Skip to content

Conversation

@edward-jones
Copy link

@edward-jones edward-jones commented Apr 18, 2019

When attempting to run ri5cy using the upstream gdb and newlib I found that the tester would fail to detect the _exit breakpoint, causing all of the benchmarks to timeout.

@edward-jones edward-jones requested a review from gmarkall April 18, 2019 09:03
@jeremybennett
Copy link
Member

I believe this is already fixed by commit e23aaf7. Set the environment variable BEEBS_TIMEOUT to adjust the timeout for each test.

Second comment. BEEBS has a ChangeLog, which you need to update.

@edward-jones edward-jones force-pushed the erj-ri5cy-fixes branch 3 times, most recently from 994b6e2 to e2cddac Compare April 24, 2019 10:35
@edward-jones edward-jones changed the title Fixes to the ri5cy verilator expect script Fixes for ri5cy verilator Apr 24, 2019
@edward-jones
Copy link
Author

I've updated this so it now contains three basic changes:

  • Correctly detects "_exit"
  • Fixes ri5cy build to target rv32imc instead of rv32i
  • Allow an alternative gdbserver to be used through an environment variable

This should work more consistently than looking for a call to
__internal_syscall, which was very library specific.

testsuite/ChangeLog:

	* config/ri5cyverilator.exp (beebs_load): Check for
	_exit breakpoint instead of __internal_syscall when trying
	to detect exit.
ChangeLog:

	* config/riscv32/chips/ri5cy/chip.cfg: Change CHIP_CFLAGS to
	match architecture of ri5cy (-march=rv32imc -mabi=ilp32)
This allows an alternative gdbserver command to be used when
running the tests.

testsuite/ChangeLog:

	* config/ri5cyverilator.exp (beebs_load): Pick up
	GDBSERVER_CMD environment variable if set.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants