Commit ec5e6814 by Thomas Preud'homme Committed by Thomas Preud'homme

[ARM] Do softfloat when -mfpu set, -mfloat-abi=softfp

FP instructions are only enabled for TARGET_32BIT and TARGET_HARD_FLOAT
but GCC only gives an error when TARGET_HARD_FLOAT is true and -mfpu is
not set. Among other things, it makes some of the cmse tests (eg.
gcc.target/arm/cmse/baseline/softfp.c) fail when targeting
-march=armv8-m.base -mcmse -mfpu=<something> -mfloat-abi=softfp. This
commit adds an extra check for TARGET_32BIT to TARGET_HARD_FLOAT such
that it is false on TARGET_THUMB1 targets even when a FPU is specified.

2018-12-19  thomas Preud'homme  <thomas.preudhomme@linaro.org>

    gcc/
    * config/arm/arm.h (TARGET_HARD_FLOAT): Restrict to TARGET_32BIT
    targets.
    * config/arm/arm.c (output_return_instruction): Only check
    TARGET_HARD_FLOAT to decide whether FP instructions are available.

    gcc/testsuite/
    * gcc.target/arm/cmse/baseline/softfp.c: Force an FPU.

From-SVN: r267270
parent a152954e
2018-12-19 Thomas Preud'homme <thomas.preudhomme@linaro.org>
* config/arm/arm.h (TARGET_HARD_FLOAT): Restrict to TARGET_32BIT
targets.
* config/arm/arm.c (output_return_instruction): Only check
TARGET_HARD_FLOAT to decide whether FP instructions are available.
2018-12-19 Tom de Vries <tom@codesourcery.com> 2018-12-19 Tom de Vries <tom@codesourcery.com>
* doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
...@@ -19872,7 +19872,7 @@ output_return_instruction (rtx operand, bool really_return, bool reverse, ...@@ -19872,7 +19872,7 @@ output_return_instruction (rtx operand, bool really_return, bool reverse,
"msr%s\tAPSR_nzcvq, %%|lr", conditional); "msr%s\tAPSR_nzcvq, %%|lr", conditional);
output_asm_insn (instr, & operand); output_asm_insn (instr, & operand);
if (TARGET_HARD_FLOAT && !TARGET_THUMB1) if (TARGET_HARD_FLOAT)
{ {
/* Clear the cumulative exception-status bits (0-4,7) and the /* Clear the cumulative exception-status bits (0-4,7) and the
condition code bits (28-31) of the FPSCR. We need to condition code bits (28-31) of the FPSCR. We need to
......
...@@ -125,7 +125,8 @@ extern tree arm_fp16_type_node; ...@@ -125,7 +125,8 @@ extern tree arm_fp16_type_node;
/* Use hardware floating point instructions. */ /* Use hardware floating point instructions. */
#define TARGET_HARD_FLOAT (arm_float_abi != ARM_FLOAT_ABI_SOFT \ #define TARGET_HARD_FLOAT (arm_float_abi != ARM_FLOAT_ABI_SOFT \
&& bitmap_bit_p (arm_active_target.isa, \ && bitmap_bit_p (arm_active_target.isa, \
isa_bit_vfpv2)) isa_bit_vfpv2) \
&& TARGET_32BIT)
#define TARGET_SOFT_FLOAT (!TARGET_HARD_FLOAT) #define TARGET_SOFT_FLOAT (!TARGET_HARD_FLOAT)
/* User has permitted use of FP instructions, if they exist for this /* User has permitted use of FP instructions, if they exist for this
target. */ target. */
......
2018-12-19 Thomas Preud'homme <thomas.preudhomme@linaro.org>
* gcc.target/arm/cmse/baseline/softfp.c: Force an FPU.
2018-12-19 Tom de Vries <tdevries@suse.de> 2018-12-19 Tom de Vries <tdevries@suse.de>
* gcc.dg/goacc/nvptx-merged-loop.c: Move to * gcc.dg/goacc/nvptx-merged-loop.c: Move to
......
/* { dg-do compile } */ /* { dg-do compile } */
/* { dg-options "-mcmse -mfloat-abi=softfp" } */ /* Force an FPU to test that it is ignored for Thumb-1 -like targets and that
no clearing of VFP register occurs. */
/* { dg-options "-mcmse -mfloat-abi=softfp -mfpu=fpv5-d16" } */
double __attribute__ ((cmse_nonsecure_call)) (*bar) (float, double); double __attribute__ ((cmse_nonsecure_call)) (*bar) (float, double);
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment