemv.cpp modified: src/library/blas/gens/trsv_trtri.cpp modified: src/library/blas/include/clblas-internal.h modified: src/library/blas/ixamax.c modified: src/library/blas/xasum.c modified: src/library/blas/xaxpy.c modified: src/library/blas/xcopy.c modified: src/library/blas/xdot.c modified: src/library/blas/xgbmv.c modified: src/library/blas/xgemv.c modified: src/library/blas/xger.c modified: src/library/blas/xhemv.c modified: src/library/blas/xher.c modified: src/library/blas/xher2.c modified: src/library/blas/xhpmv.c modified: src/library/blas/xnrm2.c modified: src/library/blas/xrot.c modified: src/library/blas/xrotm.c modified: src/library/blas/xscal.c modified: src/library/blas/xshbmv.c modified: src/library/blas/xspmv.c modified: src/library/blas/xswap.c modified: src/library/blas/xsymv.c modified: src/library/blas/xsyr.c modified: src/library/blas/xsyr2.c modified: src/library/blas/xtbmv.c modified: src/library/blas/xtbsv.c modified: src/library/blas/xtrmv.c modified: src/library/blas/xtrsv.c --- src/library/blas/functor/functor_xscal.cc | 2 +- src/library/blas/generic/solution_seq_make.c | 4 +-- src/library/blas/gens/asum.cpp | 6 ++-- src/library/blas/gens/axpy_reg.cpp | 8 ++--- src/library/blas/gens/copy_reg.cpp | 8 ++--- src/library/blas/gens/dot.cpp | 8 ++--- src/library/blas/gens/gbmv.cpp | 4 +-- src/library/blas/gens/gemv.c | 10 +++---- src/library/blas/gens/ger_lds.cpp | 4 +-- src/library/blas/gens/her2_lds.cpp | 4 +-- src/library/blas/gens/her_lds.cpp | 2 +- src/library/blas/gens/iamax.cpp | 6 ++-- src/library/blas/gens/nrm2.cpp | 6 ++-- src/library/blas/gens/rotm_reg.cpp | 8 ++--- src/library/blas/gens/scal_reg.cpp | 4 +-- src/library/blas/gens/swap_reg.cpp | 8 ++--- src/library/blas/gens/symv.c | 10 +++---- src/library/blas/gens/syr2_lds.cpp | 4 +-- src/library/blas/gens/syr_lds.cpp | 2 +- src/library/blas/gens/trmv_reg.cpp | 4 +-- src/library/blas/gens/trsv_gemv.cpp | 2 +- src/library/blas/gens/trsv_trtri.cpp | 2 +- src/library/blas/include/clblas-internal.h | 2 +- src/library/blas/ixamax.c | 2 +- src/library/blas/xasum.c | 2 +- src/library/blas/xaxpy.c | 4 +-- src/library/blas/xcopy.c | 4 +-- src/library/blas/xdot.c | 4 +-- src/library/blas/xgbmv.c | 4 +-- src/library/blas/xgemv.c | 4 +-- src/library/blas/xger.c | 4 +-- src/library/blas/xhemv.c | 4 +-- src/library/blas/xher.c | 2 +- src/library/blas/xher2.c | 8 ++--- src/library/blas/xhpmv.c | 4 +-- src/library/blas/xnrm2.c | 2 +- src/library/blas/xrot.c | 4 +-- src/library/blas/xrotm.c | 4 +-- src/library/blas/xscal.c | 2 +- src/library/blas/xshbmv.c | 4 +-- src/library/blas/xspmv.c | 4 +-- src/library/blas/xswap.c | 4 +-- src/library/blas/xsymv.c | 4 +-- src/library/blas/xsyr.c | 2 +- src/library/blas/xsyr2.c | 4 +-- src/library/blas/xtbmv.c | 4 +-- src/library/blas/xtbsv.c | 44 ++++++++++++++-------------- src/library/blas/xtrmv.c | 4 +-- src/library/blas/xtrsv.c | 2 +- 49 files changed, 126 insertions(+), 126 deletions(-) diff --git a/src/library/blas/functor/functor_xscal.cc b/src/library/blas/functor/functor_xscal.cc index 1615499..c19d268 100644 --- a/src/library/blas/functor/functor_xscal.cc +++ b/src/library/blas/functor/functor_xscal.cc @@ -81,7 +81,7 @@ doScal( kargs->N = N; kargs->A = X; kargs->offBX = offx; - kargs->ldb.vector = incx; // Will be using this as incx + kargs->ldb.Vector = incx; // Will be using this as incx if(incx < 0) { // According to Netlib - return for negative incx return clblasSuccess; diff --git a/src/library/blas/generic/solution_seq_make.c b/src/library/blas/generic/solution_seq_make.c index b92e18e..ab64869 100644 --- a/src/library/blas/generic/solution_seq_make.c +++ b/src/library/blas/generic/solution_seq_make.c @@ -709,10 +709,10 @@ clblasArgsToKextraFlags(const CLBlasKargs *args, BlasFunctionID funcID) } } if (funcID == CLBLAS_GEMV || funcID == CLBLAS_SYMV) { - if (args->ldb.vector == 1) { + if (args->ldb.Vector == 1) { flags |= KEXTRA_INCX_ONE; } - if (args->ldc.vector == 1) { + if (args->ldc.Vector == 1) { flags |= KEXTRA_INCY_ONE; } } diff --git a/src/library/blas/gens/asum.cpp b/src/library/blas/gens/asum.cpp index 06b9f54..8c377cc 100644 --- a/src/library/blas/gens/asum.cpp +++ b/src/library/blas/gens/asum.cpp @@ -137,10 +137,10 @@ setBuildOpts( printf("Setting build options ... Double... for DOUBLE PRECISION support\n"); #endif } - if( (kargs->ldb.vector) != 1) { + if( (kargs->ldb.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCX_NONUNITY"); } - if( (kargs->ldb.vector) < 1) { + if( (kargs->ldb.Vector) < 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCX_NEGATIVE"); } return; @@ -275,7 +275,7 @@ assignKargs(KernelArg *args, const void *params, const void* ) INIT_KARG(&args[1], blasArgs->D); initSizeKarg(&args[2], blasArgs->N); initSizeKarg(&args[3], blasArgs->offBX); - incx = blasArgs->ldb.vector; + incx = blasArgs->ldb.Vector; INIT_KARG(&args[4], incx); return; } diff --git a/src/library/blas/gens/axpy_reg.cpp b/src/library/blas/gens/axpy_reg.cpp index 52aab71..fc73cbd 100644 --- a/src/library/blas/gens/axpy_reg.cpp +++ b/src/library/blas/gens/axpy_reg.cpp @@ -130,10 +130,10 @@ setBuildOpts( printf("Setting build options ... Double... for DOUBLE PRECISION support\n"); #endif } - if( (kargs->ldb.vector) != 1) { + if( (kargs->ldb.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCX_NONUNITY"); } - if( (kargs->ldc.vector) != 1) { + if( (kargs->ldc.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCY_NONUNITY"); } @@ -269,10 +269,10 @@ assignKargs(KernelArg *args, const void *params, const void* ) INIT_KARG(&args[2], blasArgs->B); initSizeKarg(&args[3], blasArgs->N); initSizeKarg(&args[4], blasArgs->offBX); - incx = blasArgs->ldb.vector; + incx = blasArgs->ldb.Vector; INIT_KARG(&args[5], incx); initSizeKarg(&args[6], blasArgs->offCY); - incy = blasArgs->ldc.vector; + incy = blasArgs->ldc.Vector; INIT_KARG(&args[7], incy); return; diff --git a/src/library/blas/gens/copy_reg.cpp b/src/library/blas/gens/copy_reg.cpp index ba1ff39..28df2ee 100644 --- a/src/library/blas/gens/copy_reg.cpp +++ b/src/library/blas/gens/copy_reg.cpp @@ -130,10 +130,10 @@ setBuildOpts( printf("Setting build options ... Double... for DOUBLE PRECISION support\n"); #endif } - if( (kargs->ldb.vector) != 1) { + if( (kargs->ldb.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCX_NONUNITY"); } - if( (kargs->ldc.vector) != 1) { + if( (kargs->ldc.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCY_NONUNITY"); } @@ -264,10 +264,10 @@ assignKargs(KernelArg *args, const void *params, const void* ) INIT_KARG(&args[1], blasArgs->B); initSizeKarg(&args[2], blasArgs->N); initSizeKarg(&args[3], blasArgs->offBX); - incx = blasArgs->ldb.vector; + incx = blasArgs->ldb.Vector; INIT_KARG(&args[4], incx); initSizeKarg(&args[5], blasArgs->offCY); - incy = blasArgs->ldc.vector; + incy = blasArgs->ldc.Vector; INIT_KARG(&args[6], incy); return; diff --git a/src/library/blas/gens/dot.cpp b/src/library/blas/gens/dot.cpp index ed3e72b..ce766cb 100644 --- a/src/library/blas/gens/dot.cpp +++ b/src/library/blas/gens/dot.cpp @@ -133,10 +133,10 @@ setBuildOpts( printf("Setting build options ... Double... for DOUBLE PRECISION support\n"); #endif } - if( (kargs->ldb.vector) != 1) { + if( (kargs->ldb.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCX_NONUNITY"); } - if( (kargs->ldc.vector) != 1) { + if( (kargs->ldc.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCY_NONUNITY"); } @@ -272,10 +272,10 @@ assignKargs(KernelArg *args, const void *params, const void* ) INIT_KARG(&args[2], blasArgs->D); initSizeKarg(&args[3], blasArgs->N); initSizeKarg(&args[4], blasArgs->offBX); - incx = blasArgs->ldb.vector; + incx = blasArgs->ldb.Vector; INIT_KARG(&args[5], incx); initSizeKarg(&args[6], blasArgs->offCY); - incy = blasArgs->ldc.vector; + incy = blasArgs->ldc.Vector; INIT_KARG(&args[7], incy); doConj = blasArgs->K; INIT_KARG(&args[8], doConj); diff --git a/src/library/blas/gens/gbmv.cpp b/src/library/blas/gens/gbmv.cpp index ab8e5e2..ad63a62 100644 --- a/src/library/blas/gens/gbmv.cpp +++ b/src/library/blas/gens/gbmv.cpp @@ -389,9 +389,9 @@ assignKargs(KernelArg *args, const void *params, const void* ) initSizeKarg(&args[6], fKU); initSizeKarg(&args[7], blasArgs->lda.matrix); - inc = blasArgs->ldb.vector; + inc = blasArgs->ldb.Vector; INIT_KARG(&args[8], inc); - inc = blasArgs->ldc.vector; + inc = blasArgs->ldc.Vector; INIT_KARG(&args[9], inc); initSizeKarg(&args[10], blasArgs->offa); diff --git a/src/library/blas/gens/gemv.c b/src/library/blas/gens/gemv.c index 40293d8..9835482 100644 --- a/src/library/blas/gens/gemv.c +++ b/src/library/blas/gens/gemv.c @@ -434,12 +434,12 @@ assignKargs(KernelArg *args, const void *params, const void *extra) initSizeKarg(&args[i++], blasArgs->offCY); } if (!(kflags & KEXTRA_INCX_ONE)) { - inc = blasArgs->ldb.vector; + inc = blasArgs->ldb.Vector; INIT_KARG(&args[i], inc); i++; } if (!(kflags & KEXTRA_INCY_ONE)) { - inc = blasArgs->ldc.vector; + inc = blasArgs->ldc.Vector; INIT_KARG(&args[i], inc); i++; } @@ -479,12 +479,12 @@ fixupArgs(void *args, SubproblemDim *subdims, void *extra) else { kargs->offA += off * kargs->lda.matrix; } - if (kargs->ldc.vector < 0) { + if (kargs->ldc.Vector < 0) { // K store the original height of the matrix A - kargs->offCY += (kargs->K - off) * abs(kargs->ldc.vector); + kargs->offCY += (kargs->K - off) * abs(kargs->ldc.Vector); } else { - kargs->offCY += off * kargs->ldc.vector; + kargs->offCY += off * kargs->ldc.Vector; } } diff --git a/src/library/blas/gens/ger_lds.cpp b/src/library/blas/gens/ger_lds.cpp index f72d197..267c979 100644 --- a/src/library/blas/gens/ger_lds.cpp +++ b/src/library/blas/gens/ger_lds.cpp @@ -317,8 +317,8 @@ assignKargs(KernelArg *args, const void *params, const void*) initSizeKarg(&args[3], blasArgs->M); initSizeKarg(&args[4], blasArgs->N); - incx = blasArgs->ldb.vector; - incy = blasArgs->ldc.vector; + incx = blasArgs->ldb.Vector; + incy = blasArgs->ldc.Vector; initSizeKarg(&args[5], blasArgs->offBX); INIT_KARG(&args[6], incx); initSizeKarg(&args[7], blasArgs->offCY); diff --git a/src/library/blas/gens/her2_lds.cpp b/src/library/blas/gens/her2_lds.cpp index 5adda19..3fd2fd0 100644 --- a/src/library/blas/gens/her2_lds.cpp +++ b/src/library/blas/gens/her2_lds.cpp @@ -332,10 +332,10 @@ assignKargs(KernelArg *args, const void *params, const void*) INIT_KARG(&args[2], blasArgs->C); //Y - y vector initSizeKarg(&args[3], blasArgs->N); initSizeKarg(&args[4], blasArgs->offBX); - inc = blasArgs->ldb.vector; + inc = blasArgs->ldb.Vector; INIT_KARG(&args[5], inc); initSizeKarg(&args[6], blasArgs->offCY); - inc = blasArgs->ldc.vector; + inc = blasArgs->ldc.Vector; INIT_KARG(&args[7], inc); initSizeKarg(&args[8], blasArgs->offa); initSizeKarg(&args[9], blasArgs->lda.matrix); diff --git a/src/library/blas/gens/her_lds.cpp b/src/library/blas/gens/her_lds.cpp index 1a8365f..8748645 100644 --- a/src/library/blas/gens/her_lds.cpp +++ b/src/library/blas/gens/her_lds.cpp @@ -330,7 +330,7 @@ assignKargs(KernelArg *args, const void *params, const void*) INIT_KARG(&args[1], blasArgs->B); //x - x vector initSizeKarg(&args[2], blasArgs->N); initSizeKarg(&args[3], blasArgs->offBX); - incx = blasArgs->ldb.vector; + incx = blasArgs->ldb.Vector; INIT_KARG(&args[4], incx); initSizeKarg(&args[5], blasArgs->offa); initSizeKarg(&args[6], blasArgs->lda.matrix); diff --git a/src/library/blas/gens/iamax.cpp b/src/library/blas/gens/iamax.cpp index 7a5966d..13928a8 100644 --- a/src/library/blas/gens/iamax.cpp +++ b/src/library/blas/gens/iamax.cpp @@ -130,12 +130,12 @@ setBuildOpts( #endif } - if( (kargs->ldb.vector) != 1) + if( (kargs->ldb.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCX_NONUNITY"); } - if( (kargs->ldb.vector) < 1) + if( (kargs->ldb.Vector) < 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DRETURN_ON_INVALID"); } @@ -277,7 +277,7 @@ assignKargs(KernelArg *args, const void *params, const void* ) INIT_KARG(&args[1], blasArgs->D); initSizeKarg(&args[2], blasArgs->N); initSizeKarg(&args[3], blasArgs->offb); - incx = blasArgs->ldb.vector; + incx = blasArgs->ldb.Vector; INIT_KARG(&args[4], incx); return; diff --git a/src/library/blas/gens/nrm2.cpp b/src/library/blas/gens/nrm2.cpp index d898ffb..c2dfe91 100644 --- a/src/library/blas/gens/nrm2.cpp +++ b/src/library/blas/gens/nrm2.cpp @@ -139,10 +139,10 @@ setBuildOpts( addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DUSE_SSQ"); } - if( (kargs->ldb.vector) != 1) { + if( (kargs->ldb.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCX_NONUNITY"); } - if( (kargs->ldb.vector) < 1) { + if( (kargs->ldb.Vector) < 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DRETURN_ON_INVALID"); } return; @@ -269,7 +269,7 @@ assignKargs(KernelArg *args, const void *params, const void* ) INIT_KARG(&args[1], blasArgs->D); initSizeKarg(&args[2], blasArgs->N); initSizeKarg(&args[3], blasArgs->offBX); - incx = blasArgs->ldb.vector; + incx = blasArgs->ldb.Vector; INIT_KARG(&args[4], incx); return; diff --git a/src/library/blas/gens/rotm_reg.cpp b/src/library/blas/gens/rotm_reg.cpp index 2b87507..5bfdc0f 100644 --- a/src/library/blas/gens/rotm_reg.cpp +++ b/src/library/blas/gens/rotm_reg.cpp @@ -127,10 +127,10 @@ setBuildOpts( { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DDO_ROT"); } - if( (kargs->ldb.vector) != 1) { + if( (kargs->ldb.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCX_NONUNITY"); } - if( (kargs->ldc.vector) != 1) { + if( (kargs->ldc.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCY_NONUNITY"); } @@ -268,10 +268,10 @@ assignKargs(KernelArg *args, const void *params, const void* ) INIT_KARG(&args[1], blasArgs->B); initSizeKarg(&args[2], blasArgs->N); initSizeKarg(&args[3], blasArgs->offBX); - incx = blasArgs->ldb.vector; + incx = blasArgs->ldb.Vector; INIT_KARG(&args[4], incx); initSizeKarg(&args[5], blasArgs->offCY); - incy = blasArgs->ldc.vector; + incy = blasArgs->ldc.Vector; INIT_KARG(&args[6], incy); if(blasArgs->pigFuncID == CLBLAS_ROT) diff --git a/src/library/blas/gens/scal_reg.cpp b/src/library/blas/gens/scal_reg.cpp index 8b85310..732ecb0 100644 --- a/src/library/blas/gens/scal_reg.cpp +++ b/src/library/blas/gens/scal_reg.cpp @@ -130,7 +130,7 @@ setBuildOpts( printf("Setting build options ... Double... for DOUBLE PRECISION support\n"); #endif } - if( (kargs->ldb.vector) != 1) { + if( (kargs->ldb.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCX_NONUNITY"); } @@ -261,7 +261,7 @@ assignKargs(KernelArg *args, const void *params, const void* ) INIT_KARG(&args[1], blasArgs->A); initSizeKarg(&args[2], blasArgs->N); initSizeKarg(&args[3], blasArgs->offBX); - incx = blasArgs->ldb.vector; + incx = blasArgs->ldb.Vector; INIT_KARG(&args[4], incx); return; diff --git a/src/library/blas/gens/swap_reg.cpp b/src/library/blas/gens/swap_reg.cpp index b75e100..a93cbd0 100644 --- a/src/library/blas/gens/swap_reg.cpp +++ b/src/library/blas/gens/swap_reg.cpp @@ -130,10 +130,10 @@ setBuildOpts( printf("Setting build options ... Double... for DOUBLE PRECISION support\n"); #endif } - if( (kargs->ldb.vector) != 1) { + if( (kargs->ldb.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCX_NONUNITY"); } - if( (kargs->ldc.vector) != 1) { + if( (kargs->ldc.Vector) != 1) { addBuildOpt( buildOptStr, BUILD_OPTS_MAXLEN, "-DINCY_NONUNITY"); } @@ -265,10 +265,10 @@ assignKargs(KernelArg *args, const void *params, const void* ) INIT_KARG(&args[1], blasArgs->B); initSizeKarg(&args[2], blasArgs->N); initSizeKarg(&args[3], blasArgs->offBX); - incx = blasArgs->ldb.vector; + incx = blasArgs->ldb.Vector; INIT_KARG(&args[4], incx); initSizeKarg(&args[5], blasArgs->offCY); - incy = blasArgs->ldc.vector; + incy = blasArgs->ldc.Vector; INIT_KARG(&args[6], incy); return; diff --git a/src/library/blas/gens/symv.c b/src/library/blas/gens/symv.c index 4944892..47c8f1d 100644 --- a/src/library/blas/gens/symv.c +++ b/src/library/blas/gens/symv.c @@ -926,12 +926,12 @@ assignKargs(KernelArg *args, const void *params, const void *extra) initSizeKarg(&args[i++], blasArgs->offCY); } if (!(kflags & KEXTRA_INCX_ONE)) { - inc = blasArgs->ldb.vector; + inc = blasArgs->ldb.Vector; INIT_KARG(&args[i], inc); i++; } if (!(kflags & KEXTRA_INCY_ONE)) { - inc = blasArgs->ldc.vector; + inc = blasArgs->ldc.Vector; INIT_KARG(&args[i], inc); i++; } @@ -949,13 +949,13 @@ fixupArgs(void *args, SubproblemDim *subdims, void *extra) (void)subdims; if (kargs->offsetN) { - if (kargs->ldc.vector < 0) { + if (kargs->ldc.Vector < 0) { // K store the original height of the matrix A kargs->offCY += (kargs->K - kargs->offsetN) * - abs(kargs->ldc.vector); + abs(kargs->ldc.Vector); } else { - kargs->offCY += kargs->offsetN * kargs->ldc.vector; + kargs->offCY += kargs->offsetN * kargs->ldc.Vector; } } } diff --git a/src/library/blas/gens/syr2_lds.cpp b/src/library/blas/gens/syr2_lds.cpp index f5c20cb..6e7df1d 100644 --- a/src/library/blas/gens/syr2_lds.cpp +++ b/src/library/blas/gens/syr2_lds.cpp @@ -338,10 +338,10 @@ assignKargs(KernelArg *args, const void *params, const void*) INIT_KARG(&args[2], blasArgs->C); //Y - y vector initSizeKarg(&args[3], blasArgs->N); initSizeKarg(&args[4], blasArgs->offBX); - inc = blasArgs->ldb.vector; + inc = blasArgs->ldb.Vector; INIT_KARG(&args[5], inc); initSizeKarg(&args[6], blasArgs->offCY); - inc = blasArgs->ldc.vector; + inc = blasArgs->ldc.Vector; INIT_KARG(&args[7], inc); initSizeKarg(&args[8], blasArgs->offa); initSizeKarg(&args[9], blasArgs->lda.matrix); diff --git a/src/library/blas/gens/syr_lds.cpp b/src/library/blas/gens/syr_lds.cpp index 16911bb..15e56ef 100644 --- a/src/library/blas/gens/syr_lds.cpp +++ b/src/library/blas/gens/syr_lds.cpp @@ -337,7 +337,7 @@ assignKargs(KernelArg *args, const void *params, const void*) INIT_KARG(&args[1], blasArgs->B); //x - x vector initSizeKarg(&args[2], blasArgs->N); initSizeKarg(&args[3], blasArgs->offBX); - inc = blasArgs->ldb.vector; + inc = blasArgs->ldb.Vector; INIT_KARG(&args[4], inc); initSizeKarg(&args[5], blasArgs->offA); initSizeKarg(&args[6], blasArgs->lda.matrix); diff --git a/src/library/blas/gens/trmv_reg.cpp b/src/library/blas/gens/trmv_reg.cpp index 9cacd0f..0adcd22 100644 --- a/src/library/blas/gens/trmv_reg.cpp +++ b/src/library/blas/gens/trmv_reg.cpp @@ -426,7 +426,7 @@ assignKargs(KernelArg *args, const void *params, const void* ) INIT_KARG(&args[2], blasArgs->C); //y - scratch == _x_vector argument } initSizeKarg(&args[3], blasArgs->N); - inc = blasArgs->ldb.vector; + inc = blasArgs->ldb.Vector; INIT_KARG(&args[4], inc); unity = (blasArgs->diag == clblasUnit); INIT_KARG(&args[5], unity); @@ -444,7 +444,7 @@ assignKargs(KernelArg *args, const void *params, const void* ) // For HEMV both alpha and beta has to be passed. if( (step->funcID == CLBLAS_HEMV) || (blasArgs->pigFuncID == CLBLAS_HPMV) || (blasArgs->pigFuncID == CLBLAS_SPMV) ) { - inc = blasArgs->ldc.vector; + inc = blasArgs->ldc.Vector; INIT_KARG(&args[10], inc); initSizeKarg(&args[11], blasArgs->offCY); assignScalarKarg(&args[12], &(blasArgs->alpha), blasArgs->dtype); diff --git a/src/library/blas/gens/trsv_gemv.cpp b/src/library/blas/gens/trsv_gemv.cpp index ca73fbe..65121a1 100644 --- a/src/library/blas/gens/trsv_gemv.cpp +++ b/src/library/blas/gens/trsv_gemv.cpp @@ -474,7 +474,7 @@ assignKargs(KernelArg *args, const void *params, const void*) INIT_KARG(&args[0], blasArgs->A); //A - input matrix - argument INIT_KARG(&args[1], blasArgs->B); //x - result buffer = _xnew argument initSizeKarg(&args[2], blasArgs->N); - inc = blasArgs->ldb.vector; + inc = blasArgs->ldb.Vector; INIT_KARG(&args[3], inc); unity = (blasArgs->diag == clblasUnit); INIT_KARG(&args[4], unity); diff --git a/src/library/blas/gens/trsv_trtri.cpp b/src/library/blas/gens/trsv_trtri.cpp index 0bae0f9..db4c9a0 100644 --- a/src/library/blas/gens/trsv_trtri.cpp +++ b/src/library/blas/gens/trsv_trtri.cpp @@ -382,7 +382,7 @@ assignKargs(KernelArg *args, const void *params, const void*) INIT_KARG(&args[0], blasArgs->A); //A - input matrix - argument INIT_KARG(&args[1], blasArgs->B); //x - result buffer = _xnew argument initSizeKarg(&args[2], blasArgs->N); - inc = blasArgs->ldb.vector; + inc = blasArgs->ldb.Vector; INIT_KARG(&args[3], inc); unity = (blasArgs->diag == clblasUnit); INIT_KARG(&args[4], unity); diff --git a/src/library/blas/include/clblas-internal.h b/src/library/blas/include/clblas-internal.h index e9a2d42..e5508e4 100644 --- a/src/library/blas/include/clblas-internal.h +++ b/src/library/blas/include/clblas-internal.h @@ -61,7 +61,7 @@ typedef union ArgMultiplier { typedef union LeadingDimention { size_t matrix; /**< Positive ld value for matrixes */ - int vector; /**< Integer offset value for vectors */ + int Vector; /**< Integer offset value for vectors */ } LeadingDimention; typedef enum reductionType { diff --git a/src/library/blas/ixamax.c b/src/library/blas/ixamax.c index 18088b6..43099f6 100644 --- a/src/library/blas/ixamax.c +++ b/src/library/blas/ixamax.c @@ -108,7 +108,7 @@ doiAmax( kargs->N = N; kargs->B = X; kargs->offb = offx; - kargs->ldb.vector = incx; // Will be using this as incx + kargs->ldb.Vector = incx; // Will be using this as incx if(incx < 1) { // According to netlib, if incx<1, NRM2 will be zero kargs->N = 1; // Makeing it launch only 1 work-group } diff --git a/src/library/blas/xasum.c b/src/library/blas/xasum.c index edc718c..eafd3d8 100644 --- a/src/library/blas/xasum.c +++ b/src/library/blas/xasum.c @@ -111,7 +111,7 @@ doAsum( kargs->offA = offAsum; kargs->B = X; kargs->offBX = offx; - kargs->ldb.vector = incx; // Will be using this as incx + kargs->ldb.Vector = incx; // Will be using this as incx if(incx <1){ kargs->N = 1; } diff --git a/src/library/blas/xaxpy.c b/src/library/blas/xaxpy.c index 4eb66a3..c5e1e56 100644 --- a/src/library/blas/xaxpy.c +++ b/src/library/blas/xaxpy.c @@ -103,10 +103,10 @@ doAxpy( kargs->N = N; kargs->A = X; kargs->offBX = offx; - kargs->ldb.vector = incx; // Will be using this as incx + kargs->ldb.Vector = incx; // Will be using this as incx kargs->B = Y; kargs->offCY = offy; - kargs->ldc.vector = incy; // Will be using this as incy + kargs->ldc.Vector = incy; // Will be using this as incy #ifdef DEBUG_AXPY printf("Calling makeSolutionSeq from DoAxpy: AXPY\n"); diff --git a/src/library/blas/xcopy.c b/src/library/blas/xcopy.c index d60ba5e..5540375 100644 --- a/src/library/blas/xcopy.c +++ b/src/library/blas/xcopy.c @@ -96,10 +96,10 @@ doCopy( kargs->N = N; kargs->A = X; kargs->offBX = offx; - kargs->ldb.vector = incx; // Will be using this as incx + kargs->ldb.Vector = incx; // Will be using this as incx kargs->B = Y; kargs->offCY = offy; - kargs->ldc.vector = incy; // Will be using this as incy + kargs->ldc.Vector = incy; // Will be using this as incy #ifdef DEBUG_COPY printf("Calling makeSolutionSeq from DoCopy: COPY\n"); diff --git a/src/library/blas/xdot.c b/src/library/blas/xdot.c index ed5530e..e37d0e0 100644 --- a/src/library/blas/xdot.c +++ b/src/library/blas/xdot.c @@ -119,10 +119,10 @@ doDot( kargs->offa = offDP; kargs->B = X; kargs->offBX = offx; - kargs->ldb.vector = incx; // Will be using this as incx + kargs->ldb.Vector = incx; // Will be using this as incx kargs->C = Y; kargs->offCY = offy; - kargs->ldc.vector = incy; // Will be using this as incy + kargs->ldc.Vector = incy; // Will be using this as incy kargs->D = scratchBuff; kargs->redctnType = REDUCE_BY_SUM; kargs->K = (size_t)doConj; diff --git a/src/library/blas/xgbmv.c b/src/library/blas/xgbmv.c index 205f8eb..dd5dd54 100644 --- a/src/library/blas/xgbmv.c +++ b/src/library/blas/xgbmv.c @@ -102,10 +102,10 @@ doGbmv( kargs->lda.matrix = lda; kargs->B = x; kargs->offBX = offx; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->C = y; kargs->offCY = offy; - kargs->ldc.vector = incy; + kargs->ldc.Vector = incy; listInitHead(&seq); err = makeSolutionSeq(CLBLAS_GBMV, kargs, numCommandQueues, commandQueues, diff --git a/src/library/blas/xgemv.c b/src/library/blas/xgemv.c index 5999979..e012918 100644 --- a/src/library/blas/xgemv.c +++ b/src/library/blas/xgemv.c @@ -85,10 +85,10 @@ doGemv( kargs->lda.matrix = lda; kargs->B = x; kargs->offBX = offx; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->C = y; kargs->offCY = offy; - kargs->ldc.vector = incy; + kargs->ldc.Vector = incy; listInitHead(&seq); err = makeSolutionSeq(CLBLAS_GEMV, kargs, numCommandQueues, commandQueues, diff --git a/src/library/blas/xger.c b/src/library/blas/xger.c index 9f79420..4644bfa 100644 --- a/src/library/blas/xger.c +++ b/src/library/blas/xger.c @@ -121,10 +121,10 @@ doGer( kargs->lda.matrix = lda; kargs->B = X; kargs->offBX = offx; - kargs->ldb.vector = incx; // Will be using this as incx + kargs->ldb.Vector = incx; // Will be using this as incx kargs->C = Y; kargs->offCY = offy; - kargs->ldc.vector = incy; // Will be using this as incy + kargs->ldc.Vector = incy; // Will be using this as incy kargs->offsetM = 0; kargs->offsetN = 0; kargs->scimage[0] = 0; diff --git a/src/library/blas/xhemv.c b/src/library/blas/xhemv.c index 21011dd..655b197 100644 --- a/src/library/blas/xhemv.c +++ b/src/library/blas/xhemv.c @@ -86,10 +86,10 @@ doHemv( kargs->lda.matrix = lda; kargs->B = x; kargs->offBX = offx; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->C = y; kargs->offCY = offy; - kargs->ldc.vector = incy; + kargs->ldc.Vector = incy; kargs->transA = clblasNoTrans; kargs->diag = clblasNonUnit; diff --git a/src/library/blas/xher.c b/src/library/blas/xher.c index 0b94580..ecf8f8a 100644 --- a/src/library/blas/xher.c +++ b/src/library/blas/xher.c @@ -100,7 +100,7 @@ doher( kargs->A = A; kargs->lda.matrix = lda; kargs->B = X; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->offBX = offx; kargs->offa = offa; kargs->offA = offa; diff --git a/src/library/blas/xher2.c b/src/library/blas/xher2.c index a6de9fe..05027ad 100644 --- a/src/library/blas/xher2.c +++ b/src/library/blas/xher2.c @@ -101,20 +101,20 @@ doHer2( { kargs->uplo = (uplo == clblasUpper) ? clblasLower : clblasUpper; kargs->B = Y; - kargs->ldb.vector = incy; + kargs->ldb.Vector = incy; kargs->offBX = offy; kargs->C = X; - kargs->ldc.vector = incx; + kargs->ldc.Vector = incx; kargs->offCY = offx; } else { kargs->uplo = uplo; kargs->B = X; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->offBX = offx; kargs->C = Y; - kargs->ldc.vector = incy; + kargs->ldc.Vector = incy; kargs->offCY = offy; } kargs->N = N; diff --git a/src/library/blas/xhpmv.c b/src/library/blas/xhpmv.c index 1f0fe67..1def97c 100644 --- a/src/library/blas/xhpmv.c +++ b/src/library/blas/xhpmv.c @@ -85,10 +85,10 @@ doHpmv( kargs->lda.matrix = 0; // Set lda as zero for packed matrices kargs->B = X; kargs->offBX = offx; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->C = Y; kargs->offCY = offy; - kargs->ldc.vector = incy; + kargs->ldc.Vector = incy; kargs->transA = clblasNoTrans; kargs->diag = clblasNonUnit; diff --git a/src/library/blas/xnrm2.c b/src/library/blas/xnrm2.c index 833d855..fc3dd51 100644 --- a/src/library/blas/xnrm2.c +++ b/src/library/blas/xnrm2.c @@ -219,7 +219,7 @@ doNrm2( kargs->offa = offNRM2; kargs->B = X; kargs->offBX = offx; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; if(incx < 1) { // According to netlib, if incx<1, NRM2 will be zero kargs->N = 1; // Makeing it launch only 1 work-group } diff --git a/src/library/blas/xrot.c b/src/library/blas/xrot.c index cf9acc6..0ca5498 100644 --- a/src/library/blas/xrot.c +++ b/src/library/blas/xrot.c @@ -95,10 +95,10 @@ doRot( kargs->N = N; kargs->A = X; kargs->offBX = offx; - kargs->ldb.vector = incx; // Will be using this as incx + kargs->ldb.Vector = incx; // Will be using this as incx kargs->B = Y; kargs->offCY = offy; - kargs->ldc.vector = incy; // Will be using this as incy + kargs->ldc.Vector = incy; // Will be using this as incy kargs->pigFuncID = CLBLAS_ROT; // Using ROTM kernel for ROT. Both are similar listInitHead(&seq); diff --git a/src/library/blas/xrotm.c b/src/library/blas/xrotm.c index dad78d8..67adb04 100644 --- a/src/library/blas/xrotm.c +++ b/src/library/blas/xrotm.c @@ -103,10 +103,10 @@ doRotm( kargs->N = N; kargs->A = X; kargs->offBX = offx; - kargs->ldb.vector = incx; // Will be using this as incx + kargs->ldb.Vector = incx; // Will be using this as incx kargs->B = Y; kargs->offCY = offy; - kargs->ldc.vector = incy; // Will be using this as incy + kargs->ldc.Vector = incy; // Will be using this as incy kargs->D = param; kargs->offd = offParam; kargs->pigFuncID = CLBLAS_ROTM; diff --git a/src/library/blas/xscal.c b/src/library/blas/xscal.c index 08bcb61..1756016 100644 --- a/src/library/blas/xscal.c +++ b/src/library/blas/xscal.c @@ -87,7 +87,7 @@ doScal( kargs->N = N; kargs->A = X; kargs->offBX = offx; - kargs->ldb.vector = incx; // Will be using this as incx + kargs->ldb.Vector = incx; // Will be using this as incx if(incx < 0) { // According to Netlib - return for negative incx return clblasSuccess; diff --git a/src/library/blas/xshbmv.c b/src/library/blas/xshbmv.c index 94f733d..7657b4e 100644 --- a/src/library/blas/xshbmv.c +++ b/src/library/blas/xshbmv.c @@ -100,10 +100,10 @@ doSHbmv( kargs->lda.matrix = lda; kargs->B = x; kargs->offBX = offx; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->C = y; kargs->offCY = offy; - kargs->ldc.vector = incy; + kargs->ldc.Vector = incy; listInitHead(&seq); err = makeSolutionSeq(CLBLAS_GBMV, kargs, numCommandQueues, commandQueues, diff --git a/src/library/blas/xspmv.c b/src/library/blas/xspmv.c index b40e026..2ce1015 100644 --- a/src/library/blas/xspmv.c +++ b/src/library/blas/xspmv.c @@ -85,10 +85,10 @@ doSpmv( kargs->lda.matrix = 0; // Set lda as zero for packed matrices kargs->B = X; kargs->offBX = offx; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->C = Y; kargs->offCY = offy; - kargs->ldc.vector = incy; + kargs->ldc.Vector = incy; kargs->transA = clblasNoTrans; kargs->diag = clblasNonUnit; diff --git a/src/library/blas/xswap.c b/src/library/blas/xswap.c index 006053b..9216312 100644 --- a/src/library/blas/xswap.c +++ b/src/library/blas/xswap.c @@ -96,10 +96,10 @@ doSwap( kargs->N = N; kargs->A = X; kargs->offBX = offx; - kargs->ldb.vector = incx; // Will be using this as incx + kargs->ldb.Vector = incx; // Will be using this as incx kargs->B = Y; kargs->offCY = offy; - kargs->ldc.vector = incy; // Will be using this as incy + kargs->ldc.Vector = incy; // Will be using this as incy #ifdef DEBUG_SWAP printf("Calling makeSolutionSeq from DoSwap: SWAP\n"); diff --git a/src/library/blas/xsymv.c b/src/library/blas/xsymv.c index 790e872..e65a1e5 100644 --- a/src/library/blas/xsymv.c +++ b/src/library/blas/xsymv.c @@ -84,10 +84,10 @@ doSymv( kargs->lda.matrix = lda; kargs->B = x; kargs->offBX = offx; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->C = y; kargs->offCY = offy; - kargs->ldc.vector = incy; + kargs->ldc.Vector = incy; #ifndef USE_SYMV diff --git a/src/library/blas/xsyr.c b/src/library/blas/xsyr.c index 99b47b4..fc3d9e5 100644 --- a/src/library/blas/xsyr.c +++ b/src/library/blas/xsyr.c @@ -104,7 +104,7 @@ doSyr( kargs->A = A; kargs->lda.matrix = lda; kargs->B = X; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->offBX = offx; kargs->offa = offa; kargs->offA = offa; diff --git a/src/library/blas/xsyr2.c b/src/library/blas/xsyr2.c index 37a31ee..0ca71e8 100644 --- a/src/library/blas/xsyr2.c +++ b/src/library/blas/xsyr2.c @@ -109,10 +109,10 @@ doSyr2( kargs->A = A; kargs->lda.matrix = lda; kargs->B = X; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->offBX = offx; kargs->C = Y; - kargs->ldc.vector = incy; + kargs->ldc.Vector = incy; kargs->offCY = offy; kargs->offa = offa; kargs->offA = offa; diff --git a/src/library/blas/xtbmv.c b/src/library/blas/xtbmv.c index 83322e5..a927b81 100644 --- a/src/library/blas/xtbmv.c +++ b/src/library/blas/xtbmv.c @@ -143,9 +143,9 @@ doTbmv( kargs->A = A; kargs->lda.matrix = lda; kargs->B = y; // Now it becomes x = A * y - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->C = x; - kargs->ldc.vector = incx; + kargs->ldc.Vector = incx; kargs->offBX = 0; // Not used by assignKargs(); Just for clarity kargs->offCY = offx; kargs->offa = offa; diff --git a/src/library/blas/xtbsv.c b/src/library/blas/xtbsv.c index a24d74c..44f00bd 100644 --- a/src/library/blas/xtbsv.c +++ b/src/library/blas/xtbsv.c @@ -153,15 +153,15 @@ if (err == CL_SUCCESS) gbmv->args.offA += kargs->offA; gbmv->args.offa = gbmv->args.offA; - if(kargs->ldb.vector < 0) + if(kargs->ldb.Vector < 0) { - gbmv->args.offBX = kargs->offBX + ((i-1) * TARGET_ROWS) * abs(kargs->ldb.vector); - gbmv->args.offCY = kargs->offBX + ((i * TARGET_ROWS) ) * abs(kargs->ldb.vector); + gbmv->args.offBX = kargs->offBX + ((i-1) * TARGET_ROWS) * abs(kargs->ldb.Vector); + gbmv->args.offCY = kargs->offBX + ((i * TARGET_ROWS) ) * abs(kargs->ldb.Vector); } else { - gbmv->args.offBX = kargs->offBX + (trtri->args.startRow) * kargs->ldb.vector; - gbmv->args.offCY = kargs->offBX + (gbmv->args.startRow) * kargs->ldb.vector; + gbmv->args.offBX = kargs->offBX + (trtri->args.startRow) * kargs->ldb.Vector; + gbmv->args.offCY = kargs->offBX + (gbmv->args.startRow) * kargs->ldb.Vector; } } else { @@ -176,15 +176,15 @@ if (err == CL_SUCCESS) gbmv->args.offA -= gbmv->args.KL; gbmv->args.offA += kargs->offA; gbmv->args.offa = gbmv->args.offA; - if(kargs->ldb.vector < 0) + if(kargs->ldb.Vector < 0) { - gbmv->args.offBX = kargs->offBX + (kargs->N - gbmv->args.startRow) * abs(kargs->ldb.vector); - gbmv->args.offCY = kargs->offBX + (kargs->N - (gbmv->args.startRow + gbmv->args.M) ) * abs(kargs->ldb.vector); + gbmv->args.offBX = kargs->offBX + (kargs->N - gbmv->args.startRow) * abs(kargs->ldb.Vector); + gbmv->args.offCY = kargs->offBX + (kargs->N - (gbmv->args.startRow + gbmv->args.M) ) * abs(kargs->ldb.Vector); } else { - gbmv->args.offBX = kargs->offBX + (trtri->args.startRow) * kargs->ldb.vector; - gbmv->args.offCY = kargs->offBX + (gbmv->args.startRow) * kargs->ldb.vector; + gbmv->args.offBX = kargs->offBX + (trtri->args.startRow) * kargs->ldb.Vector; + gbmv->args.offCY = kargs->offBX + (gbmv->args.startRow) * kargs->ldb.Vector; } } @@ -388,15 +388,15 @@ offa = r * lda + col - (r - k); gbmv->args.offA += kargs->offA; gbmv->args.offa = gbmv->args.offA; - if(kargs->ldb.vector < 0) + if(kargs->ldb.Vector < 0) { - gbmv->args.offBX = kargs->offBX + (kargs->N - (gbmv->args.endRow)) * abs(kargs->ldb.vector); - gbmv->args.offCY = kargs->offBX + (kargs->N - (gbmv->args.endRow + gbmv->args.N) ) * abs(kargs->ldb.vector); + gbmv->args.offBX = kargs->offBX + (kargs->N - (gbmv->args.endRow)) * abs(kargs->ldb.Vector); + gbmv->args.offCY = kargs->offBX + (kargs->N - (gbmv->args.endRow + gbmv->args.N) ) * abs(kargs->ldb.Vector); } else { - gbmv->args.offBX = kargs->offBX + (gbmv->args.startRow) * kargs->ldb.vector; - gbmv->args.offCY = kargs->offBX + (gbmv->args.endRow) * kargs->ldb.vector; + gbmv->args.offBX = kargs->offBX + (gbmv->args.startRow) * kargs->ldb.Vector; + gbmv->args.offCY = kargs->offBX + (gbmv->args.endRow) * kargs->ldb.Vector; } @@ -416,15 +416,15 @@ offa = r * lda + col - (r - k); gbmv->args.offA -= gbmv->args.KL; gbmv->args.offA += kargs->offA; gbmv->args.offa = gbmv->args.offA; - if(kargs->ldb.vector < 0) + if(kargs->ldb.Vector < 0) { - gbmv->args.offBX = kargs->offBX + (kargs->N - gbmv->args.endRow) * abs(kargs->ldb.vector); - gbmv->args.offCY = kargs->offBX + (kargs->N - (gbmv->args.startRow) ) * abs(kargs->ldb.vector); + gbmv->args.offBX = kargs->offBX + (kargs->N - gbmv->args.endRow) * abs(kargs->ldb.Vector); + gbmv->args.offCY = kargs->offBX + (kargs->N - (gbmv->args.startRow) ) * abs(kargs->ldb.Vector); } else { - gbmv->args.offBX = kargs->offBX + (gbmv->args.startRow) * kargs->ldb.vector; - gbmv->args.offCY = kargs->offBX + (gbmv->args.startRow - gbmv->args.N) * kargs->ldb.vector; + gbmv->args.offBX = kargs->offBX + (gbmv->args.startRow) * kargs->ldb.Vector; + gbmv->args.offCY = kargs->offBX + (gbmv->args.startRow - gbmv->args.N) * kargs->ldb.Vector; } } @@ -626,13 +626,13 @@ doTbsv( kargs->A = A; kargs->lda.matrix = lda; kargs->B = x; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->offBX = offx; kargs->offa = offa; kargs->offA = offa; kargs->C = x; kargs->offCY = offx; - kargs->ldc.vector = incx; + kargs->ldc.Vector = incx; kargs->startRow = 0; if(trans == clblasNoTrans) diff --git a/src/library/blas/xtrmv.c b/src/library/blas/xtrmv.c index 6ff205b..b412694 100644 --- a/src/library/blas/xtrmv.c +++ b/src/library/blas/xtrmv.c @@ -132,9 +132,9 @@ doTrmv( kargs->A = A; kargs->lda.matrix = lda; kargs->B = x; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->C = y; - kargs->ldc.vector = incx; + kargs->ldc.Vector = incx; kargs->offBX = offx; kargs->offCY = 0; // Not used by assignKargs(); Just for clarity kargs->offa = offa; diff --git a/src/library/blas/xtrsv.c b/src/library/blas/xtrsv.c index 1e48349..c75dadb 100644 --- a/src/library/blas/xtrsv.c +++ b/src/library/blas/xtrsv.c @@ -406,7 +406,7 @@ doTrsv( kargs->A = A; kargs->lda.matrix = lda; kargs->B = x; - kargs->ldb.vector = incx; + kargs->ldb.Vector = incx; kargs->offBX = offx; kargs->offa = offa; kargs->offA = offa; -- 2.7.0
From ef1b194971d9e4ea1043e668940c83ee60eebc2d Mon Sep 17 00:00:00 2001 From: Thierry Fauck <tfa...@free.fr> Date: Tue, 26 Apr 2016 05:50:03 -0400 Subject: [PATCH 2/2] ppc64le altivec support P8+ powerpc64 processor add altivec capabilities known as -maltivec flag in gcc. bool defintion is not supported properly as such we properly redefined it. modified: src/include/defbool.h --- src/include/defbool.h | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/src/include/defbool.h b/src/include/defbool.h index 26caf6a..741e6c5 100644 --- a/src/include/defbool.h +++ b/src/include/defbool.h @@ -18,6 +18,11 @@ #ifndef DEFBOOL_H_ #define DEFBOOL_H_ +#if defined(__powerpc64__) && defined(__ALTIVEC__) +#include "altivec.h" +#undef bool +#endif + #if defined(_MSC_VER) && _MSC_VER <= 1700 /* -- 2.7.0
0xFCC181CE.asc
Description: application/pgp-keys
-- debian-science-maintainers mailing list debian-science-maintainers@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/debian-science-maintainers