diff mbox series

[committed,AArch64] Use unspecs for SVE conversions involving floats

Message ID mpt7e7gp2ia.fsf@arm.com
State New
Headers show
Series [committed,AArch64] Use unspecs for SVE conversions involving floats | expand

Commit Message

Richard Sandiford Aug. 14, 2019, 8:36 a.m. UTC 
This patch changes the SVE FP<->FP and FP<->INT patterns so that
they use unspecs rather than rtx codes, continuing the series
to make the patterns work with predicates that might not be all-true.

Tested on aarch64-linux-gnu (with and without SVE) and aarch64_be-elf.
Applied as r274423.

Richard


2019-08-14  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
	* config/aarch64/aarch64.md (UNSPEC_FLOAT_CONVERT): Delete.
	* config/aarch64/iterators.md (UNSPEC_COND_FCVT, UNSPEC_COND_FCVTZS)
	(UNSPEC_COND_FCVTZU, UNSPEC_COND_SCVTF, UNSPEC_COND_UCVTF): New
	unspecs.
	(optab, su): Handle them.
	(SVE_COND_FCVTI, SVE_COND_ICVTF): New int iterators.
	* config/aarch64/aarch64-sve.md
	(<fix_trunc_optab><SVE_F:mode><v_int_equiv>2): Replace with...
	(<SVE_COND_FCVTI:optab><SVE_F:mode><v_int_equiv>2): ...this.
	(*<fix_trunc_optab>v16hsf<:SVE_HSDImode>2): Replace with...
	(*<SVE_COND_FCVTI:optab>v16hsf<SVE_F:mode>2): ...this.
	(*<fix_trunc_optab>vnx4sf<SVE_SDI:mode>2): Replace with...
	(*<SVE_COND_FCVTI:optab>vnx4sf<SVE_SDI:mode>2): ...this.
	(*<fix_trunc_optab>vnx2df<SVE_SDI:mode>2): Replace with...
	(*<SVE_COND_FCVTI:optab>vnx2df<SVE_SDI:mode>2): ...this.
	(vec_pack_<su>fix_trunc_vnx2df): Use SVE_COND_FCVTI instead of
	FIXUORS.
	(<FLOATUORS:optab><v_int_equiv><SVE_F:mode>2): Replace with...
	(<SVE_COND_ICVTF:optab><v_int_equiv><SVE_F:mode>2): ...this.
	(*<FLOATUORS:optab><SVE_HSDI:mode>vnx8hf2): Replace with...
	(*<SVE_COND_ICVTF:optab><SVE_HSDI:mode>vnx8hf2): ...this.
	(*<FLOATUORS:optab><SVE_SDI:mode>vnx4sf2): Replace with...
	(*<SVE_COND_ICVTF:optab><SVE_SDI:mode>vnx4sf2): ...this.
	(aarch64_sve_<FLOATUORS:optab><SVE_SDI:mode>vnx2df2): Replace with...
	(aarch64_sve_<SVE_COND_ICVTF:optab><SVE_SDI:mode>vnx2df2): ...this.
	(vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Pass a GP strictness
	operand to aarch64_sve_<SVE_COND_ICVTF:optab><SVE_SDI:mode>vnx2df2.
	(vec_pack_trunc_<SVE_HSF:Vwide>, *trunc<Vwide><SVE_HSF:mode>2)
	(aarch64_sve_extend<mode><Vwide>2): Use UNSPEC_COND_FCVT instead
	of UNSPEC_FLOAT_CONVERT.
	(vec_unpacks_<perm_hilo>_<mode>): Pass a GP strictness operand to
	aarch64_sve_extend<mode><Vwide>2.
diff mbox series

Patch

Index: gcc/config/aarch64/aarch64.md
===================================================================
--- gcc/config/aarch64/aarch64.md	2019-08-14 09:15:57.617827961 +0100
+++ gcc/config/aarch64/aarch64.md	2019-08-14 09:32:41.846404780 +0100
@@ -226,7 +226,6 @@  (define_c_enum "unspec" [
     UNSPEC_UNPACKSLO
     UNSPEC_UNPACKULO
     UNSPEC_PACK
-    UNSPEC_FLOAT_CONVERT
     UNSPEC_WHILE_LO
     UNSPEC_LDN
     UNSPEC_STN
Index: gcc/config/aarch64/iterators.md
===================================================================
--- gcc/config/aarch64/iterators.md	2019-08-14 09:29:52.871653684 +0100
+++ gcc/config/aarch64/iterators.md	2019-08-14 09:32:41.846404780 +0100
@@ -480,6 +480,9 @@  (define_c_enum "unspec"
     UNSPEC_COND_FCMLT	; Used in aarch64-sve.md.
     UNSPEC_COND_FCMNE	; Used in aarch64-sve.md.
     UNSPEC_COND_FCMUO	; Used in aarch64-sve.md.
+    UNSPEC_COND_FCVT	; Used in aarch64-sve.md.
+    UNSPEC_COND_FCVTZS	; Used in aarch64-sve.md.
+    UNSPEC_COND_FCVTZU	; Used in aarch64-sve.md.
     UNSPEC_COND_FDIV	; Used in aarch64-sve.md.
     UNSPEC_COND_FMAXNM	; Used in aarch64-sve.md.
     UNSPEC_COND_FMINNM	; Used in aarch64-sve.md.
@@ -498,6 +501,8 @@  (define_c_enum "unspec"
     UNSPEC_COND_FRINTZ	; Used in aarch64-sve.md.
     UNSPEC_COND_FSQRT	; Used in aarch64-sve.md.
     UNSPEC_COND_FSUB	; Used in aarch64-sve.md.
+    UNSPEC_COND_SCVTF	; Used in aarch64-sve.md.
+    UNSPEC_COND_UCVTF	; Used in aarch64-sve.md.
     UNSPEC_LASTB	; Used in aarch64-sve.md.
     UNSPEC_FCADD90	; Used in aarch64-simd.md.
     UNSPEC_FCADD270	; Used in aarch64-simd.md.
@@ -1642,6 +1647,9 @@  (define_int_iterator SVE_COND_FP_UNARY [
 					UNSPEC_COND_FRINTZ
 					UNSPEC_COND_FSQRT])
 
+(define_int_iterator SVE_COND_FCVTI [UNSPEC_COND_FCVTZS UNSPEC_COND_FCVTZU])
+(define_int_iterator SVE_COND_ICVTF [UNSPEC_COND_SCVTF UNSPEC_COND_UCVTF])
+
 (define_int_iterator SVE_COND_FP_BINARY [UNSPEC_COND_FADD
 					 UNSPEC_COND_FDIV
 					 UNSPEC_COND_FMAXNM
@@ -1715,6 +1723,9 @@  (define_int_attr optab [(UNSPEC_ANDF "an
 			(UNSPEC_FMINV "smin_nan")
 			(UNSPEC_COND_FABS "abs")
 			(UNSPEC_COND_FADD "add")
+			(UNSPEC_COND_FCVT "fcvt")
+			(UNSPEC_COND_FCVTZS "fix_trunc")
+			(UNSPEC_COND_FCVTZU "fixuns_trunc")
 			(UNSPEC_COND_FDIV "div")
 			(UNSPEC_COND_FMAXNM "smax")
 			(UNSPEC_COND_FMINNM "smin")
@@ -1732,7 +1743,9 @@  (define_int_attr optab [(UNSPEC_ANDF "an
 			(UNSPEC_COND_FRINTX "rint")
 			(UNSPEC_COND_FRINTZ "btrunc")
 			(UNSPEC_COND_FSQRT "sqrt")
-			(UNSPEC_COND_FSUB "sub")])
+			(UNSPEC_COND_FSUB "sub")
+			(UNSPEC_COND_SCVTF "float")
+			(UNSPEC_COND_UCVTF "floatuns")])
 
 (define_int_attr  maxmin_uns [(UNSPEC_UMAXV "umax")
 			      (UNSPEC_UMINV "umin")
@@ -1773,7 +1786,11 @@  (define_int_attr su [(UNSPEC_UNPACKSHI "
 		     (UNSPEC_UNPACKSLO "s")
 		     (UNSPEC_UNPACKULO "u")
 		     (UNSPEC_SMUL_HIGHPART "s")
-		     (UNSPEC_UMUL_HIGHPART "u")])
+		     (UNSPEC_UMUL_HIGHPART "u")
+		     (UNSPEC_COND_FCVTZS "s")
+		     (UNSPEC_COND_FCVTZU "u")
+		     (UNSPEC_COND_SCVTF "s")
+		     (UNSPEC_COND_UCVTF "u")])
 
 (define_int_attr sur [(UNSPEC_SHADD "s") (UNSPEC_UHADD "u")
 		      (UNSPEC_SRHADD "sr") (UNSPEC_URHADD "ur")
Index: gcc/config/aarch64/aarch64-sve.md
===================================================================
--- gcc/config/aarch64/aarch64-sve.md	2019-08-14 09:29:52.863653741 +0100
+++ gcc/config/aarch64/aarch64-sve.md	2019-08-14 09:32:41.842404810 +0100
@@ -3643,13 +3643,13 @@  (define_insn "aarch64_sve_<su>unpk<perm_
 
 ;; Unpredicated conversion of floats to integers of the same size (HF to HI,
 ;; SF to SI or DF to DI).
-(define_expand "<fix_trunc_optab><mode><v_int_equiv>2"
+(define_expand "<optab><mode><v_int_equiv>2"
   [(set (match_operand:<V_INT_EQUIV> 0 "register_operand")
 	(unspec:<V_INT_EQUIV>
 	  [(match_dup 2)
-	   (FIXUORS:<V_INT_EQUIV>
-	     (match_operand:SVE_F 1 "register_operand"))]
-	  UNSPEC_MERGE_PTRUE))]
+	   (const_int SVE_RELAXED_GP)
+	   (match_operand:SVE_F 1 "register_operand")]
+	  SVE_COND_FCVTI))]
   "TARGET_SVE"
   {
     operands[2] = aarch64_ptrue_reg (<VPRED>mode);
@@ -3657,37 +3657,37 @@  (define_expand "<fix_trunc_optab><mode><
 )
 
 ;; Conversion of SF to DI, SI or HI, predicated with a PTRUE.
-(define_insn "*<fix_trunc_optab>v16hsf<mode>2"
+(define_insn "*<optab>v16hsf<mode>2"
   [(set (match_operand:SVE_HSDI 0 "register_operand" "=w")
 	(unspec:SVE_HSDI
 	  [(match_operand:<VPRED> 1 "register_operand" "Upl")
-	   (FIXUORS:SVE_HSDI
-	     (match_operand:VNx8HF 2 "register_operand" "w"))]
-	  UNSPEC_MERGE_PTRUE))]
+	   (match_operand:SI 3 "aarch64_sve_gp_strictness")
+	   (match_operand:VNx8HF 2 "register_operand" "w")]
+	  SVE_COND_FCVTI))]
   "TARGET_SVE"
   "fcvtz<su>\t%0.<Vetype>, %1/m, %2.h"
 )
 
 ;; Conversion of SF to DI or SI, predicated with a PTRUE.
-(define_insn "*<fix_trunc_optab>vnx4sf<mode>2"
+(define_insn "*<optab>vnx4sf<mode>2"
   [(set (match_operand:SVE_SDI 0 "register_operand" "=w")
 	(unspec:SVE_SDI
 	  [(match_operand:<VPRED> 1 "register_operand" "Upl")
-	   (FIXUORS:SVE_SDI
-	     (match_operand:VNx4SF 2 "register_operand" "w"))]
-	  UNSPEC_MERGE_PTRUE))]
+	   (match_operand:SI 3 "aarch64_sve_gp_strictness")
+	   (match_operand:VNx4SF 2 "register_operand" "w")]
+	  SVE_COND_FCVTI))]
   "TARGET_SVE"
   "fcvtz<su>\t%0.<Vetype>, %1/m, %2.s"
 )
 
 ;; Conversion of DF to DI or SI, predicated with a PTRUE.
-(define_insn "*<fix_trunc_optab>vnx2df<mode>2"
+(define_insn "*<optab>vnx2df<mode>2"
   [(set (match_operand:SVE_SDI 0 "register_operand" "=w")
 	(unspec:SVE_SDI
 	  [(match_operand:VNx2BI 1 "register_operand" "Upl")
-	   (FIXUORS:SVE_SDI
-	     (match_operand:VNx2DF 2 "register_operand" "w"))]
-	  UNSPEC_MERGE_PTRUE))]
+	   (match_operand:SI 3 "aarch64_sve_gp_strictness")
+	   (match_operand:VNx2DF 2 "register_operand" "w")]
+	  SVE_COND_FCVTI))]
   "TARGET_SVE"
   "fcvtz<su>\t%0.<Vetype>, %1/m, %2.d"
 )
@@ -3703,13 +3703,15 @@  (define_expand "vec_pack_<su>fix_trunc_v
   [(set (match_dup 4)
 	(unspec:VNx4SI
 	  [(match_dup 3)
-	   (FIXUORS:VNx4SI (match_operand:VNx2DF 1 "register_operand"))]
-	  UNSPEC_MERGE_PTRUE))
+	   (const_int SVE_RELAXED_GP)
+	   (match_operand:VNx2DF 1 "register_operand")]
+	  SVE_COND_FCVTI))
    (set (match_dup 5)
 	(unspec:VNx4SI
 	  [(match_dup 3)
-	   (FIXUORS:VNx4SI (match_operand:VNx2DF 2 "register_operand"))]
-	  UNSPEC_MERGE_PTRUE))
+	   (const_int SVE_RELAXED_GP)
+	   (match_operand:VNx2DF 2 "register_operand")]
+	  SVE_COND_FCVTI))
    (set (match_operand:VNx4SI 0 "register_operand")
 	(unspec:VNx4SI [(match_dup 4) (match_dup 5)] UNSPEC_UZP1))]
   "TARGET_SVE"
@@ -3740,9 +3742,9 @@  (define_expand "<optab><v_int_equiv><mod
   [(set (match_operand:SVE_F 0 "register_operand")
 	(unspec:SVE_F
 	  [(match_dup 2)
-	   (FLOATUORS:SVE_F
-	     (match_operand:<V_INT_EQUIV> 1 "register_operand"))]
-	  UNSPEC_MERGE_PTRUE))]
+	   (const_int SVE_RELAXED_GP)
+	   (match_operand:<V_INT_EQUIV> 1 "register_operand")]
+	  SVE_COND_ICVTF))]
   "TARGET_SVE"
   {
     operands[2] = aarch64_ptrue_reg (<VPRED>mode);
@@ -3755,11 +3757,11 @@  (define_insn "*<optab><mode>vnx8hf2"
   [(set (match_operand:VNx8HF 0 "register_operand" "=w")
 	(unspec:VNx8HF
 	  [(match_operand:<VPRED> 1 "register_operand" "Upl")
-	   (FLOATUORS:VNx8HF
-	     (match_operand:SVE_HSDI 2 "register_operand" "w"))]
-	  UNSPEC_MERGE_PTRUE))]
+	   (match_operand:SI 3 "aarch64_sve_gp_strictness")
+	   (match_operand:SVE_HSDI 2 "register_operand" "w")]
+	  SVE_COND_ICVTF))]
   "TARGET_SVE"
-  "<su_optab>cvtf\t%0.h, %1/m, %2.<Vetype>"
+  "<su>cvtf\t%0.h, %1/m, %2.<Vetype>"
 )
 
 ;; Conversion of DI or SI to the same number of SFs, predicated with a PTRUE.
@@ -3767,11 +3769,11 @@  (define_insn "*<optab><mode>vnx4sf2"
   [(set (match_operand:VNx4SF 0 "register_operand" "=w")
 	(unspec:VNx4SF
 	  [(match_operand:<VPRED> 1 "register_operand" "Upl")
-	   (FLOATUORS:VNx4SF
-	     (match_operand:SVE_SDI 2 "register_operand" "w"))]
-	  UNSPEC_MERGE_PTRUE))]
+	   (match_operand:SI 3 "aarch64_sve_gp_strictness")
+	   (match_operand:SVE_SDI 2 "register_operand" "w")]
+	  SVE_COND_ICVTF))]
   "TARGET_SVE"
-  "<su_optab>cvtf\t%0.s, %1/m, %2.<Vetype>"
+  "<su>cvtf\t%0.s, %1/m, %2.<Vetype>"
 )
 
 ;; Conversion of DI or SI to DF, predicated with a PTRUE.
@@ -3779,11 +3781,11 @@  (define_insn "aarch64_sve_<optab><mode>v
   [(set (match_operand:VNx2DF 0 "register_operand" "=w")
 	(unspec:VNx2DF
 	  [(match_operand:VNx2BI 1 "register_operand" "Upl")
-	   (FLOATUORS:VNx2DF
-	     (match_operand:SVE_SDI 2 "register_operand" "w"))]
-	  UNSPEC_MERGE_PTRUE))]
+	   (match_operand:SI 3 "aarch64_sve_gp_strictness")
+	   (match_operand:SVE_SDI 2 "register_operand" "w")]
+	  SVE_COND_ICVTF))]
   "TARGET_SVE"
-  "<su_optab>cvtf\t%0.d, %1/m, %2.<Vetype>"
+  "<su>cvtf\t%0.d, %1/m, %2.<Vetype>"
 )
 
 ;; -------------------------------------------------------------------------
@@ -3818,8 +3820,9 @@  (define_expand "vec_unpack<su_optab>_flo
 		: gen_aarch64_sve_zip1vnx4si)
 	       (temp, operands[1], operands[1]));
     rtx ptrue = aarch64_ptrue_reg (VNx2BImode);
-    emit_insn (gen_aarch64_sve_<FLOATUORS:optab>vnx4sivnx2df2 (operands[0],
-							       ptrue, temp));
+    rtx strictness = gen_int_mode (SVE_RELAXED_GP, SImode);
+    emit_insn (gen_aarch64_sve_<FLOATUORS:optab>vnx4sivnx2df2
+	       (operands[0], ptrue, temp, strictness));
     DONE;
   }
 )
@@ -3837,15 +3840,15 @@  (define_expand "vec_pack_trunc_<Vwide>"
   [(set (match_dup 4)
 	(unspec:SVE_HSF
 	  [(match_dup 3)
-	   (unspec:SVE_HSF [(match_operand:<VWIDE> 1 "register_operand")]
-			   UNSPEC_FLOAT_CONVERT)]
-	  UNSPEC_MERGE_PTRUE))
+	   (const_int SVE_RELAXED_GP)
+	   (match_operand:<VWIDE> 1 "register_operand")]
+	  UNSPEC_COND_FCVT))
    (set (match_dup 5)
 	(unspec:SVE_HSF
 	  [(match_dup 3)
-	   (unspec:SVE_HSF [(match_operand:<VWIDE> 2 "register_operand")]
-			   UNSPEC_FLOAT_CONVERT)]
-	  UNSPEC_MERGE_PTRUE))
+	   (const_int SVE_RELAXED_GP)
+	   (match_operand:<VWIDE> 2 "register_operand")]
+	  UNSPEC_COND_FCVT))
    (set (match_operand:SVE_HSF 0 "register_operand")
 	(unspec:SVE_HSF [(match_dup 4) (match_dup 5)] UNSPEC_UZP1))]
   "TARGET_SVE"
@@ -3862,10 +3865,9 @@  (define_insn "*trunc<Vwide><mode>2"
   [(set (match_operand:SVE_HSF 0 "register_operand" "=w")
 	(unspec:SVE_HSF
 	  [(match_operand:<VWIDE_PRED> 1 "register_operand" "Upl")
-	   (unspec:SVE_HSF
-	     [(match_operand:<VWIDE> 2 "register_operand" "w")]
-	     UNSPEC_FLOAT_CONVERT)]
-	  UNSPEC_MERGE_PTRUE))]
+	   (match_operand:SI 3 "aarch64_sve_gp_strictness")
+	   (match_operand:<VWIDE> 2 "register_operand" "w")]
+	  UNSPEC_COND_FCVT))]
   "TARGET_SVE"
   "fcvt\t%0.<Vetype>, %1/m, %2.<Vewtype>"
 )
@@ -3896,8 +3898,9 @@  (define_expand "vec_unpacks_<perm_hilo>_
 		: gen_aarch64_sve_zip1<mode>)
 		(temp, operands[1], operands[1]));
     rtx ptrue = aarch64_ptrue_reg (<VWIDE_PRED>mode);
-    emit_insn (gen_aarch64_sve_extend<mode><Vwide>2 (operands[0],
-						     ptrue, temp));
+    rtx strictness = gen_int_mode (SVE_RELAXED_GP, SImode);
+    emit_insn (gen_aarch64_sve_extend<mode><Vwide>2
+	       (operands[0], ptrue, temp, strictness));
     DONE;
   }
 )
@@ -3908,10 +3911,9 @@  (define_insn "aarch64_sve_extend<mode><V
   [(set (match_operand:<VWIDE> 0 "register_operand" "=w")
 	(unspec:<VWIDE>
 	  [(match_operand:<VWIDE_PRED> 1 "register_operand" "Upl")
-	   (unspec:<VWIDE>
-	     [(match_operand:SVE_HSF 2 "register_operand" "w")]
-	     UNSPEC_FLOAT_CONVERT)]
-	  UNSPEC_MERGE_PTRUE))]
+	   (match_operand:SI 3 "aarch64_sve_gp_strictness")
+	   (match_operand:SVE_HSF 2 "register_operand" "w")]
+	  UNSPEC_COND_FCVT))]
   "TARGET_SVE"
   "fcvt\t%0.<Vewtype>, %1/m, %2.<Vetype>"
 )