Understanding GCC's Target Specific Options

Overview

In addition to the common options, GCC also supports a set of -m<value> target specific options. Each of them configures a particular feature or an extension. For example, -mno-code-density option disables generating code density instructions.

-mcpu=<core> selects a particular ISA and CPU family and enables/disables a set of other -m options. Each -mcpu value leads to linking with a prebuilt library which corresponds to this particular -mcpu=<value> and is built using this -mcpu=<value>.

For example, option combination -mcpu=em4 -mno-code-density will generate code that doesn't use code density instructions, however it will be linked with standard library that has been built with just -mcpu=em4, which uses code density instructions - therefore a final application still may use code density instructions.

Table of Target Options

There are several configurations files in GCC's source tree which describe target options for all targets. All of them may be found in gcc/config/arc directory:

• arc-options.def, arc-opts.h, arc.opt - definitions and descriptions for all target options.
• arc-arches.def - definitions of CPU families with all available and default target options.
• arc-cpus.def - definitions of particular -mcpu=<core> targets.

Here is a comparison table for all ARCv1 and ARCv2 CPU families.

Options	ARC HS	ARC EM	ARC 700	ARC 600
-mbarrel-shifter	Default	Yes	Default	Yes
-mnorm	Default	Yes	Default	Yes
-mswap	Default	Yes	Default	Yes
-rf16	Yes	Yes	Yes	Yes
-mfpu=<...>	Yes	Yes	—	—
-mmpy-option=<...>	Yes	Yes	—	—
-mcode-density	Default	Yes	—	—
-mdiv-rem	Yes	Yes	—	—
-matomic	Default	—	Yes	—
-mll64	Yes	—	—	—
-mspfp	—	Yes	Yes	Yes
-mdpfp	—	Yes	Yes	Yes
-msimd	—	Yes	Yes	—
-mmargonaut	—	—	Yes	Yes
-mea	—	—	Yes	—
-mmul64	—	—	—	Yes
-mmul32x16	—	—	—	Yes

Floating Point Options for ARCv2

Here is a table which matches -mfpu= values to corresponding features for ARC EM and ARC HS as they defined in Architect.

-mfpu=	Families	-has_fpu	-fpu_dp_option	-fpu_div_option	-fpu_fma_option	-fpu_dp_assist
fpus	EM, HS	On	—	—	—	—
fpus_div	EM, HS	On	—	On	—	—
fpus_fma	EM, HS	On	—	—	On	—
fpus_all	EM, HS	On	—	On	On	—
fpud	HS	On	On	—	—	—
fpud_div	HS	On	On	On	—	—
fpud_fma	HS	On	On	—	On	—
fpud_all	HS	On	On	On	On	—
fpuda	EM	On	—	—	—	On
fpuda_div	EM	On	—	On	—	On
fpuda_fma	EM	On	—	—	On	On
fpuda_all	EM	On	—	On	On	On

Multiplication Options for ARCv2

Here is a table which matches -mmpy= values to corresponding features for ARC EM and ARC HS as they defined in Architect.

-mcpu=	Aliases	Architect option for ARC HS	Architect option for ARC EM
0	none	-mpy_option=none	-mpy_option=none
1	w	—	—
2	wlh1, mpy	-mpy_option=mpy	-mpy_option=wlh1
3	wlh2	—	-mpy_option=wlh2
4	wlh3	—	-mpy_option=wlh3
5	wlh4	—	-mpy_option=wlh4
6	wlh5	—	-mpy_option=wlh5
7	plus_dmpy	-mpy_option=plus_dmpy	—
8	plus_macd	-mpy_option=plus_macd	—
9	plus_qmacw	-mpy_option=plus_qmacw	—

Values of -mcpu for ARC HS3x and HS4x Families

-mcpu=	-mdiv-rem	-matomic	-mll64	-mmpy-option=	-mfpu=
hs		Y
hs34		Y		mpy
archs	Y	Y	Y	mpy
hs38	Y	Y	Y	plus_qmacw
hs4x	Y	Y	Y	plus_qmacw
hs4xd	Y	Y	Y	plus_qmacw
hs38_linux	Y	Y	Y	plus_qmacw	fpud_all

Values of -mcpu for ARC HS5x and ARC HS6x Families

-mcpu=	-msimd	-m128	-mll64
hs5x	Y	N/A
hs58	Y	N/A	Y
hs6x	Y		N/A
hs68	Y	Y	N/A

Values of -mcpu for ARC EM Family

-mcpu	-mcode-density	-mnorm	-mswap	-mbarrel-shifter	-mdiv-rem	-mmpy-option=	-mfpu=	-mrf16
em
em_mini								Y
em4	Y
arcem	Y			Y		wlh1
em4_dmips	Y	Y	Y	Y	Y	wlh1
em4_fpus	Y	Y	Y	Y	Y	wlh1	fpus
em4_fpuda	Y	Y	Y	Y	Y	wlh1	fpuda
quarkse_em	Y	Y	Y	Y	Y	wlh2	quark

Values of -mcpu for ARC 600 and 700 Families

-mcpu=	-mnorm	-mswap	-mbarrel-shifter	Multiplier
arc700	Y	Y	Y	-mmpy
arc600			Y
arc600_norm	Y		Y
arc600_mul64	Y		Y	-mmul64
arc600_mul32x16	Y		Y	-mmul32x16
arc601
arc601_norm	Y
arc601_mul64	Y			-mmul64
arc601_mul32x16	Y			-mmul32x16