[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <87fura1qsh.fsf@eliezer.anholt.net>
Date: Fri, 15 Jul 2016 15:26:38 -0700
From: Eric Anholt <eric@...olt.net>
To: dri-devel@...ts.freedesktop.org
Cc: linux-kernel@...r.kernel.org
Subject: Re: [PATCH 3/7] drm/vc4: Add a bitmap of branch targets during shader validation.
Eric Anholt <eric@...olt.net> writes:
> This isn't used yet, it's just a first step toward loop validation.
> During the main parsing of instructions, we need to know when we hit a
> new basic block so that we can reset validated state.
>
> Signed-off-by: Eric Anholt <eric@...olt.net>
> ---
> drivers/gpu/drm/vc4/vc4_qpu_defines.h | 12 +++
> drivers/gpu/drm/vc4/vc4_validate_shaders.c | 114 ++++++++++++++++++++++++++++-
> 2 files changed, 124 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/vc4/vc4_qpu_defines.h b/drivers/gpu/drm/vc4/vc4_qpu_defines.h
> index d5c2f3c85ebb..82ef0e525d55 100644
> --- a/drivers/gpu/drm/vc4/vc4_qpu_defines.h
> +++ b/drivers/gpu/drm/vc4/vc4_qpu_defines.h
> + /* The actual branch target is the instruction after the delay
> + * slots, plus whatever byte offset is in the low 32 bits of
> + * the instruction. Make sure we're not branching beyond the
> + * end of the shader object.
> + */
> + if (branch_imm % sizeof(inst) != 0) {
> + DRM_ERROR("branch target not aligned\n");
> + return false;
> + };
Last change before pull request: I dropped the stray ';' that kbuild
test robot caught.
Download attachment "signature.asc" of type "application/pgp-signature" (819 bytes)
Powered by blists - more mailing lists