On U85, both NPU_SET_IFM_BROADCAST and NPU_SET_IFM2_BROADCAST must be
emitted for elementwise operations, matching Vela's GenerateInputBroadcast.
Add calc_broadcast_mode() matching Vela's CalculateBroadcast(): broadcasts
a dimension of shape1 when it is 1 and shape2 is larger, producing a
broadcast_mode bitmask (H=1, W=2, C=4, SCALAR=8).
Split emit_ifm2_broadcast into U65 (legacy bitfields) and U85 paths.
The U85 path emits both IFM_BROADCAST and IFM2_BROADCAST using
calc_broadcast_mode in each direction.
Also fix emit_eltwise to call emit_ifm2_precision instead of
emit_ifm_broadcast for U85, which was emitting 0 instead of the
required IFM2_PRECISION register.
Signed-off-by: Tomeu Vizoso <tomeu@tomeuvizoso.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39611>
Map DRM buffer objects once at resource_create and unmap at
resource_destroy, instead of mapping them in buffer_map where they
were never unmapped. This fixes a virtual memory leak that caused
SIGBUS under heavy workloads by exhausting CMA.
Also remove unused phys_addr and obj_addr fields from ethosu_resource,
and add asserts on pipe_buffer_create return values.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39611>
For U85-256 with 8-bit IFM, Vela's _uBlockToOpTable restricts which
microblocks are valid per operation type:
{2,2,8} and {4,1,8}: conv, matmul, vectorprod, reducesum, eltwise, resize
{2,1,16}: depthwise, pool, eltwise, reduceminmax, argmax, resize
Mesa's find_ublock() was not enforcing these constraints, allowing
{4,1,8} or {2,2,8} to be selected for depthwise/pooling based on
minimum waste. For depthwise ops with OFM shapes that aligned better
to {4,1,8}, the wrong ublock was chosen, causing incorrect weight
encoding and NPU hangs.
Fix by skipping {4,1,8} and {2,2,8} for depthwise/pooling operations,
matching Vela's operation-validity table.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39611>
_mesa_sha1_format has a few remaining uses, so it's moved to build_id.c,
which is its last user.
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40383>
Ethos-U can support 16-bit tensors. So far the driver just assumed 8-bit
tensors.
There's a few cases where 32-bit tensors are supported, but exactly what
those are hasn't been determined, so just reject them for now.
Signed-off-by: Rob Herring (Arm) <robh@kernel.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40525>
The output comparison expects number of elements, not bytes. Adjusting the
size was missed when adding UInt16/Int16 output comparisons.
Signed-off-by: Rob Herring (Arm) <robh@kernel.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40525>
We are going to disallow continue statements without
loop continue constructs.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39942>
We are going to disallow continue statements without
loop continue constructs.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39942>
with the unlowering pass, there is no longer a separate gl_LastFragData variable,
so this workaround just breaks color outputs
fixes dEQP-GLES31.functional.shaders.framebuffer_fetch.basic.last_frag_data
cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40437>
Previously the driver decided when the backend should use
LD_VAR_BUF[_IMM] instructions based on the total number of varyings
read, falling back to LD_VAR[_IMM] + descriptors when the varying index
could overflow the immediate index in the instructions. That means that
even adding a single varying read could overflow the index and make
everything fall back to LD_VAR.
With this patch the backend decides when to use LD_VAR_BUF for each
varying load, reporting that decision to the driver. This helps with
index overflows because only the instruction that actually overflow the
immediate use the LD_VAR fallback, leaving all other instructions on the
fast path.
Signed-off-by: Lorenzo Rossi <lorenzo.rossi@collabora.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40515>
