VCS Compiling problem for cv32e40p and PULP

[yingjiezero]

Issue Description

Hello, I encountered an issue while running the cv32e40p project. When I attempt to execute the following command in the cv32e40p/sim/core directory:

make vcs-custom TEST=pulp_general_alu CUSTOM=. CV_SIMULATOR=vcs

The simulation run fails, producing multiple errors, including a UVM_FATAL log entry. Below is part of the error output:

ucli% run
[cv32e40p_core]: PULP_XPULP =           0, PULP_CLUSTER =           0, FPU           0, PULP_ZFINX           0, NUM_MHPMCOUNTERS           1
UVM_INFO /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(265) @ 0ns: reporter [RNDSTALL] INSTR OBI stall enable: 1
UVM_INFO /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(266) @ 0ns: reporter [RNDSTALL] INSTR OBI stall mode:   1
UVM_INFO /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(267) @ 0ns: reporter [RNDSTALL] INSTR OBI stall gnt:    1
UVM_INFO /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(268) @ 0ns: reporter [RNDSTALL] INSTR OBI stall valid:  2
UVM_INFO /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(269) @ 0ns: reporter [RNDSTALL] INSTR OBI stall max:    3
UVM_INFO /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(270) @ 0ns: reporter [RNDSTALL] DATA  OBI stall enable: 0
UVM_INFO /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(271) @ 0ns: reporter [RNDSTALL] DATA  OBI stall mode:   0
UVM_INFO /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(272) @ 0ns: reporter [RNDSTALL] DATA  OBI stall gnt:    0
UVM_INFO /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(273) @ 0ns: reporter [RNDSTALL] DATA  OBI stall valid:  0
UVM_INFO /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(274) @ 0ns: reporter [RNDSTALL] DATA  OBI stall max:    0
UVM_ERROR /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(487) @ 6140ns: reporter [MM_RAM] out of bounds read from ccda43ac (RAM_ADDR_WIDTH=22; dm_halt_addri=1a110800, DBG_ADDR_WIDTH=14)
UVM_ERROR /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(487) @ 15980ns: reporter [MM_RAM] out of bounds read from 0550107b (RAM_ADDR_WIDTH=22; dm_halt_addri=1a110800, DBG_ADDR_WIDTH=14)
UVM_ERROR /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(487) @ 15980ns: reporter [MM_RAM] out of bounds read from 0550107c (RAM_ADDR_WIDTH=22; dm_halt_addri=1a110800, DBG_ADDR_WIDTH=14)
UVM_ERROR /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(487) @ 15990ns: reporter [MM_RAM] out of bounds read from 0550107c (RAM_ADDR_WIDTH=22; dm_halt_addri=1a110800, DBG_ADDR_WIDTH=14)
UVM_ERROR /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(487) @ 15990ns: reporter [MM_RAM] out of bounds read from 0550107f (RAM_ADDR_WIDTH=22; dm_halt_addri=1a110800, DBG_ADDR_WIDTH=14)
UVM_ERROR /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(487) @ 16080ns: reporter [MM_RAM] out of bounds read from 055010d3 (RAM_ADDR_WIDTH=22; dm_halt_addri=1a110800, DBG_ADDR_WIDTH=14)
UVM_ERROR /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(487) @ 16080ns: reporter [MM_RAM] out of bounds read from 055010d4 (RAM_ADDR_WIDTH=22; dm_halt_addri=1a110800, DBG_ADDR_WIDTH=14)
UVM_ERROR /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(487) @ 16090ns: reporter [MM_RAM] out of bounds read from 055010d4 (RAM_ADDR_WIDTH=22; dm_halt_addri=1a110800, DBG_ADDR_WIDTH=14)
UVM_ERROR /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(487) @ 16090ns: reporter [MM_RAM] out of bounds read from 055010d7 (RAM_ADDR_WIDTH=22; dm_halt_addri=1a110800, DBG_ADDR_WIDTH=14)
"/home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv", 445: tb_top.cv32e40p_tb_wrapper_i.ram_i.out_of_bounds_write: started at 16120000ps failed at 16120000ps
        Offending '(((((((((((((data_addr_i < (2 ** RAM_ADDR_WIDTH)) || ((data_addr_i >= dm_halt_addr_i) && (data_addr_i < (dm_halt_addr_i + (2 ** DBG_ADDR_WIDTH))))) || (data_addr_i == MMADDR_PRINT)) || (data_addr_i == MMADDR_TIMERREG)) || (data_addr_i == MMADDR_TIMERVAL)) || (data_addr_i == MMADDR_DBG)) || (data_addr_i == MMADDR_TESTSTATUS)) || (data_addr_i == MMADDR_EXIT)) || (data_addr_i == MMADDR_SIGBEGIN)) || (data_addr_i == MMADDR_SIGEND)) || (data_addr_i == MMADDR_SIGDUMP)) || (data_addr_i == MMADDR_TICKS)) || (data_addr_i[31:16] == MMADDR_RNDSTALL))'
UVM_FATAL /home/gk/Desktop/Package/gk/core-v-verif/cv32e40p/tb/core/mm_ram.sv(462) @ 16120ns: reporter [MM_RAM] out of bounds write to 0550107c with 00000020

--- UVM Report Summary ---

** Report counts by severity
UVM_INFO :   10
UVM_WARNING :    0
UVM_ERROR :    9
UVM_FATAL :    1
** Report counts by id
[MM_RAM]    10
[RNDSTALL]    10
$finish called from file "/home/gk/Desktop/Package/vcs2018/vcs/O-2018.09-SP2/etc/uvm/base/uvm_report_object.svh", line 292.

---

Changes I Made

To enable PULP functionality and try to resolve the issue, I made the following changes:

1. Modified the PULP_XPULP parameter in cv32e40p_core.sv:

   parameter PULP_XPULP = 1;

2. Uncommented the following lines (327–329) in Common.mk:

   ifeq ($(firstword $(subst _, ,$(TEST))),pulp)
   CFLAGS = -Os -g -D__riscv__=1 -D__LITTLE_ENDIAN__=1 -march=rv32imcxpulpv2 -Wa,-march=rv32imcxpulpv2 -fdata-sections -ffunction-sections -fdiagnostics-color=always
   endif

3. Configured toolchain environment variables for PULP GCC:

   export CV_SW_TOOLCHAIN=/home/gk/Desktop/Package/v1.0.16-pulp-riscv-gcc-ubuntu-16
   export CV_SW_PREFIX=riscv32-unknown-elf-
   export CV_SW_MARCH=rv32gc

---

Request for Help

I am seeking guidance on how to properly configure the project to enable support for PULP-based tests (TEST=pulp_general_alu). A similar setup with corev-openhw-gcc-ubuntu2004-20241020 as the toolchain was able to successfully run TEST=hello-world, but it also failed when running TEST=pulp_general_alu.

---

Additional Information
Host system: Ubuntu 20.04
VCS version: O-2018.09-SP2_Full64
Hardware compiler: PULP GCC v1.0.16 riscv

---

Attachment Details

I have attached a file named full_terminal_output.log to this issue. The file contains the complete terminal logs that resulted from running the failing command. Please review this file for further insights into the error.

full_terminal_output.log

[MikeOpenHWGroup]

Hi @yingjiezero, thanks for your interest in OpenHW and core-v-verif. And a double-thank-you for providing such a detailed Issue report. It really helps!

I can see three problems here:

1. You are attempting to run pulp test-programs on the core testbench. This is not supported and would require a substantial update to the memory model (MM_RAM). If you want to run the pulp test-programs, you should be using the UVM environment.
2. Your version of VCS is very old and there may compatibility issues, especially with the UVM environment and the functional coverage code (see Issue Standardize on a VCS version and remove the "UNSUPPORTED_WITH" workaround #2404 for an example).
3. We do not use the PULP toolchain. I strongly recommend using one of the prebuilt toolchains generated by Embecosm.

[yingjiezero]

Thank you for your guidance and insights. I have made some changes based on your recommendations and have attempted to run the tests in a different setup. However, I'm encountering a new set of errors. Here's a summary of what I've done:

Current Attempt

I am now running the tests in the cv32e40p/sim/uvmt directory with the following command:

make test TEST=pulp_general_alu CV_SIMULATOR=vcs USE_ISS=NO COV=YES

New Error Details

While the setup is slightly different, the simulation still results in errors, but with different messages:

UVM_INFO @ 16.800 ns : uvmt_cv32e40p_base_test.sv(281) uvm_test_top [BASE TEST] set load_instr_mem
UVM_INFO @ 144.300 ns : uvmt_cv32e40p_firmware_test.sv(120) uvm_test_top [TEST] Started RUN
779.000 ns: Illegal instruction (core 0) at PC 0x0000019a:
"/home/gk/Desktop/Package/gk/core-v-verif/lib/uvm_agents/uvma_obi_memory/src/uvma_obi_memory_assert.sv", 234: uvmt_cv32e40p_tb.dut_wrap.obi_data_memory_assert_i.u_assert.a_addr_be_consistent: started at 2100300ps failed at 2100300ps
	Offending '(get_min_be(addr) <= be)'
UVM_ERROR @ 2100.300 ns : uvma_obi_memory_assert.sv(236) reporter [OBIMEMASRT] be of 0x1 not consistent with addr 0x0550107c
"/home/gk/Desktop/Package/gk/core-v-verif/lib/uvm_agents/uvma_obi_memory/src/uvma_obi_memory_assert.sv", 234: uvmt_cv32e40p_tb.dut_wrap.obi_data_memory_assert_i.u_assert.a_addr_be_consistent: started at 2118300ps failed at 2118300ps
	Offending '(get_min_be(addr) <= be)'
UVM_ERROR @ 2118.300 ns : uvma_obi_memory_assert.sv(236) reporter [OBIMEMASRT] be of 0x7 not consistent with addr 0x055010d4
"/home/gk/Desktop/Package/gk/core-v-verif/lib/uvm_agents/uvma_obi_memory/src/uvma_obi_memory_assert.sv", 234: uvmt_cv32e40p_tb.dut_wrap.obi_data_memory_assert_i.u_assert.a_addr_be_consistent: started at 2127300ps failed at 2127300ps
	Offending '(get_min_be(addr) <= be)'
UVM_ERROR @ 2127.300 ns : uvma_obi_memory_assert.sv(236) reporter [OBIMEMASRT] be of 0x1 not consistent with addr 0x0550107c
"/home/gk/Desktop/Package/gk/core-v-verif/lib/uvm_agents/uvma_obi_memory/src/uvma_obi_memory_assert.sv", 234: uvmt_cv32e40p_tb.dut_wrap.obi_data_memory_assert_i.u_assert.a_addr_be_consistent: started at 2142300ps failed at 2142300ps
	Offending '(get_min_be(addr) <= be)'
UVM_ERROR @ 2142.300 ns : uvma_obi_memory_assert.sv(236) reporter [OBIMEMASRT] be of 0x7 not consistent with addr 0x055010d4
"/home/gk/Desktop/Package/gk/core-v-verif/lib/uvm_agents/uvma_obi_memory/src/uvma_obi_memory_assert.sv", 234: uvmt_cv32e40p_tb.dut_wrap.obi_data_memory_assert_i.u_assert.a_addr_be_consistent: started at 2175300ps failed at 2175300ps
	Offending '(get_min_be(addr) <= be)'
UVM_ERROR @ 2175.300 ns : uvma_obi_memory_assert.sv(236) reporter [OBIMEMASRT] be of 0x1 not consistent with addr 0x0550107c
UVM_INFO @ 2175.300 ns : uvm_report_server.svh(894) reporter [UVM/REPORT/SERVER] 
--- UVM Report Summary ---

Quit count reached!
Quit count :     5 of     5
** Report counts by severity
UVM_INFO :   50
UVM_WARNING :    0
UVM_ERROR :    5
UVM_FATAL :    0
** Report counts by id
[BASE TEST]     4
[CLKNRST]     4
[CORE_CNTRL_IF]     1
[DEBUGCOVG]     1
[INTERRUPTCOVG]     1
[MEMREADMEMH]     1
[OBIMEMASRT]     5
[OBIVPSEQ]    22
[OBI_MEMORY_MON]     4
[OVPWRAP]     1
[RNTST]     1
[RST_VSEQ]     3
[Step-andCompare]     1
[TEST]     2
[TEST_CFG]     1
[UVM/RELNOTES]     1
[UVME_CV32E40P_ENV]     1
[rv32isa_covg]     1

$finish called from file "/home/gk/Desktop/Package/vcs2018/vcs/O-2018.09-SP2/etc/uvm-1.2/base/uvm_root.svh", line 135.

uvmt_cv32e40p_tb.end_of_test: *** Test Summary ***

    FFFFFFFF   AAAAAA   IIIIII  LL        EEEEEEEE  DDDDDDD       
    FF        AA    AA    II    LL        EE        DD    DD      
    FF        AA    AA    II    LL        EE        DD    DD      
    FFFFF     AAAAAAAA    II    LL        EEEEE     DD    DD      
    FF        AA    AA    II    LL        EE        DD    DD      
    FF        AA    AA    II    LL        EE        DD    DD      
    FF        AA    AA  IIIIII  LLLLLLLL  EEEEEEEE  DDDDDDD       
    --------------------------------------------------------
                   SIMULATION FAILED - ABORTED              
    --------------------------------------------------------
$finish at simulation time 2175.300 ns

Changes Made

1. Directory Change: I have shifted to running tests in the cv32e40p/sim/uvmt directory as per your suggestion to use the UVM environment.
2. Toolchain Update: Switched to using the recommended GCC toolchain:

   export CV_SW_TOOLCHAIN=/home/gk/Desktop/Package/pulp-gcc-ubuntu2004-20200913
   export CV_SW_PREFIX=riscv32-unknown-elf-
   export CV_SW_MARCH=rv32gc

3. VCS Version: Although I attempted to install newer versions of VCS (2024 and 2022), they were not compatible with my Ubuntu 2004 setup, resulting in failures to run. Therefore, I am currently using VCS version 2018 to see if it can work.

I have attached a file named make_test.log to this issue. The file contains the complete terminal logs that resulted from running the failing command. Please review this file for further insights into the error.

make_test.log

[MikeOpenHWGroup]

Ahahaha! You have encountered a bug that we just recently discovered as well. (Embarrassing!) If you are sign-up for an Eclipse account, you can see the discussion about this on our MatterMost chat-board (link). We will create a GitHub issue about this soon.

The assertion that is firing is in uvma_obi_memory_assert.sv. You can do one or more of these things:

1. Comment out the assertion. We will be fixing it sometime next week.
2. Patch the property to this:

  property p_addr_be_consistent;
    disable iff (DATA_WIDTH !== 32) req |-> addr[1:0] <= get_addr_lsb(be);  // is that right ??
  endproperty : p_addr_be_consistent

3. Use the cv32e40p/dev branch of this repo. This is strongly recommended as this is the most up-to-date version of the repo for the CV32E40P core.

Lastly, you should not use rv32gc for your CV_SW_MARCH shell env variable. The CV32E40P does not support the rv32gc profile. The CV32E40P test-program environment configuration defaults this var to rv32imc_zicsr_zifencei, so you can either unset your CV_SW_MARCH, or use rv32imc_zicsr_zifencei. If you want to learn about how this all works in core-v-verif, please read the TOOLCHAIN documentation.

[yingjiezero]

Thank you for your previous guidance and recommendations. I have made several adjustments based on your input and attempted to run the tests, but I am still encountering errors with the make test TEST=pulp_general_alu CV_SIMULATOR=vcs USE_ISS=NO COV=YES command. Below are the details of the current issues and the changes I have made:

---

Current Errors

1. While compiling uvme_cv32e40p_fp_instr_covg.sv:

   Error-[FCIWCS] Invalid 'with' clause specification
   /home/gk/Desktop/Package/core-v-verif/cv32e40p/env/uvme/cov/uvme_cv32e40p_fp_instr_covg.sv, 396
   uvme_cv32e40p_pkg, "cp_curr_fpu_inst_rd_for_0_lat_apu_result"
     The 'with' clause specified on the coverpoint bin does not have any impact 
     and is invalid; it should contain at least one occurence of the 'item' 
     keyword.
     Please remove or correct the clause and compile
   

2. Errors during assembly of pulp_general_alu.S:

   /home/gk/Desktop/Package/core-v-verif/cv32e40p/tests/programs/custom/pulp_general_alu/pulp_general_alu.S: Assembler messages:
   /home/gk/Desktop/Package/core-v-verif/cv32e40p/tests/programs/custom/pulp_general_alu/pulp_general_alu.S:67: 错误： unrecognized opcode `cv.abs x18,x20'
   /home/gk/Desktop/Package/core-v-verif/cv32e40p/tests/programs/custom/pulp_general_alu/pulp_general_alu.S:73: 错误： unrecognized opcode `cv.abs x18,x20'
   /home/gk/Desktop/Package/core-v-verif/cv32e40p/tests/programs/custom/pulp_general_alu/pulp_general_alu.S:79: 错误： unrecognized opcode `cv.abs x18,x20'
   /home/gk/Desktop/Package/core-v-verif/cv32e40p/tests/programs/custom/pulp_general_alu/pulp_general_alu.S:85: 错误： unrecognized opcode `cv.abs x18,x20'
   /home/gk/Desktop/Package/core-v-verif/cv32e40p/tests/programs/custom/pulp_general_alu/pulp_general_alu.S:91: 错误： unrecognized opcode `cv.abs x18,x20'
   /home/gk/Desktop/Package/core-v-verif/cv32e40p/tests/programs/custom/pulp_general_alu/pulp_general_alu.S:97: 错误： unrecognized opcode `cv.abs x18,x20'
   

······
/home/gk/Desktop/Package/core-v-verif/cv32e40p/tests/programs/custom/pulp_general_alu/pulp_general_alu.S:1423: 错误： unrecognized opcode cv.suburnr x18,x20,x21' /home/gk/Desktop/Package/core-v-verif/cv32e40p/tests/programs/custom/pulp_general_alu/pulp_general_alu.S:1431: 错误： unrecognized opcode cv.suburnr x18,x20,x21'
/home/gk/Desktop/Package/core-v-verif/cv32e40p/tests/programs/custom/pulp_general_alu/pulp_general_alu.S:1439: 错误： unrecognized opcode `cv.suburnr x18,x20,x21'

---

### Changes Made

To align as closely as possible with the setup you suggested, I have made the following modifications:

1. **Branch Update**: I am now using the `cv32e40p/dev` branch of the repository.
2. **Updated Toolchain**: Switched to GCC toolchain with these configurations:
```bash
export CV_SW_TOOLCHAIN=/home/gk/Desktop/Package/pulp-gcc-ubuntu2004-20200913
export CV_SW_PREFIX=riscv32-unknown-elf-
export CV_SW_MARCH=rv32imc_zicsr_zifencei

3. VCS Version: Upgraded to VCS version T-2022.06-SP2-10, as the previously used version was outdated and caused compatibility issues.

[MikeOpenHWGroup]

Your compilation error is due to Issue #2404, which I referred to before. If you look at that Issue you will see that we are recommending VCS version V-2023.12-1 or newer. This was a bug in VCS, so we will not change our code, since it works with all other commercial SystemVerilog simulators and the newer versions of VCS have solved the problem.

As for the compiler issue, since you are overriding the CV_SW_MARCH with a shell environment variable and setting it to rv32imc_zicsr_zifencei, you will not be able to compile test-programs with instructions from the PULP ISA. As stated earlier, it is recommended to unset your CV_SW_MARCH and let the CFG scripts do it for you as explained in TOOLCHAIN.

[yingjiezero]

I attempted to build the PULP toolchain from source following the instructions in TOOLCHAIN.md. However, during the sudo make step, I encountered compilation errors related to unresolved symbols like isl_id_free and others.

Here are the details of the error:

• The environment variables were set as below:

  export PATH=/home/gk/Desktop/Package/pulp/bin:$PATH
  export PULP="YES"

  The library versions I used were from pulp-gcc-ubuntu2004-20200913.
• Error Details

g++ -fno-PIE -c   -g -O2 -DIN_GCC  -DCROSS_DIRECTORY_STRUCTURE   -fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wmissing-format-attribute -Woverloaded-virtual -pedantic -Wno-long-long -Wno-variadic-macros -Wno-overlength-strings -fno-common  -DHAVE_CONFIG_H -I. -I. -I/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc -I/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/. -I/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/../include -I/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/../libcpp/include  -I/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/../libdecnumber -I/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/../libdecnumber/dpd -I../libdecnumber -I/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/../libbacktrace   -o graphite-isl-ast-to-gimple.o -MT graphite-isl-ast-to-gimple.o -MMD -MP -MF ./.deps/graphite-isl-ast-to-gimple.TPo /home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c: In function ‘void ivs_params_clear(ivs_params&)’:
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:98:7: error: ‘isl_id_free’ was not declared in this scope; did you mean ‘isl_aff_free’?
   98 |       isl_id_free (it->first);
      |       ^~~~~~~~~~~
      |       isl_aff_free
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c: In member function ‘tree_node* translate_isl_ast_to_gimple::gcc_expression_from_isl_ast_expr_id(tree, isl_ast_expr*, ivs_params&)’:
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:292:3: error: ‘isl_id_free’ was not declared in this scope; did you mean ‘isl_aff_free’?
  292 |   isl_id_free (tmp_isl_id);
      |   ^~~~~~~~~~~
      |   isl_aff_free
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c: In member function ‘loop* translate_isl_ast_to_gimple::graphite_create_new_loop(edge, isl_ast_node*, loop_p, tree, tree, tree, ivs_params&)’:
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:628:5: error: ‘isl_id_free’ was not declared in this scope; did you mean ‘isl_aff_free’?
  628 |     isl_id_free (res->first);
      |     ^~~~~~~~~~~
      |     isl_aff_free
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c: In member function ‘edge_def* translate_isl_ast_to_gimple::translate_isl_ast_for_loop(loop_p, isl_ast_node*, edge, tree, tree, tree, ivs_params&)’:
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:668:53: error: ‘isl_id_get_user’ was not declared in this scope; did you mean ‘isl_arg_user’?
  668 |       ast_build_info *for_info = (ast_build_info *) isl_id_get_user (id);
      |                                                     ^~~~~~~~~~~~~~~
      |                                                     isl_arg_user
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:671:7: error: ‘isl_id_free’ was not declared in this scope; did you mean ‘isl_aff_free’?
  671 |       isl_id_free (id);
      |       ^~~~~~~~~~~
      |       isl_aff_free
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c: In member function ‘edge_def* translate_isl_ast_to_gimple::translate_isl_ast_node_user(isl_ast_node*, edge, ivs_params&)’:
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:875:31: error: ‘isl_id_get_user’ was not declared in this scope; did you mean ‘isl_arg_user’?
  875 |   poly_bb_p pbb = (poly_bb_p) isl_id_get_user (name_id);
      |                               ^~~~~~~~~~~~~~~
      |                               isl_arg_user
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:881:3: error: ‘isl_id_free’ was not declared in this scope; did you mean ‘isl_aff_free’?
  881 |   isl_id_free (name_id);
      |   ^~~~~~~~~~~
      |   isl_aff_free
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c: In function ‘isl_id* ast_build_before_for(isl_ast_build*, void*)’:
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:2876:19: error: ‘isl_space_dim’ was not declared in this scope; did you mean ‘isl_aff_dim’?
 2876 |   int dimension = isl_space_dim (schedule_space, isl_dim_out);
      |                   ^~~~~~~~~~~~~
      |                   isl_aff_dim
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:2880:3: error: ‘isl_space_free’ was not declared in this scope; did you mean ‘isl_aff_free’?
 2880 |   isl_space_free (schedule_space);
      |   ^~~~~~~~~~~~~~
      |   isl_aff_free
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:2881:16: error: ‘isl_id_alloc’ was not declared in this scope; did you mean ‘isl_mat_alloc’?
 2881 |   isl_id *id = isl_id_alloc (isl_ast_build_get_ctx (build), "", for_info);
      |                ^~~~~~~~~~~~
      |                isl_mat_alloc
In file included from /home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/hash-table.h:236,
                 from /home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/coretypes.h:369,
                 from /home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:29:
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/vec.h: In instantiation of ‘bool vec<T, va_heap, vl_ptr>::reserve(unsigned int, bool) [with T = std::pair<gphi*, gphi*>]’:
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/vec.h:1554:3:   required from ‘T* vec<T, va_heap, vl_ptr>::safe_push(const T&) [with T = std::pair<gphi*, gphi*>]’
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:1859:72:   required from here
/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/vec.h:1448:14: warning: ‘void* memcpy(void*, const void*, size_t)’ writing to an object of type ‘struct std::pair<gphi*, gphi*>’ with no trivial copy-assignment; use copy-assignment or copy-initialization instead [-Wclass-memaccess]
 1448 |       memcpy (m_vec->address (), oldvec->address (), sizeof (T) * oldsize);
      |       ~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from /usr/include/c++/9/bits/stl_algobase.h:64,
                 from /usr/include/c++/9/bits/stl_tree.h:63,
                 from /usr/include/c++/9/map:60,
                 from /home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/system.h:227,
                 from /home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/riscv-gcc/gcc/graphite-isl-ast-to-gimple.c:28:
/usr/include/c++/9/bits/stl_pair.h:208:12: note: ‘struct std::pair<gphi*, gphi*>’ declared here
  208 |     struct pair
      |            ^~~~
make[2]: *** [Makefile:1099：graphite-isl-ast-to-gimple.o] 错误 1
make[2]: 离开目录“/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/build-gcc-newlib-stage1/gcc”
make[1]: *** [Makefile:4220：all-gcc] 错误 2
make[1]: 离开目录“/home/gk/Desktop/Package/pulp-riscv-gnu-toolchain/master/build-gcc-newlib-stage1”
make: *** [Makefile:227：stamps/build-gcc-newlib-stage1] 错误 2

Attempts Made

1. Installed the latest version of libisl-dev to address potential compatibility issues and attempted recompilation.

[MikeOpenHWGroup]

I have never attempted to build the PULP toolchain from source, so I am not much help there. Have you tried the prebuilt PULP toolchain from Embecosm? If you are still having trouble, I would recommend reaching out to either the PULP Platform team at ETH Zurich or Embecosm.

Also, FYI, the Issue related to the assertion in uvma_obi_memory_assert.sv is #2599.

[MikeOpenHWGroup]

Hi @yingjiezero. The bug in uvma_obi_memory_assert.sv as reported in #2599 was resolved with #2601.

[yingjiezero]

I've been working on configuring the environment and running tests as discussed. Here are the steps and issues encountered:

Setup and Configuration

I followed the recommendation to use the PULP toolchain from here, specifically for xpulpv2. The environment is set up as follows:

export CV_SW_TOOLCHAIN=/home/host/Desktop/package/xpulpv2/riscv-pulp
export CV_SW_PREFIX=riscv32-unknown-elf-
export RISCV=/home/host/Desktop/package/xpulpv2/riscv-pulp
export PATH=$RISCV/bin:$PATH

Issues Encountered

1. When using the cv32e40p/dev branch and running make test TEST=pulp_general_alu CV_SIMULATOR=vcs USE_ISS=NO CFG=pulp, an error occurs:

   cc1: error: -march=rv32imc_zicsr_zifencei_xcvhwlp_xcvmem_xcvmac_xcvbi_xcvalu_xcvsimd_xcvbitmanip: unsupported ISA substring '_ZICSR_ZIFENCEI_XCVHWLP_XCVMEM_XCVMAC_XCVBI_XCVALU_XCVSIMD_XCVBITMANIP'
   

   I modified pulp.yaml to use -march=rv32imcxpulpv2, but additional errors arose due to unrecognized opcodes like cv.start, cv.end, and cv.count.

make -C /home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/sim/uvmt/vcs_results/pulp/pulp_general_alu/0/test_program/bsp \
        VPATH=/home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/bsp \
        RISCV=/home/host/Desktop/package/xpulpv2/riscv-pulp \
        RISCV_PREFIX=riscv32-unknown-elf- \
        RISCV_EXE_PREFIX=/home/host/Desktop/package/xpulpv2/riscv-pulp/bin/riscv32-unknown-elf- \
        RISCV_MARCH=rv32imcxpulpv2 \
        RISCV_CC=gcc \
        RISCV_CFLAGS="" \
        CFG_CFLAGS="-DPULP" \
        all
make[2]: Entering directory '/home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/sim/uvmt/vcs_results/pulp/pulp_general_alu/0/test_program/bsp'
/home/host/Desktop/package/xpulpv2/riscv-pulp/bin/riscv32-unknown-elf-gcc -Os -g -static -mabi=ilp32 -march=rv32imcxpulpv2 -Wall -pedantic -DPULP -c /home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/bsp/crt0.S -o crt0.o
/home/host/Desktop/package/xpulpv2/riscv-pulp/bin/riscv32-unknown-elf-gcc -Os -g -static -mabi=ilp32 -march=rv32imcxpulpv2 -Wall -pedantic -DPULP -c /home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/bsp/handlers.S -o handlers.o
/home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/bsp/handlers.S: Assembler messages:
/home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/bsp/handlers.S:204: Error: unrecognized opcode `cv.start 0,t6'
/home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/bsp/handlers.S:206: Error: unrecognized opcode `cv.end 0,t6'
/home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/bsp/handlers.S:208: Error: unrecognized opcode `cv.count 0,t6'

2. Switching to the master branch and modifying uvma_obi_memory_assert.sv as per OBI ASSERT : Fix issue #2599 #2601 resulted in continuous "Illegal instruction" errors during simulation:

   Illegal instruction (core 0) at PC 0x00000002:
   

3. I also tried installing xpulpv3 via the riscv-gnu-toolchain URL with similar issues persisting.

4. However, I confirmed my xpulpv2 and xpulpv3 setup seems sound since a simple C program utilizing PULP instructions compiles successfully:
   riscv32-unknown-elf-gcc -march=rv32imcxpulpv2 -o test_pulp test_pulp.c

   #include <stdio.h>
   
   int main() {
       int a = 3, b = 0;
       asm volatile(
           "p.addN x0, x1, x2, 1\n"
           "p.subN x0, x1, x2, 1\n"
       );
   }

[MikeOpenHWGroup]

At this point, I am not sure, so let me dig a little deeper for you. In the meantime:

• try CV_SW_MARCH=rv32imcxpulpv2_xcv
• don't waste your time on the master branch. :-)

[yingjiezero]

cv32e40p/dev branch , try CV_SW_MARCH=rv32imcxpulpv2_xcv，an error occurs:

/home/host/Desktop/package/xpulpv2/riscv-pulp/bin/riscv32-unknown-elf-gcc \
        -DNO_PULP \
        -Os -g -static -mabi=ilp32 -march=rv32imcxpulpv2_xcv -Wall -pedantic  \
         \
        -I ../../tests/asm \
        -I /home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/bsp \
        -o /home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/sim/uvmt/vcs_results/default/pulp_general_alu/0/test_program/pulp_general_alu.elf \
        -nostartfiles \
        /home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/tests/programs/custom/pulp_general_alu/pulp_general_alu.S \
        -T /home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/bsp/link.ld \
        -Xlinker -Map=/home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/sim/uvmt/vcs_results/default/pulp_general_alu/0/test_program/pulp_general_alu.map \
         -L /home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/sim/uvmt/vcs_results/default/pulp_general_alu/0/test_program/bsp \
        -lcv-verif
cc1: error: -march=rv32imcxpulpv2_xcv: unsupported ISA substring '_XCV'
make: *** [/home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/mk/Common.mk:525: /home/host/Desktop/Coverage/CV32E40P/DEV/core-v-verif/cv32e40p/sim/uvmt/vcs_results/default/pulp_general_alu/0/test_program/pulp_general_alu.elf] Error 1