Fix AMO stall falling through when write data phase should proceed to error phase
This commit is contained in:
parent
8bfc089660
commit
29c5c8ca7f
|
@ -301,11 +301,13 @@ wire x_stall_on_exclusive_overlap = |EXTENSION_A && (
|
|||
// 0-3 are read/write address/data phases. Phase 4 is error, due to HRESP or
|
||||
// due to low HEXOKAY response to read.
|
||||
|
||||
// Also need to clear AMO if it follows an excepting instruction.
|
||||
// Also need to clear AMO if it follows an excepting instruction. Note we
|
||||
// still stall on phase 3 when hready is high if hresp is also high, since we
|
||||
// then proceed to phase 4 for the error response.
|
||||
|
||||
reg [2:0] x_amo_phase;
|
||||
wire x_stall_on_amo = |EXTENSION_A && d_memop_is_amo && !m_trap_enter_soon && (
|
||||
x_amo_phase < 3'h3 || (x_amo_phase == 3'h3 && !bus_dph_ready_d)
|
||||
x_amo_phase < 3'h3 || (x_amo_phase == 3'h3 && (!bus_dph_ready_d || bus_dph_err_d))
|
||||
);
|
||||
|
||||
// Read-after-write hazard detection (e.g. load-use)
|
||||
|
|
Loading…
Reference in New Issue