Espresso Tests Failing

[stffrdhrn]

When running the or1k-tests with ESPRESSO pipeline I am seeing many tests fail.

< shorne@lianli ~/work/openrisc/or1k-tests/native > ./runtests.sh 
Running or1k-alignillegalinsn                               TIME OUT
Running or1k-backtoback_jmp                                 TIME OUT
Running or1k-basic                                          TIME OUT
Running or1k-cmov                                           PASS
Running or1k-cy                                             PASS
Running or1k-dsx                                            TIME OUT
Running or1k-dsxinsn                                        TIME OUT
Running or1k-ext                                            PASS
Running or1k-ffl1                                           TIME OUT
Running or1k-icache                                         TIME OUT
Running or1k-illegalinsn                                    TIME OUT
Running or1k-illegalinsndelayslot                           TIME OUT
Running or1k-insnfetchalign                                 TIME OUT
Running or1k-insnfetcherror                                 TIME OUT
Running or1k-intloop                                        TIME OUT
Running or1k-intmulticycle                                  TIME OUT
Running or1k-intsyscall                                     TIME OUT
Running or1k-inttickloop                                    TIME OUT
Running or1k-jmp                                            FAIL
Running or1k-jr                                             FAIL
Running or1k-lsu                                            TIME OUT
Running or1k-lsualign                                       TIME OUT
Running or1k-lsualigndelayslot                              TIME OUT
Running or1k-lsuerror                                       TIME OUT
Running or1k-lsuerrordelayslot                              TIME OUT
Running or1k-lwjr                                           TIME OUT
Running or1k-msync                                          TIME OUT
Running or1k-mul-basic                                      FAIL
Running or1k-ov                                             FAIL
Running or1k-regjmp                                         FAIL
Running or1k-rfe                                            TIME OUT
Running or1k-sf                                             PASS
Running or1k-sfbf                                           TIME OUT
Running or1k-shiftopts                                      PASS
Running or1k-shortbranch                                    FAIL
Running or1k-shortjump                                      FAIL
Running or1k-systemcall                                     TIME OUT
Running or1k-tickloop                                       TIME OUT
Running or1k-tickrfforward                                  TIME OUT
Running or1k-ticksyscall                                    TIME OUT
Running or1k-timer                                          TIME OUT
Running or1k-trap                                           TIME OUT
Running or1k-trapdelayslot                                  TIME OUT
Results                                                     Total:  43  FAIL:  38

[stffrdhrn]

Looking at the first test I found the decode_op_rfe_o is not connected to anything, meaning we don't seem to support l.rfe in the espresso pipeline.

[edit] ... Nevermind, the espresso pipeline tries to handle RFE insize the mor1kx_ctrl_espresso module, bypassing the decoder. But still something is wrong and its not working as expected.

[stffrdhrn]

This patch fixes the first issue that RFE doesn't return. But there seems to be other issues with jump and delay slots.

diff --git a/rtl/verilog/mor1kx_ctrl_espresso.v b/rtl/verilog/mor1kx_ctrl_espresso.v
index 7eb3dcf..36bb296 100644
--- a/rtl/verilog/mor1kx_ctrl_espresso.v
+++ b/rtl/verilog/mor1kx_ctrl_espresso.v
@@ -469,7 +469,7 @@ module mor1kx_ctrl_espresso
      if (rst)
        take_exception <= 0;
      else
-       take_exception <= (exception_pending | exception_r | doing_rfe_r) &
+       take_exception <= (exception_pending | exception_r) &
                          (fetch_advance |
                           // Cause exception to always be 'taken' if stepping
                           (stepping & execute_done)

[stffrdhrn]

Example trace of running fusesoc sim mor1kx-generic --elf-load ~/work/openrisc/or1k-tests/native/build/or1k/or1k-basic --pipeline ESPRESSO --trace_to_screen --trace_enable. Trace show espresso pipeline gets stuck jumping to 0x0.

https://gist.github.com/stffrdhrn/88509c84587f88a4fc756afb62c745e7

[stffrdhrn]

See also #60