Implement cublasSgetrsBatched.

zluda_blas/src/cublas.rs

@@ -4337,7 +4337,19 @@ pub unsafe extern "system" fn cublasSgetrsBatched(
     info: *mut ::std::os::raw::c_int,
     batchSize: ::std::os::raw::c_int,
 ) -> cublasStatus_t {
-    crate::unsupported()
+    crate::sgetrs_batched(
+        handle,
+        trans,
+        n,
+        nrhs,
+        Aarray,
+        lda,
+        devIpiv,
+        Barray,
+        ldb,
+        info,
+        batchSize,
+    )
 }
 
 #[no_mangle]

zluda_blas/src/lib.rs

@@ -6,8 +6,7 @@ pub use cublas::*;
 use cuda_types::*;
 use rocblas_sys::*;
 use rocsolver_sys::{
-    rocsolver_cgetrf_batched, rocsolver_cgetri_batched, rocsolver_cgetri_outofplace_batched,
-    rocsolver_zgetrf_batched, rocsolver_zgetri_batched, rocsolver_zgetri_outofplace_batched,
+    rocsolver_cgetrf_batched, rocsolver_cgetri_batched, rocsolver_cgetri_outofplace_batched, rocsolver_sgetrs_batched, rocsolver_zgetrf_batched, rocsolver_zgetri_batched, rocsolver_zgetri_outofplace_batched
 };
 use std::{mem, ptr};
 
@@ -83,6 +82,15 @@ fn op_from_cuda(trans: cublasOperation_t) -> rocblas_operation {
     }
 }
 
+fn op_from_cuda_for_solver(trans: cublasOperation_t) -> rocsolver_sys::rocblas_operation {
+    match trans {
+        cublasOperation_t::CUBLAS_OP_N => rocsolver_sys::rocblas_operation::rocblas_operation_none,
+        cublasOperation_t::CUBLAS_OP_T => rocsolver_sys::rocblas_operation::rocblas_operation_transpose,
+        cublasOperation_t::CUBLAS_OP_C => rocsolver_sys::rocblas_operation::rocblas_operation_conjugate_transpose,
+        _ => panic!(),
+    }
+}
+
 unsafe fn destroy(handle: cublasHandle_t) -> cublasStatus_t {
     to_cuda(rocblas_destroy_handle(handle as _))
 }
@@ -656,6 +664,36 @@ unsafe fn cgetri_batched(
     ))
 }
 
+unsafe fn sgetrs_batched(
+    handle: *mut cublasContext,
+    trans: cublasOperation_t,
+    n: i32,
+    nrhs: i32,
+    a: *const *const f32,
+    lda: i32,
+    dev_ipiv: *const i32,
+    b: *const *mut f32,
+    ldb: i32,
+    info: *mut i32,
+    batch_size: i32,
+) -> cublasStatus_t {
+    let trans = op_from_cuda_for_solver(trans);
+    let stride = n * nrhs;
+    to_cuda_solver(rocsolver_sgetrs_batched(
+        handle.cast(),
+        trans,
+        n,
+        nrhs,
+        a.cast(),
+        lda,
+        dev_ipiv,
+        stride as _,
+        b.cast(),
+        ldb,
+        batch_size,
+    ))
+}
+
 unsafe fn dtrmm_v2(
     handle: *mut cublasContext,
     side: cublasSideMode_t,

zluda_dnn/src/lib.rs

@@ -20,7 +20,7 @@ use types::*;
 
 use hip_runtime_sys::*;
 use miopen_sys::*;
-use std::{mem, ptr::{self, null, null_mut}, alloc::{self, Layout}};
+use std::{mem, ptr, alloc::{self, Layout}};
 
 macro_rules! call {
     ($expr:expr) => {{