From: "Li, Pan2" <pan2.li@intel.com>
To: Richard Biener <richard.guenther@gmail.com>
Cc: "gcc-patches@gcc.gnu.org" <gcc-patches@gcc.gnu.org>,
"juzhe.zhong@rivai.ai" <juzhe.zhong@rivai.ai>,
"Wang, Yanzhang" <yanzhang.wang@intel.com>,
"kito.cheng@gmail.com" <kito.cheng@gmail.com>,
"jeffreyalaw@gmail.com" <jeffreyalaw@gmail.com>
Subject: RE: [PATCH v3] RISC-V: Bugfix for doesn't honor no-signed-zeros option
Date: Tue, 9 Jan 2024 01:22:06 +0000 [thread overview]
Message-ID: <MW5PR11MB5908A17D1E8497ACFE6BB323A96A2@MW5PR11MB5908.namprd11.prod.outlook.com> (raw)
In-Reply-To: <CAFiYyc1Uw99w7SXTaM_oyGzVCVqd3vZD89-yF0Fe-7vCKvVmWQ@mail.gmail.com>
Thanks Richard B for comments.
> We don't really expect targets to do this. The small testcase above
> is somewhat ill-formed with -fno-signed-zeros. Note there's no
> -0.0 in pr30957-1.c so why does that one fail for you? Does
> the -fvariable-expansion-in-unroller code maybe not trigger for
> riscv?
Sorry this confused me a little about the sematics of the option -fno-signed-zeros.
I wonder what the target/backend need to do for this option.
About the failure, it comes from below code in pr30957-1.c. The 0.0 / -5.0 is initialized to -0.0 in riscv but +0.0 in aarch64.
if (__builtin_copysignf (1.0, foo (0.0 / -5.0, 10)) != 1.0)
abort ();
If my understanding is correct, the loop will be vectorized during vect_transform_loop with a variable factor.
It won't benefit from unrolling/peeling and mark the loop->unroll as 1, and then we have tree-vect log similar to below:
Disabling unrolling due to variable-length vectorization factor.
> I think we should go to PR30957 and see what that was filed originally
> for, the testcase doesn't make much sense to me.
Sure thing, will take a look and back to you later.
Pan
-----Original Message-----
From: Richard Biener <richard.guenther@gmail.com>
Sent: Monday, January 8, 2024 6:45 PM
To: Li, Pan2 <pan2.li@intel.com>
Cc: gcc-patches@gcc.gnu.org; juzhe.zhong@rivai.ai; Wang, Yanzhang <yanzhang.wang@intel.com>; kito.cheng@gmail.com; jeffreyalaw@gmail.com
Subject: Re: [PATCH v3] RISC-V: Bugfix for doesn't honor no-signed-zeros option
On Tue, Jan 2, 2024 at 2:37 PM <pan2.li@intel.com> wrote:
>
> From: Pan Li <pan2.li@intel.com>
>
> According to the sematics of no-signed-zeros option, the backend
> like RISC-V should treat the minus zero -0.0f as plus zero 0.0f.
>
> Consider below example with option -fno-signed-zeros.
>
> void
> test (float *a)
> {
> *a = -0.0;
> }
>
> We will generate code as below, which doesn't treat the minus zero
> as plus zero.
>
> test:
> lui a5,%hi(.LC0)
> flw fa5,%lo(.LC0)(a5)
> fsw fa5,0(a0)
> ret
>
> .LC0:
> .word -2147483648 // aka -0.0 (0x80000000 in hex)
>
> This patch would like to fix the bug and treat the minus zero -0.0
> as plus zero, aka +0.0. Thus after this patch we will have asm code
> as below for the above sampe code.
>
> test:
> sw zero,0(a0)
> ret
>
> This patch also fix the run failure of the test case pr30957-1.c. The
> below tests are passed for this patch.
We don't really expect targets to do this. The small testcase above
is somewhat ill-formed with -fno-signed-zeros. Note there's no
-0.0 in pr30957-1.c so why does that one fail for you? Does
the -fvariable-expansion-in-unroller code maybe not trigger for
riscv?
I think we should go to PR30957 and see what that was filed originally
for, the testcase doesn't make much sense to me.
> * The riscv regression tests.
> * The pr30957-1.c run tests.
>
> gcc/ChangeLog:
>
> * config/riscv/constraints.md: Leverage func riscv_float_const_zero_rtx_p
> for predicating the rtx is const zero float or not.
> * config/riscv/predicates.md: Ditto.
> * config/riscv/riscv.cc (riscv_const_insns): Ditto.
> (riscv_float_const_zero_rtx_p): New func impl for predicating the rtx is
> const zero float or not.
> (riscv_const_zero_rtx_p): New func impl for predicating the rtx
> is const zero (both int and fp) or not.
> * config/riscv/riscv-protos.h (riscv_float_const_zero_rtx_p):
> New func decl.
> (riscv_const_zero_rtx_p): Ditto.
> * config/riscv/riscv.md: Making sure the operand[1] of movfp is
> CONST0_RTX when the operand[1] is const zero float.
>
> gcc/testsuite/ChangeLog:
>
> * gcc.target/riscv/no-signed-zeros-0.c: New test.
> * gcc.target/riscv/no-signed-zeros-1.c: New test.
> * gcc.target/riscv/no-signed-zeros-2.c: New test.
> * gcc.target/riscv/no-signed-zeros-3.c: New test.
> * gcc.target/riscv/no-signed-zeros-4.c: New test.
> * gcc.target/riscv/no-signed-zeros-5.c: New test.
> * gcc.target/riscv/no-signed-zeros-run-0.c: New test.
> * gcc.target/riscv/no-signed-zeros-run-1.c: New test.
>
> Signed-off-by: Pan Li <pan2.li@intel.com>
> ---
> gcc/config/riscv/constraints.md | 2 +-
> gcc/config/riscv/predicates.md | 2 +-
> gcc/config/riscv/riscv-protos.h | 2 +
> gcc/config/riscv/riscv.cc | 35 ++++++++++++-
> gcc/config/riscv/riscv.md | 49 ++++++++++++++++---
> .../gcc.target/riscv/no-signed-zeros-0.c | 26 ++++++++++
> .../gcc.target/riscv/no-signed-zeros-1.c | 28 +++++++++++
> .../gcc.target/riscv/no-signed-zeros-2.c | 26 ++++++++++
> .../gcc.target/riscv/no-signed-zeros-3.c | 28 +++++++++++
> .../gcc.target/riscv/no-signed-zeros-4.c | 26 ++++++++++
> .../gcc.target/riscv/no-signed-zeros-5.c | 28 +++++++++++
> .../gcc.target/riscv/no-signed-zeros-run-0.c | 36 ++++++++++++++
> .../gcc.target/riscv/no-signed-zeros-run-1.c | 36 ++++++++++++++
> 13 files changed, 314 insertions(+), 10 deletions(-)
> create mode 100644 gcc/testsuite/gcc.target/riscv/no-signed-zeros-0.c
> create mode 100644 gcc/testsuite/gcc.target/riscv/no-signed-zeros-1.c
> create mode 100644 gcc/testsuite/gcc.target/riscv/no-signed-zeros-2.c
> create mode 100644 gcc/testsuite/gcc.target/riscv/no-signed-zeros-3.c
> create mode 100644 gcc/testsuite/gcc.target/riscv/no-signed-zeros-4.c
> create mode 100644 gcc/testsuite/gcc.target/riscv/no-signed-zeros-5.c
> create mode 100644 gcc/testsuite/gcc.target/riscv/no-signed-zeros-run-0.c
> create mode 100644 gcc/testsuite/gcc.target/riscv/no-signed-zeros-run-1.c
>
> diff --git a/gcc/config/riscv/constraints.md b/gcc/config/riscv/constraints.md
> index de4359af00d..db1d5e1385f 100644
> --- a/gcc/config/riscv/constraints.md
> +++ b/gcc/config/riscv/constraints.md
> @@ -108,7 +108,7 @@ (define_constraint "DnS"
> (define_constraint "G"
> "@internal"
> (and (match_code "const_double")
> - (match_test "op == CONST0_RTX (mode)")))
> + (match_test "riscv_float_const_zero_rtx_p (op)")))
>
> (define_memory_constraint "A"
> "An address that is held in a general-purpose register."
> diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md
> index b87a6900841..b428d842101 100644
> --- a/gcc/config/riscv/predicates.md
> +++ b/gcc/config/riscv/predicates.md
> @@ -78,7 +78,7 @@ (define_predicate "sleu_operand"
>
> (define_predicate "const_0_operand"
> (and (match_code "const_int,const_wide_int,const_double,const_vector")
> - (match_test "op == CONST0_RTX (GET_MODE (op))")))
> + (match_test "riscv_const_zero_rtx_p (op)")))
>
> (define_predicate "const_1_operand"
> (and (match_code "const_int,const_wide_int,const_vector")
> diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h
> index 31049ef7523..fcf30e084a3 100644
> --- a/gcc/config/riscv/riscv-protos.h
> +++ b/gcc/config/riscv/riscv-protos.h
> @@ -131,6 +131,8 @@ extern void riscv_asm_output_external (FILE *, const tree, const char *);
> extern bool
> riscv_zcmp_valid_stack_adj_bytes_p (HOST_WIDE_INT, int);
> extern void riscv_legitimize_poly_move (machine_mode, rtx, rtx, rtx);
> +extern bool riscv_float_const_zero_rtx_p (rtx);
> +extern bool riscv_const_zero_rtx_p (rtx);
>
> #ifdef RTX_CODE
> extern void riscv_expand_int_scc (rtx, enum rtx_code, rtx, rtx, bool *invert_ptr = 0);
> diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc
> index 0d1cbc5cb5f..a8ad86b7068 100644
> --- a/gcc/config/riscv/riscv.cc
> +++ b/gcc/config/riscv/riscv.cc
> @@ -1635,7 +1635,7 @@ riscv_const_insns (rtx x)
> return 1;
>
> /* We can use x0 to load floating-point zero. */
> - return x == CONST0_RTX (GET_MODE (x)) ? 1 : 0;
> + return riscv_float_const_zero_rtx_p (x) ? 1 : 0;
> case CONST_VECTOR:
> {
> /* TODO: This is not accurate, we will need to
> @@ -9481,6 +9481,39 @@ riscv_zcmp_valid_stack_adj_bytes_p (HOST_WIDE_INT total, int regs_num)
> || additioanl_bytes == ZCMP_MAX_SPIMM * ZCMP_SP_INC_STEP;
> }
>
> +/* Return true if rtx op is const zero of the floating-point. */
> +bool
> +riscv_float_const_zero_rtx_p (rtx op)
> +{
> + machine_mode mode = GET_MODE (op);
> +
> + if (GET_MODE_CLASS (mode) != MODE_FLOAT)
> + return false;
> +
> + if (GET_CODE (op) != CONST_DOUBLE)
> + return false;
> +
> + const struct real_value *op_rvalue = CONST_DOUBLE_REAL_VALUE (op);
> +
> + if (REAL_VALUE_MINUS_ZERO (*op_rvalue))
> + return !HONOR_SIGNED_ZEROS (mode);
> +
> + return real_equal (op_rvalue, &dconst0);
> +}
> +
> +/* Return true if rtx op is const zero, include both the integer
> + and floating-point. */
> +bool
> +riscv_const_zero_rtx_p (rtx op)
> +{
> + machine_mode mode = GET_MODE (op);
> +
> + if (GET_MODE_CLASS (mode) == MODE_FLOAT)
> + return riscv_float_const_zero_rtx_p (op);
> +
> + return op == CONST0_RTX (mode);
> +}
> +
> /* Return true if it's valid gpr_save pattern. */
>
> bool
> diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md
> index 68f7203b676..cd429f6dcdd 100644
> --- a/gcc/config/riscv/riscv.md
> +++ b/gcc/config/riscv/riscv.md
> @@ -1878,7 +1878,12 @@ (define_insn "*movhf_hardfloat"
> "TARGET_ZFHMIN
> && (register_operand (operands[0], HFmode)
> || reg_or_0_operand (operands[1], HFmode))"
> - { return riscv_output_move (operands[0], operands[1]); }
> + {
> + if (riscv_float_const_zero_rtx_p (operands[1]))
> + operands[1] = CONST0_RTX (GET_MODE (operands[1]));
> +
> + return riscv_output_move (operands[0], operands[1]);
> + }
> [(set_attr "move_type" "fmove,fmove,mtc,fpload,fpstore,store,mtc,mfc,move,load,store")
> (set_attr "type" "fmove")
> (set_attr "mode" "HF")])
> @@ -1889,7 +1894,12 @@ (define_insn "*movhf_softfloat"
> "!TARGET_ZFHMIN
> && (register_operand (operands[0], HFmode)
> || reg_or_0_operand (operands[1], HFmode))"
> - { return riscv_output_move (operands[0], operands[1]); }
> + {
> + if (riscv_float_const_zero_rtx_p (operands[1]))
> + operands[1] = CONST0_RTX (GET_MODE (operands[1]));
> +
> + return riscv_output_move (operands[0], operands[1]);
> + }
> [(set_attr "move_type" "fmove,move,load,store,mtc,mfc")
> (set_attr "type" "fmove")
> (set_attr "mode" "HF")])
> @@ -2243,7 +2253,12 @@ (define_insn "*movsf_hardfloat"
> "TARGET_HARD_FLOAT
> && (register_operand (operands[0], SFmode)
> || reg_or_0_operand (operands[1], SFmode))"
> - { return riscv_output_move (operands[0], operands[1]); }
> + {
> + if (riscv_float_const_zero_rtx_p (operands[1]))
> + operands[1] = CONST0_RTX (GET_MODE (operands[1]));
> +
> + return riscv_output_move (operands[0], operands[1]);
> + }
> [(set_attr "move_type" "fmove,fmove,mtc,fpload,fpstore,store,mtc,mfc,move,load,store")
> (set_attr "type" "fmove")
> (set_attr "mode" "SF")])
> @@ -2254,7 +2269,12 @@ (define_insn "*movsf_softfloat"
> "!TARGET_HARD_FLOAT
> && (register_operand (operands[0], SFmode)
> || reg_or_0_operand (operands[1], SFmode))"
> - { return riscv_output_move (operands[0], operands[1]); }
> + {
> + if (riscv_float_const_zero_rtx_p (operands[1]))
> + operands[1] = CONST0_RTX (GET_MODE (operands[1]));
> +
> + return riscv_output_move (operands[0], operands[1]);
> + }
> [(set_attr "move_type" "move,load,store")
> (set_attr "type" "fmove")
> (set_attr "mode" "SF")])
> @@ -2279,7 +2299,12 @@ (define_insn "*movdf_hardfloat_rv32"
> "!TARGET_64BIT && TARGET_DOUBLE_FLOAT
> && (register_operand (operands[0], DFmode)
> || reg_or_0_operand (operands[1], DFmode))"
> - { return riscv_output_move (operands[0], operands[1]); }
> + {
> + if (riscv_float_const_zero_rtx_p (operands[1]))
> + operands[1] = CONST0_RTX (GET_MODE (operands[1]));
> +
> + return riscv_output_move (operands[0], operands[1]);
> + }
> [(set_attr "move_type" "fmove,fmove,mtc,fpload,fpstore,store,mtc,mfc,move,load,store")
> (set_attr "type" "fmove")
> (set_attr "mode" "DF")])
> @@ -2290,7 +2315,12 @@ (define_insn "*movdf_hardfloat_rv64"
> "TARGET_64BIT && TARGET_DOUBLE_FLOAT
> && (register_operand (operands[0], DFmode)
> || reg_or_0_operand (operands[1], DFmode))"
> - { return riscv_output_move (operands[0], operands[1]); }
> + {
> + if (riscv_float_const_zero_rtx_p (operands[1]))
> + operands[1] = CONST0_RTX (GET_MODE (operands[1]));
> +
> + return riscv_output_move (operands[0], operands[1]);
> + }
> [(set_attr "move_type" "fmove,fmove,mtc,fpload,fpstore,store,mtc,mfc,move,load,store")
> (set_attr "type" "fmove")
> (set_attr "mode" "DF")])
> @@ -2301,7 +2331,12 @@ (define_insn "*movdf_softfloat"
> "!TARGET_DOUBLE_FLOAT
> && (register_operand (operands[0], DFmode)
> || reg_or_0_operand (operands[1], DFmode))"
> - { return riscv_output_move (operands[0], operands[1]); }
> + {
> + if (riscv_float_const_zero_rtx_p (operands[1]))
> + operands[1] = CONST0_RTX (GET_MODE (operands[1]));
> +
> + return riscv_output_move (operands[0], operands[1]);
> + }
> [(set_attr "move_type" "move,load,store")
> (set_attr "type" "fmove")
> (set_attr "mode" "DF")])
> diff --git a/gcc/testsuite/gcc.target/riscv/no-signed-zeros-0.c b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-0.c
> new file mode 100644
> index 00000000000..1eda13a3406
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-0.c
> @@ -0,0 +1,26 @@
> +/* { dg-do compile } */
> +/* { dg-options "-march=rv64gcv -mabi=lp64d -O3 -fno-signed-zeros -fno-schedule-insns -fno-schedule-insns2" } */
> +/* { dg-skip-if "" { *-*-* } { "-flto" } } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> +
> +/*
> +** test_float_plus_zero_assign:
> +** sw\s+zero,0\([atx][0-9]+\)
> +** ret
> +*/
> +void
> +test_float_plus_zero_assign (float *a)
> +{
> + *a = +0.0;
> +}
> +
> +/*
> +** test_float_plus_zero_assign:
> +** sw\s+zero,0\([atx][0-9]+\)
> +** ret
> +*/
> +void
> +test_float_minus_zero_assign (float *a)
> +{
> + *a = -0.0;
> +}
> diff --git a/gcc/testsuite/gcc.target/riscv/no-signed-zeros-1.c b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-1.c
> new file mode 100644
> index 00000000000..8041ec3ea95
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-1.c
> @@ -0,0 +1,28 @@
> +/* { dg-do compile } */
> +/* { dg-options "-march=rv64gcv -mabi=lp64d -O3 -fno-signed-zeros -fno-schedule-insns -fno-schedule-insns2" } */
> +/* { dg-skip-if "" { *-*-* } { "-flto" } } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> +
> +extern void float_assign (float *a, float b);
> +
> +/*
> +** test_float_plus_zero_assign:
> +** fmv\.s\.x\s+fa[0-9]+,\s*zero
> +** tail\s+float_assign
> +*/
> +void
> +test_float_plus_zero_assign (float *a)
> +{
> + float_assign (a, +0.0);
> +}
> +
> +/*
> +** test_float_minus_zero_assign:
> +** fmv\.s\.x\s+fa[0-9]+,\s*zero
> +** tail\s+float_assign
> +*/
> +void
> +test_float_minus_zero_assign (float *a)
> +{
> + float_assign (a, -0.0);
> +}
> diff --git a/gcc/testsuite/gcc.target/riscv/no-signed-zeros-2.c b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-2.c
> new file mode 100644
> index 00000000000..a6996dae4de
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-2.c
> @@ -0,0 +1,26 @@
> +/* { dg-do compile } */
> +/* { dg-options "-march=rv64gcv -mabi=lp64d -O3 -fno-signed-zeros -fno-schedule-insns -fno-schedule-insns2" } */
> +/* { dg-skip-if "" { *-*-* } { "-flto" } } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> +
> +/*
> +** test_float_plus_zero_assign:
> +** sd\s+zero,0\([atx][0-9]+\)
> +** ret
> +*/
> +void
> +test_float_plus_zero_assign (double *a)
> +{
> + *a = +0.0;
> +}
> +
> +/*
> +** test_float_plus_zero_assign:
> +** sd\s+zero,0\([atx][0-9]+\)
> +** ret
> +*/
> +void
> +test_float_minus_zero_assign (double *a)
> +{
> + *a = -0.0;
> +}
> diff --git a/gcc/testsuite/gcc.target/riscv/no-signed-zeros-3.c b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-3.c
> new file mode 100644
> index 00000000000..b4ba8a247df
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-3.c
> @@ -0,0 +1,28 @@
> +/* { dg-do compile } */
> +/* { dg-options "-march=rv64gcv -mabi=lp64d -O3 -fno-signed-zeros -fno-schedule-insns -fno-schedule-insns2" } */
> +/* { dg-skip-if "" { *-*-* } { "-flto" } } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> +
> +extern void float_assign (double *a, double b);
> +
> +/*
> +** test_float_plus_zero_assign:
> +** fmv\.d\.x\s+fa[0-9]+,\s*zero
> +** tail\s+float_assign
> +*/
> +void
> +test_float_plus_zero_assign (double *a)
> +{
> + float_assign (a, +0.0);
> +}
> +
> +/*
> +** test_float_minus_zero_assign:
> +** fmv\.d\.x\s+fa[0-9]+,\s*zero
> +** tail\s+float_assign
> +*/
> +void
> +test_float_minus_zero_assign (double *a)
> +{
> + float_assign (a, -0.0);
> +}
> diff --git a/gcc/testsuite/gcc.target/riscv/no-signed-zeros-4.c b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-4.c
> new file mode 100644
> index 00000000000..60acf7155d3
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-4.c
> @@ -0,0 +1,26 @@
> +/* { dg-do compile } */
> +/* { dg-options "-march=rv64gcv_zfh -mabi=lp64d -O3 -fno-signed-zeros -fno-schedule-insns -fno-schedule-insns2" } */
> +/* { dg-skip-if "" { *-*-* } { "-flto" } } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> +
> +/*
> +** test_float_plus_zero_assign:
> +** sh\s+zero,0\([atx][0-9]+\)
> +** ret
> +*/
> +void
> +test_float_plus_zero_assign (_Float16 *a)
> +{
> + *a = +0.0;
> +}
> +
> +/*
> +** test_float_plus_zero_assign:
> +** sh\s+zero,0\([atx][0-9]+\)
> +** ret
> +*/
> +void
> +test_float_minus_zero_assign (_Float16 *a)
> +{
> + *a = -0.0;
> +}
> diff --git a/gcc/testsuite/gcc.target/riscv/no-signed-zeros-5.c b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-5.c
> new file mode 100644
> index 00000000000..d10efbeb37b
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-5.c
> @@ -0,0 +1,28 @@
> +/* { dg-do compile } */
> +/* { dg-options "-march=rv64gcv_zfh -mabi=lp64d -O3 -fno-signed-zeros -fno-schedule-insns -fno-schedule-insns2" } */
> +/* { dg-skip-if "" { *-*-* } { "-flto" } } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> +
> +extern void float_assign (_Float16 *a, _Float16 b);
> +
> +/*
> +** test_float_plus_zero_assign:
> +** fmv\.h\.x\s+fa[0-9]+,\s*zero
> +** tail\s+float_assign
> +*/
> +void
> +test_float_plus_zero_assign (_Float16 *a)
> +{
> + float_assign (a, +0.0);
> +}
> +
> +/*
> +** test_float_minus_zero_assign:
> +** fmv\.h\.x\s+fa[0-9]+,\s*zero
> +** tail\s+float_assign
> +*/
> +void
> +test_float_minus_zero_assign (_Float16 *a)
> +{
> + float_assign (a, -0.0);
> +}
> diff --git a/gcc/testsuite/gcc.target/riscv/no-signed-zeros-run-0.c b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-run-0.c
> new file mode 100644
> index 00000000000..347e4b0ff74
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-run-0.c
> @@ -0,0 +1,36 @@
> +/* { dg-do run { target { riscv_v } } } */
> +/* { dg-additional-options "-std=c99 -O3 -fno-signed-zeros" } */
> +
> +float f_val;
> +
> +void
> +__attribute__ ((noinline))
> +test_float_plus_zero_assign (float *a)
> +{
> + *a = +0.0;
> +}
> +
> +void
> +__attribute__ ((noinline))
> +test_float_minus_zero_assign (float *a)
> +{
> + *a = -0.0;
> +}
> +
> +int
> +main ()
> +{
> + f_val = -1.0;
> + test_float_plus_zero_assign (&f_val);
> +
> + if (__builtin_copysignf (1.0, f_val) != 1.0)
> + __builtin_abort ();
> +
> + f_val = -1.0;
> + test_float_minus_zero_assign (&f_val);
> +
> + if (__builtin_copysignf (1.0, f_val) != 1.0)
> + __builtin_abort ();
> +
> + return 0;
> +}
> diff --git a/gcc/testsuite/gcc.target/riscv/no-signed-zeros-run-1.c b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-run-1.c
> new file mode 100644
> index 00000000000..1eb1edba457
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/no-signed-zeros-run-1.c
> @@ -0,0 +1,36 @@
> +/* { dg-do run { target { riscv_v } } } */
> +/* { dg-additional-options "-std=c99 -O3 -fno-signed-zeros" } */
> +
> +double f_val;
> +
> +void
> +__attribute__ ((noinline))
> +test_float_plus_zero_assign (double *a)
> +{
> + *a = +0.0;
> +}
> +
> +void
> +__attribute__ ((noinline))
> +test_float_minus_zero_assign (double *a)
> +{
> + *a = -0.0;
> +}
> +
> +int
> +main ()
> +{
> + f_val = -1.0;
> + test_float_plus_zero_assign (&f_val);
> +
> + if (__builtin_copysignf (1.0, f_val) != 1.0)
> + __builtin_abort ();
> +
> + f_val = -1.0;
> + test_float_minus_zero_assign (&f_val);
> +
> + if (__builtin_copysignf (1.0, f_val) != 1.0)
> + __builtin_abort ();
> +
> + return 0;
> +}
> --
> 2.34.1
>
next prev parent reply other threads:[~2024-01-09 1:22 UTC|newest]
Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-12-23 11:07 [PATCH v1] RISC-V: XFAIL pr30957-1.c when loop vectorized with variable factor pan2.li
2023-12-23 17:19 ` Jeff Law
2023-12-24 2:01 ` Li, Pan2
2023-12-26 9:34 ` [PATCH v2] " pan2.li
2023-12-28 16:39 ` Jeff Law
2023-12-29 0:42 ` Li, Pan2
2023-12-29 1:03 ` Jeff Law
2023-12-29 5:56 ` Li, Pan2
2023-12-30 3:13 ` Jeff Law
2024-01-01 8:56 ` Li, Pan2
2024-01-02 11:55 ` [PATCH v3] RISC-V: Bugfix for doesn't honor no-signed-zeros option pan2.li
2024-01-08 10:45 ` Richard Biener
2024-01-09 1:22 ` Li, Pan2 [this message]
2024-01-09 7:17 ` Li, Pan2
2024-01-09 13:08 ` Richard Biener
2024-01-09 17:46 ` Jeff Law
2024-01-10 4:28 ` Li, Pan2
2024-01-11 1:38 ` [PATCH v4] LOOP-UNROLL: Leverage HAS_SIGNED_ZERO for var expansion pan2.li
2024-01-11 8:33 ` Richard Biener
2024-01-11 8:48 ` Li, Pan2
2024-01-11 8:50 ` [PATCH v5] " pan2.li
2024-01-11 9:21 ` Richard Biener
2024-01-11 10:35 ` Li, Pan2
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=MW5PR11MB5908A17D1E8497ACFE6BB323A96A2@MW5PR11MB5908.namprd11.prod.outlook.com \
--to=pan2.li@intel.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=jeffreyalaw@gmail.com \
--cc=juzhe.zhong@rivai.ai \
--cc=kito.cheng@gmail.com \
--cc=richard.guenther@gmail.com \
--cc=yanzhang.wang@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).