Hi,
This patch is part of a series that adds support for Armv8.6-A
(Matrix Multiply and BFloat16 extensions).
This patch contains some general refactoring of the atof_ieee
function, exposing a function that allows a higher level of control
over the format of IEEE-like floating point numbers.
This has been done in order to be able to add a directive for assembling
floating point literals in the bfloat16 format in the following patches.
Committed on behalf of Mihail Ionescu.
Tested on arm-none-eabi, arm-none-linux-gnueabihf, aarch64-none-elf
and aarch64-none-linux-gnuwith no issues.
gas/ChangeLog:
2019-10-21 Mihail Ionescu <mihail.ionescu@arm.com>
2019-10-21 Barnaby Wilks <barnaby.wilks@arm.com>
* as.h (atof_ieee_detail): Add prototype for atof_ieee_detail function.
(atof_ieee): Move some code into the atof_ieee_detail function.
(atof_ieee_detail): Add function that provides a higher level of control over generating
IEEE-like numbers.
Is it ok for trunk?
Regards,
Mihail
Half precision floating point numbers will be encoded using the IEEE 754
half precision floating point format - 16 bits in total, 1 for sign, 5
for exponent and 10 bits of mantissa.
This patch implements the float16 directive for both the IEEE 754 format
and the Arm alternative format for the Arm backend.
The syntax of the directive is:
.float16 <0-n decimal numbers>
e.g.
.float16 12.0
.float16 0.23, 433.1, 0.06
The Arm alternative format is almost identical to the IEEE 754 format,
except that it doesn't encode for NaNs or Infinity (instead an exponent
of 0x1F represents a normalized number in the range 65536 to 131008).
The alternative format is documented in the reference manual:
https://static.docs.arm.com/ddi0487/db/DDI0487D_b_armv8_arm.pdf?_ga=2.72318806.49764181.1561632697-999473562.1560847439
Which format is used is controlled by the .float16_format <format>
directive, where if <format> = ieee, then use the IEEE 754
half-precision format else if <format> = alternative, then use the
Arm alternative format
Or the format can be set on the command line via the -mfp16-format
option that has a similar syntax. -mfp16-format=<ieee|alternative>.
This also fixes the format and it cannot be changed by any directives.
Once the format has been set (either by the command line option or a directive) it cannot be changed,
and any attempts to change it (i.e. with the float16_format directive) will result in a warning and the
line being ignored.
For ELF targets the appropriate EABI attribute will be written out at the end of assembling
if the format has been explicitly specified. If no format has been explicitly specified then no
EABI attributes will be written.
If the format is not explicitly specified then any float16 directives are encoding using the IEEE 754-2008
format by default until the format is fixed or changed with the float16_format directive.
gas * config/tc-arm.c (enum fp_16bit_format): Add enum to represent the 2 float16 encodings.
(md_atof): Set precision for float16 type.
(arm_is_largest_exponent_ok): Check for whether to encode with the IEEE or alternative
format.
(set_fp16_format): Parse a float16_format directive.
(arm_parse_fp16_opt): Parse the fp16-format command line option.
(aeabi_set_public_attributes): For ELF encode the FP16 format EABI attribute.
* config/tc-arm.h (TC_LARGEST_EXPONENT_IS_NORMAL): Macro that expands to
arm_is_largest_exponent_ok.
(arm_is_largest_exponent_ok): Add prototype for arm_is_largest_exponent_ok function.
* doc/c-arm.texi: Add documentation for .float16, .float16_format and -mfp16-format=
* testsuite/gas/arm/float16-bad.d: New test.
* testsuite/gas/arm/float16-bad.l: New test.
* testsuite/gas/arm/float16-bad.s: New test.
* testsuite/gas/arm/float16-be.d: New test.
* testsuite/gas/arm/float16-format-bad.d: New test.
* testsuite/gas/arm/float16-format-bad.l: New test.
* testsuite/gas/arm/float16-format-bad.s: New test.
* testsuite/gas/arm/float16-format-opt-bad.d: New test.
* testsuite/gas/arm/float16-format-opt-bad.l: New test.
* testsuite/gas/arm/float16-le.d: New test.
* testsuite/gas/arm/float16.s: New test.
* testsuite/gas/arm/float16-eabi-alternative-format.d: New test.
* testsuite/gas/arm/float16-eabi-ieee-format.d: New test.
* testsuite/gas/arm/float16-eabi-no-format.d: New test.
* testsuite/gas/arm/float16-eabi.s: New test.
* config/atof-ieee.c (H_PRECISION): Macro for precision of float16
type.
(atof_ieee): Set precision and exponent bits for encoding float16
types.
(gen_to_words): NaN and Infinity encoding for float16.
(ieee_md_atof): Set precision for encoding float16 type.
PR 24538
gas * macro.c (get_any_string): Increase size of buffer used to hold
decimal value of expression result.
* dw2gencfi.c (get_debugseg_name): Handle an empty name.
* dwarf2dbg.c (get_filenum): Catch integer wraparound when
extending allocate file array.
(dwarf2_directive_filename): Add extra checks of the computed file
number.
* config/tc-arm.c (arm_tc_equal_in_insn): Insert copy of name into
warning hash table.
(s_arm_eabi_attribute): Check for obj_elf_vendor_attribute
returning -1.
* config/tc-i386.c (i386_output_nops): Catch an attempt to
generate nops of negative lengths.
* as.h (MAX_LITTLENUMS): Move definition to here from...
* config/atof-ieee.c: ...here.
* config/tc-aarch64.c: ...here.
* config/tc-arc.c: ...here.
* config/tc-arm.c: ...here.
* config/tc-epiphany.c: ...here.
* config/tc-i386.c: ...here.
* config/tc-ia64.c: ...here. (And correct the value).
* config/tc-m32c.c: ...here.
* config/tc-m32r.c: ...here.
* config/tc-metag.c: ...here.
* config/tc-microblaze.c: ...here.
* config/tc-nds32.c: ...here.
* config/tc-or1k.c: ...here.
* config/tc-score.c: ...here.
* config/tc-score7.c: ...here.
* config/tc-tic4x.c: ...here.
* config/tc-tilegx.c: ...here.
* config/tc-tilepro.c: ...here.
* config/tc-visium.c: ...here.
* config/tc-sh.c (md_assemble): Add check for an instruction with
no opcodes.
* config/tc-mips.c (mips_lookup_insn): Add check for very short
instruction name.
* config/tc-tic54x.c: Use unsigned chars to access is_end_of_line
array.
(tic54x_start_line_hook): Check for an empty line.
(next_line_shows_parallel): Do not walk off the end of the string.
(tic54x_macro_start): Check for too much macro nesting.
(tic54x_start_label): Add label_start parameter. Use this
parameter to check the first character of the label.
* config/tc-tic54x.h (TC_START_LABEL_WITHOUT_COLON): Pass
line_start variable to tic54x_start_label.
PR 24538
opcodes * ia64-opc.c (ia64_find_matching_opcode): Check for reaching the
end of the table prematurely.
Providing declarations in tc.h points out that alpha wasn't properly marking
FLT_CHARS as const. We can also get rid of the confusing redefinition of
X_CHARS to mmix_x_chars. Finally we can get rid of some now redundant
declarations of these constants.
gas/ChangeLog:
2016-04-03 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-alpha.c: Const qualify FLT_CHARS.
* config/atof-ieee.c: Remove declarations of FLT_CHARS and EXP_CHARS.
* config/tc-cris.h: Likewise.
* expr.c: Likewise.
* config/tc-mmix.c (md_atof): Adjust comment.
* config/tc-mmix.h: Stop defining FLT_CHARS and EXP_CHARS as macros.
* tc.h: Declare FLT_CHARS and EXP_CHARS.
* config/atof-ieee.c (atof_ieee): Change what_kind to int.
* config/atof-vax.c (flonum_gen2vax): Change format_letter to
int.
(md_atof): Return NULL rather than 0.
* config/tc-i386.c (md_atof): Change type to int.
Alan Modra
Matthew Malcomson
Barnaby Wilks
Nick Clifton
Trevor Saunders
H.J. Lu
Ben Elliston
Richard Sandiford
Kazu Hirata
Ian Lance Taylor
Andreas Schwab
Richard Henderson