cython wrapper for acados (#22784)

* cython wrapper for acados

* fix building

* sconscript cleanup

* no cython numpy

* cleanup

* upgrade build script

* try without slices

* new acados commit

* c3 update acados libs

* c2 libs

* make faster

* undo profiling

* fix build

* somewhat faster

* tryout cost_set_slice

* Revert "tryout cost_set_slice"

This reverts commit d358d93a133270e4edab9e7c07ffb6f577c52bd6.

* cleanup

* undo t_renderer change

Co-authored-by: Comma Device <device@comma.ai>
old-commit-hash: 89d0a52d16872c403c69426ab32e5788a41ee2ec

SConstruct

@@ -68,6 +68,7 @@ lenv = {
   "PYTHONPATH": Dir("#").abspath + ":" + Dir("#pyextra/").abspath,
 
   "ACADOS_SOURCE_DIR": Dir("#third_party/acados/acados").abspath,
+  "ACADOS_PYTHON_INTERFACE_PATH": Dir("#pyextra/acados_template").abspath,
   "TERA_PATH": Dir("#").abspath + f"/third_party/acados/{arch}/t_renderer",
 }
 

selfdrive/controls/lib/lateral_mpc_lib/SConscript

@@ -1,4 +1,4 @@
-Import('env', 'arch')
+Import('env', 'envCython', 'arch', 'common')
 
 gen = "c_generated_code"
 
@@ -33,6 +33,7 @@ generated_files = [
 
   f'{gen}/main_lat.c',
   f'{gen}/acados_solver_lat.h',
+  f'{gen}/acados_solver.pxd',
 
   f'{gen}/lat_model/lat_expl_vde_adj.c',
 
@@ -53,6 +54,24 @@ lenv["CFLAGS"].append("-DACADOS_WITH_QPOASES")
 lenv["CXXFLAGS"].append("-DACADOS_WITH_QPOASES")
 lenv["CCFLAGS"].append("-Wno-unused")
 lenv["LINKFLAGS"].append("-Wl,--disable-new-dtags")
-lenv.SharedLibrary(f"{gen}/acados_ocp_solver_lat",
-                   build_files,
-                   LIBS=['m', 'acados', 'hpipm', 'blasfeo', 'qpOASES_e'])
+lib_solver = lenv.SharedLibrary(f"{gen}/acados_ocp_solver_lat",
+                                build_files,
+                                LIBS=['m', 'acados', 'hpipm', 'blasfeo', 'qpOASES_e'])
+
+# generate cython stuff
+acados_ocp_solver_pyx = File("#pyextra/acados_template/acados_ocp_solver_pyx.pyx")
+acados_ocp_solver_common = File("#pyextra/acados_template/acados_solver_common.pxd")
+libacados_ocp_solver_pxd = File(f'{gen}/acados_solver.pxd')
+libacados_ocp_solver_c = File(f'{gen}/acados_ocp_solver_pyx.c')
+
+lenv2 = envCython.Clone()
+lenv2["LINKFLAGS"] += [lib_solver[0].get_labspath()]
+lenv2.Command(libacados_ocp_solver_c,
+  [acados_ocp_solver_pyx, acados_ocp_solver_common, libacados_ocp_solver_pxd],
+  f'cython' + \
+  f' -o {libacados_ocp_solver_c.get_labspath()}' + \
+  f' -I {libacados_ocp_solver_pxd.get_dir().get_labspath()}' + \
+  f' -I {acados_ocp_solver_common.get_dir().get_labspath()}' + \
+  f' {acados_ocp_solver_pyx.get_labspath()}')
+lib_cython = lenv2.Program(f'{gen}/acados_ocp_solver_pyx.so', [libacados_ocp_solver_c])
+lenv2.Depends(lib_cython, lib_solver)

selfdrive/controls/lib/lateral_mpc_lib/lat_mpc.py

@@ -5,8 +5,12 @@ import numpy as np
 from casadi import SX, vertcat, sin, cos
 from selfdrive.controls.lib.drive_helpers import LAT_MPC_N as N
 from selfdrive.controls.lib.drive_helpers import T_IDXS
-from pyextra.acados_template import AcadosModel, AcadosOcp, AcadosOcpSolver
 
+if __name__ == '__main__':  # generating code
+  from pyextra.acados_template import AcadosModel, AcadosOcp, AcadosOcpSolver
+else:
+  # from pyextra.acados_template import AcadosOcpSolverFast
+  from selfdrive.controls.lib.lateral_mpc_lib.c_generated_code.acados_ocp_solver_pyx import AcadosOcpSolverFast  # pylint: disable=no-name-in-module, import-error
 
 LAT_MPC_DIR = os.path.dirname(os.path.abspath(__file__))
 EXPORT_DIR = os.path.join(LAT_MPC_DIR, "c_generated_code")
@@ -110,17 +114,16 @@ def gen_lat_mpc_solver():
 
 class LateralMpc():
   def __init__(self, x0=np.zeros(X_DIM)):
-    self.solver = AcadosOcpSolver('lat', N, EXPORT_DIR)
+    self.solver = AcadosOcpSolverFast('lat', N, EXPORT_DIR)
     self.reset(x0)
 
   def reset(self, x0=np.zeros(X_DIM)):
     self.x_sol = np.zeros((N+1, X_DIM))
     self.u_sol = np.zeros((N, 1))
     self.yref = np.zeros((N+1, 3))
-    self.solver.cost_set_slice(0, N, "yref", self.yref[:N])
+    for i in range(N):
+      self.solver.cost_set(i, "yref", self.yref[i])
     self.solver.cost_set(N, "yref", self.yref[N][:2])
-    W = np.eye(3)
-    self.Ws = np.tile(W[None], reps=(N,1,1))
 
     # Somehow needed for stable init
     for i in range(N+1):
@@ -132,12 +135,11 @@ class LateralMpc():
     self.cost = 0
 
   def set_weights(self, path_weight, heading_weight, steer_rate_weight):
-    self.Ws[:,0,0] = path_weight
-    self.Ws[:,1,1] = heading_weight
-    self.Ws[:,2,2] = steer_rate_weight
-    self.solver.cost_set_slice(0, N, 'W', self.Ws, api='old')
+    W = np.asfortranarray(np.diag([path_weight, heading_weight, steer_rate_weight]))
+    for i in range(N):
+      self.solver.cost_set(i, 'W', W)
     #TODO hacky weights to keep behavior the same
-    self.solver.cost_set(N, 'W', (3/20.)*self.Ws[0,:2,:2])
+    self.solver.cost_set(N, 'W', (3/20.)*W[:2,:2])
 
   def run(self, x0, v_ego, car_rotation_radius, y_pts, heading_pts):
     x0_cp = np.copy(x0)
@@ -145,12 +147,15 @@ class LateralMpc():
     self.solver.constraints_set(0, "ubx", x0_cp)
     self.yref[:,0] = y_pts
     self.yref[:,1] = heading_pts*(v_ego+5.0)
-    self.solver.cost_set_slice(0, N, "yref", self.yref[:N])
+    for i in range(N):
+      self.solver.cost_set(i, "yref", self.yref[i])
     self.solver.cost_set(N, "yref", self.yref[N][:2])
 
     self.solution_status = self.solver.solve()
-    self.solver.fill_in_slice(0, N+1, 'x', self.x_sol)
-    self.solver.fill_in_slice(0, N, 'u', self.u_sol)
+    for i in range(N+1):
+      self.x_sol[i] = self.solver.get(i, 'x')
+    for i in range(N):
+      self.u_sol[i] = self.solver.get(i, 'u')
     self.cost = self.solver.get_cost()
 
 

selfdrive/controls/lib/longitudinal_mpc_lib/SConscript

@@ -1,4 +1,4 @@
-Import('env', 'arch')
+Import('env', 'envCython', 'arch', 'common')
 
 gen = "c_generated_code"
 
@@ -41,6 +41,7 @@ generated_files = [
 
   f'{gen}/main_long.c',
   f'{gen}/acados_solver_long.h',
+  f'{gen}/acados_solver.pxd',
 
   f'{gen}/long_model/long_expl_vde_adj.c',
 
@@ -63,6 +64,24 @@ lenv["CFLAGS"].append("-DACADOS_WITH_QPOASES")
 lenv["CXXFLAGS"].append("-DACADOS_WITH_QPOASES")
 lenv["CCFLAGS"].append("-Wno-unused")
 lenv["LINKFLAGS"].append("-Wl,--disable-new-dtags")
-lenv.SharedLibrary(f"{gen}/acados_ocp_solver_long",
-                   build_files,
-                   LIBS=['m', 'acados', 'hpipm', 'blasfeo', 'qpOASES_e'])
+lib_solver = lenv.SharedLibrary(f"{gen}/acados_ocp_solver_long",
+                                build_files,
+                                LIBS=['m', 'acados', 'hpipm', 'blasfeo', 'qpOASES_e'])
+
+# generate cython stuff
+acados_ocp_solver_pyx = File("#pyextra/acados_template/acados_ocp_solver_pyx.pyx")
+acados_ocp_solver_common = File("#pyextra/acados_template/acados_solver_common.pxd")
+libacados_ocp_solver_pxd = File(f'{gen}/acados_solver.pxd')
+libacados_ocp_solver_c = File(f'{gen}/acados_ocp_solver_pyx.c')
+
+lenv2 = envCython.Clone()
+lenv2["LINKFLAGS"] += [lib_solver[0].get_labspath()]
+lenv2.Command(libacados_ocp_solver_c,
+  [acados_ocp_solver_pyx, acados_ocp_solver_common, libacados_ocp_solver_pxd],
+  f'cython' + \
+  f' -o {libacados_ocp_solver_c.get_labspath()}' + \
+  f' -I {libacados_ocp_solver_pxd.get_dir().get_labspath()}' + \
+  f' -I {acados_ocp_solver_common.get_dir().get_labspath()}' + \
+  f' {acados_ocp_solver_pyx.get_labspath()}')
+lib_cython = lenv2.Program(f'{gen}/acados_ocp_solver_pyx.so', [libacados_ocp_solver_c])
+lenv2.Depends(lib_cython, lib_solver)

selfdrive/controls/lib/longitudinal_mpc_lib/long_mpc.py

@@ -8,7 +8,12 @@ from selfdrive.swaglog import cloudlog
 from selfdrive.modeld.constants import index_function
 from selfdrive.controls.lib.radar_helpers import _LEAD_ACCEL_TAU
 
-from pyextra.acados_template import AcadosModel, AcadosOcp, AcadosOcpSolver
+if __name__ == '__main__':  # generating code
+  from pyextra.acados_template import AcadosModel, AcadosOcp, AcadosOcpSolver
+else:
+  # from pyextra.acados_template import AcadosOcpSolver as AcadosOcpSolverFast
+  from selfdrive.controls.lib.longitudinal_mpc_lib.c_generated_code.acados_ocp_solver_pyx import AcadosOcpSolverFast  # pylint: disable=no-name-in-module, import-error
+
 from casadi import SX, vertcat
 
 LONG_MPC_DIR = os.path.dirname(os.path.abspath(__file__))
@@ -190,13 +195,14 @@ class LongitudinalMpc():
     self.source = SOURCES[2]
 
   def reset(self):
-    self.solver = AcadosOcpSolver('long', N, EXPORT_DIR)
+    self.solver = AcadosOcpSolverFast('long', N, EXPORT_DIR)
     self.v_solution = [0.0 for i in range(N+1)]
     self.a_solution = [0.0 for i in range(N+1)]
     self.j_solution = [0.0 for i in range(N)]
     self.yref = np.zeros((N+1, COST_DIM))
-    self.solver.cost_set_slice(0, N, "yref", self.yref[:N])
-    self.solver.set(N, "yref", self.yref[N][:COST_E_DIM])
+    for i in range(N):
+      self.solver.cost_set(i, "yref", self.yref[i])
+    self.solver.cost_set(N, "yref", self.yref[N][:COST_E_DIM])
     self.x_sol = np.zeros((N+1, X_DIM))
     self.u_sol = np.zeros((N,1))
     self.params = np.zeros((N+1,3))
@@ -216,30 +222,30 @@ class LongitudinalMpc():
       self.set_weights_for_lead_policy()
 
   def set_weights_for_lead_policy(self):
-    W = np.diag([X_EGO_OBSTACLE_COST, X_EGO_COST, V_EGO_COST, A_EGO_COST, J_EGO_COST])
-    Ws = np.tile(W[None], reps=(N,1,1))
-    self.solver.cost_set_slice(0, N, 'W', Ws, api='old')
+    W = np.asfortranarray(np.diag([X_EGO_OBSTACLE_COST, X_EGO_COST, V_EGO_COST, A_EGO_COST, J_EGO_COST]))
+    for i in range(N):
+      self.solver.cost_set(i, 'W', W)
     # Setting the slice without the copy make the array not contiguous,
     # causing issues with the C interface.
     self.solver.cost_set(N, 'W', np.copy(W[:COST_E_DIM, :COST_E_DIM]))
 
     # Set L2 slack cost on lower bound constraints
     Zl = np.array([LIMIT_COST, LIMIT_COST, LIMIT_COST, DANGER_ZONE_COST])
-    Zls = np.tile(Zl[None], reps=(N+1,1,1))
-    self.solver.cost_set_slice(0, N+1, 'Zl', Zls, api='old')
+    for i in range(N):
+      self.solver.cost_set(i, 'Zl', Zl)
 
   def set_weights_for_xva_policy(self):
-    W = np.diag([0., 10., 1., 10., 1.])
-    Ws = np.tile(W[None], reps=(N,1,1))
-    self.solver.cost_set_slice(0, N, 'W', Ws, api='old')
+    W = np.asfortranarray(np.diag([0., 10., 1., 10., 1.]))
+    for i in range(N):
+      self.solver.cost_set(i, 'W', W)
     # Setting the slice without the copy make the array not contiguous,
     # causing issues with the C interface.
     self.solver.cost_set(N, 'W', np.copy(W[:COST_E_DIM, :COST_E_DIM]))
 
     # Set L2 slack cost on lower bound constraints
     Zl = np.array([LIMIT_COST, LIMIT_COST, LIMIT_COST, 0.0])
-    Zls = np.tile(Zl[None], reps=(N+1,1,1))
-    self.solver.cost_set_slice(0, N+1, 'Zl', Zls, api='old')
+    for i in range(N):
+      self.solver.cost_set(i, 'Zl', Zl)
 
   def set_cur_state(self, v, a):
     if abs(self.x0[1] - v) > 1.:
@@ -326,8 +332,9 @@ class LongitudinalMpc():
     self.yref[:,1] = x
     self.yref[:,2] = v
     self.yref[:,3] = a
-    self.solver.cost_set_slice(0, N, "yref", self.yref[:N], api='old')
-    self.solver.set(N, "yref", self.yref[N][:COST_E_DIM])
+    for i in range(N):
+      self.solver.cost_set(i, "yref", self.yref[i])
+    self.solver.cost_set(N, "yref", self.yref[N][:COST_E_DIM])
     self.accel_limit_arr[:,0] = -10.
     self.accel_limit_arr[:,1] = 10.
     x_obstacle = 1e5*np.ones((N+1))
@@ -338,12 +345,14 @@ class LongitudinalMpc():
 
   def run(self):
     for i in range(N+1):
-      self.solver.set_param(i, self.params[i])
+      self.solver.set(i, 'p', self.params[i])
     self.solver.constraints_set(0, "lbx", self.x0)
     self.solver.constraints_set(0, "ubx", self.x0)
     self.solution_status = self.solver.solve()
-    self.solver.fill_in_slice(0, N+1, 'x', self.x_sol)
-    self.solver.fill_in_slice(0, N, 'u', self.u_sol)
+    for i in range(N+1):
+      self.x_sol[i] = self.solver.get(i, 'x')
+    for i in range(N):
+      self.u_sol[i] = self.solver.get(i, 'u')
 
     self.v_solution = self.x_sol[:,1]
     self.a_solution = self.x_sol[:,2]

third_party/acados/build.sh

@@ -18,7 +18,7 @@ if [ ! -d acados_repo/ ]; then
 fi
 cd acados_repo
 git fetch
-git checkout 43ba28e95062f9ac9b48facd3b45698d57666fa3
+git checkout 79e9e3e76f2751198858adf382c97837833ad31f
 git submodule update --recursive --init
 
 # build

third_party/acados/include/acados/ocp_nlp/ocp_nlp_common.h

@@ -135,6 +135,8 @@ typedef struct ocp_nlp_dims
     int *nz;  // number of algebraic variables
     int *ns;  // number of slack variables
     int N;    // number of shooting nodes
+
+    void *raw_memory; // Pointer to allocated memory, to be used for freeing
 } ocp_nlp_dims;
 
 //
@@ -203,6 +205,9 @@ typedef struct ocp_nlp_in
     /// Pointers to constraints functions (TBC).
     void **constraints;
 
+    /// Pointer to allocated memory, to be used for freeing.
+    void *raw_memory;
+
 } ocp_nlp_in;
 
 //
@@ -235,6 +240,8 @@ typedef struct ocp_nlp_out
     double inf_norm_res;
     double total_time;
 
+    void *raw_memory; // Pointer to allocated memory, to be used for freeing
+
 } ocp_nlp_out;
 
 //

third_party/acados/include/acados/sim/sim_collocation_utils.h

@@ -43,44 +43,53 @@ extern "C" {
 
 
 
-enum Newton_type_collocation
+// enum Newton_type_collocation
+// {
+//     exact = 0,
+//     simplified_in,
+//     simplified_inis
+// };
+
+
+
+// typedef struct
+// {
+//     enum Newton_type_collocation type;
+//     double *eig;
+//     double *low_tria;
+//     bool single;
+//     bool freeze;
+
+//     double *transf1;
+//     double *transf2;
+
+//     double *transf1_T;
+//     double *transf2_T;
+// } Newton_scheme;
+
+
+typedef enum
 {
-    exact = 0,
-    simplified_in,
-    simplified_inis
-};
-
-
-
-typedef struct
-{
-    enum Newton_type_collocation type;
-    double *eig;
-    double *low_tria;
-    bool single;
-    bool freeze;
-
-    double *transf1;
-    double *transf2;
-
-    double *transf1_T;
-    double *transf2_T;
-} Newton_scheme;
-
+    GAUSS_LEGENDRE,
+    GAUSS_RADAU_IIA,
+} sim_collocation_type;
 
 
 //
-acados_size_t gauss_nodes_work_calculate_size(int ns);
+// acados_size_t gauss_legendre_nodes_work_calculate_size(int ns);
 //
-void gauss_nodes(int ns, double *nodes, void *raw_memory);
+// void gauss_legendre_nodes(int ns, double *nodes, void *raw_memory);
 //
-acados_size_t gauss_simplified_work_calculate_size(int ns);
+// acados_size_t gauss_simplified_work_calculate_size(int ns);
+// //
+// void gauss_simplified(int ns, Newton_scheme *scheme, void *work);
 //
-void gauss_simplified(int ns, Newton_scheme *scheme, void *work);
+acados_size_t butcher_tableau_work_calculate_size(int ns);
 //
-acados_size_t butcher_table_work_calculate_size(int ns);
+// void calculate_butcher_tableau_from_nodes(int ns, double *nodes, double *b, double *A, void *work);
 //
-void butcher_table(int ns, double *nodes, double *b, double *A, void *work);
+void calculate_butcher_tableau(int ns, sim_collocation_type collocation_type, double *c_vec,
+                               double *b_vec, double *A_mat, void *work);
 
 
 

third_party/acados/include/acados/sim/sim_common.h

@@ -141,12 +141,13 @@ typedef struct
     bool output_z;        // 1 -- if zn should be computed
     bool sens_algebraic;  // 1 -- if S_algebraic should be computed
     bool exact_z_output;  // 1 -- if z, S_algebraic should be computed exactly, extra Newton iterations
+    sim_collocation_type collocation_type;
 
     // for explicit integrators: newton_iter == 0 && scheme == NULL
     // && jac_reuse=false
     int newton_iter;
     bool jac_reuse;
-    Newton_scheme *scheme;
+    // Newton_scheme *scheme;
 
     // workspace
     void *work;

third_party/acados/include/acados/utils/math.h

@@ -40,7 +40,7 @@ extern "C" {
 
 #include "acados/utils/types.h"
 
-#if defined(__DSPACE__)
+#if defined(__MABX2__)
 double fmax(double a, double b);
 #endif
 

third_party/acados/include/acados/utils/timing.h

@@ -67,7 +67,7 @@ typedef struct acados_timer_
     mach_timebase_info_data_t tinfo;
 } acados_timer;
 
-#elif defined(__DSPACE__)
+#elif defined(__MABX2__)
 
 #include <brtenv.h>
 

third_party/acados/include/acados_c/ocp_nlp_interface.h

@@ -227,10 +227,8 @@ int ocp_nlp_dynamics_model_set(ocp_nlp_config *config, ocp_nlp_dims *dims, ocp_n
 /// \param field The name of the field, either nls_res_jac,
 ///     y_ref, W (others TBC)
 /// \param value Cost values.
-int ocp_nlp_cost_model_set_slice(ocp_nlp_config *config, ocp_nlp_dims *dims, ocp_nlp_in *in,
-        int start_stage, int end_stage, const char *field, void *value, int dim);
 int ocp_nlp_cost_model_set(ocp_nlp_config *config, ocp_nlp_dims *dims, ocp_nlp_in *in,
-        int start_stage, const char *field, void *value);
+        int stage, const char *field, void *value);
 
 
 /// Sets the function pointers to the constraints functions for the given stage.
@@ -241,8 +239,6 @@ int ocp_nlp_cost_model_set(ocp_nlp_config *config, ocp_nlp_dims *dims, ocp_nlp_i
 /// \param stage Stage number.
 /// \param field The name of the field, either lb, ub (others TBC)
 /// \param value Constraints function or values.
-int ocp_nlp_constraints_model_set_slice(ocp_nlp_config *config, ocp_nlp_dims *dims, ocp_nlp_in *in,
-        int start_stage, int end_stage, const char *field, void *value, int dim);
 int ocp_nlp_constraints_model_set(ocp_nlp_config *config, ocp_nlp_dims *dims,
         ocp_nlp_in *in, int stage, const char *field, void *value);
 
@@ -283,9 +279,6 @@ void ocp_nlp_out_set(ocp_nlp_config *config, ocp_nlp_dims *dims, ocp_nlp_out *ou
 void ocp_nlp_out_get(ocp_nlp_config *config, ocp_nlp_dims *dims, ocp_nlp_out *out,
         int stage, const char *field, void *value);
 
-void ocp_nlp_out_get_slice(ocp_nlp_config *config, ocp_nlp_dims *dims, ocp_nlp_out *out,
-        int start_stage, int end_stage, const char *field, void *value);
-
 //
 void ocp_nlp_get_at_stage(ocp_nlp_config *config, ocp_nlp_dims *dims, ocp_nlp_solver *solver,
         int stage, const char *field, void *value);
@@ -300,6 +293,8 @@ void ocp_nlp_constraint_dims_get_from_attr(ocp_nlp_config *config, ocp_nlp_dims
 void ocp_nlp_cost_dims_get_from_attr(ocp_nlp_config *config, ocp_nlp_dims *dims, ocp_nlp_out *out,
         int stage, const char *field, int *dims_out);
 
+void ocp_nlp_dynamics_dims_get_from_attr(ocp_nlp_config *config, ocp_nlp_dims *dims, ocp_nlp_out *out,
+        int stage, const char *field, int *dims_out);
 
 /* opts */
 
@@ -374,6 +369,8 @@ int ocp_nlp_precompute(ocp_nlp_solver *solver, ocp_nlp_in *nlp_in, ocp_nlp_out *
 /// \param nlp_out The output struct.
 void ocp_nlp_eval_cost(ocp_nlp_solver *solver, ocp_nlp_in *nlp_in, ocp_nlp_out *nlp_out);
 
+//
+void ocp_nlp_eval_residuals(ocp_nlp_solver *solver, ocp_nlp_in *nlp_in, ocp_nlp_out *nlp_out);
 
 //
 void ocp_nlp_eval_param_sens(ocp_nlp_solver *solver, char *field, int stage, int index, ocp_nlp_out *sens_nlp_out);

third_party/acados/include/acados_c/sim_interface.h

@@ -45,11 +45,11 @@ extern "C" {
 
 typedef enum
 {
-	ERK,
-	IRK,
-	GNSF,
-	LIFTED_IRK,
-	INVALID_SIM_SOLVER,
+    ERK,
+    IRK,
+    GNSF,
+    LIFTED_IRK,
+    INVALID_SIM_SOLVER,
 } sim_solver_t;
 
 

third_party/acados/include/blasfeo/include/blasfeo_block_size.h

@@ -43,7 +43,28 @@
 
 
 
-#if defined( TARGET_X64_INTEL_HASWELL )
+#if defined( TARGET_X64_INTEL_SKYLAKE_X )
+// common
+#define CACHE_LINE_SIZE 64 // data cache size: 64 bytes
+#define L1_CACHE_SIZE (32*1024) // L1 data cache size: 32 kB, 8-way
+#define L2_CACHE_SIZE (256*1024) //(1024*1024) // L2 data cache size: 1 MB ; DTLB1 64*4 kB = 256 kB
+#define LLC_CACHE_SIZE (6*1024*1024) //(8*1024*1024) // LLC cache size: 8 MB ; TLB 1536*4 kB = 6 MB
+// double
+#define D_PS 8 // panel size
+#define D_PLD 8 // 4 // GCD of panel length
+#define D_M_KERNEL 24 // max kernel size
+#define D_KC 128 //256 // 192
+#define D_NC 144 //72 //96 //72 // 120 // 512
+#define D_MC 2400 // 6000
+// single
+#define S_PS 16 // panel size
+#define S_PLD 4 // GCD of panel length TODO probably 16 when writing assebly
+#define S_M_KERNEL 32 // max kernel size
+#define S_KC 128 //256
+#define S_NC 128 //144
+#define S_MC 3000
+
+#elif defined( TARGET_X64_INTEL_HASWELL )
 // common
 #define CACHE_LINE_SIZE 64 // data cache size: 64 bytes
 #define L1_CACHE_SIZE (32*1024) // L1 data cache size: 32 kB, 8-way

third_party/acados/include/blasfeo/include/blasfeo_d_blasfeo_api.h

@@ -101,18 +101,18 @@ void blasfeo_dtrsv_ltu(int m, struct blasfeo_dmat *sA, int ai, int aj, struct bl
 void blasfeo_dtrsv_unn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
 // z <= inv( A^T ) * x, A (m)x(m) upper, transposed, not_unit
 void blasfeo_dtrsv_utn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
+// z <= A * x ; A lower triangular
+void blasfeo_dtrmv_lnn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
+// z <= A * x ; A lower triangular, unit diagonal
+void blasfeo_dtrmv_lnu(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
+// z <= A^T * x ; A lower triangular
+void blasfeo_dtrmv_ltn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
+// z <= A^T * x ; A lower triangular, unit diagonal
+void blasfeo_dtrmv_ltu(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
 // z <= beta * y + alpha * A * x ; A upper triangular
 void blasfeo_dtrmv_unn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
 // z <= A^T * x ; A upper triangular
 void blasfeo_dtrmv_utn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
-// z <= A * x ; A lower triangular
-void blasfeo_dtrmv_lnn(int m, int n, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
-// z <= A^T * x ; A lower triangular
-void blasfeo_dtrmv_ltn(int m, int n, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
-// z <= A * x ; A lower triangular, unit diagonal
-void blasfeo_dtrmv_lnu(int m, int n, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
-// z <= A^T * x ; A lower triangular, unit diagonal
-void blasfeo_dtrmv_ltu(int m, int n, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
 // z_n <= beta_n * y_n + alpha_n * A  * x_n
 // z_t <= beta_t * y_t + alpha_t * A^T * x_t
 void blasfeo_dgemv_nt(int m, int n, double alpha_n, double alpha_t, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx_n, int xi_n, struct blasfeo_dvec *sx_t, int xi_t, double beta_n, double beta_t, struct blasfeo_dvec *sy_n, int yi_n, struct blasfeo_dvec *sy_t, int yi_t, struct blasfeo_dvec *sz_n, int zi_n, struct blasfeo_dvec *sz_t, int zi_t);

third_party/acados/include/blasfeo/include/blasfeo_d_blasfeo_ref_api.h

@@ -101,23 +101,24 @@ void blasfeo_ref_dtrsv_ltu(int m, struct blasfeo_dmat *sA, int ai, int aj, struc
 void blasfeo_ref_dtrsv_unn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
 // z <= inv( A' ) * x, A (m)x(m) upper, transposed, not_unit
 void blasfeo_ref_dtrsv_utn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
+// z <= A * x ; A lower triangular
+void blasfeo_ref_dtrmv_lnn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
+// z <= A * x ; A lower triangular, unit diagonal
+void blasfeo_ref_dtrmv_lnu(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
+// z <= A' * x ; A lower triangular
+void blasfeo_ref_dtrmv_ltn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
+// z <= A' * x ; A lower triangular, unit diagonal
+void blasfeo_ref_dtrmv_ltu(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
 // z <= beta * y + alpha * A * x ; A upper triangular
 void blasfeo_ref_dtrmv_unn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
 // z <= A' * x ; A upper triangular
 void blasfeo_ref_dtrmv_utn(int m, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
-// z <= A * x ; A lower triangular
-void blasfeo_ref_dtrmv_lnn(int m, int n, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
-// z <= A' * x ; A lower triangular
-void blasfeo_ref_dtrmv_ltn(int m, int n, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
-// z <= A * x ; A lower triangular, unit diagonal
-void blasfeo_ref_dtrmv_lnu(int m, int n, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
-// z <= A' * x ; A lower triangular, unit diagonal
-void blasfeo_ref_dtrmv_ltu(int m, int n, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, struct blasfeo_dvec *sz, int zi);
 // z_n <= beta_n * y_n + alpha_n * A  * x_n
 // z_t <= beta_t * y_t + alpha_t * A' * x_t
 void blasfeo_ref_dgemv_nt(int m, int n, double alpha_n, double alpha_t, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx_n, int xi_n, struct blasfeo_dvec *sx_t, int xi_t, double beta_n, double beta_t, struct blasfeo_dvec *sy_n, int yi_n, struct blasfeo_dvec *sy_t, int yi_t, struct blasfeo_dvec *sz_n, int zi_n, struct blasfeo_dvec *sz_t, int zi_t);
 // z <= beta * y + alpha * A * x, where A is symmetric and only the lower triangular patr of A is accessed
-void blasfeo_ref_dsymv_l(int m, int n, double alpha, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, double beta, struct blasfeo_dvec *sy, int yi, struct blasfeo_dvec *sz, int zi);
+void blasfeo_ref_dsymv_l(int m, double alpha, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, double beta, struct blasfeo_dvec *sy, int yi, struct blasfeo_dvec *sz, int zi);
+void blasfeo_ref_dsymv_l_mn(int m, int n, double alpha, struct blasfeo_dmat *sA, int ai, int aj, struct blasfeo_dvec *sx, int xi, double beta, struct blasfeo_dvec *sy, int yi, struct blasfeo_dvec *sz, int zi);
 
 // diagonal
 

third_party/acados/include/blasfeo/include/blasfeo_d_kernel.h

@@ -54,6 +54,158 @@ void blasfeo_align_4096_byte(void *ptr, void **ptr_align);
 void blasfeo_align_64_byte(void *ptr, void **ptr_align);
 
 
+//
+// lib8
+//
+
+// 24x8
+void kernel_dgemm_nt_24x8_lib8(int k, double *alpha, double *A, int sda, double *B, double *beta, double *C, int sdc, double *D, int sdd); //
+void kernel_dgemm_nt_24x8_vs_lib8(int k, double *alpha, double *A, int sda, double *B, double *beta, double *C, int sdc, double *D, int sdd, int m1, int n1); //
+void kernel_dtrsm_nt_rl_inv_24x8_lib8(int k, double *A, int sda, double *B, double *beta, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E); //
+void kernel_dpotrf_nt_l_24x8_lib8(int k, double *A, int sda, double *B, double *C, int sdc, double *D, int sdd, double *inv_diag_D);
+void kernel_dpotrf_nt_l_24x8_vs_lib8(int k, double *A, int sda, double *B, double *C, int sdc, double *D, int sdd, double *inv_diag_D, int m1, int n1);
+void kernel_dtrsm_nt_rl_inv_24x8_vs_lib8(int k, double *A, int sda, double *B, double *beta, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E, int m1, int n1); //
+void kernel_dgemm_dtrsm_nt_rl_inv_24x8_lib8(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E);
+void kernel_dgemm_dtrsm_nt_rl_inv_24x8_vs_lib8(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E, int m1, int n1);
+void kernel_dsyrk_dpotrf_nt_l_24x8_lib8(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *inv_diag_D);
+void kernel_dsyrk_dpotrf_nt_l_24x8_vs_lib8(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *inv_diag_D, int m1, int n1);
+void kernel_dlarfb8_rn_24_lib8(int kmax, double *pV, double *pT, double *pD, int sdd);
+// 16x8
+void kernel_dgemm_nt_16x8_lib8(int k, double *alpha, double *A, int sda, double *B, double *beta, double *C, int sdc, double *D, int sdd); //
+void kernel_dgemm_nt_16x8_vs_lib8(int k, double *alpha, double *A, int sda, double *B, double *beta, double *C, int sdc, double *D, int sdd, int m1, int n1); //
+void kernel_dgemm_nt_16x8_gen_lib8(int k, double *alpha, double *A, int sda, double *B, double *beta, int offC, double *C, int sdc, int offD, double *D, int sdd, int m0, int m1, int n0, int n1); //
+void kernel_dgemm_nn_16x8_lib8(int k, double *alpha, double *A, int sda, int offB, double *B, int sdb, double *beta, double *C, int sdc, double *D, int sdd); //
+void kernel_dgemm_nn_16x8_vs_lib8(int k, double *alpha, double *A, int sda, int offB, double *B, int sdb, double *beta, double *C, int sdc, double *D, int sdd, int m1, int n1); //
+void kernel_dgemm_nn_16x8_gen_lib8(int k, double *alpha, double *A, int sda, int offB, double *B, int sdb, double *beta, int offC, double *C, int sdc, int offD, double *D, int sdd, int m0, int m1, int n0, int n1); //
+void kernel_dsyrk_nt_l_16x8_lib8(int k, double *alpha, double *A, int sda, double *B, double *beta, double *C, int sdc, double *D, int sdd); //
+void kernel_dsyrk_nt_l_16x8_vs_lib8(int k, double *alpha, double *A, int sda, double *B, double *beta, double *C, int sdc, double *D, int sdd, int m1, int n1); //
+void kernel_dsyrk_nt_l_16x8_gen_lib8(int k, double *alpha, double *A, int sda, double *B, double *beta, int offC, double *C, int sdc, int offD, double *D, int sdd, int m0, int m1, int n0, int n1); //
+void kernel_dtrmm_nn_rl_16x8_lib8(int k, double *alpha, double *A, int sda, int offsetB, double *B, int sdb, double *D, int sdd);
+void kernel_dtrmm_nn_rl_16x8_vs_lib8(int k, double *alpha, double *A, int sda, int offsetB, double *B, int sdb, double *D, int sdd, int m1, int n1);
+void kernel_dtrmm_nn_rl_16x8_gen_lib8(int k, double *alpha, double *A, int sda, int offsetB, double *B, int sdb, int offD, double *D, int sdd, int m0, int m1, int n0, int n1);
+void kernel_dtrsm_nt_rl_inv_16x8_lib8(int k, double *A, int sda, double *B, double *beta, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E); //
+void kernel_dtrsm_nt_rl_inv_16x8_vs_lib8(int k, double *A, int sda, double *B, double *beta, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E, int m1, int n1); //
+void kernel_dpotrf_nt_l_16x8_lib8(int k, double *A, int sda, double *B, double *C, int sdc, double *D, int sdd, double *inv_diag_D);
+void kernel_dpotrf_nt_l_16x8_vs_lib8(int k, double *A, int sda, double *B, double *C, int sdc, double *D, int sdd, double *inv_diag_D, int m1, int n1);
+void kernel_dgemm_dtrsm_nt_rl_inv_16x8_lib8(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E);
+void kernel_dgemm_dtrsm_nt_rl_inv_16x8_vs_lib8(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E, int m1, int n1);
+void kernel_dsyrk_dpotrf_nt_l_16x8_lib8(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *inv_diag_D);
+void kernel_dsyrk_dpotrf_nt_l_16x8_vs_lib8(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *inv_diag_D, int m1, int n1);
+void kernel_dlarfb8_rn_16_lib8(int kmax, double *pV, double *pT, double *pD, int sdd);
+void kernel_dlarfb8_rn_la_16_lib8(int n1, double *pVA, double *pT, double *pD, int sdd, double *pA, int sda);
+void kernel_dlarfb8_rn_lla_16_lib8(int n0, int n1, double *pVL, double *pVA, double *pT, double *pD, int sdd, double *pL, int sdl, double *pA, int sda);
+// 8x16
+void kernel_dgemm_tt_8x16_lib8(int k, double *alpha, int offA, double *A, int sda, double *B, int sdb, double *beta, double *C, double *D); //
+void kernel_dgemm_tt_8x16_vs_lib8(int k, double *alpha, int offA, double *A, int sda, double *B, int sdb, double *beta, double *C, double *D, int m1, int n1); //
+void kernel_dgemm_tt_8x16_gen_lib8(int k, double *alpha, int offA, double *A, int sda, double *B, int sdb, double *beta, int offC, double *C, int sdc, int offD, double *D, int sdd, int m0, int m1, int n0, int n1); //
+void kernel_dgemm_nt_8x16_lib8(int k, double *alpha, double *A, double *B, int sdb, double *beta, double *C, double *D); //
+void kernel_dgemm_nt_8x16_vs_lib8(int k, double *alpha, double *A, double *B, int sdb, double *beta, double *C, double *D, int m1, int n1); //
+// 8x8
+void kernel_dgemm_nt_8x8_lib8(int k, double *alpha, double *A, double *B, double *beta, double *C, double *D); //
+void kernel_dgemm_nt_8x8_vs_lib8(int k, double *alpha, double *A, double *B, double *beta, double *C, double *D, int m1, int n1); //
+void kernel_dgemm_nt_8x8_gen_lib8(int k, double *alpha, double *A, double *B, double *beta, int offC, double *C, int sdc, int offD, double *D, int sdd, int m0, int m1, int n0, int n1); //
+void kernel_dgemm_nn_8x8_lib8(int k, double *alpha, double *A, int offB, double *B, int sdb, double *beta, double *C, double *D); //
+void kernel_dgemm_nn_8x8_vs_lib8(int k, double *alpha, double *A, int offB, double *B, int sdb, double *beta, double *C, double *D, int m1, int n1); //
+void kernel_dgemm_nn_8x8_gen_lib8(int k, double *alpha, double *A, int offB, double *B, int sdb, double *beta, int offC, double *C, int sdc, int offD, double *D, int sdd, int m0, int m1, int n0, int n1); //
+void kernel_dgemm_tt_8x8_lib8(int k, double *alpha, int offA, double *A, int sda, double *B, double *beta, double *C, double *D); //
+void kernel_dgemm_tt_8x8_vs_lib8(int k, double *alpha, int offA, double *A, int sda, double *B, double *beta, double *C, double *D, int m1, int n1); //
+void kernel_dgemm_tt_8x8_gen_lib8(int k, double *alpha, int offA, double *A, int sda, double *B, double *beta, int offc, double *C, int sdc, int offD, double *D, int sdd, int m0, int m1, int n0, int n1); //
+void kernel_dsyrk_nt_l_8x8_lib8(int k, double *alpha, double *A, double *B, double *beta, double *C, double *D); //
+void kernel_dsyrk_nt_l_8x8_vs_lib8(int k, double *alpha, double *A, double *B, double *beta, double *C, double *D, int m1, int n1); //
+void kernel_dsyrk_nt_l_8x8_gen_lib8(int k, double *alpha, double *A, double *B, double *beta, int offC, double *C, int sdc, int offD, double *D, int sdd, int m0, int m1, int n0, int n1); //
+void kernel_dtrmm_nn_rl_8x8_lib8(int k, double *alpha, double *A, int offsetB, double *B, int sdb, double *D);
+void kernel_dtrmm_nn_rl_8x8_vs_lib8(int k, double *alpha, double *A, int offsetB, double *B, int sdb, double *D, int m1, int n1);
+void kernel_dtrmm_nn_rl_8x8_gen_lib8(int k, double *alpha, double *A, int offsetB, double *B, int sdb, int offD, double *D, int sdd, int m0, int m1, int n0, int n1);
+void kernel_dtrsm_nt_rl_inv_8x8_lib8(int k, double *A, double *B, double *beta, double *C, double *D, double *E, double *inv_diag_E);
+void kernel_dtrsm_nt_rl_inv_8x8_vs_lib8(int k, double *A, double *B, double *beta, double *C, double *D, double *E, double *inv_diag_E, int m1, int n1);
+void kernel_dpotrf_nt_l_8x8_lib8(int k, double *A, double *B, double *C, double *D, double *inv_diag_D);
+void kernel_dpotrf_nt_l_8x8_vs_lib8(int k, double *A, double *B, double *C, double *D, double *inv_diag_D, int m1, int n1);
+void kernel_dgemm_dtrsm_nt_rl_inv_8x8_lib8(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *E, double *inv_diag_E);
+void kernel_dgemm_dtrsm_nt_rl_inv_8x8_vs_lib8(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *E, double *inv_diag_E, int m1, int n1);
+void kernel_dsyrk_dpotrf_nt_l_8x8_lib8(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *inv_diag_D);
+void kernel_dsyrk_dpotrf_nt_l_8x8_vs_lib8(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *inv_diag_D, int m1, int n1);
+void kernel_dgelqf_vs_lib8(int m, int n, int k, int offD, double *pD, int sdd, double *dD);
+void kernel_dgelqf_pd_vs_lib8(int m, int n, int k, int offD, double *pD, int sdd, double *dD);
+void kernel_dgelqf_8_lib8(int kmax, double *pD, double *dD);
+void kernel_dgelqf_pd_8_lib8(int kmax, double *pD, double *dD);
+void kernel_dlarft_8_lib8(int kmax, double *pD, double *dD, double *pT);
+void kernel_dlarfb8_rn_8_lib8(int kmax, double *pV, double *pT, double *pD);
+void kernel_dlarfb8_rn_8_vs_lib8(int kmax, double *pV, double *pT, double *pD, int m1);
+void kernel_dlarfb8_rn_1_lib8(int kmax, double *pV, double *pT, double *pD);
+void kernel_dgelqf_dlarft8_8_lib8(int kmax, double *pD, double *dD, double *pT);
+void kernel_dgelqf_pd_dlarft8_8_lib8(int kmax, double *pD, double *dD, double *pT);
+void kernel_dgelqf_pd_la_vs_lib8(int m, int n1, int k, int offD, double *pD, int sdd, double *dD, int offA, double *pA, int sda);
+void kernel_dgelqf_pd_la_dlarft8_8_lib8(int kmax, double *pD, double *dD, double *pA, double *pT);
+void kernel_dlarft_la_8_lib8(int n1, double *dD, double *pA, double *pT);
+void kernel_dlarfb8_rn_la_8_lib8(int n1, double *pVA, double *pT, double *pD, double *pA);
+void kernel_dlarfb8_rn_la_8_vs_lib8(int n1, double *pVA, double *pT, double *pD, double *pA, int m1);
+void kernel_dlarfb8_rn_la_1_lib8(int n1, double *pVA, double *pT, double *pD, double *pA);
+void kernel_dgelqf_pd_lla_vs_lib8(int m, int n0, int n1, int k, int offD, double *pD, int sdd, double *dD, int offL, double *pL, int sdl, int offA, double *pA, int sda);
+void kernel_dgelqf_pd_lla_dlarft8_8_lib8(int n0, int n1, double *pD, double *dD, double *pL, double *pA, double *pT);
+void kernel_dlarft_lla_8_lib8(int n0, int n1, double *dD, double *pL, double *pA, double *pT);
+void kernel_dlarfb8_rn_lla_8_lib8(int n0, int n1, double *pVL, double *pVA, double *pT, double *pD, double *pL, double *pA);
+void kernel_dlarfb8_rn_lla_8_vs_lib8(int n0, int n1, double *pVL, double *pVA, double *pT, double *pD, double *pL, double *pA, int m1);
+void kernel_dlarfb8_rn_lla_1_lib8(int n0, int n1, double *pVL, double *pVA, double *pT, double *pD, double *pL, double *pA);
+
+// panel copy / pack
+// 24
+void kernel_dpack_nn_24_lib8(int kmax, double *A, int lda, double *C, int sdc);
+void kernel_dpack_nn_24_vs_lib8(int kmax, double *A, int lda, double *C, int sdc, int m1);
+// 16
+void kernel_dpacp_nn_16_lib8(int kmax, int offsetA, double *A, int sda, double *B, int sdb);
+void kernel_dpacp_nn_16_vs_lib8(int kmax, int offsetA, double *A, int sda, double *B, int sdb, int m1);
+void kernel_dpack_nn_16_lib8(int kmax, double *A, int lda, double *C, int sdc);
+void kernel_dpack_nn_16_vs_lib8(int kmax, double *A, int lda, double *C, int sdc, int m1);
+// 8
+void kernel_dpacp_nn_8_lib8(int kmax, int offsetA, double *A, int sda, double *B);
+void kernel_dpacp_nn_8_vs_lib8(int kmax, int offsetA, double *A, int sda, double *B, int m1);
+void kernel_dpacp_tn_8_lib8(int kmax, int offsetA, double *A, int sda, double *B);
+void kernel_dpacp_tn_8_vs_lib8(int kmax, int offsetA, double *A, int sda, double *B, int m1);
+void kernel_dpacp_l_nn_8_lib8(int kmax, int offsetA, double *A, int sda, double *B);
+void kernel_dpacp_l_nn_8_vs_lib8(int kmax, int offsetA, double *A, int sda, double *B, int m1);
+void kernel_dpacp_l_tn_8_lib8(int kmax, int offsetA, double *A, int sda, double *B);
+void kernel_dpacp_l_tn_8_vs_lib8(int kmax, int offsetA, double *A, int sda, double *B, int m1);
+void kernel_dpaad_nn_8_lib8(int kmax, double *alpha, int offsetA, double *A, int sda, double *B);
+void kernel_dpaad_nn_8_vs_lib8(int kmax, double *alpha, int offsetA, double *A, int sda, double *B, int m1);
+void kernel_dpack_nn_8_lib8(int kmax, double *A, int lda, double *C);
+void kernel_dpack_nn_8_vs_lib8(int kmax, double *A, int lda, double *C, int m1);
+void kernel_dpack_tn_8_lib8(int kmax, double *A, int lda, double *C);
+void kernel_dpack_tn_8_vs_lib8(int kmax, double *A, int lda, double *C, int m1);
+// 4
+void kernel_dpack_tt_4_lib8(int kmax, double *A, int lda, double *C, int sdc); // TODO offsetC 
+void kernel_dpack_tt_4_vs_lib8(int kmax, double *A, int lda, double *C, int sdc, int m1); // TODO offsetC 
+
+// level 2 BLAS
+// 16
+void kernel_dgemv_n_16_lib8(int k, double *alpha, double *A, int sda, double *x, double *beta, double *y, double *z);
+// 8
+void kernel_dgemv_n_8_lib8(int k, double *alpha, double *A, double *x, double *beta, double *y, double *z);
+void kernel_dgemv_n_8_vs_lib8(int k, double *alpha, double *A, double *x, double *beta, double *y, double *z, int m1);
+//void kernel_dgemv_n_8_gen_lib8(int k, double *alpha, double *A, double *x, double *beta, double *y, double *z, int m0, int m1);
+void kernel_dgemv_n_8_gen_lib8(int k, double *alpha, int offsetA, double *A, double *x, double *beta, double *y, double *z, int m1);
+void kernel_dgemv_t_8_lib8(int k, double *alpha, int offsetA, double *A, int sda, double *x, double *beta, double *y, double *z);
+void kernel_dgemv_t_8_vs_lib8(int k, double *alpha, int offsetA, double *A, int sda, double *x, double *beta, double *y, double *z, int n1);
+void kernel_dgemv_nt_8_lib8(int kmax, double *alpha_n, double *alpha_t, int offsetA, double *A, int sda, double *x_n, double *x_t, double *beta_t, double *y_t, double *z_n, double *z_t);
+void kernel_dgemv_nt_8_vs_lib8(int kmax, double *alpha_n, double *alpha_t, int offsetA, double *A, int sda, double *x_n, double *x_t, double *beta_t, double *y_t, double *z_n, double *z_t, int n1);
+void kernel_dsymv_l_8_lib8(int kmax, double *alpha, double *A, int sda, double *x, double *z);
+void kernel_dsymv_l_8_vs_lib8(int kmax, double *alpha, double *A, int sda, double *x, double *z, int n1);
+void kernel_dsymv_l_8_gen_lib8(int kmax, double *alpha, int offsetA, double *A, int sda, double *x, double *z, int n1);
+void kernel_dtrmv_n_ln_8_lib8(int k, double *A, double *x, double *z);
+void kernel_dtrmv_n_ln_8_vs_lib8(int k, double *A, double *x, double *z, int m1);
+void kernel_dtrmv_n_ln_8_gen_lib8(int k, int offsetA, double *A, double *x, double *z, int m1);
+void kernel_dtrmv_t_ln_8_lib8(int k, double *A, int sda, double *x, double *z);
+void kernel_dtrmv_t_ln_8_vs_lib8(int k, double *A, int sda, double *x, double *z, int n1);
+void kernel_dtrmv_t_ln_8_gen_lib8(int k, int offsetA, double *A, int sda, double *x, double *z, int n1);
+void kernel_dtrsv_n_l_inv_8_lib8(int k, double *A, double *inv_diag_A, double *x, double *z);
+void kernel_dtrsv_n_l_inv_8_vs_lib8(int k, double *A, double *inv_diag_A, double *x, double *z, int m1, int n1);
+void kernel_dtrsv_t_l_inv_8_lib8(int k, double *A, int sda, double *inv_diag_A, double *x, double *z);
+void kernel_dtrsv_t_l_inv_8_vs_lib8(int k, double *A, int sda, double *inv_diag_A, double *x, double *z, int m1, int n1);
+
+
+
+//
+// lib4
+//
 
 // level 2 BLAS
 // 12
@@ -413,10 +565,10 @@ void kernel_drowsw_lib4(int kmax, double *pA, double *pC);
 
 // merged routines
 // 12x4
-void kernel_dgemm_dtrsm_nt_rl_inv_12x4_vs_lib4(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E, int km, int kn);
 void kernel_dgemm_dtrsm_nt_rl_inv_12x4_lib4(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E);
-void kernel_dsyrk_dpotrf_nt_l_12x4_vs_lib4(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *inv_diag_D, int km, int kn);
+void kernel_dgemm_dtrsm_nt_rl_inv_12x4_vs_lib4(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E, int km, int kn);
 void kernel_dsyrk_dpotrf_nt_l_12x4_lib4(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *inv_diag_D);
+void kernel_dsyrk_dpotrf_nt_l_12x4_vs_lib4(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *inv_diag_D, int km, int kn);
 // 4x12
 void kernel_dgemm_dtrsm_nt_rl_inv_4x12_vs_lib4(int kp, double *Ap, double *Bp, int sdbp, int km_, double *Am, double *Bm, int sdbm, double *C, double *D, double *E, int sde, double *inv_diag_E, int km, int kn);
 // 8x8
@@ -425,8 +577,8 @@ void kernel_dsyrk_dpotrf_nt_l_8x8_vs_lib4(int kp, double *Ap, int sdap, double *
 void kernel_dgemm_dtrsm_nt_rl_inv_8x8l_vs_lib4(int kp, double *Ap, int sdap, double *Bp, int sdb, int km_, double *Am, int sdam, double *Bm, int sdbm, double *C, int sdc, double *D, int sdd, double *E, int sde, double *inv_diag_E, int km, int kn);
 void kernel_dgemm_dtrsm_nt_rl_inv_8x8u_vs_lib4(int kp, double *Ap, int sdap, double *Bp, int sdb, int km_, double *Am, int sdam, double *Bm, int sdbm, double *C, int sdc, double *D, int sdd, double *E, int sde, double *inv_diag_E, int km, int kn);
 // 8x4
-void kernel_dgemm_dtrsm_nt_rl_inv_8x4_vs_lib4(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E, int km, int kn);
 void kernel_dgemm_dtrsm_nt_rl_inv_8x4_lib4(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E);
+void kernel_dgemm_dtrsm_nt_rl_inv_8x4_vs_lib4(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *E, double *inv_diag_E, int km, int kn);
 void kernel_dsyrk_dpotrf_nt_l_8x4_lib4(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *inv_diag_D);
 void kernel_dsyrk_dpotrf_nt_l_8x4_vs_lib4(int kp, double *Ap, int sdap, double *Bp, int km_, double *Am, int sdam, double *Bm, double *C, int sdc, double *D, int sdd, double *inv_diag_D, int km, int kn);
 // 4x8
@@ -434,16 +586,16 @@ void kernel_dgemm_dtrsm_nt_rl_inv_4x8_vs_lib4(int kp, double *Ap, double *Bp, in
 // 4x4
 void kernel_dgemm_dtrsm_nt_rl_inv_4x4_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *E, double *inv_diag_E);
 void kernel_dgemm_dtrsm_nt_rl_inv_4x4_vs_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *E, double *inv_diag_E, int km, int kn);
-void kernel_dsyrk_dpotrf_nt_l_4x4_vs_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *inv_diag_D, int km, int kn);
 void kernel_dsyrk_dpotrf_nt_l_4x4_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *inv_diag_D);
+void kernel_dsyrk_dpotrf_nt_l_4x4_vs_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *inv_diag_D, int km, int kn);
 // 4x2
 void kernel_dgemm_dtrsm_nt_rl_inv_4x2_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *E, double *inv_diag_E);
 void kernel_dgemm_dtrsm_nt_rl_inv_4x2_vs_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *E, double *inv_diag_E, int km, int kn);
-void kernel_dsyrk_dpotrf_nt_l_4x2_vs_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *inv_diag_D, int km, int kn);
 void kernel_dsyrk_dpotrf_nt_l_4x2_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *inv_diag_D);
+void kernel_dsyrk_dpotrf_nt_l_4x2_vs_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *inv_diag_D, int km, int kn);
 // 2x2
-void kernel_dsyrk_dpotrf_nt_l_2x2_vs_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *inv_diag_D, int km, int kn);
 void kernel_dsyrk_dpotrf_nt_l_2x2_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *inv_diag_D);
+void kernel_dsyrk_dpotrf_nt_l_2x2_vs_lib4(int kp, double *Ap, double *Bp, int km_, double *Am, double *Bm, double *C, double *D, double *inv_diag_D, int km, int kn);
 
 /*
  *
@@ -1034,6 +1186,53 @@ void kernel_dgemm_nt_8xn_p0_lib44cc(int n, int k, double *alpha, double *A, int
 
 
 
+// A, B panel-major bs=8; C, D column-major
+// 24x8
+void kernel_dgemm_nt_24x8_lib88cc(int kmax, double *alpha, double *A, int sda, double *B, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nt_24x8_vs_lib88cc(int kmax, double *alpha, double *A, int sda, double *B, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+// 16x8
+void kernel_dgemm_nt_16x8_lib88cc(int kmax, double *alpha, double *A, int sda, double *B, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nt_16x8_vs_lib88cc(int kmax, double *alpha, double *A, int sda, double *B, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+// 8x8
+void kernel_dgemm_nt_8x8_lib88cc(int kmax, double *alpha, double *A, double *B, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nt_8x8_vs_lib88cc(int kmax, double *alpha, double *A, double *B, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+
+// A, panel-major bs=8; B, C, D column-major
+// 24x8
+void kernel_dgemm_nt_24x8_lib8ccc(int kmax, double *alpha, double *A, int sda, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nt_24x8_vs_lib8ccc(int kmax, double *alpha, double *A, int sda, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+void kernel_dgemm_nn_24x8_lib8ccc(int kmax, double *alpha, double *A, int sda, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nn_24x8_vs_lib8ccc(int kmax, double *alpha, double *A, int sda, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+// 16x8
+void kernel_dgemm_nt_16x8_lib8ccc(int kmax, double *alpha, double *A, int sda, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nt_16x8_vs_lib8ccc(int kmax, double *alpha, double *A, int sda, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+void kernel_dgemm_nn_16x8_lib8ccc(int kmax, double *alpha, double *A, int sda, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nn_16x8_vs_lib8ccc(int kmax, double *alpha, double *A, int sda, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+// 8x8
+void kernel_dgemm_nt_8x8_lib8ccc(int kmax, double *alpha, double *A, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nt_8x8_vs_lib8ccc(int kmax, double *alpha, double *A, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+void kernel_dgemm_nn_8x8_lib8ccc(int kmax, double *alpha, double *A, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nn_8x8_vs_lib8ccc(int kmax, double *alpha, double *A, double *B, int ldb, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+
+// B, panel-major bs=8; A, C, D column-major
+// 8x24
+void kernel_dgemm_nt_8x24_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, int sdb, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nt_8x24_vs_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, int sdb, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+void kernel_dgemm_tt_8x24_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, int sdb, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_tt_8x24_vs_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, int sdb, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+// 8x16
+void kernel_dgemm_nt_8x16_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, int sdb, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nt_8x16_vs_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, int sdb, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+void kernel_dgemm_tt_8x16_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, int sdb, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_tt_8x16_vs_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, int sdb, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+// 8x8
+void kernel_dgemm_nt_8x8_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_nt_8x8_vs_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+void kernel_dgemm_tt_8x8_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, double *beta, double *C, int ldc, double *D, int ldd);
+void kernel_dgemm_tt_8x8_vs_libc8cc(int kmax, double *alpha, double *A, int lda, double *B, double *beta, double *C, int ldc, double *D, int ldd, int m1, int n1);
+
+
+
 // aux
 void kernel_dvecld_inc1(int kmax, double *x);
 void kernel_dveccp_inc1(int kmax, double *x, double *y);

third_party/acados/include/blasfeo/include/blasfeo_s_blasfeo_api.h

@@ -97,14 +97,18 @@ void blasfeo_strsv_ltu(int m, struct blasfeo_smat *sA, int ai, int aj, struct bl
 void blasfeo_strsv_unn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
 // z <= inv( A' ) * x, A (m)x(m) upper, transposed, not_unit
 void blasfeo_strsv_utn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
+// z <= A * x ; A lower triangular
+void blasfeo_strmv_lnn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
+// z <= A * x ; A lower triangular, unit diagonal
+void blasfeo_strmv_lnu(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
+// z <= A' * x ; A lower triangular
+void blasfeo_strmv_ltn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
+// z <= A' * x ; A lower triangular, unit diagonal
+void blasfeo_strmv_ltu(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
 // z <= beta * y + alpha * A * x ; A upper triangular
 void blasfeo_strmv_unn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
 // z <= A' * x ; A upper triangular
 void blasfeo_strmv_utn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
-// z <= A * x ; A lower triangular
-void blasfeo_strmv_lnn(int m, int n, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
-// z <= A' * x ; A lower triangular
-void blasfeo_strmv_ltn(int m, int n, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
 // z_n <= beta_n * y_n + alpha_n * A  * x_n
 // z_t <= beta_t * y_t + alpha_t * A' * x_t
 void blasfeo_sgemv_nt(int m, int n, float alpha_n, float alpha_t, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx_n, int xi_n, struct blasfeo_svec *sx_t, int xi_t, float beta_n, float beta_t, struct blasfeo_svec *sy_n, int yi_n, struct blasfeo_svec *sy_t, int yi_t, struct blasfeo_svec *sz_n, int zi_n, struct blasfeo_svec *sz_t, int zi_t);

third_party/acados/include/blasfeo/include/blasfeo_s_blasfeo_ref_api.h

@@ -97,19 +97,24 @@ void blasfeo_ref_strsv_ltu(int m, struct blasfeo_smat *sA, int ai, int aj, struc
 void blasfeo_ref_strsv_unn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
 // z <= inv( A' ) * x, A (m)x(m) upper, transposed, not_unit
 void blasfeo_ref_strsv_utn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
+// z <= A * x ; A lower triangular
+void blasfeo_ref_strmv_lnn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
+// z <= A * x ; A lower triangular, unit diagonal
+void blasfeo_ref_strmv_lnu(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
+// z <= A' * x ; A lower triangular
+void blasfeo_ref_strmv_ltn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
+// z <= A' * x ; A lower triangular, unit diagonal
+void blasfeo_ref_strmv_ltu(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
 // z <= beta * y + alpha * A * x ; A upper triangular
 void blasfeo_ref_strmv_unn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
 // z <= A' * x ; A upper triangular
 void blasfeo_ref_strmv_utn(int m, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
-// z <= A * x ; A lower triangular
-void blasfeo_ref_strmv_lnn(int m, int n, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
-// z <= A' * x ; A lower triangular
-void blasfeo_ref_strmv_ltn(int m, int n, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, struct blasfeo_svec *sz, int zi);
 // z_n <= beta_n * y_n + alpha_n * A  * x_n
 // z_t <= beta_t * y_t + alpha_t * A' * x_t
 void blasfeo_ref_sgemv_nt(int m, int n, float alpha_n, float alpha_t, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx_n, int xi_n, struct blasfeo_svec *sx_t, int xi_t, float beta_n, float beta_t, struct blasfeo_svec *sy_n, int yi_n, struct blasfeo_svec *sy_t, int yi_t, struct blasfeo_svec *sz_n, int zi_n, struct blasfeo_svec *sz_t, int zi_t);
 // z <= beta * y + alpha * A * x, where A is symmetric and only the lower triangular patr of A is accessed
-void blasfeo_ref_ssymv_l(int m, int n, float alpha, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, float beta, struct blasfeo_svec *sy, int yi, struct blasfeo_svec *sz, int zi);
+void blasfeo_ref_ssymv_l(int m, float alpha, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, float beta, struct blasfeo_svec *sy, int yi, struct blasfeo_svec *sz, int zi);
+void blasfeo_ref_ssymv_l_mn(int m, int n, float alpha, struct blasfeo_smat *sA, int ai, int aj, struct blasfeo_svec *sx, int xi, float beta, struct blasfeo_svec *sy, int yi, struct blasfeo_svec *sz, int zi);
 
 // diagonal
 

third_party/acados/include/blasfeo/include/blasfeo_s_kernel.h

@@ -550,6 +550,16 @@ void kernel_spotrf_nt_l_4x4_lib44cc(int kmax, float *A, float *B, float *C, int
 void kernel_spotrf_nt_l_4x4_vs_lib44cc(int kmax, float *A, float *B, float *C, int ldc, float *D, int ldd, float *dD, int m1, int n1);
 
 // B panel-major bs=8; A, C, D column-major
+// 4x24
+void kernel_sgemm_nt_4x24_libc8cc(int kmax, float *alpha, float *A, int lda, float *B, int sdb, float *beta, float *C, int ldc, float *D, int ldd);
+void kernel_sgemm_nt_4x24_vs_libc8cc(int kmax, float *alpha, float *A, int lda, float *B, int sdb, float *beta, float *C, int ldc, float *D, int ldd, int m1, int n1);
+void kernel_sgemm_tt_4x24_libc8cc(int kmax, float *alpha, float *A, int lda, float *B, int sdb, float *beta, float *C, int ldc, float *D, int ldd);
+void kernel_sgemm_tt_4x24_vs_libc8cc(int kmax, float *alpha, float *A, int lda, float *B, int sdb, float *beta, float *C, int ldc, float *D, int ldd, int m1, int n1);
+// 4x16
+void kernel_sgemm_nt_4x16_libc8cc(int kmax, float *alpha, float *A, int lda, float *B, int sdb, float *beta, float *C, int ldc, float *D, int ldd);
+void kernel_sgemm_nt_4x16_vs_libc8cc(int kmax, float *alpha, float *A, int lda, float *B, int sdb, float *beta, float *C, int ldc, float *D, int ldd, int m1, int n1);
+void kernel_sgemm_tt_4x16_libc8cc(int kmax, float *alpha, float *A, int lda, float *B, int sdb, float *beta, float *C, int ldc, float *D, int ldd);
+void kernel_sgemm_tt_4x16_vs_libc8cc(int kmax, float *alpha, float *A, int lda, float *B, int sdb, float *beta, float *C, int ldc, float *D, int ldd, int m1, int n1);
 // 8x8
 void kernel_sgemm_nt_8x8_libc8cc(int kmax, float *alpha, float *A, int lda, float *B, float *beta, float *C, int ldc, float *D, int ldd);
 void kernel_sgemm_nt_8x8_vs_libc8cc(int kmax, float *alpha, float *A, int lda, float *B, float *beta, float *C, int ldc, float *D, int ldd, int m1, int n1);