Skip to content

libclc: clspv: update gen_convert.cl for clspv #66902

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Mar 14, 2024
Merged

libclc: clspv: update gen_convert.cl for clspv #66902

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 13 additions & 2 deletions libclc/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -174,6 +174,12 @@ add_custom_command(
DEPENDS ${script_loc} )
add_custom_target( "generate_convert.cl" DEPENDS convert.cl )

add_custom_command(
OUTPUT clspv-convert.cl
COMMAND ${Python3_EXECUTABLE} ${script_loc} --clspv > clspv-convert.cl
DEPENDS ${script_loc} )
add_custom_target( "clspv-generate_convert.cl" DEPENDS clspv-convert.cl )

enable_testing()

foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
Expand Down Expand Up @@ -218,11 +224,14 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
# Add the generated convert.cl here to prevent adding
# the one listed in SOURCES
if( NOT ${ARCH} STREQUAL "spirv" AND NOT ${ARCH} STREQUAL "spirv64" )
set( rel_files convert.cl )
set( objects convert.cl )
if( NOT ENABLE_RUNTIME_SUBNORMAL AND NOT ${ARCH} STREQUAL "clspv" AND
NOT ${ARCH} STREQUAL "clspv64" )
set( rel_files convert.cl )
set( objects convert.cl )
list( APPEND rel_files generic/lib/subnormal_use_default.ll )
elseif(${ARCH} STREQUAL "clspv" OR ${ARCH} STREQUAL "clspv64")
set( rel_files clspv-convert.cl )
set( objects clspv-convert.cl )
endif()
else()
set( rel_files )
Expand Down Expand Up @@ -286,6 +295,8 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
# multiple invocations
add_dependencies( builtins.link.${arch_suffix}
generate_convert.cl )
add_dependencies( builtins.link.${arch_suffix}
clspv-generate_convert.cl )
# CMake will turn this include into absolute path
target_include_directories( builtins.link.${arch_suffix} PRIVATE
"generic/include" )
Expand Down
129 changes: 103 additions & 26 deletions libclc/generic/lib/gen_convert.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
#
# Copyright (c) 2013 Victor Oliveira <[email protected]>
# Copyright (c) 2013 Jesse Towner <[email protected]>
# Copyright (c) 2024 Romaric Jodin <[email protected]>
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
Expand All @@ -26,6 +27,16 @@
#
# convert_<destTypen><_sat><_roundingMode>(<sourceTypen>)

import argparse

parser = argparse.ArgumentParser()
parser.add_argument(
"--clspv", action="store_true", help="Generate the clspv variant of the code"
)
args = parser.parse_args()

clspv = args.clspv

types = [
"char",
"uchar",
Expand Down Expand Up @@ -251,13 +262,19 @@ def generate_default_conversion(src, dst, mode):
print("#endif")


for src in types:
for dst in types:
generate_default_conversion(src, dst, "")
# Do not generate default conversion for clspv as they are handled natively
if not clspv:
for src in types:
for dst in types:
generate_default_conversion(src, dst, "")

for src in int_types:
for dst in int_types:
for mode in rounding_modes:
# Do not generate "_rte" conversion for clspv as they are handled
# natively
if clspv and mode == "_rte":
continue
generate_default_conversion(src, dst, mode)

#
Expand Down Expand Up @@ -304,21 +321,38 @@ def generate_saturated_conversion(src, dst, size):

elif src in float_types:

# Conversion from float to int
print(
""" {DST}{N} y = convert_{DST}{N}(x);
y = select(y, ({DST}{N}){DST_MIN}, {BP}(x < ({SRC}{N}){DST_MIN}){BS});
y = select(y, ({DST}{N}){DST_MAX}, {BP}(x > ({SRC}{N}){DST_MAX}){BS});
return y;""".format(
SRC=src,
DST=dst,
N=size,
DST_MIN=limit_min[dst],
DST_MAX=limit_max[dst],
BP=bool_prefix,
BS=bool_suffix,
if clspv:
# Conversion from float to int
print(
""" {DST}{N} y = convert_{DST}{N}(x);
y = select(y, ({DST}{N}){DST_MIN}, {BP}(x <= ({SRC}{N}){DST_MIN}){BS});
y = select(y, ({DST}{N}){DST_MAX}, {BP}(x >= ({SRC}{N}){DST_MAX}){BS});
return y;""".format(
SRC=src,
DST=dst,
N=size,
DST_MIN=limit_min[dst],
DST_MAX=limit_max[dst],
BP=bool_prefix,
BS=bool_suffix,
)
)
else:
# Conversion from float to int
print(
""" {DST}{N} y = convert_{DST}{N}(x);
y = select(y, ({DST}{N}){DST_MIN}, {BP}(x < ({SRC}{N}){DST_MIN}){BS});
y = select(y, ({DST}{N}){DST_MAX}, {BP}(x > ({SRC}{N}){DST_MAX}){BS});
return y;""".format(
SRC=src,
DST=dst,
N=size,
DST_MIN=limit_min[dst],
DST_MAX=limit_max[dst],
BP=bool_prefix,
BS=bool_suffix,
)
)
)

else:

Expand Down Expand Up @@ -432,7 +466,10 @@ def generate_float_conversion(src, dst, size, mode, sat):
print(" return convert_{DST}{N}(x);".format(DST=dst, N=size))
else:
print(" {DST}{N} r = convert_{DST}{N}(x);".format(DST=dst, N=size))
print(" {SRC}{N} y = convert_{SRC}{N}(r);".format(SRC=src, N=size))
if clspv:
print(" {SRC}{N} y = convert_{SRC}{N}_sat(r);".format(SRC=src, N=size))
else:
print(" {SRC}{N} y = convert_{SRC}{N}(r);".format(SRC=src, N=size))
if mode == "_rtz":
if src in int_types:
print(
Expand All @@ -448,23 +485,59 @@ def generate_float_conversion(src, dst, size, mode, sat):
else:
print(" {SRC}{N} abs_x = fabs(x);".format(SRC=src, N=size))
print(" {SRC}{N} abs_y = fabs(y);".format(SRC=src, N=size))
print(
" return select(r, nextafter(r, sign(r) * ({DST}{N})-INFINITY), convert_{BOOL}{N}(abs_y > abs_x));".format(
DST=dst, N=size, BOOL=bool_type[dst]
if clspv:
print(
" {BOOL}{N} c = convert_{BOOL}{N}(abs_y > abs_x);".format(
BOOL=bool_type[dst], N=size
)
)
if sizeof_type[src] >= 4 and src in int_types:
print(
" c = c || convert_{BOOL}{N}(({SRC}{N}){SRC_MAX} == x);".format(
BOOL=bool_type[dst], N=size, SRC=src, SRC_MAX=limit_max[src]
)
)
print(
" return select(r, nextafter(r, sign(r) * ({DST}{N})-INFINITY), c);".format(
DST=dst, N=size, BOOL=bool_type[dst], SRC=src
)
)
else:
print(
" return select(r, nextafter(r, sign(r) * ({DST}{N})-INFINITY), convert_{BOOL}{N}(abs_y > abs_x));".format(
DST=dst, N=size, BOOL=bool_type[dst]
)
)
)
if mode == "_rtp":
print(
" return select(r, nextafter(r, ({DST}{N})INFINITY), convert_{BOOL}{N}(y < x));".format(
DST=dst, N=size, BOOL=bool_type[dst]
)
)
if mode == "_rtn":
print(
" return select(r, nextafter(r, ({DST}{N})-INFINITY), convert_{BOOL}{N}(y > x));".format(
DST=dst, N=size, BOOL=bool_type[dst]
if clspv:
print(
" {BOOL}{N} c = convert_{BOOL}{N}(y > x);".format(
BOOL=bool_type[dst], N=size
)
)
if sizeof_type[src] >= 4 and src in int_types:
print(
" c = c || convert_{BOOL}{N}(({SRC}{N}){SRC_MAX} == x);".format(
BOOL=bool_type[dst], N=size, SRC=src, SRC_MAX=limit_max[src]
)
)
print(
" return select(r, nextafter(r, ({DST}{N})-INFINITY), c);".format(
DST=dst, N=size, BOOL=bool_type[dst], SRC=src
)
)
else:
print(
" return select(r, nextafter(r, ({DST}{N})-INFINITY), convert_{BOOL}{N}(y > x));".format(
DST=dst, N=size, BOOL=bool_type[dst]
)
)
)

# Footer
print("}")
Expand All @@ -484,4 +557,8 @@ def generate_float_conversion(src, dst, size, mode, sat):
for dst in float_types:
for size in vector_sizes:
for mode in rounding_modes:
# Do not generate "_rte" conversion for clspv as they are
# handled natively
if clspv and mode == "_rte":
continue
generate_float_conversion(src, dst, size, mode, "")