Re: [Mesa-dev] [PATCH 0/9] spirv: Improve logging and error handling

2017-09-04 Thread Tapani Pälli 



On 08/29/2017 01:55 PM, Tapani Pälli wrote:

LGTM, patches 1-5

Reviewed-by: Tapani Pälli 

With these changes we can add a nir_spirv_debug_callback that has 
instance and shader object in private data so that we can call 
VK_EXT_debug_report functionality from anv_pipeline code.




Went through the rest of it, series is:
Reviewed-by: Tapani Pälli 


On 08/17/2017 08:22 PM, Jason Ekstrand wrote:

This series has two objectives:

  1) Improve logging to provide more detail and provide hooks so we can
 plumb errors and warnings through to debug_report extensions.

  2) Improve error handling so that not all errors result in killing the
 process.  This is done by adding new spv_fail and spv_assert helpers
 which log the error and longjump back to a point where we can 
clean up

 and return NULL without crashing.

There is still quite a ways to go with error handling if we want to be 
able

to guarantee that we won't crash given an arbitrary stream of bytes.
However, this should at least be a step in the right direction.  There 
are
probably at least a couple of cases where I could have left an 
assert() as

an actual "kill the process" assert because it's testing for internal
consistency.  However, the vast majority of them are validation checks so
it seemed better to just search+replace them all and we can convert 
the few

that we want back to real asserts later.

Cc: "Ian Romanick" 

Jason Ekstrand (9):
   ralloc: Allow reparenting to a NULL context
   spirv: Parent the nir_shader to the builder while building
   spirv: Re-arrange vtn_builder initialization
   spirv: Rework logging
   spirv: Do something useful with OpSource
   util: Add a NORETURN macro
   spirv: Add vtn_fail and vtn_assert helpers
   spirv: Replace assert with vtn_assert
   spirv: Replace unreachable with vtn_fail

  configure.ac   |   1 +
  src/amd/vulkan/radv_pipeline.c |   2 +-
  src/compiler/spirv/nir_spirv.h |  17 +-
  src/compiler/spirv/spirv2nir.c |   3 +-
  src/compiler/spirv/spirv_to_nir.c  | 496 
+++--

  src/compiler/spirv/vtn_alu.c   |  46 ++--
  src/compiler/spirv/vtn_cfg.c   |  59 ++---
  src/compiler/spirv/vtn_glsl450.c   |  18 +-
  src/compiler/spirv/vtn_private.h   |  56 -
  src/compiler/spirv/vtn_variables.c | 253 +--
  src/intel/vulkan/anv_pipeline.c|   2 +-
  src/util/macros.h  |   6 +
  src/util/ralloc.c  |   2 +-
  13 files changed, 578 insertions(+), 383 deletions(-)


___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev

___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev

Re: [Mesa-dev] [PATCH 0/9] spirv: Improve logging and error handling

2017-08-29 Thread Tapani Pälli 

LGTM, patches 1-5

Reviewed-by: Tapani Pälli 

With these changes we can add a nir_spirv_debug_callback that has 
instance and shader object in private data so that we can call 
VK_EXT_debug_report functionality from anv_pipeline code.



On 08/17/2017 08:22 PM, Jason Ekstrand wrote:

This series has two objectives:

  1) Improve logging to provide more detail and provide hooks so we can
 plumb errors and warnings through to debug_report extensions.

  2) Improve error handling so that not all errors result in killing the
 process.  This is done by adding new spv_fail and spv_assert helpers
 which log the error and longjump back to a point where we can clean up
 and return NULL without crashing.

There is still quite a ways to go with error handling if we want to be able
to guarantee that we won't crash given an arbitrary stream of bytes.
However, this should at least be a step in the right direction.  There are
probably at least a couple of cases where I could have left an assert() as
an actual "kill the process" assert because it's testing for internal
consistency.  However, the vast majority of them are validation checks so
it seemed better to just search+replace them all and we can convert the few
that we want back to real asserts later.

Cc: "Ian Romanick" 

Jason Ekstrand (9):
   ralloc: Allow reparenting to a NULL context
   spirv: Parent the nir_shader to the builder while building
   spirv: Re-arrange vtn_builder initialization
   spirv: Rework logging
   spirv: Do something useful with OpSource
   util: Add a NORETURN macro
   spirv: Add vtn_fail and vtn_assert helpers
   spirv: Replace assert with vtn_assert
   spirv: Replace unreachable with vtn_fail

  configure.ac   |   1 +
  src/amd/vulkan/radv_pipeline.c |   2 +-
  src/compiler/spirv/nir_spirv.h |  17 +-
  src/compiler/spirv/spirv2nir.c |   3 +-
  src/compiler/spirv/spirv_to_nir.c  | 496 +++--
  src/compiler/spirv/vtn_alu.c   |  46 ++--
  src/compiler/spirv/vtn_cfg.c   |  59 ++---
  src/compiler/spirv/vtn_glsl450.c   |  18 +-
  src/compiler/spirv/vtn_private.h   |  56 -
  src/compiler/spirv/vtn_variables.c | 253 +--
  src/intel/vulkan/anv_pipeline.c|   2 +-
  src/util/macros.h  |   6 +
  src/util/ralloc.c  |   2 +-
  13 files changed, 578 insertions(+), 383 deletions(-)


___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev

[Mesa-dev] [PATCH 0/9] spirv: Improve logging and error handling

2017-08-17 Thread Jason Ekstrand 
This series has two objectives:

 1) Improve logging to provide more detail and provide hooks so we can
plumb errors and warnings through to debug_report extensions.

 2) Improve error handling so that not all errors result in killing the
process.  This is done by adding new spv_fail and spv_assert helpers
which log the error and longjump back to a point where we can clean up
and return NULL without crashing.

There is still quite a ways to go with error handling if we want to be able
to guarantee that we won't crash given an arbitrary stream of bytes.
However, this should at least be a step in the right direction.  There are
probably at least a couple of cases where I could have left an assert() as
an actual "kill the process" assert because it's testing for internal
consistency.  However, the vast majority of them are validation checks so
it seemed better to just search+replace them all and we can convert the few
that we want back to real asserts later.

Cc: "Ian Romanick" 

Jason Ekstrand (9):
  ralloc: Allow reparenting to a NULL context
  spirv: Parent the nir_shader to the builder while building
  spirv: Re-arrange vtn_builder initialization
  spirv: Rework logging
  spirv: Do something useful with OpSource
  util: Add a NORETURN macro
  spirv: Add vtn_fail and vtn_assert helpers
  spirv: Replace assert with vtn_assert
  spirv: Replace unreachable with vtn_fail

 configure.ac   |   1 +
 src/amd/vulkan/radv_pipeline.c |   2 +-
 src/compiler/spirv/nir_spirv.h |  17 +-
 src/compiler/spirv/spirv2nir.c |   3 +-
 src/compiler/spirv/spirv_to_nir.c  | 496 +++--
 src/compiler/spirv/vtn_alu.c   |  46 ++--
 src/compiler/spirv/vtn_cfg.c   |  59 ++---
 src/compiler/spirv/vtn_glsl450.c   |  18 +-
 src/compiler/spirv/vtn_private.h   |  56 -
 src/compiler/spirv/vtn_variables.c | 253 +--
 src/intel/vulkan/anv_pipeline.c|   2 +-
 src/util/macros.h  |   6 +
 src/util/ralloc.c  |   2 +-
 13 files changed, 578 insertions(+), 383 deletions(-)

-- 
2.5.0.400.gff86faf

___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev