BRIG FE testsuite: Fix dump scan patterns in packed.hsail test

BRIG FE testsuite: Fix dump scan patterns in packed.hsail test

Starting with r11-165-eb72dc663e9 which converted DECL_GIMPLE_REG_P to
DECL_NOT_GIMPLE_REG_P we have failing BRIG testcase:

-PASS: packed.hsail.brig scan-tree-dump gimple "_[0-9]+ = q2 \\+ q3;"
-PASS: packed.hsail.brig scan-tree-dump gimple "= VEC_PERM_EXPR <new_output.[0-9]+_[0-9]+, [a-z0-9_]+, { 0, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 }>;"
+FAIL: packed.hsail.brig scan-tree-dump gimple "_[0-9]+ = q2 \\+ q3;"
+FAIL: packed.hsail.brig scan-tree-dump gimple "= VEC_PERM_EXPR <new_output.[0-9]+_[0-9]+, [a-z0-9_]+, { 0, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 }>;"

because the gimplifier is now smarter and generates nicer code, which
however, does not match the regexp in the testsuite:

--- before/packed.hsail.brig.005t.gimple        2020-05-12 17:59:26.434305513 +0200
+++ after/packed.hsail.brig.005t.gimple 2020-05-12 17:52:34.477055987 +0200
@@ -109,277 +109,267 @@
   q2 = q1 + _24;
   _25 = VEC_PERM_EXPR <q1, q1, { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }>;
   q3 = q2 + _25;
-  _26 = q2 + q3;
-  new_output.11 = _26;
-  new_output.21_27 = new_output.11;
-  _28 = VEC_PERM_EXPR <new_output.21_27, q4, { 0, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 }>;
-  s_output.12 = _28;
+  new_output.11 = q2 + q3;
+  s_output.12 = VEC_PERM_EXPR <new_output.11, q4, { 0, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 }>;
   q4 = s_output.12;

I have looked at the SSA dump and verified that the variable in
question is a gimple register because it gets its SSA name.  I have
not looked into why the gimplifier previously though it had to go
through the additional temporaries though.

Tested with make -k check-brig.

	* brig.dg/test/gimple/packed.hsail: Fix scan dump patterns.