| Message ID | DB6PR0801MB2053F830C8FA9359EBC8946283380@DB6PR0801MB2053.eurprd08.prod.outlook.com |
|---|---|
| State | New |
| Headers | show |
| Series | [AArch64] Fix address printing on ILP32 | expand |
On Thu, 2017-11-30 at 17:27 +0000, Wilco Dijkstra wrote: > > This fixes the ICE in https://gcc.gnu.org/ml/gcc-patches/2017-11/msg02509.html. > > > ChangeLog: > 2017-11-30 Wilco Dijkstra <wdijkstr@arm.com> > > gcc/ > * config/aarch64/aarch64.md (call_insn): Use %c rather than %a. > (call_value_insn): Likewise. > (sibcall_insn): Likewise. > (sibcall_value_insn): Likewise. > (movsi_aarch64): Likewise. > (movdi_aarch64): Likewise. > (add_losym_): Likewise. > (ldr_got_small_): Likewise. > (ldr_got_small_sidi): Likewise. > (ldr_got_small_28k_): Likewise. > (ldr_got_small_28k_sidi): Likewise. > * config/aarch64/aarch64.c (aarch64_print_address_internal): > Move output_addr_const to symbolic case. Add error check. > testsuite/ > * gcc.dg/asm-4.c: Skip on AArch64 with ILP32 as test is incorrect. > This fixed the build problems on my system with ILP32. Steve Ellcey sellcey@cavium.com
On Thu, Nov 30, 2017 at 05:27:47PM +0000, Wilco Dijkstra wrote: > Fix address printing for ILP32. The md file uses 'a' in assembler > templates for symbolic addresses in adrp/add, which end up calling > aarch64_print_operand_address. However in ILP32 these are not valid > memory addresses (being ptr_mode rather than Pmode), so the assert > triggers. Since it is incorrect to use symbols in memory addresses > (besides literal pool accesses), change the 'a' to 'c' in the md file. > > Skip one failing test in ILP32 which combines the 'p' modifier with the 'a' > assembler template to fake a memory reference. > > This fixes the ICE in https://gcc.gnu.org/ml/gcc-patches/2017-11/msg02509.html. OK. Thanks, James > ChangeLog: > 2017-11-30 Wilco Dijkstra <wdijkstr@arm.com> > > gcc/ > * config/aarch64/aarch64.md (call_insn): Use %c rather than %a. > (call_value_insn): Likewise. > (sibcall_insn): Likewise. > (sibcall_value_insn): Likewise. > (movsi_aarch64): Likewise. > (movdi_aarch64): Likewise. > (add_losym_): Likewise. > (ldr_got_small_): Likewise. > (ldr_got_small_sidi): Likewise. > (ldr_got_small_28k_): Likewise. > (ldr_got_small_28k_sidi): Likewise. > * config/aarch64/aarch64.c (aarch64_print_address_internal): > Move output_addr_const to symbolic case. Add error check. > testsuite/ > * gcc.dg/asm-4.c: Skip on AArch64 with ILP32 as test is incorrect. >
diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c index 6d89b164275c0b31d65a7c5c0361a6d55700215c..7b50ab43dbc075e6b6d4541c3fb71e5cc872c88b 100644 --- a/gcc/config/aarch64/aarch64.c +++ b/gcc/config/aarch64/aarch64.c @@ -5936,10 +5936,11 @@ aarch64_print_address_internal (FILE *f, machine_mode mode, rtx x, RTX_CODE op) return; case ADDRESS_SYMBOLIC: - break; + output_addr_const (f, x); + return; } - output_addr_const (f, x); + output_operand_lossage ("invalid operand for '%%%c'", op); } /* Print address 'x' of a LDP/STP with mode 'mode'. */ diff --git a/gcc/config/aarch64/aarch64.md b/gcc/config/aarch64/aarch64.md index 8744722b8daf4f89aca65bbce75ff743df025f47..d14b57b0ef7f4eeca40bfdcaf3ebb02a1031cb99 100644 --- a/gcc/config/aarch64/aarch64.md +++ b/gcc/config/aarch64/aarch64.md @@ -749,7 +749,7 @@ (define_insn "*call_insn" "" "@ blr\\t%0 - bl\\t%a0" + bl\\t%c0" [(set_attr "type" "call, call")] ) @@ -775,7 +775,7 @@ (define_insn "*call_value_insn" "" "@ blr\\t%1 - bl\\t%a1" + bl\\t%c1" [(set_attr "type" "call, call")] ) @@ -811,7 +811,7 @@ (define_insn "*sibcall_insn" "SIBLING_CALL_P (insn)" "@ br\\t%0 - b\\t%a0" + b\\t%c0" [(set_attr "type" "branch, branch")] ) @@ -824,7 +824,7 @@ (define_insn "*sibcall_value_insn" "SIBLING_CALL_P (insn)" "@ br\\t%1 - b\\t%a1" + b\\t%c1" [(set_attr "type" "branch, branch")] ) @@ -946,7 +946,7 @@ (define_insn_and_split "*movsi_aarch64" ldr\\t%s0, %1 str\\t%w1, %0 str\\t%s1, %0 - adr\\t%x0, %a1 + adr\\t%x0, %c1 adrp\\t%x0, %A1 fmov\\t%s0, %w1 fmov\\t%w0, %s1 @@ -981,7 +981,7 @@ (define_insn_and_split "*movdi_aarch64" ldr\\t%d0, %1 str\\t%x1, %0 str\\t%d1, %0 - adr\\t%x0, %a1 + adr\\t%x0, %c1 adrp\\t%x0, %A1 fmov\\t%d0, %x1 fmov\\t%x0, %d1 @@ -5401,7 +5401,7 @@ (define_insn "add_losym_<mode>" (lo_sum:P (match_operand:P 1 "register_operand" "r") (match_operand 2 "aarch64_valid_symref" "S")))] "" - "add\\t%<w>0, %<w>1, :lo12:%a2" + "add\\t%<w>0, %<w>1, :lo12:%c2" [(set_attr "type" "alu_imm")] ) @@ -5412,7 +5412,7 @@ (define_insn "ldr_got_small_<mode>" (match_operand:PTR 2 "aarch64_valid_symref" "S")))] UNSPEC_GOTSMALLPIC))] "" - "ldr\\t%<w>0, [%1, #:got_lo12:%a2]" + "ldr\\t%<w>0, [%1, #:got_lo12:%c2]" [(set_attr "type" "load_<ldst_sz>")] ) @@ -5424,7 +5424,7 @@ (define_insn "ldr_got_small_sidi" (match_operand:DI 2 "aarch64_valid_symref" "S")))] UNSPEC_GOTSMALLPIC)))] "TARGET_ILP32" - "ldr\\t%w0, [%1, #:got_lo12:%a2]" + "ldr\\t%w0, [%1, #:got_lo12:%c2]" [(set_attr "type" "load_4")] ) @@ -5435,7 +5435,7 @@ (define_insn "ldr_got_small_28k_<mode>" (match_operand:PTR 2 "aarch64_valid_symref" "S")))] UNSPEC_GOTSMALLPIC28K))] "" - "ldr\\t%<w>0, [%1, #:<got_modifier>:%a2]" + "ldr\\t%<w>0, [%1, #:<got_modifier>:%c2]" [(set_attr "type" "load_<ldst_sz>")] ) @@ -5447,7 +5447,7 @@ (define_insn "ldr_got_small_28k_sidi" (match_operand:DI 2 "aarch64_valid_symref" "S")))] UNSPEC_GOTSMALLPIC28K)))] "TARGET_ILP32" - "ldr\\t%w0, [%1, #:gotpage_lo14:%a2]" + "ldr\\t%w0, [%1, #:gotpage_lo14:%c2]" [(set_attr "type" "load_4")] ) diff --git a/gcc/testsuite/gcc.dg/asm-4.c b/gcc/testsuite/gcc.dg/asm-4.c index 1e6a538aeabba00cd1bf3d0d636ceec4ed50cf4a..f142daac8c65688fae8b8f3eadaa2066cc8fbd86 100644 --- a/gcc/testsuite/gcc.dg/asm-4.c +++ b/gcc/testsuite/gcc.dg/asm-4.c @@ -1,6 +1,9 @@ /* { dg-do compile } */ /* { dg-options "" } */ +/* "p" modifier can't be used to generate a valid memory address with ILP32. */ +/* { dg-skip-if "" { aarch64*-*-* && ilp32 } } */ + int main() { int x, y, z;
Fix address printing for ILP32. The md file uses 'a' in assembler templates for symbolic addresses in adrp/add, which end up calling aarch64_print_operand_address. However in ILP32 these are not valid memory addresses (being ptr_mode rather than Pmode), so the assert triggers. Since it is incorrect to use symbols in memory addresses (besides literal pool accesses), change the 'a' to 'c' in the md file. Skip one failing test in ILP32 which combines the 'p' modifier with the 'a' assembler template to fake a memory reference. This fixes the ICE in https://gcc.gnu.org/ml/gcc-patches/2017-11/msg02509.html. ChangeLog: 2017-11-30 Wilco Dijkstra <wdijkstr@arm.com> gcc/ * config/aarch64/aarch64.md (call_insn): Use %c rather than %a. (call_value_insn): Likewise. (sibcall_insn): Likewise. (sibcall_value_insn): Likewise. (movsi_aarch64): Likewise. (movdi_aarch64): Likewise. (add_losym_): Likewise. (ldr_got_small_): Likewise. (ldr_got_small_sidi): Likewise. (ldr_got_small_28k_): Likewise. (ldr_got_small_28k_sidi): Likewise. * config/aarch64/aarch64.c (aarch64_print_address_internal): Move output_addr_const to symbolic case. Add error check. testsuite/ * gcc.dg/asm-4.c: Skip on AArch64 with ILP32 as test is incorrect. --