CAMLprim value spoc_cublasDgemm(value transa, value transb,
value m, value n, value k,
value alpha, value a, value lda,
value b, value ldb, value beta, value c, value ldc, value dev){
CAMLparam5(transa, transb, m, n, k);
CAMLxparam5(alpha, a, lda, b, ldb);
CAMLxparam4(beta, c, ldc, dev);
CAMLlocal3(dev_vec_array, dev_vec, gi);
CUdeviceptr d_A;
CUdeviceptr d_B;
CUdeviceptr d_C;
int id;
gi = Field(dev, 0);
id = Int_val(Field(gi, 7));
GET_VEC(a, d_A);
GET_VEC(b, d_B);
GET_VEC(c, d_C);
//CUBLAS_GET_CONTEXT;
CUBLAS_GET_CONTEXT;
cublasDgemm (Int_val(transa), Int_val(transb), Int_val(m), Int_val(n),
Int_val(k), (double)Double_val(alpha), (double*) d_A, Int_val(lda),
(double*) d_B, Int_val(ldb), (double) Double_val(beta),
(double *)d_C, Int_val(ldc));
CUDA_RESTORE_CONTEXT;
CAMLreturn(Val_unit);
}
value f_i15_caml(value i0, value i1, value i2, value i3, value i4, value i5, value i6, value i7, value i8, value i9, value i10, value i11, value i12, value i13, value i14) {
CAMLparam5(i0,i1,i2,i3,i4);
CAMLxparam5(i5,i6,i7,i8,i9);
CAMLxparam5(i10,i11,i12,i13,i14);
int ii0 = Int_val(i0);
int ii1 = Int_val(i1);
int ii2 = Int_val(i2);
int ii3 = Int_val(i3);
int ii4 = Int_val(i4);
int ii5 = Int_val(i5);
int ii6 = Int_val(i6);
int ii7 = Int_val(i7);
int ii8 = Int_val(i8);
int ii9 = Int_val(i9);
int ii10= Int_val(i10);
int ii11= Int_val(i11);
int ii12= Int_val(i12);
int ii13= Int_val(i13);
int ii14= Int_val(i14);
CAMLreturn(Val_int(f_i15(ii0,ii1,ii2,ii3,ii4,ii5,ii6,ii7,ii8,ii9,ii10,ii11,ii12,ii13,ii14)));
}
CAMLprim value stub_xc_linux_build_native(value xc_handle, value domid,
value mem_max_mib, value mem_start_mib,
value image_name, value ramdisk_name,
value cmdline, value features,
value flags, value store_evtchn,
value console_evtchn)
{
CAMLparam5(xc_handle, domid, mem_max_mib, mem_start_mib, image_name);
CAMLxparam5(ramdisk_name, cmdline, features, flags, store_evtchn);
CAMLxparam1(console_evtchn);
CAMLlocal1(result);
unsigned long store_mfn;
unsigned long console_mfn;
int r;
struct xc_dom_image *dom;
char c_protocol[64];
/* Copy the ocaml values into c-land before dropping the mutex */
xc_interface *xch = _H(xc_handle);
unsigned int c_mem_start_mib = Int_val(mem_start_mib);
uint32_t c_domid = _D(domid);
char *c_image_name = strdup(String_val(image_name));
char *c_ramdisk_name = ramdisk_name == None_val ? NULL : strdup(String_val(Field(ramdisk_name, 0)));
unsigned long c_flags = Int_val(flags);
unsigned int c_store_evtchn = Int_val(store_evtchn);
unsigned int c_console_evtchn = Int_val(console_evtchn);
struct flags f;
get_flags(&f,c_domid);
xc_dom_loginit(xch);
dom = xc_dom_allocate(xch, String_val(cmdline), String_val(features));
if (!dom)
failwith_oss_xc(xch, "xc_dom_allocate");
configure_vcpus(xch, c_domid, f);
configure_tsc(xch, c_domid, f);
#ifdef XC_HAVE_DECOMPRESS_LIMITS
if ( xc_dom_kernel_max_size(dom, f.kernel_max_size) )
failwith_oss_xc(xch, "xc_dom_kernel_max_size");
if ( xc_dom_ramdisk_max_size(dom, f.ramdisk_max_size) )
failwith_oss_xc(xch, "xc_dom_ramdisk_max_size");
#else
if ( f.kernel_max_size || f.ramdisk_max_size ) {
syslog(LOG_WARNING|LOG_DAEMON,"Kernel/Ramdisk limits set, but no support compiled in");
}
#endif
caml_enter_blocking_section();
r = xc_dom_linux_build(xch, dom, c_domid, c_mem_start_mib,
c_image_name, c_ramdisk_name, c_flags,
c_store_evtchn, &store_mfn,
c_console_evtchn, &console_mfn);
caml_leave_blocking_section();
#ifndef XEN_UNSTABLE
strncpy(c_protocol, xc_dom_get_native_protocol(dom), 64);
#else
memset(c_protocol, '\0', 64);
#endif
free(c_image_name);
free(c_ramdisk_name);
xc_dom_release(dom);
if (r != 0)
failwith_oss_xc(xch, "xc_dom_linux_build");
result = caml_alloc_tuple(3);
Store_field(result, 0, caml_copy_nativeint(store_mfn));
Store_field(result, 1, caml_copy_nativeint(console_mfn));
Store_field(result, 2, caml_copy_string(c_protocol));
CAMLreturn(result);
}