* [PATCH, rs6000] Fix PR83332 (missing vcond patterns)
@ 2017-12-11 21:55 Bill Schmidt
2017-12-12 8:02 ` Richard Biener
2017-12-12 17:13 ` Segher Boessenkool
0 siblings, 2 replies; 5+ messages in thread
From: Bill Schmidt @ 2017-12-11 21:55 UTC (permalink / raw)
To: GCC Patches; +Cc: Segher Boessenkool, David Edelsohn
Hi,
A new test case introduced for PR81303 failed on powerpc64 (BE, LE). This
turns out to be due to a missing standard pattern (vcondv2div2df). This
and a couple of other patterns are easy to support with existing logic
by just adding new patterns with appropriate modes. That's all this patch
does. That's sufficient to cause the failing test to pass.
Bootstrapped and tested on powerpc64le-linux-gnu with no regressions. Is
this okay for trunk?
Thanks,
Bill
2017-12-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
PR target/83332
* config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
(vcondv2div2df): Likewise.
(vconduv2dfv2di): Likewise.
Index: gcc/config/rs6000/vector.md
===================================================================
--- gcc/config/rs6000/vector.md (revision 255539)
+++ gcc/config/rs6000/vector.md (working copy)
@@ -455,6 +455,44 @@
FAIL;
}")
+(define_expand "vcondv2dfv2di"
+ [(set (match_operand:V2DF 0 "vfloat_operand" "")
+ (if_then_else:V2DF
+ (match_operator 3 "comparison_operator"
+ [(match_operand:V2DI 4 "vint_operand" "")
+ (match_operand:V2DI 5 "vint_operand" "")])
+ (match_operand:V2DF 1 "vfloat_operand" "")
+ (match_operand:V2DF 2 "vfloat_operand" "")))]
+ "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
+ && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
+ "
+{
+ if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
+ operands[3], operands[4], operands[5]))
+ DONE;
+ else
+ FAIL;
+}")
+
+(define_expand "vcondv2div2df"
+ [(set (match_operand:V2DI 0 "vint_operand" "")
+ (if_then_else:V2DI
+ (match_operator 3 "comparison_operator"
+ [(match_operand:V2DF 4 "vfloat_operand" "")
+ (match_operand:V2DF 5 "vfloat_operand" "")])
+ (match_operand:V2DI 1 "vint_operand" "")
+ (match_operand:V2DI 2 "vint_operand" "")))]
+ "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
+ && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
+ "
+{
+ if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
+ operands[3], operands[4], operands[5]))
+ DONE;
+ else
+ FAIL;
+}")
+
(define_expand "vcondu<mode><mode>"
[(set (match_operand:VEC_I 0 "vint_operand")
(if_then_else:VEC_I
@@ -492,6 +530,25 @@
FAIL;
}")
+(define_expand "vconduv2dfv2di"
+ [(set (match_operand:V2DF 0 "vfloat_operand" "")
+ (if_then_else:V2DF
+ (match_operator 3 "comparison_operator"
+ [(match_operand:V2DI 4 "vint_operand" "")
+ (match_operand:V2DI 5 "vint_operand" "")])
+ (match_operand:V2DF 1 "vfloat_operand" "")
+ (match_operand:V2DF 2 "vfloat_operand" "")))]
+ "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
+ && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
+ "
+{
+ if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
+ operands[3], operands[4], operands[5]))
+ DONE;
+ else
+ FAIL;
+}")
+
(define_expand "vector_eq<mode>"
[(set (match_operand:VEC_C 0 "vlogical_operand" "")
(eq:VEC_C (match_operand:VEC_C 1 "vlogical_operand" "")
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH, rs6000] Fix PR83332 (missing vcond patterns)
2017-12-11 21:55 [PATCH, rs6000] Fix PR83332 (missing vcond patterns) Bill Schmidt
@ 2017-12-12 8:02 ` Richard Biener
2017-12-12 13:46 ` Bill Schmidt
2017-12-12 17:13 ` Segher Boessenkool
1 sibling, 1 reply; 5+ messages in thread
From: Richard Biener @ 2017-12-12 8:02 UTC (permalink / raw)
To: Bill Schmidt; +Cc: GCC Patches, Segher Boessenkool, David Edelsohn
On Mon, Dec 11, 2017 at 10:55 PM, Bill Schmidt
<wschmidt@linux.vnet.ibm.com> wrote:
> Hi,
>
> A new test case introduced for PR81303 failed on powerpc64 (BE, LE). This
> turns out to be due to a missing standard pattern (vcondv2div2df). This
> and a couple of other patterns are easy to support with existing logic
> by just adding new patterns with appropriate modes. That's all this patch
> does. That's sufficient to cause the failing test to pass.
>
> Bootstrapped and tested on powerpc64le-linux-gnu with no regressions. Is
> this okay for trunk?
>
> Thanks,
> Bill
>
>
> 2017-12-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
>
> PR target/83332
> * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
> (vcondv2div2df): Likewise.
> (vconduv2dfv2di): Likewise.
>
> Index: gcc/config/rs6000/vector.md
> ===================================================================
> --- gcc/config/rs6000/vector.md (revision 255539)
> +++ gcc/config/rs6000/vector.md (working copy)
> @@ -455,6 +455,44 @@
> FAIL;
> }")
>
> +(define_expand "vcondv2dfv2di"
given you already have vcondv4siv4sf this asks for macroization, no? I see you
have vcond<mode><mode> already, you can use
vcond<VEC_I:mode><VEC_F:mode>
and guard with GET_MODE_NUNITS () == GET_MODE_NUNINTS ()?
> + [(set (match_operand:V2DF 0 "vfloat_operand" "")
> + (if_then_else:V2DF
> + (match_operator 3 "comparison_operator"
> + [(match_operand:V2DI 4 "vint_operand" "")
> + (match_operand:V2DI 5 "vint_operand" "")])
> + (match_operand:V2DF 1 "vfloat_operand" "")
> + (match_operand:V2DF 2 "vfloat_operand" "")))]
> + "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
> + && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
> + "
> +{
> + if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
> + operands[3], operands[4], operands[5]))
> + DONE;
> + else
> + FAIL;
> +}")
> +
> +(define_expand "vcondv2div2df"
> + [(set (match_operand:V2DI 0 "vint_operand" "")
> + (if_then_else:V2DI
> + (match_operator 3 "comparison_operator"
> + [(match_operand:V2DF 4 "vfloat_operand" "")
> + (match_operand:V2DF 5 "vfloat_operand" "")])
> + (match_operand:V2DI 1 "vint_operand" "")
> + (match_operand:V2DI 2 "vint_operand" "")))]
> + "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
> + && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
> + "
> +{
> + if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
> + operands[3], operands[4], operands[5]))
> + DONE;
> + else
> + FAIL;
> +}")
> +
> (define_expand "vcondu<mode><mode>"
> [(set (match_operand:VEC_I 0 "vint_operand")
> (if_then_else:VEC_I
> @@ -492,6 +530,25 @@
> FAIL;
> }")
>
> +(define_expand "vconduv2dfv2di"
> + [(set (match_operand:V2DF 0 "vfloat_operand" "")
> + (if_then_else:V2DF
> + (match_operator 3 "comparison_operator"
> + [(match_operand:V2DI 4 "vint_operand" "")
> + (match_operand:V2DI 5 "vint_operand" "")])
> + (match_operand:V2DF 1 "vfloat_operand" "")
> + (match_operand:V2DF 2 "vfloat_operand" "")))]
> + "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
> + && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
> + "
> +{
> + if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
> + operands[3], operands[4], operands[5]))
> + DONE;
> + else
> + FAIL;
> +}")
> +
> (define_expand "vector_eq<mode>"
> [(set (match_operand:VEC_C 0 "vlogical_operand" "")
> (eq:VEC_C (match_operand:VEC_C 1 "vlogical_operand" "")
>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH, rs6000] Fix PR83332 (missing vcond patterns)
2017-12-12 8:02 ` Richard Biener
@ 2017-12-12 13:46 ` Bill Schmidt
2017-12-12 14:04 ` Richard Biener
0 siblings, 1 reply; 5+ messages in thread
From: Bill Schmidt @ 2017-12-12 13:46 UTC (permalink / raw)
To: Richard Biener; +Cc: GCC Patches, Segher Boessenkool, David Edelsohn
> On Dec 12, 2017, at 2:02 AM, Richard Biener <richard.guenther@gmail.com> wrote:
>
> On Mon, Dec 11, 2017 at 10:55 PM, Bill Schmidt
> <wschmidt@linux.vnet.ibm.com> wrote:
>> Hi,
>>
>> A new test case introduced for PR81303 failed on powerpc64 (BE, LE). This
>> turns out to be due to a missing standard pattern (vcondv2div2df). This
>> and a couple of other patterns are easy to support with existing logic
>> by just adding new patterns with appropriate modes. That's all this patch
>> does. That's sufficient to cause the failing test to pass.
>>
>> Bootstrapped and tested on powerpc64le-linux-gnu with no regressions. Is
>> this okay for trunk?
>>
>> Thanks,
>> Bill
>>
>>
>> 2017-12-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
>>
>> PR target/83332
>> * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
>> (vcondv2div2df): Likewise.
>> (vconduv2dfv2di): Likewise.
>>
>> Index: gcc/config/rs6000/vector.md
>> ===================================================================
>> --- gcc/config/rs6000/vector.md (revision 255539)
>> +++ gcc/config/rs6000/vector.md (working copy)
>> @@ -455,6 +455,44 @@
>> FAIL;
>> }")
>>
>> +(define_expand "vcondv2dfv2di"
>
> given you already have vcondv4siv4sf this asks for macroization, no? I see you
> have vcond<mode><mode> already, you can use
>
> vcond<VEC_I:mode><VEC_F:mode>
>
> and guard with GET_MODE_NUNITS () == GET_MODE_NUNINTS ()?
Yes, that can be done -- although the guard conditions start to get kind of ugly
now, since the v4siv4sf patterns require VECTOR_UNIT_ALTIVEC_P (V4SImode)
while the v2div2df ones require VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode).
What I have may be more straightforward to understand. I can implement
whichever is considered preferable.
Thanks,
Bill
>
>> + [(set (match_operand:V2DF 0 "vfloat_operand" "")
>> + (if_then_else:V2DF
>> + (match_operator 3 "comparison_operator"
>> + [(match_operand:V2DI 4 "vint_operand" "")
>> + (match_operand:V2DI 5 "vint_operand" "")])
>> + (match_operand:V2DF 1 "vfloat_operand" "")
>> + (match_operand:V2DF 2 "vfloat_operand" "")))]
>> + "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
>> + && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
>> + "
>> +{
>> + if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
>> + operands[3], operands[4], operands[5]))
>> + DONE;
>> + else
>> + FAIL;
>> +}")
>> +
>> +(define_expand "vcondv2div2df"
>> + [(set (match_operand:V2DI 0 "vint_operand" "")
>> + (if_then_else:V2DI
>> + (match_operator 3 "comparison_operator"
>> + [(match_operand:V2DF 4 "vfloat_operand" "")
>> + (match_operand:V2DF 5 "vfloat_operand" "")])
>> + (match_operand:V2DI 1 "vint_operand" "")
>> + (match_operand:V2DI 2 "vint_operand" "")))]
>> + "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
>> + && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
>> + "
>> +{
>> + if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
>> + operands[3], operands[4], operands[5]))
>> + DONE;
>> + else
>> + FAIL;
>> +}")
>> +
>> (define_expand "vcondu<mode><mode>"
>> [(set (match_operand:VEC_I 0 "vint_operand")
>> (if_then_else:VEC_I
>> @@ -492,6 +530,25 @@
>> FAIL;
>> }")
>>
>> +(define_expand "vconduv2dfv2di"
>> + [(set (match_operand:V2DF 0 "vfloat_operand" "")
>> + (if_then_else:V2DF
>> + (match_operator 3 "comparison_operator"
>> + [(match_operand:V2DI 4 "vint_operand" "")
>> + (match_operand:V2DI 5 "vint_operand" "")])
>> + (match_operand:V2DF 1 "vfloat_operand" "")
>> + (match_operand:V2DF 2 "vfloat_operand" "")))]
>> + "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
>> + && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
>> + "
>> +{
>> + if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
>> + operands[3], operands[4], operands[5]))
>> + DONE;
>> + else
>> + FAIL;
>> +}")
>> +
>> (define_expand "vector_eq<mode>"
>> [(set (match_operand:VEC_C 0 "vlogical_operand" "")
>> (eq:VEC_C (match_operand:VEC_C 1 "vlogical_operand" "")
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH, rs6000] Fix PR83332 (missing vcond patterns)
2017-12-12 13:46 ` Bill Schmidt
@ 2017-12-12 14:04 ` Richard Biener
0 siblings, 0 replies; 5+ messages in thread
From: Richard Biener @ 2017-12-12 14:04 UTC (permalink / raw)
To: Bill Schmidt; +Cc: GCC Patches, Segher Boessenkool, David Edelsohn
On Tue, Dec 12, 2017 at 2:46 PM, Bill Schmidt
<wschmidt@linux.vnet.ibm.com> wrote:
>
>> On Dec 12, 2017, at 2:02 AM, Richard Biener <richard.guenther@gmail.com> wrote:
>>
>> On Mon, Dec 11, 2017 at 10:55 PM, Bill Schmidt
>> <wschmidt@linux.vnet.ibm.com> wrote:
>>> Hi,
>>>
>>> A new test case introduced for PR81303 failed on powerpc64 (BE, LE). This
>>> turns out to be due to a missing standard pattern (vcondv2div2df). This
>>> and a couple of other patterns are easy to support with existing logic
>>> by just adding new patterns with appropriate modes. That's all this patch
>>> does. That's sufficient to cause the failing test to pass.
>>>
>>> Bootstrapped and tested on powerpc64le-linux-gnu with no regressions. Is
>>> this okay for trunk?
>>>
>>> Thanks,
>>> Bill
>>>
>>>
>>> 2017-12-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
>>>
>>> PR target/83332
>>> * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
>>> (vcondv2div2df): Likewise.
>>> (vconduv2dfv2di): Likewise.
>>>
>>> Index: gcc/config/rs6000/vector.md
>>> ===================================================================
>>> --- gcc/config/rs6000/vector.md (revision 255539)
>>> +++ gcc/config/rs6000/vector.md (working copy)
>>> @@ -455,6 +455,44 @@
>>> FAIL;
>>> }")
>>>
>>> +(define_expand "vcondv2dfv2di"
>>
>> given you already have vcondv4siv4sf this asks for macroization, no? I see you
>> have vcond<mode><mode> already, you can use
>>
>> vcond<VEC_I:mode><VEC_F:mode>
>>
>> and guard with GET_MODE_NUNITS () == GET_MODE_NUNINTS ()?
>
> Yes, that can be done -- although the guard conditions start to get kind of ugly
> now, since the v4siv4sf patterns require VECTOR_UNIT_ALTIVEC_P (V4SImode)
> while the v2div2df ones require VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode).
> What I have may be more straightforward to understand. I can implement
> whichever is considered preferable.
Up to the target maintainers of course.
Richard.
> Thanks,
> Bill
>
>>
>>> + [(set (match_operand:V2DF 0 "vfloat_operand" "")
>>> + (if_then_else:V2DF
>>> + (match_operator 3 "comparison_operator"
>>> + [(match_operand:V2DI 4 "vint_operand" "")
>>> + (match_operand:V2DI 5 "vint_operand" "")])
>>> + (match_operand:V2DF 1 "vfloat_operand" "")
>>> + (match_operand:V2DF 2 "vfloat_operand" "")))]
>>> + "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
>>> + && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
>>> + "
>>> +{
>>> + if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
>>> + operands[3], operands[4], operands[5]))
>>> + DONE;
>>> + else
>>> + FAIL;
>>> +}")
>>> +
>>> +(define_expand "vcondv2div2df"
>>> + [(set (match_operand:V2DI 0 "vint_operand" "")
>>> + (if_then_else:V2DI
>>> + (match_operator 3 "comparison_operator"
>>> + [(match_operand:V2DF 4 "vfloat_operand" "")
>>> + (match_operand:V2DF 5 "vfloat_operand" "")])
>>> + (match_operand:V2DI 1 "vint_operand" "")
>>> + (match_operand:V2DI 2 "vint_operand" "")))]
>>> + "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
>>> + && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
>>> + "
>>> +{
>>> + if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
>>> + operands[3], operands[4], operands[5]))
>>> + DONE;
>>> + else
>>> + FAIL;
>>> +}")
>>> +
>>> (define_expand "vcondu<mode><mode>"
>>> [(set (match_operand:VEC_I 0 "vint_operand")
>>> (if_then_else:VEC_I
>>> @@ -492,6 +530,25 @@
>>> FAIL;
>>> }")
>>>
>>> +(define_expand "vconduv2dfv2di"
>>> + [(set (match_operand:V2DF 0 "vfloat_operand" "")
>>> + (if_then_else:V2DF
>>> + (match_operator 3 "comparison_operator"
>>> + [(match_operand:V2DI 4 "vint_operand" "")
>>> + (match_operand:V2DI 5 "vint_operand" "")])
>>> + (match_operand:V2DF 1 "vfloat_operand" "")
>>> + (match_operand:V2DF 2 "vfloat_operand" "")))]
>>> + "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
>>> + && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
>>> + "
>>> +{
>>> + if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
>>> + operands[3], operands[4], operands[5]))
>>> + DONE;
>>> + else
>>> + FAIL;
>>> +}")
>>> +
>>> (define_expand "vector_eq<mode>"
>>> [(set (match_operand:VEC_C 0 "vlogical_operand" "")
>>> (eq:VEC_C (match_operand:VEC_C 1 "vlogical_operand" "")
>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH, rs6000] Fix PR83332 (missing vcond patterns)
2017-12-11 21:55 [PATCH, rs6000] Fix PR83332 (missing vcond patterns) Bill Schmidt
2017-12-12 8:02 ` Richard Biener
@ 2017-12-12 17:13 ` Segher Boessenkool
1 sibling, 0 replies; 5+ messages in thread
From: Segher Boessenkool @ 2017-12-12 17:13 UTC (permalink / raw)
To: Bill Schmidt; +Cc: GCC Patches, David Edelsohn
Hi!
On Mon, Dec 11, 2017 at 03:55:20PM -0600, Bill Schmidt wrote:
> A new test case introduced for PR81303 failed on powerpc64 (BE, LE). This
> turns out to be due to a missing standard pattern (vcondv2div2df). This
> and a couple of other patterns are easy to support with existing logic
> by just adding new patterns with appropriate modes. That's all this patch
> does. That's sufficient to cause the failing test to pass.
Nitpicking:
> +(define_expand "vcondv2dfv2di"
> + [(set (match_operand:V2DF 0 "vfloat_operand" "")
No empty default "" arguments please.
> + (if_then_else:V2DF
> + (match_operator 3 "comparison_operator"
> + [(match_operand:V2DI 4 "vint_operand" "")
> + (match_operand:V2DI 5 "vint_operand" "")])
> + (match_operand:V2DF 1 "vfloat_operand" "")
> + (match_operand:V2DF 2 "vfloat_operand" "")))]
> + "VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DFmode)
> + && VECTOR_UNIT_ALTIVEC_OR_VSX_P (V2DImode)"
> + "
> +{
> + if (rs6000_emit_vector_cond_expr (operands[0], operands[1], operands[2],
> + operands[3], operands[4], operands[5]))
> + DONE;
> + else
> + FAIL;
> +}")
And no "" around the block.
Okay for trunk with that fixed up. Thanks!
Segher
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2017-12-12 17:13 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-12-11 21:55 [PATCH, rs6000] Fix PR83332 (missing vcond patterns) Bill Schmidt
2017-12-12 8:02 ` Richard Biener
2017-12-12 13:46 ` Bill Schmidt
2017-12-12 14:04 ` Richard Biener
2017-12-12 17:13 ` Segher Boessenkool
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).