On Wed, Mar 17, 2021 at 6:14 AM Uros Bizjak <ubiz...@gmail.com> wrote: > > On Wed, Mar 17, 2021 at 1:45 PM H.J. Lu <hjl.to...@gmail.com> wrote: > > > > On Wed, Mar 17, 2021 at 12:40 AM Uros Bizjak <ubiz...@gmail.com> wrote: > > > > > > On Tue, Mar 16, 2021 at 4:59 PM H.J. Lu <hjl.to...@gmail.com> wrote: > > > > > > > > On Sun, Mar 14, 2021 at 1:31 PM H.J. Lu <hjl.to...@gmail.com> wrote: > > > > > > > > > > On Sun, Mar 14, 2021 at 12:43 PM Uros Bizjak <ubiz...@gmail.com> > > > > > wrote: > > > > > > > > > > > > On Sun, Mar 14, 2021 at 8:14 PM H.J. Lu <hjl.to...@gmail.com> wrote: > > > > > > > > > Done. Here is the updated patch. Tested on Linux/x86-64. > > > > > > > > > OK for master? > > > > > > > > > > > > > > > > I don't understand the purpose of the current_output_insn check > > > > > > > > and I > > > > > > > > don't know if the usage of current_output_insn is correct. The > > > > > > > > comments are not helpful either, and no other target uses this > > > > > > > > variable in the way you propose. Can you please elaborate the > > > > > > > > reason > > > > > > > > and the purpose of the check a bit more? > > > > > > > > > > > > > > > > Uros. > > > > > > > > > > > > > > Originally, ix86_force_load_from_GOT_p is only for non-PIC. My > > > > > > > patch extended > > > > > > > it to inline assembly statements where current_output_insn == > > > > > > > NULL and PIC is > > > > > > > allowed in 64-bit. > > > > > > > > > > > > I can see this from the patch, but this explanation didn't answer > > > > > > my question. > > > > > > > > > > > > > > > > The purpose of current_output_insn == NULL is to allow PIC for inline > > > > > asm statements in 64-bit mode. Is there a better way to check if > > > > > ix86_print_operand () is called on inline asm statements? > > > > > > > > > > > > > Here is the v4 patch to check this_is_asm_operands for inline > > > > asm statements. OK for master? > > > > > > -ENOPATCH. > > > > > > Uros. > > > > Here is the patch. > > -/* True if operand X should be loaded from GOT. */ > +/* True if the function symbol operand X should be loaded from GOT. > > vertical space here
Fixed. > + NB: For inline assembly statement, allow PIC/non-PIC in 64-bit mode > + and in 32-bit mode, only non-PIC is allowed since we don't have > + PIC register at call site. Otherwise, only PIC is allowed. */ > NB: in 32-bit mode, only non-PIC is allowed in inline assembly > statements, since a PIC register could not be available at the call site. Fixed. > bool > > + /* For inline assembly statement, load function address > + from GOT with 'P' operand modifier to avoid PLT. > + NB: This works only with call or jmp. */ > > Please remove this NB, you can't expect from users to read source > code. This should be in the documentation. Fixed. > LGTM otherwise. This is the patch I am checking in. Thanks. -- H.J.
From 3e0bb7c313f1c13b5a01af9ef57c6d0d766085fd Mon Sep 17 00:00:00 2001 From: "H.J. Lu" <hjl.to...@gmail.com> Date: Thu, 11 Mar 2021 06:48:24 -0800 Subject: [PATCH v5] x86: Update 'P' operand modifier for -fno-plt Update 'P' operand modifier for -fno-plt to support inline assembly statements. In 64-bit, we can always load function address with @GOTPCREL. In 32-bit, we load function address with @GOT only for non-PIC since PIC register may not be available at call site. gcc/ PR target/99504 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support inline assembly statements. (ix86_print_operand): Update 'P' handling for -fno-plt. gcc/testsuite/ PR target/99504 * gcc.target/i386/pr99530-1.c: New test. * gcc.target/i386/pr99530-2.c: Likewise. * gcc.target/i386/pr99530-3.c: Likewise. * gcc.target/i386/pr99530-4.c: Likewise. * gcc.target/i386/pr99530-5.c: Likewise. * gcc.target/i386/pr99530-6.c: Likewise. --- gcc/config/i386/i386.c | 32 +++++++++++++++++++---- gcc/testsuite/gcc.target/i386/pr99530-1.c | 11 ++++++++ gcc/testsuite/gcc.target/i386/pr99530-2.c | 11 ++++++++ gcc/testsuite/gcc.target/i386/pr99530-3.c | 11 ++++++++ gcc/testsuite/gcc.target/i386/pr99530-4.c | 11 ++++++++ gcc/testsuite/gcc.target/i386/pr99530-5.c | 11 ++++++++ gcc/testsuite/gcc.target/i386/pr99530-6.c | 11 ++++++++ 7 files changed, 93 insertions(+), 5 deletions(-) create mode 100644 gcc/testsuite/gcc.target/i386/pr99530-1.c create mode 100644 gcc/testsuite/gcc.target/i386/pr99530-2.c create mode 100644 gcc/testsuite/gcc.target/i386/pr99530-3.c create mode 100644 gcc/testsuite/gcc.target/i386/pr99530-4.c create mode 100644 gcc/testsuite/gcc.target/i386/pr99530-5.c create mode 100644 gcc/testsuite/gcc.target/i386/pr99530-6.c diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c index 2603333f87b..540d4f44517 100644 --- a/gcc/config/i386/i386.c +++ b/gcc/config/i386/i386.c @@ -10265,15 +10265,20 @@ darwin_local_data_pic (rtx disp) && XINT (disp, 1) == UNSPEC_MACHOPIC_OFFSET); } -/* True if operand X should be loaded from GOT. */ +/* True if the function symbol operand X should be loaded from GOT. + + NB: In 32-bit mode, only non-PIC is allowed in inline assembly + statements, since a PIC register could not be available at the + call site. */ bool ix86_force_load_from_GOT_p (rtx x) { - return ((TARGET_64BIT || HAVE_AS_IX86_GOT32X) + return ((TARGET_64BIT || (!flag_pic && HAVE_AS_IX86_GOT32X)) && !TARGET_PECOFF && !TARGET_MACHO - && !flag_pic + && (!flag_pic || this_is_asm_operands) && ix86_cmodel != CM_LARGE + && ix86_cmodel != CM_LARGE_PIC && GET_CODE (x) == SYMBOL_REF && SYMBOL_REF_FUNCTION_P (x) && (!flag_plt @@ -12701,7 +12706,8 @@ print_reg (rtx x, int code, FILE *file) y -- print "st(0)" instead of "st" as a register. d -- print duplicated register operand for AVX instruction. D -- print condition for SSE cmp instruction. - P -- if PIC, print an @PLT suffix. + P -- if PIC, print an @PLT suffix. For -fno-plt, load function + address from GOT. p -- print raw symbol name. X -- don't print any sort of PIC '@' suffix for a symbol. & -- print some in-use local-dynamic symbol name. @@ -13445,7 +13451,23 @@ ix86_print_operand (FILE *file, rtx x, int code) x = const0_rtx; } - if (code != 'P' && code != 'p') + if (code == 'P') + { + if (ix86_force_load_from_GOT_p (x)) + { + /* For inline assembly statement, load function address + from GOT with 'P' operand modifier to avoid PLT. */ + x = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, x), + (TARGET_64BIT + ? UNSPEC_GOTPCREL + : UNSPEC_GOT)); + x = gen_rtx_CONST (Pmode, x); + x = gen_const_mem (Pmode, x); + ix86_print_operand (file, x, 'A'); + return; + } + } + else if (code != 'p') { if (CONST_INT_P (x)) { diff --git a/gcc/testsuite/gcc.target/i386/pr99530-1.c b/gcc/testsuite/gcc.target/i386/pr99530-1.c new file mode 100644 index 00000000000..080d7cc9399 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr99530-1.c @@ -0,0 +1,11 @@ +/* { dg-do compile { target { i?86-*-linux* x86_64-*-linux* } } } */ +/* { dg-require-effective-target lp64 } */ +/* { dg-options "-O2 -fpic -mcmodel=large -fno-plt" } */ +/* { dg-final { scan-assembler-not "foo@GOTPCREL" } } */ + +extern void foo (void); +void +bar (void) +{ + asm ("call %P0" : : "X" (foo)); +} diff --git a/gcc/testsuite/gcc.target/i386/pr99530-2.c b/gcc/testsuite/gcc.target/i386/pr99530-2.c new file mode 100644 index 00000000000..9808957d624 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr99530-2.c @@ -0,0 +1,11 @@ +/* { dg-do compile { target { i?86-*-linux* x86_64-*-linux* } } } */ +/* { dg-require-effective-target lp64 } */ +/* { dg-options "-O2 -fno-pic -mcmodel=large -fno-plt" } */ +/* { dg-final { scan-assembler-not "foo@GOTPCREL" } } */ + +extern void foo (void); +void +bar (void) +{ + asm ("call %P0" : : "X" (foo)); +} diff --git a/gcc/testsuite/gcc.target/i386/pr99530-3.c b/gcc/testsuite/gcc.target/i386/pr99530-3.c new file mode 100644 index 00000000000..22fe81b25f2 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr99530-3.c @@ -0,0 +1,11 @@ +/* { dg-do compile { target { i?86-*-linux* x86_64-*-linux* } } } */ +/* { dg-options "-O2 -fno-pic -fno-plt" } */ +/* { dg-final { scan-assembler "call\[ \t\]+\[*\]foo@GOTPCREL" { target { ! ia32 } } } } */ +/* { dg-final { scan-assembler "call\[ \t\]+\[*\]foo@GOT" { target { ia32 && got32x_reloc } } } } */ + +extern void foo (void); +void +bar (void) +{ + asm ("call %P0" : : "X" (foo)); +} diff --git a/gcc/testsuite/gcc.target/i386/pr99530-4.c b/gcc/testsuite/gcc.target/i386/pr99530-4.c new file mode 100644 index 00000000000..aba0b5d7892 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr99530-4.c @@ -0,0 +1,11 @@ +/* { dg-do compile { target { i?86-*-linux* x86_64-*-linux* } } } */ +/* { dg-options "-O2 -fpic -fno-plt" } */ +/* { dg-final { scan-assembler "call\[ \t\]+\[*\]foo@GOTPCREL" { target { ! ia32 } } } } */ +/* { dg-final { scan-assembler-not "foo@GOT" { target { ia32 } } } } */ + +extern void foo (void); +void +bar (void) +{ + asm ("call %P0" : : "X" (foo)); +} diff --git a/gcc/testsuite/gcc.target/i386/pr99530-5.c b/gcc/testsuite/gcc.target/i386/pr99530-5.c new file mode 100644 index 00000000000..02c5799d273 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr99530-5.c @@ -0,0 +1,11 @@ +/* { dg-do compile { target { i?86-*-linux* x86_64-*-linux* } } } */ +/* { dg-options "-O2 -fno-pic" } */ +/* { dg-final { scan-assembler "call\[ \t\]+\[*\]foo@GOTPCREL" { target { ! ia32 } } } } */ +/* { dg-final { scan-assembler "call\[ \t\]+\[*\]foo@GOT" { target { ia32 && got32x_reloc } } } } */ + +extern void foo (void) __attribute__ ((noplt)); +void +bar (void) +{ + asm ("call %P0" : : "X" (foo)); +} diff --git a/gcc/testsuite/gcc.target/i386/pr99530-6.c b/gcc/testsuite/gcc.target/i386/pr99530-6.c new file mode 100644 index 00000000000..0dfdd3752f7 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr99530-6.c @@ -0,0 +1,11 @@ +/* { dg-do compile { target { i?86-*-linux* x86_64-*-linux* } } } */ +/* { dg-options "-O2 -fpic" } */ +/* { dg-final { scan-assembler "call\[ \t\]+\[*\]foo@GOTPCREL" { target { ! ia32 } } } } */ +/* { dg-final { scan-assembler-not "foo@GOT" { target { ia32 } } } } */ + +extern void foo (void) __attribute__ ((noplt)); +void +bar (void) +{ + asm ("call %P0" : : "X" (foo)); +} -- 2.30.2