OpenHarmony/third_party_mesa3d
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

nir: Split NIR_INTRINSIC_TYPE into separate src/dest indices

Browse Source 
We're about to introduce conversion ops which are going to want two
different types.  We may as well just split the one we have rather than
end up with three.  There are a couple places where this is mildly
inconvenient but most of the time I find it to actually be nicer.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6945>
This commit is contained in:
Jason Ekstrand
2020-09-30 21:20:53 -05:00
committed by Marge Bot
parent 4c70f1ba2f
commit 0aa08ae2f6
21 changed files with 93 additions and 61 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
src/compiler/nir/nir_intrinsics.py
View File

@@ -124,8 +124,10 @@ ALIGN_MUL = "NIR_INTRINSIC_ALIGN_MUL"
ALIGN_OFFSET = "NIR_INTRINSIC_ALIGN_OFFSET"
# The vulkan descriptor type for vulkan_resource_index
DESC_TYPE = "NIR_INTRINSIC_DESC_TYPE"
# The nir_alu_type of a uniform/input/output
TYPE = "NIR_INTRINSIC_TYPE"
# The nir_alu_type of input data to a store
SRC_TYPE = "NIR_INTRINSIC_SRC_TYPE"
# The nir_alu_type of the data output from a load
DEST_TYPE = "NIR_INTRINSIC_DEST_TYPE"
# The swizzle mask for quad_swizzle_amd & masked_swizzle_amd
SWIZZLE_MASK = "NIR_INTRINSIC_SWIZZLE_MASK"
# Driver location of attribute
@@ -399,8 +401,8 @@ def image(name, src_comp=[], extra_indices=[], **kwargs):
intrinsic("bindless_image_" + name, src_comp=[1] + src_comp,
indices=[IMAGE_DIM, IMAGE_ARRAY, FORMAT, ACCESS] + extra_indices, **kwargs)
image("load", src_comp=[4, 1, 1], extra_indices=[TYPE], dest_comp=0, flags=[CAN_ELIMINATE])
image("store", src_comp=[4, 1, 0, 1], extra_indices=[TYPE])
image("load", src_comp=[4, 1, 1], extra_indices=[DEST_TYPE], dest_comp=0, flags=[CAN_ELIMINATE])
image("store", src_comp=[4, 1, 0, 1], extra_indices=[SRC_TYPE])
image("atomic_add", src_comp=[4, 1, 1], dest_comp=1)
image("atomic_imin", src_comp=[4, 1, 1], dest_comp=1)
image("atomic_umin", src_comp=[4, 1, 1], dest_comp=1)
@@ -746,15 +748,15 @@ def load(name, src_comp, indices=[], flags=[]):
flags=flags)
# src[] = { offset }.
load("uniform", [1], [BASE, RANGE, TYPE], [CAN_ELIMINATE, CAN_REORDER])
load("uniform", [1], [BASE, RANGE, DEST_TYPE], [CAN_ELIMINATE, CAN_REORDER])
# src[] = { buffer_index, offset }.
load("ubo", [-1, 1], [ACCESS, ALIGN_MUL, ALIGN_OFFSET, RANGE_BASE, RANGE], flags=[CAN_ELIMINATE, CAN_REORDER])
# src[] = { buffer_index, offset in vec4 units }
load("ubo_vec4", [-1, 1], [ACCESS, COMPONENT], flags=[CAN_ELIMINATE, CAN_REORDER])
# src[] = { offset }.
load("input", [1], [BASE, COMPONENT, TYPE, IO_SEMANTICS], [CAN_ELIMINATE, CAN_REORDER])
load("input", [1], [BASE, COMPONENT, DEST_TYPE, IO_SEMANTICS], [CAN_ELIMINATE, CAN_REORDER])
# src[] = { vertex_id, offset }.
load("input_vertex", [1, 1], [BASE, COMPONENT, TYPE, IO_SEMANTICS], [CAN_ELIMINATE, CAN_REORDER])
load("input_vertex", [1, 1], [BASE, COMPONENT, DEST_TYPE, IO_SEMANTICS], [CAN_ELIMINATE, CAN_REORDER])
# src[] = { vertex, offset }.
load("per_vertex_input", [1, 1], [BASE, COMPONENT, IO_SEMANTICS], [CAN_ELIMINATE, CAN_REORDER])
# src[] = { barycoord, offset }.
@@ -794,7 +796,7 @@ def store(name, srcs, indices=[], flags=[]):
intrinsic("store_" + name, [0] + srcs, indices=indices, flags=flags)
# src[] = { value, offset }.
store("output", [1], [BASE, WRMASK, COMPONENT, TYPE, IO_SEMANTICS])
store("output", [1], [BASE, WRMASK, COMPONENT, SRC_TYPE, IO_SEMANTICS])
# src[] = { value, vertex, offset }.
store("per_vertex_output", [1, 1], [BASE, WRMASK, COMPONENT, IO_SEMANTICS])
# src[] = { value, block_index, offset }
@@ -928,7 +930,7 @@ load("tlb_color_v3d", [1], [BASE, COMPONENT], [])
#
# src[] = { value, render_target }
# BASE = sample index
store("tlb_sample_color_v3d", [1], [BASE, COMPONENT, TYPE], [])
store("tlb_sample_color_v3d", [1], [BASE, COMPONENT, SRC_TYPE], [])
# V3D-specific intrinsic to load the number of layers attached to
# the target framebuffer