| Message ID | 1426814664-5702-1-git-send-email-rodrigo.vivi@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Thu, Mar 19, 2015 at 06:24:24PM -0700, Rodrigo Vivi wrote: > This will allow manual tests when crc isn't available. > > v2: Remove unused and non-sense buf->size and decrease buf->stride a bit as suggested by Daniel. buf->size isn't nonsense really, it does seem to match what we've allocated. But it also seems unused in the rendercpy functions I've checked. Imo you should keep that line. -Daniel > > Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> > --- > tests/kms_psr_sink_crc.c | 5 ++--- > 1 file changed, 2 insertions(+), 3 deletions(-) > > diff --git a/tests/kms_psr_sink_crc.c b/tests/kms_psr_sink_crc.c > index 0a56705..46e04c8 100644 > --- a/tests/kms_psr_sink_crc.c > +++ b/tests/kms_psr_sink_crc.c > @@ -137,9 +137,8 @@ static void fill_blt(data_t *data, uint32_t handle, unsigned char color) > static void scratch_buf_init(struct igt_buf *buf, drm_intel_bo *bo) > { > buf->bo = bo; > - buf->stride = 4096; > + buf->stride = 1024; > buf->tiling = I915_TILING_X; > - buf->size = 4096; > } > > static void fill_render(data_t *data, uint32_t handle, unsigned char color) > @@ -167,7 +166,7 @@ static void fill_render(data_t *data, uint32_t handle, unsigned char color) > igt_assert(batch); > > rendercopy(batch, NULL, > - &src_buf, 0, 0, 1, 1, > + &src_buf, 0, 0, 0xff, 0xff, > &dst_buf, 0, 0); > > intel_batchbuffer_free(batch); > -- > 2.1.0 > > _______________________________________________ > Intel-gfx mailing list > Intel-gfx@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
On Fri, 2015-03-20 at 10:54 +0100, Daniel Vetter wrote: > On Thu, Mar 19, 2015 at 06:24:24PM -0700, Rodrigo Vivi wrote: > > This will allow manual tests when crc isn't available. > > > > v2: Remove unused and non-sense buf->size and decrease buf->stride a bit as suggested by Daniel. > > buf->size isn't nonsense really, it does seem to match what we've > allocated. But it also seems unused in the rendercpy functions I've > checked. Imo you should keep that line oh sure, bad phrase.... the value 4 was nonsense. But if I increases that it fails. Without setting it works. What do you think? > -Daniel > > > > > Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> > > --- > > tests/kms_psr_sink_crc.c | 5 ++--- > > 1 file changed, 2 insertions(+), 3 deletions(-) > > > > diff --git a/tests/kms_psr_sink_crc.c b/tests/kms_psr_sink_crc.c > > index 0a56705..46e04c8 100644 > > --- a/tests/kms_psr_sink_crc.c > > +++ b/tests/kms_psr_sink_crc.c > > @@ -137,9 +137,8 @@ static void fill_blt(data_t *data, uint32_t handle, unsigned char color) > > static void scratch_buf_init(struct igt_buf *buf, drm_intel_bo *bo) > > { > > buf->bo = bo; > > - buf->stride = 4096; > > + buf->stride = 1024; > > buf->tiling = I915_TILING_X; > > - buf->size = 4096; > > } > > > > static void fill_render(data_t *data, uint32_t handle, unsigned char color) > > @@ -167,7 +166,7 @@ static void fill_render(data_t *data, uint32_t handle, unsigned char color) > > igt_assert(batch); > > > > rendercopy(batch, NULL, > > - &src_buf, 0, 0, 1, 1, > > + &src_buf, 0, 0, 0xff, 0xff, > > &dst_buf, 0, 0); > > > > intel_batchbuffer_free(batch); > > -- > > 2.1.0 > > > > _______________________________________________ > > Intel-gfx mailing list > > Intel-gfx@lists.freedesktop.org > > http://lists.freedesktop.org/mailman/listinfo/intel-gfx >
On Fri, Mar 20, 2015 at 05:43:40PM +0000, Vivi, Rodrigo wrote: > On Fri, 2015-03-20 at 10:54 +0100, Daniel Vetter wrote: > > On Thu, Mar 19, 2015 at 06:24:24PM -0700, Rodrigo Vivi wrote: > > > This will allow manual tests when crc isn't available. > > > > > > v2: Remove unused and non-sense buf->size and decrease buf->stride a bit as suggested by Daniel. > > > > buf->size isn't nonsense really, it does seem to match what we've > > allocated. But it also seems unused in the rendercpy functions I've > > checked. Imo you should keep that line > > oh sure, bad phrase.... the value 4 was nonsense. > But if I increases that it fails. Without setting it works. > What do you think? If just keeping that line makes the test work there's something _really_ fishy going on. If it persists after git clean -dfx and full recompile I'd dig into it, since this really doesn't make sense. -Daniel
On Mon, Mar 23, 2015 at 09:35:27AM +0100, Daniel Vetter wrote: > On Fri, Mar 20, 2015 at 05:43:40PM +0000, Vivi, Rodrigo wrote: > > On Fri, 2015-03-20 at 10:54 +0100, Daniel Vetter wrote: > > > On Thu, Mar 19, 2015 at 06:24:24PM -0700, Rodrigo Vivi wrote: > > > > This will allow manual tests when crc isn't available. > > > > > > > > v2: Remove unused and non-sense buf->size and decrease buf->stride a bit as suggested by Daniel. > > > > > > buf->size isn't nonsense really, it does seem to match what we've > > > allocated. But it also seems unused in the rendercpy functions I've > > > checked. Imo you should keep that line > > > > oh sure, bad phrase.... the value 4 was nonsense. > > But if I increases that it fails. Without setting it works. > > What do you think? > > If just keeping that line makes the test work there's something _really_ > fishy going on. If it persists after git clean -dfx and full recompile I'd > dig into it, since this really doesn't make sense. valgrined might be the best tool for a first attempt at figuring out what's wrong. -Daniel
On Mon, 2015-03-23 at 09:36 +0100, Daniel Vetter wrote: > On Mon, Mar 23, 2015 at 09:35:27AM +0100, Daniel Vetter wrote: > > On Fri, Mar 20, 2015 at 05:43:40PM +0000, Vivi, Rodrigo wrote: > > > On Fri, 2015-03-20 at 10:54 +0100, Daniel Vetter wrote: > > > > On Thu, Mar 19, 2015 at 06:24:24PM -0700, Rodrigo Vivi wrote: > > > > > This will allow manual tests when crc isn't available. > > > > > > > > > > v2: Remove unused and non-sense buf->size and decrease buf->stride a bit as suggested by Daniel. > > > > > > > > buf->size isn't nonsense really, it does seem to match what we've > > > > allocated. But it also seems unused in the rendercpy functions I've > > > > checked. Imo you should keep that line > > > > > > oh sure, bad phrase.... the value 4 was nonsense. > > > But if I increases that it fails. Without setting it works. > > > What do you think? > > > > If just keeping that line makes the test work there's something _really_ > > fishy going on. If it persists after git clean -dfx and full recompile I'd > > dig into it, since this really doesn't make sense. But what are the values of size and stride you believe it should work in a actual visible way? > > valgrined might be the best tool for a first attempt at figuring out > what's wrong. > -Daniel
On Mon, Mar 23, 2015 at 09:53:44PM +0000, Vivi, Rodrigo wrote: > On Mon, 2015-03-23 at 09:36 +0100, Daniel Vetter wrote: > > On Mon, Mar 23, 2015 at 09:35:27AM +0100, Daniel Vetter wrote: > > > On Fri, Mar 20, 2015 at 05:43:40PM +0000, Vivi, Rodrigo wrote: > > > > On Fri, 2015-03-20 at 10:54 +0100, Daniel Vetter wrote: > > > > > On Thu, Mar 19, 2015 at 06:24:24PM -0700, Rodrigo Vivi wrote: > > > > > > This will allow manual tests when crc isn't available. > > > > > > > > > > > > v2: Remove unused and non-sense buf->size and decrease buf->stride a bit as suggested by Daniel. > > > > > > > > > > buf->size isn't nonsense really, it does seem to match what we've > > > > > allocated. But it also seems unused in the rendercpy functions I've > > > > > checked. Imo you should keep that line > > > > > > > > oh sure, bad phrase.... the value 4 was nonsense. > > > > But if I increases that it fails. Without setting it works. > > > > What do you think? > > > > > > If just keeping that line makes the test work there's something _really_ > > > fishy going on. If it persists after git clean -dfx and full recompile I'd > > > dig into it, since this really doesn't make sense. > > But what are the values of size and stride you believe it should work in > a actual visible way? Size should obviously match the size of the bo allocated, and stride should be whatever you need, but necessarily must be smaller than the buffer to make sense. I didn't read the test too closely and don't have an edp psr machine at hand to try out what would really be needed here. Ben has played around a lot with the rendercpy stuff, probably best to ask him for a small favour and look into this a bit ;-) Cheers, Daniel
diff --git a/tests/kms_psr_sink_crc.c b/tests/kms_psr_sink_crc.c index 0a56705..46e04c8 100644 --- a/tests/kms_psr_sink_crc.c +++ b/tests/kms_psr_sink_crc.c @@ -137,9 +137,8 @@ static void fill_blt(data_t *data, uint32_t handle, unsigned char color) static void scratch_buf_init(struct igt_buf *buf, drm_intel_bo *bo) { buf->bo = bo; - buf->stride = 4096; + buf->stride = 1024; buf->tiling = I915_TILING_X; - buf->size = 4096; } static void fill_render(data_t *data, uint32_t handle, unsigned char color) @@ -167,7 +166,7 @@ static void fill_render(data_t *data, uint32_t handle, unsigned char color) igt_assert(batch); rendercopy(batch, NULL, - &src_buf, 0, 0, 1, 1, + &src_buf, 0, 0, 0xff, 0xff, &dst_buf, 0, 0); intel_batchbuffer_free(batch);
This will allow manual tests when crc isn't available. v2: Remove unused and non-sense buf->size and decrease buf->stride a bit as suggested by Daniel. Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> --- tests/kms_psr_sink_crc.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)