[v2] xtensa: Optimize boolean evaluation or branching when EQ/NE to zero in S[IF]mode
Checks
Commit Message
This patch optimizes the boolean evaluation of EQ/NE against zero
by adding two insn_and_split patterns similar to SImode conditional
store:
"eq_zero":
op0 = (op1 == 0) ? 1 : 0;
op0 = clz(op1) >> 5; /* optimized (requires TARGET_NSA) */
"movsicc_ne0_reg_0":
op0 = (op1 != 0) ? op2 : 0;
op0 = op2; if (op1 == 0) ? op0 = op1; /* optimized */
/* example #1 */
int bool_eqSI(int x) {
return x == 0;
}
int bool_neSI(int x) {
return x != 0;
}
;; after (TARGET_NSA)
bool_eqSI:
nsau a2, a2
srli a2, a2, 5
ret.n
bool_neSI:
mov.n a9, a2
movi.n a2, 1
moveqz a2, a9, a9
ret.n
These also work in SFmode by ignoring their sign bits, and further-
more, the branch if EQ/NE against zero in SFmode is also done in the
same manner.
The reasons for this optimization in SFmode are:
- Only zero values (negative or non-negative) contain no bits of 1
with both the exponent and the mantissa.
- EQ/NE comparisons involving NaNs produce no signal even if they
are signaling.
- Even if the use of IEEE 754 single-precision floating-point co-
processor is configured (TARGET_HARD_FLOAT is true):
1. Load zero value to FP register
2. Possibly, additional FP move if the comparison target is
an address register
3. FP equality check instruction
4. Read the boolean register containing the result, or condi-
tional branch
As noted above, a considerable number of instructions are still
generated.
/* example #2 */
int bool_eqSF(float x) {
return x == 0;
}
int bool_neSF(float x) {
return x != 0;
}
int bool_ltSF(float x) {
return x < 0;
}
extern void foo(void);
void cb_eqSF(float x) {
if(x != 0)
foo();
}
void cb_neSF(float x) {
if(x == 0)
foo();
}
void cb_geSF(float x) {
if(x < 0)
foo();
}
;; after
;; (TARGET_NSA, TARGET_BOOLEANS and TARGET_HARD_FLOAT)
bool_eqSF:
add.n a2, a2, a2
nsau a2, a2
srli a2, a2, 5
ret.n
bool_neSF:
add.n a9, a2, a2
movi.n a2, 1
moveqz a2, a9, a9
ret.n
bool_ltSF:
movi.n a9, 0
wfr f0, a2
wfr f1, a9
olt.s b0, f0, f1
movi.n a9, 0
movi.n a2, 1
movf a2, a9, b0
ret.n
cb_eqSF:
add.n a2, a2, a2
beqz.n a2, .L6
j.l foo, a9
.L6:
ret.n
cb_neSF:
add.n a2, a2, a2
bnez.n a2, .L8
j.l foo, a9
.L8:
ret.n
cb_geSF:
addi sp, sp, -16
movi.n a3, 0
s32i.n a12, sp, 8
s32i.n a0, sp, 12
mov.n a12, a2
call0 __unordsf2
bnez.n a2, .L10
movi.n a3, 0
mov.n a2, a12
call0 __gesf2
bnei a2, -1, .L10
l32i.n a0, sp, 12
l32i.n a12, sp, 8
addi sp, sp, 16
j.l foo, a9
.L10:
l32i.n a0, sp, 12
l32i.n a12, sp, 8
addi sp, sp, 16
ret.n
gcc/ChangeLog:
* config/xtensa/predicates.md (const_float_0_operand):
Rename from obsolete "const_float_1_operand" and change the
constant to compare.
(cstoresf_cbranchsf_operand, cstoresf_cbranchsf_operator):
New.
* config/xtensa/xtensa.cc (xtensa_expand_conditional_branch):
Add code for EQ/NE comparison with constant zero in SFmode.
(xtensa_expand_scc): Added code to derive boolean evaluation
of EQ/NE with constant zero for comparison in SFmode.
(xtensa_rtx_costs): Change cost of CONST_DOUBLE with value
zero inside "cbranchsf4" to 0.
* config/xtensa/xtensa.md (cbranchsf4, cstoresf4):
Change "match_operator" and the third "match_operand" to the
ones mentioned above.
(movsicc_ne0_reg_zero, eq_zero): New.
---
gcc/config/xtensa/predicates.md | 17 +++++++++--
gcc/config/xtensa/xtensa.cc | 45 ++++++++++++++++++++++++++++
gcc/config/xtensa/xtensa.md | 53 +++++++++++++++++++++++++++++----
3 files changed, 106 insertions(+), 9 deletions(-)
Comments
Hi Suwa-san,
On Mon, Jun 5, 2023 at 2:37 AM Takayuki 'January June' Suwa
<jjsuwa_sys3175@yahoo.co.jp> wrote:
>
> This patch optimizes the boolean evaluation of EQ/NE against zero
> by adding two insn_and_split patterns similar to SImode conditional
> store:
>
> "eq_zero":
> op0 = (op1 == 0) ? 1 : 0;
> op0 = clz(op1) >> 5; /* optimized (requires TARGET_NSA) */
>
> "movsicc_ne0_reg_0":
> op0 = (op1 != 0) ? op2 : 0;
> op0 = op2; if (op1 == 0) ? op0 = op1; /* optimized */
>
> /* example #1 */
> int bool_eqSI(int x) {
> return x == 0;
> }
> int bool_neSI(int x) {
> return x != 0;
> }
>
> ;; after (TARGET_NSA)
> bool_eqSI:
> nsau a2, a2
> srli a2, a2, 5
> ret.n
> bool_neSI:
> mov.n a9, a2
> movi.n a2, 1
> moveqz a2, a9, a9
> ret.n
>
> These also work in SFmode by ignoring their sign bits, and further-
> more, the branch if EQ/NE against zero in SFmode is also done in the
> same manner.
>
> The reasons for this optimization in SFmode are:
>
> - Only zero values (negative or non-negative) contain no bits of 1
> with both the exponent and the mantissa.
> - EQ/NE comparisons involving NaNs produce no signal even if they
> are signaling.
> - Even if the use of IEEE 754 single-precision floating-point co-
> processor is configured (TARGET_HARD_FLOAT is true):
> 1. Load zero value to FP register
> 2. Possibly, additional FP move if the comparison target is
> an address register
> 3. FP equality check instruction
> 4. Read the boolean register containing the result, or condi-
> tional branch
> As noted above, a considerable number of instructions are still
> generated.
>
> /* example #2 */
> int bool_eqSF(float x) {
> return x == 0;
> }
> int bool_neSF(float x) {
> return x != 0;
> }
> int bool_ltSF(float x) {
> return x < 0;
> }
> extern void foo(void);
> void cb_eqSF(float x) {
> if(x != 0)
> foo();
> }
> void cb_neSF(float x) {
> if(x == 0)
> foo();
> }
> void cb_geSF(float x) {
> if(x < 0)
> foo();
> }
>
> ;; after
> ;; (TARGET_NSA, TARGET_BOOLEANS and TARGET_HARD_FLOAT)
> bool_eqSF:
> add.n a2, a2, a2
> nsau a2, a2
> srli a2, a2, 5
> ret.n
> bool_neSF:
> add.n a9, a2, a2
> movi.n a2, 1
> moveqz a2, a9, a9
> ret.n
> bool_ltSF:
> movi.n a9, 0
> wfr f0, a2
> wfr f1, a9
> olt.s b0, f0, f1
> movi.n a9, 0
> movi.n a2, 1
> movf a2, a9, b0
> ret.n
> cb_eqSF:
> add.n a2, a2, a2
> beqz.n a2, .L6
> j.l foo, a9
> .L6:
> ret.n
> cb_neSF:
> add.n a2, a2, a2
> bnez.n a2, .L8
> j.l foo, a9
> .L8:
> ret.n
> cb_geSF:
> addi sp, sp, -16
> movi.n a3, 0
> s32i.n a12, sp, 8
> s32i.n a0, sp, 12
> mov.n a12, a2
> call0 __unordsf2
> bnez.n a2, .L10
> movi.n a3, 0
> mov.n a2, a12
> call0 __gesf2
> bnei a2, -1, .L10
> l32i.n a0, sp, 12
> l32i.n a12, sp, 8
> addi sp, sp, 16
> j.l foo, a9
> .L10:
> l32i.n a0, sp, 12
> l32i.n a12, sp, 8
> addi sp, sp, 16
> ret.n
>
> gcc/ChangeLog:
>
> * config/xtensa/predicates.md (const_float_0_operand):
> Rename from obsolete "const_float_1_operand" and change the
> constant to compare.
> (cstoresf_cbranchsf_operand, cstoresf_cbranchsf_operator):
> New.
> * config/xtensa/xtensa.cc (xtensa_expand_conditional_branch):
> Add code for EQ/NE comparison with constant zero in SFmode.
> (xtensa_expand_scc): Added code to derive boolean evaluation
> of EQ/NE with constant zero for comparison in SFmode.
> (xtensa_rtx_costs): Change cost of CONST_DOUBLE with value
> zero inside "cbranchsf4" to 0.
> * config/xtensa/xtensa.md (cbranchsf4, cstoresf4):
> Change "match_operator" and the third "match_operand" to the
> ones mentioned above.
> (movsicc_ne0_reg_zero, eq_zero): New.
> ---
> gcc/config/xtensa/predicates.md | 17 +++++++++--
> gcc/config/xtensa/xtensa.cc | 45 ++++++++++++++++++++++++++++
> gcc/config/xtensa/xtensa.md | 53 +++++++++++++++++++++++++++++----
> 3 files changed, 106 insertions(+), 9 deletions(-)
This version performs much better than v1, but there's still new
testsuite failure in the gcc.c-torture/execute/bitfld-3.c
and the following change in the generated code
from:
l32i.n a11, a7, 8
l8ui a9, a7, 12
movi a10, 0xff
add.n a9, a9, a10
addi.n a7, a11, -1
movi.n a10, 1
movi.n a6, 0
moveqz a10, a6, a11
to:
l32i.n a10, a7, 8
l8ui a9, a7, 12
movi a11, 0xff
add.n a9, a9, a11
addi.n a7, a10, -1
movi.n a11, 1
mov.n a10, a11
movnez a10, a11, a11
suggests that the pattern movsicc_ne0_reg_zero does not work correctly
when its operands overlap.
On 2023/06/06 0:15, Max Filippov wrote:
> Hi Suwa-san,
Hi! Thanks for your regtest every time.
>
> On Mon, Jun 5, 2023 at 2:37 AM Takayuki 'January June' Suwa
> <jjsuwa_sys3175@yahoo.co.jp> wrote:
>>
>> This patch optimizes the boolean evaluation of EQ/NE against zero
>> by adding two insn_and_split patterns similar to SImode conditional
>> store:
>>
>> "eq_zero":
>> op0 = (op1 == 0) ? 1 : 0;
>> op0 = clz(op1) >> 5; /* optimized (requires TARGET_NSA) */
>>
>> "movsicc_ne0_reg_0":
>> op0 = (op1 != 0) ? op2 : 0;
>> op0 = op2; if (op1 == 0) ? op0 = op1; /* optimized */
>>
>> /* example #1 */
>> int bool_eqSI(int x) {
>> return x == 0;
>> }
>> int bool_neSI(int x) {
>> return x != 0;
>> }
>>
>> ;; after (TARGET_NSA)
>> bool_eqSI:
>> nsau a2, a2
>> srli a2, a2, 5
>> ret.n
>> bool_neSI:
>> mov.n a9, a2
>> movi.n a2, 1
>> moveqz a2, a9, a9
>> ret.n
>>
>> These also work in SFmode by ignoring their sign bits, and further-
>> more, the branch if EQ/NE against zero in SFmode is also done in the
>> same manner.
>>
>> The reasons for this optimization in SFmode are:
>>
>> - Only zero values (negative or non-negative) contain no bits of 1
>> with both the exponent and the mantissa.
>> - EQ/NE comparisons involving NaNs produce no signal even if they
>> are signaling.
>> - Even if the use of IEEE 754 single-precision floating-point co-
>> processor is configured (TARGET_HARD_FLOAT is true):
>> 1. Load zero value to FP register
>> 2. Possibly, additional FP move if the comparison target is
>> an address register
>> 3. FP equality check instruction
>> 4. Read the boolean register containing the result, or condi-
>> tional branch
>> As noted above, a considerable number of instructions are still
>> generated.
>>
>> /* example #2 */
>> int bool_eqSF(float x) {
>> return x == 0;
>> }
>> int bool_neSF(float x) {
>> return x != 0;
>> }
>> int bool_ltSF(float x) {
>> return x < 0;
>> }
>> extern void foo(void);
>> void cb_eqSF(float x) {
>> if(x != 0)
>> foo();
>> }
>> void cb_neSF(float x) {
>> if(x == 0)
>> foo();
>> }
>> void cb_geSF(float x) {
>> if(x < 0)
>> foo();
>> }
>>
>> ;; after
>> ;; (TARGET_NSA, TARGET_BOOLEANS and TARGET_HARD_FLOAT)
>> bool_eqSF:
>> add.n a2, a2, a2
>> nsau a2, a2
>> srli a2, a2, 5
>> ret.n
>> bool_neSF:
>> add.n a9, a2, a2
>> movi.n a2, 1
>> moveqz a2, a9, a9
>> ret.n
>> bool_ltSF:
>> movi.n a9, 0
>> wfr f0, a2
>> wfr f1, a9
>> olt.s b0, f0, f1
>> movi.n a9, 0
>> movi.n a2, 1
>> movf a2, a9, b0
>> ret.n
>> cb_eqSF:
>> add.n a2, a2, a2
>> beqz.n a2, .L6
>> j.l foo, a9
>> .L6:
>> ret.n
>> cb_neSF:
>> add.n a2, a2, a2
>> bnez.n a2, .L8
>> j.l foo, a9
>> .L8:
>> ret.n
>> cb_geSF:
>> addi sp, sp, -16
>> movi.n a3, 0
>> s32i.n a12, sp, 8
>> s32i.n a0, sp, 12
>> mov.n a12, a2
>> call0 __unordsf2
>> bnez.n a2, .L10
>> movi.n a3, 0
>> mov.n a2, a12
>> call0 __gesf2
>> bnei a2, -1, .L10
>> l32i.n a0, sp, 12
>> l32i.n a12, sp, 8
>> addi sp, sp, 16
>> j.l foo, a9
>> .L10:
>> l32i.n a0, sp, 12
>> l32i.n a12, sp, 8
>> addi sp, sp, 16
>> ret.n
>>
>> gcc/ChangeLog:
>>
>> * config/xtensa/predicates.md (const_float_0_operand):
>> Rename from obsolete "const_float_1_operand" and change the
>> constant to compare.
>> (cstoresf_cbranchsf_operand, cstoresf_cbranchsf_operator):
>> New.
>> * config/xtensa/xtensa.cc (xtensa_expand_conditional_branch):
>> Add code for EQ/NE comparison with constant zero in SFmode.
>> (xtensa_expand_scc): Added code to derive boolean evaluation
>> of EQ/NE with constant zero for comparison in SFmode.
>> (xtensa_rtx_costs): Change cost of CONST_DOUBLE with value
>> zero inside "cbranchsf4" to 0.
>> * config/xtensa/xtensa.md (cbranchsf4, cstoresf4):
>> Change "match_operator" and the third "match_operand" to the
>> ones mentioned above.
>> (movsicc_ne0_reg_zero, eq_zero): New.
>> ---
>> gcc/config/xtensa/predicates.md | 17 +++++++++--
>> gcc/config/xtensa/xtensa.cc | 45 ++++++++++++++++++++++++++++
>> gcc/config/xtensa/xtensa.md | 53 +++++++++++++++++++++++++++++----
>> 3 files changed, 106 insertions(+), 9 deletions(-)
>
> This version performs much better than v1, but there's still new
> testsuite failure in the gcc.c-torture/execute/bitfld-3.c
> and the following change in the generated code
> from:
>
> l32i.n a11, a7, 8
> l8ui a9, a7, 12
> movi a10, 0xff
> add.n a9, a9, a10
> addi.n a7, a11, -1
> movi.n a10, 1
> movi.n a6, 0
> moveqz a10, a6, a11
>
> to:
>
> l32i.n a10, a7, 8
> l8ui a9, a7, 12
> movi a11, 0xff
> add.n a9, a9, a11
> addi.n a7, a10, -1
> movi.n a11, 1
> mov.n a10, a11
> movnez a10, a11, a11
>
> suggests that the pattern movsicc_ne0_reg_zero does not work correctly
> when its operands overlap.
>
The reason I wanted to implement the movsicc_ne0_reg_zero pattern in the first place is the following output
that xtensa_expand_scc() emits in its generic procedure:
(set (reg:SI 10)
(const_int 1))
(set (reg:SI 6)
(const_int 0))
(set (reg:SI 10)
(if_then_else:SI (ne (reg:SI 11)
(const_int 0))
(reg:SI 10)
(reg:SI 6])))
movi.n a10, 1
movi.n a6, 0
moveqz a10, a6, a11
I realized that A11 should be used instead of A6 (and A6 will be removed later) since it is clear that A11 is
zero if the above moveqz is executed.
Maybe it is better to tweak xtensa_expand_scc() a bit than to implement the pattern separately.
On Mon, Jun 5, 2023 at 8:15 AM Max Filippov <jcmvbkbc@gmail.com> wrote:
>
> Hi Suwa-san,
>
> On Mon, Jun 5, 2023 at 2:37 AM Takayuki 'January June' Suwa
> <jjsuwa_sys3175@yahoo.co.jp> wrote:
> >
> > This patch optimizes the boolean evaluation of EQ/NE against zero
> > by adding two insn_and_split patterns similar to SImode conditional
> > store:
> >
> > "eq_zero":
> > op0 = (op1 == 0) ? 1 : 0;
> > op0 = clz(op1) >> 5; /* optimized (requires TARGET_NSA) */
> >
> > "movsicc_ne0_reg_0":
> > op0 = (op1 != 0) ? op2 : 0;
> > op0 = op2; if (op1 == 0) ? op0 = op1; /* optimized */
> >
> > /* example #1 */
> > int bool_eqSI(int x) {
> > return x == 0;
> > }
> > int bool_neSI(int x) {
> > return x != 0;
> > }
> >
> > ;; after (TARGET_NSA)
> > bool_eqSI:
> > nsau a2, a2
> > srli a2, a2, 5
> > ret.n
> > bool_neSI:
> > mov.n a9, a2
> > movi.n a2, 1
> > moveqz a2, a9, a9
> > ret.n
> >
> > These also work in SFmode by ignoring their sign bits, and further-
> > more, the branch if EQ/NE against zero in SFmode is also done in the
> > same manner.
> >
> > The reasons for this optimization in SFmode are:
> >
> > - Only zero values (negative or non-negative) contain no bits of 1
> > with both the exponent and the mantissa.
> > - EQ/NE comparisons involving NaNs produce no signal even if they
> > are signaling.
> > - Even if the use of IEEE 754 single-precision floating-point co-
> > processor is configured (TARGET_HARD_FLOAT is true):
> > 1. Load zero value to FP register
> > 2. Possibly, additional FP move if the comparison target is
> > an address register
> > 3. FP equality check instruction
> > 4. Read the boolean register containing the result, or condi-
> > tional branch
> > As noted above, a considerable number of instructions are still
> > generated.
> >
> > /* example #2 */
> > int bool_eqSF(float x) {
> > return x == 0;
> > }
> > int bool_neSF(float x) {
> > return x != 0;
> > }
> > int bool_ltSF(float x) {
> > return x < 0;
> > }
> > extern void foo(void);
> > void cb_eqSF(float x) {
> > if(x != 0)
> > foo();
> > }
> > void cb_neSF(float x) {
> > if(x == 0)
> > foo();
> > }
> > void cb_geSF(float x) {
> > if(x < 0)
> > foo();
> > }
> >
> > ;; after
> > ;; (TARGET_NSA, TARGET_BOOLEANS and TARGET_HARD_FLOAT)
> > bool_eqSF:
> > add.n a2, a2, a2
> > nsau a2, a2
> > srli a2, a2, 5
> > ret.n
> > bool_neSF:
> > add.n a9, a2, a2
> > movi.n a2, 1
> > moveqz a2, a9, a9
> > ret.n
> > bool_ltSF:
> > movi.n a9, 0
> > wfr f0, a2
> > wfr f1, a9
> > olt.s b0, f0, f1
> > movi.n a9, 0
> > movi.n a2, 1
> > movf a2, a9, b0
> > ret.n
> > cb_eqSF:
> > add.n a2, a2, a2
> > beqz.n a2, .L6
> > j.l foo, a9
> > .L6:
> > ret.n
> > cb_neSF:
> > add.n a2, a2, a2
> > bnez.n a2, .L8
> > j.l foo, a9
> > .L8:
> > ret.n
> > cb_geSF:
> > addi sp, sp, -16
> > movi.n a3, 0
> > s32i.n a12, sp, 8
> > s32i.n a0, sp, 12
> > mov.n a12, a2
> > call0 __unordsf2
> > bnez.n a2, .L10
> > movi.n a3, 0
> > mov.n a2, a12
> > call0 __gesf2
> > bnei a2, -1, .L10
> > l32i.n a0, sp, 12
> > l32i.n a12, sp, 8
> > addi sp, sp, 16
> > j.l foo, a9
> > .L10:
> > l32i.n a0, sp, 12
> > l32i.n a12, sp, 8
> > addi sp, sp, 16
> > ret.n
> >
> > gcc/ChangeLog:
> >
> > * config/xtensa/predicates.md (const_float_0_operand):
> > Rename from obsolete "const_float_1_operand" and change the
> > constant to compare.
> > (cstoresf_cbranchsf_operand, cstoresf_cbranchsf_operator):
> > New.
> > * config/xtensa/xtensa.cc (xtensa_expand_conditional_branch):
> > Add code for EQ/NE comparison with constant zero in SFmode.
> > (xtensa_expand_scc): Added code to derive boolean evaluation
> > of EQ/NE with constant zero for comparison in SFmode.
> > (xtensa_rtx_costs): Change cost of CONST_DOUBLE with value
> > zero inside "cbranchsf4" to 0.
> > * config/xtensa/xtensa.md (cbranchsf4, cstoresf4):
> > Change "match_operator" and the third "match_operand" to the
> > ones mentioned above.
> > (movsicc_ne0_reg_zero, eq_zero): New.
> > ---
> > gcc/config/xtensa/predicates.md | 17 +++++++++--
> > gcc/config/xtensa/xtensa.cc | 45 ++++++++++++++++++++++++++++
> > gcc/config/xtensa/xtensa.md | 53 +++++++++++++++++++++++++++++----
> > 3 files changed, 106 insertions(+), 9 deletions(-)
>
> This version performs much better than v1, but there's still new
> testsuite failure in the gcc.c-torture/execute/bitfld-3.c
And on the config with FPU there's one more new failure
in the g++.dg/opt/pr58864.C with the following ICE:
gcc/testsuite/g++.dg/opt/pr58864.C:21:1: error: unrecognizable insn:
(insn 13 12 14 2 (set (reg:CC 18 b0)
(eq:CC (reg/v:SF 43 [ c ])
(const_double:SF 0.0 [0x0.0p+0]))) -1
(nil))
during RTL pass: vregs
@@ -155,11 +155,11 @@
&& CONSTANT_P (op)
&& GET_MODE_SIZE (mode) % UNITS_PER_WORD == 0")))))
-;; Accept the floating point constant 1 in the appropriate mode.
-(define_predicate "const_float_1_operand"
+;; Accept the floating point constant 0 in the appropriate mode.
+(define_predicate "const_float_0_operand"
(match_code "const_double")
{
- return real_equal (CONST_DOUBLE_REAL_VALUE (op), &dconst1);
+ return real_equal (CONST_DOUBLE_REAL_VALUE (op), &dconst0);
})
(define_predicate "fpmem_offset_operand"
@@ -179,6 +179,11 @@
return false;
})
+(define_predicate "cstoresf_cbranchsf_operand"
+ (ior (and (match_test "TARGET_HARD_FLOAT")
+ (match_operand 0 "register_operand"))
+ (match_operand 0 "const_float_0_operand")))
+
(define_predicate "branch_operator"
(match_code "eq,ne,lt,ge"))
@@ -197,6 +202,12 @@
(define_predicate "xtensa_cstoresi_operator"
(match_code "eq,ne,gt,ge,lt,le"))
+(define_predicate "cstoresf_cbranchsf_operator"
+ (ior (and (match_test "TARGET_HARD_FLOAT")
+ (and (match_operand 0 "comparison_operator")
+ (match_test "register_operand (XEXP (op, 1), SFmode)")))
+ (match_operand 0 "boolean_operator")))
+
(define_predicate "xtensa_shift_per_byte_operator"
(match_code "ashift,ashiftrt,lshiftrt"))
@@ -865,6 +865,16 @@ xtensa_expand_conditional_branch (rtx *operands, machine_mode mode)
switch (mode)
{
case E_SFmode:
+ if ((test_code == EQ || test_code == NE)
+ && const_float_0_operand (cmp1, SFmode))
+ {
+ emit_move_insn (cmp1 = gen_reg_rtx (SImode),
+ simplify_gen_subreg (SImode, cmp0, SFmode, 0));
+ emit_insn (gen_addsi3 (cmp1, cmp1, cmp1));
+ cmp = gen_int_relational (test_code, cmp1, const0_rtx);
+ break;
+ }
+
if (TARGET_HARD_FLOAT)
{
cmp = gen_float_relational (test_code, cmp0, cmp1);
@@ -996,6 +1006,36 @@ xtensa_expand_scc (rtx operands[4], machine_mode cmp_mode)
rtx one_tmp, zero_tmp;
rtx (*gen_fn) (rtx, rtx, rtx, rtx, rtx);
+ if (cmp_mode == SFmode)
+ {
+ if (const_float_0_operand (operands[3], SFmode))
+ switch (GET_CODE (operands[1]))
+ {
+ case EQ:
+ emit_move_insn (cmp = gen_reg_rtx (SImode),
+ simplify_gen_subreg (SImode, operands[2],
+ SFmode, 0));
+ emit_insn (gen_addsi3 (cmp, cmp, cmp));
+ emit_insn (gen_eq_zero (dest, cmp));
+ return 1;
+
+ case NE:
+ emit_move_insn (cmp = gen_reg_rtx (SImode),
+ simplify_gen_subreg (SImode, operands[2],
+ SFmode, 0));
+ emit_insn (gen_addsi3 (cmp, cmp, cmp));
+ one_tmp = force_reg (SImode, const1_rtx);
+ emit_insn (gen_movsicc_ne0_reg_zero (dest, cmp, one_tmp));
+ return 1;
+
+ default:
+ return 0;
+ }
+
+ if (! register_operand (operands[3], SFmode))
+ return 0;
+ }
+
if (!(cmp = gen_conditional_move (GET_CODE (operands[1]), cmp_mode,
operands[2], operands[3])))
return 0;
@@ -4438,6 +4478,11 @@ xtensa_rtx_costs (rtx x, machine_mode mode, int outer_code,
return true;
case CONST_DOUBLE:
+ if (outer_code == COMPARE && const_float_0_operand (x, SFmode))
+ {
+ *total = 0;
+ return true;
+ }
if (TARGET_CONST16)
*total = COSTS_N_INSNS (4);
else
@@ -1906,11 +1906,11 @@
})
(define_expand "cbranchsf4"
- [(match_operator 0 "comparison_operator"
+ [(match_operator 0 "cstoresf_cbranchsf_operator"
[(match_operand:SF 1 "register_operand")
- (match_operand:SF 2 "register_operand")])
+ (match_operand:SF 2 "cstoresf_cbranchsf_operand")])
(match_operand 3 "")]
- "TARGET_HARD_FLOAT"
+ ""
{
xtensa_expand_conditional_branch (operands, SFmode);
DONE;
@@ -2395,10 +2395,10 @@
(define_expand "cstoresf4"
[(match_operand:SI 0 "register_operand")
- (match_operator:SI 1 "comparison_operator"
+ (match_operator:SI 1 "cstoresf_cbranchsf_operator"
[(match_operand:SF 2 "register_operand")
- (match_operand:SF 3 "register_operand")])]
- "TARGET_HARD_FLOAT"
+ (match_operand:SF 3 "cstoresf_cbranchsf_operand")])]
+ ""
{
if (!xtensa_expand_scc (operands, SFmode))
FAIL;
@@ -2463,6 +2463,30 @@
(set_attr "mode" "SI")
(set_attr "length" "3,3")])
+(define_insn_and_split "movsicc_ne0_reg_zero"
+ [(set (match_operand:SI 0 "register_operand" "=a")
+ (if_then_else:SI (ne (match_operand:SI 1 "register_operand" "r")
+ (const_int 0))
+ (match_operand:SI 2 "register_operand" "r")
+ (const_int 0)))]
+ ""
+ "#"
+ ""
+ [(set (match_dup 0)
+ (match_dup 2))
+ (set (match_dup 0)
+ (if_then_else:SI (ne (match_dup 1)
+ (const_int 0))
+ (match_dup 0)
+ (match_dup 1)))]
+ ""
+ [(set_attr "type" "move")
+ (set_attr "mode" "SI")
+ (set (attr "length")
+ (if_then_else (match_test "TARGET_DENSITY")
+ (const_int 5)
+ (const_int 6)))])
+
(define_insn "movsfcc_internal0"
[(set (match_operand:SF 0 "register_operand" "=a,a,f,f")
(if_then_else:SF (match_operator 4 "branch_operator"
@@ -3222,6 +3246,23 @@
(const_int 5)
(const_int 6))))])
+(define_insn_and_split "eq_zero"
+ [(set (match_operand:SI 0 "register_operand" "=a")
+ (eq:SI (match_operand:SI 1 "register_operand" "r")
+ (const_int 0)))]
+ "TARGET_NSA"
+ "#"
+ "&& 1"
+ [(set (match_dup 0)
+ (clz:SI (match_dup 1)))
+ (set (match_dup 0)
+ (lshiftrt:SI (match_dup 0)
+ (const_int 5)))]
+ ""
+ [(set_attr "type" "move")
+ (set_attr "mode" "SI")
+ (set_attr "length" "6")])
+
(define_peephole2
[(set (match_operand:SI 0 "register_operand")
(match_operand:SI 6 "reload_operand"))