nir: Add load_output_u8_as_fp16_pan intrinsic

This is a single opcode, at least on newer Midgard chips. It's easier to
have this represented in NIR rather than trying to optimize out the
conversions, so let's add the intrinsic.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>

diff --git a/src/compiler/nir/nir_intrinsics.py b/src/compiler/nir/nir_intrinsics.py
index 4658187ea16907fd0358042fa243414c44791ac7..ce10dd22d6cd0d72cdfcae82a27042623627c26e 100644 (file)
--- a/src/compiler/nir/nir_intrinsics.py
+++ b/src/compiler/nir/nir_intrinsics.py
@@ -836,10 +836,16 @@ intrinsic("load_global_ir3", [2, 1], dest_comp=0, indices=[ACCESS, ALIGN_MUL, AL
 # One notable divergence is sRGB, which is asymmetric: raw_input_pan requires
 # an sRGB->linear conversion, but linear values should be written to
 # raw_output_pan and the hardware handles linear->sRGB.
+#
+# We also have format-specific Midgard intrinsics. There are rather
+# here-be-dragons. load_output_u8_as_fp16_pan does the equivalent of
+# load_raw_out_pan on an RGBA8 UNORM framebuffer followed by u2u16 -> fp16 ->
+# division by 255.
 
 # src[] = { value }
 store("raw_output_pan", 1, [])
 load("raw_output_pan", 0, [], [CAN_ELIMINATE, CAN_REORDER])
+load("output_u8_as_fp16_pan", 0, [], [CAN_ELIMINATE, CAN_REORDER])
 
 # V3D-specific instrinc for tile buffer color reads.
 #