[AArch64] Generate permute patterns using rtx builders

This patch replaces switch statements that call specific generator
functions with code that constructs the rtl pattern directly.
This seemed to scale better to SVE and also seems less error-prone.

As a side-effect, the patch fixes the REV handling for diff==1,
vmode==E_V4HFmode and adds missing support for diff==3,
vmode==E_V4HFmode.

To compensate for the lack of switches that check for specific modes,
the patch makes aarch64_expand_vec_perm_const_1 reject permutes on
single-element vectors (specifically V1DImode).

2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
	    Alan Hayward  <alan.hayward@arm.com>
	    David Sherwood  <david.sherwood@arm.com>

gcc/
	* config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
	(aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
	(aarch64_evpc_dup): Generate rtl direcly, rather than using
	named expanders.
	(aarch64_expand_vec_perm_const_1): Explicitly check for permutes
	of a single element.
	* config/aarch64/iterators.md: Add a comment above the permute
	unspecs to say that they are generated directly by
	aarch64_expand_vec_perm_const.
	* config/aarch64/aarch64-simd.md: Likewise the permute instructions.

Reviewed-by: James Greenhalgh <james.greenhalgh@arm.com>

Co-Authored-By: Alan Hayward <alan.hayward@arm.com>
Co-Authored-By: David Sherwood <david.sherwood@arm.com>

From-SVN: r254324

gcc/ChangeLog

+2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
+	    Alan Hayward  <alan.hayward@arm.com>
+	    David Sherwood  <david.sherwood@arm.com>
+
+	* config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
+	(aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
+	(aarch64_evpc_dup): Generate rtl direcly, rather than using
+	named expanders.
+	(aarch64_expand_vec_perm_const_1): Explicitly check for permutes
+	of a single element.
+	* config/aarch64/iterators.md: Add a comment above the permute
+	unspecs to say that they are generated directly by
+	aarch64_expand_vec_perm_const.
+	* config/aarch64/aarch64-simd.md: Likewise the permute instructions.
+
 2017-11-01  Nathan Sidwell  <nathan@acm.org>
 
 	* tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.

gcc/config/aarch64/aarch64-simd.md

@@ -5369,6 +5369,9 @@
 [(set_attr "type" "multiple")]
 )
 
+;; This instruction's pattern is generated directly by
+;; aarch64_expand_vec_perm_const, so any changes to the pattern would
+;; need corresponding changes there.
 (define_insn "aarch64_<PERMUTE:perm_insn><PERMUTE:perm_hilo><mode>"
   [(set (match_operand:VALL_F16 0 "register_operand" "=w")
 	(unspec:VALL_F16 [(match_operand:VALL_F16 1 "register_operand" "w")
@@ -5379,7 +5382,10 @@
   [(set_attr "type" "neon_permute<q>")]
 )
 
-;; Note immediate (third) operand is lane index not byte index.
+;; This instruction's pattern is generated directly by
+;; aarch64_expand_vec_perm_const, so any changes to the pattern would
+;; need corresponding changes there.  Note that the immediate (third)
+;; operand is a lane index not a byte index.
 (define_insn "aarch64_ext<mode>"
   [(set (match_operand:VALL_F16 0 "register_operand" "=w")
         (unspec:VALL_F16 [(match_operand:VALL_F16 1 "register_operand" "w")
@@ -5395,6 +5401,9 @@
   [(set_attr "type" "neon_ext<q>")]
 )
 
+;; This instruction's pattern is generated directly by
+;; aarch64_expand_vec_perm_const, so any changes to the pattern would
+;; need corresponding changes there.
 (define_insn "aarch64_rev<REVERSE:rev_op><mode>"
   [(set (match_operand:VALL_F16 0 "register_operand" "=w")
 	(unspec:VALL_F16 [(match_operand:VALL_F16 1 "register_operand" "w")]

gcc/config/aarch64/iterators.md

@@ -322,16 +322,21 @@
     UNSPEC_TBL		; Used in vector permute patterns.
     UNSPEC_TBX		; Used in vector permute patterns.
     UNSPEC_CONCAT	; Used in vector permute patterns.
+
+    ;; The following permute unspecs are generated directly by
+    ;; aarch64_expand_vec_perm_const, so any changes to the underlying
+    ;; instructions would need a corresponding change there.
     UNSPEC_ZIP1		; Used in vector permute patterns.
     UNSPEC_ZIP2		; Used in vector permute patterns.
     UNSPEC_UZP1		; Used in vector permute patterns.
     UNSPEC_UZP2		; Used in vector permute patterns.
     UNSPEC_TRN1		; Used in vector permute patterns.
     UNSPEC_TRN2		; Used in vector permute patterns.
-    UNSPEC_EXT		; Used in aarch64-simd.md.
+    UNSPEC_EXT		; Used in vector permute patterns.
     UNSPEC_REV64	; Used in vector reverse patterns (permute).
     UNSPEC_REV32	; Used in vector reverse patterns (permute).
     UNSPEC_REV16	; Used in vector reverse patterns (permute).
+
     UNSPEC_AESE		; Used in aarch64-simd.md.
     UNSPEC_AESD         ; Used in aarch64-simd.md.
     UNSPEC_AESMC        ; Used in aarch64-simd.md.