Re: [PATCH v3 04/10] GCN machine description

[Thomas Schwinge <thomas@codesourcery.com>]

[-- Attachment #1: Type: text/plain, Size: 2732 bytes --]

Hi!

On 2018-12-12T11:52:23+0000, Andrew Stubbs <ams@codesourcery.com> wrote:
> This patch contains the machine description portion of the GCN
> back-end.

> --- /dev/null
> +++ b/gcc/config/gcn/gcn.md

> +(define_insn_and_split "*mov<mode>_insn"
> +  [(set (match_operand:DIDF 0 "nonimmediate_operand"
> +                       "=SD,SD,SD,RS,Sm,v, v,Sg, v, v,RF,RLRG,   v, v,RM")
> +     (match_operand:DIDF 1 "general_operand"
> +                       "SSA, C,DB,Sm,RS,v,DB, v,Sv,RF, v,   v,RLRG,RM, v"))]
> +  "GET_CODE(operands[1]) != SYMBOL_REF"
> +  "@
> +  s_mov_b64\t%0, %1
> +  s_mov_b64\t%0, %1
> +  #
> +  s_store_dwordx2\t%1, %A0\;s_waitcnt\tlgkmcnt(0)
> +  s_load_dwordx2\t%0, %A1\;s_waitcnt\tlgkmcnt(0)
> +  #
> +  #
> +  #
> +  #
> +  flat_load_dwordx2\t%0, %A1%O1%g1\;s_waitcnt\t0
> +  flat_store_dwordx2\t%A0, %1%O0%g0\;s_waitcnt\t0
> +  ds_write_b64\t%A0, %1%O0\;s_waitcnt\tlgkmcnt(0)
> +  ds_read_b64\t%0, %A1%O1\;s_waitcnt\tlgkmcnt(0)
> +  global_load_dwordx2\t%0, %A1%O1%g1\;s_waitcnt\tvmcnt(0)
> +  global_store_dwordx2\t%A0, %1%O0%g0\;s_waitcnt\tvmcnt(0)"
> +  "(reload_completed && !MEM_P (operands[0]) && !MEM_P (operands[1])
> +    && !gcn_sgpr_move_p (operands[0], operands[1]))
> +   || (GET_CODE (operands[1]) == CONST_INT && !gcn_constant64_p (operands[1]))"
> +  [(set (match_dup 0) (match_dup 1))
> +   (set (match_dup 2) (match_dup 3))]
> +  {
> +    rtx inlo = gen_lowpart (SImode, operands[1]);
> +    rtx inhi = gen_highpart_mode (SImode, <MODE>mode, operands[1]);
> +    rtx outlo = gen_lowpart (SImode, operands[0]);
> +    rtx outhi = gen_highpart_mode (SImode, <MODE>mode, operands[0]);
> +
> +    /* Ensure that overlapping registers aren't corrupted.  */
> +    if (REGNO (outlo) == REGNO (inhi))

As discussed in PR94248 "[amdgcn] Doesn't build with RTL checking", I've
tested Jakub's patch, and now pushed to master branch in commit
ccf93cd0b21e9c0ff0a1d4ace59899fca25ac157 "[gcn] Fix build with RTL
checking [PR94248]".

> +      {
> +     operands[0] = outhi;
> +     operands[1] = inhi;
> +     operands[2] = outlo;
> +     operands[3] = inlo;
> +      }
> +    else
> +      {
> +     operands[0] = outlo;
> +     operands[1] = inlo;
> +     operands[2] = outhi;
> +     operands[3] = inhi;
> +      }
> +  }
> +  [(set_attr "type" "sop1,sop1,mult,smem,smem,vmult,vmult,vmult,vmult,flat,
> +                  flat,ds,ds,flat,flat")
> +   (set_attr "length" "4,8,*,12,12,*,*,*,*,12,12,12,12,12,12")])


Grüße
 Thomas


-----------------
Mentor Graphics (Deutschland) GmbH, Arnulfstraße 201, 80634 München / Germany
Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Alexander Walter

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-gcn-Fix-build-with-RTL-checking-PR94248.patch --]
[-- Type: text/x-diff, Size: 2132 bytes --]

From ccf93cd0b21e9c0ff0a1d4ace59899fca25ac157 Mon Sep 17 00:00:00 2001
From: Jakub Jelinek <jakub@redhat.com>
Date: Sat, 21 Mar 2020 14:39:56 +0000
Subject: [PATCH] [gcn] Fix build with RTL checking [PR94248]

Building (for offloading) a '--target=amdgcn-amdhsa' GCC with
'--enable-checking=yes,extra,rtl' fails:

    during RTL pass: split2
    [...]/source-gcc/libgcc/libgcc2.c: In function '__absvdi2':
    [...]/source-gcc/libgcc/libgcc2.c:271:1: internal compiler error: RTL check: expected code 'reg', have 'const_int' in rhs_regno, at rtl.h:1923
      271 | }
	  | ^
    0x565847 ???
	    [...]/source-gcc/gcc/rtl.c:881
    0x59a8a4 ???
	    [...]/source-gcc/gcc/rtl.h:1923
    0x12e3a5c ???
	    [...]/source-gcc/gcc/config/gcn/gcn.md:631
    [...]
    Makefile:501: recipe for target '_absvdi2.o' failed
    make[4]: *** [_absvdi2.o] Error 1
    make[4]: Leaving directory '[...]/build-gcc-offload-amdgcn-amdhsa/amdgcn-amdhsa/gfx900/libgcc'

	gcc/
	PR target/94248
	* config/gcn/gcn.md (*mov<mode>_insn): Use
	'reg_overlap_mentioned_p' to check for overlap.

Tested-by: Thomas Schwinge <thomas@codesourcery.com>
---
 gcc/ChangeLog         | 4 ++++
 gcc/config/gcn/gcn.md | 2 +-
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 5541d9c62d9f..3321e8839707 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,5 +1,9 @@
 2020-04-29  Jakub Jelinek  <jakub@redhat.com>
 
+	PR target/94248
+	* config/gcn/gcn.md (*mov<mode>_insn): Use
+	'reg_overlap_mentioned_p' to check for overlap.
+
 	PR target/94706
 	* config/ia64/ia64.c (hfa_element_mode): Use DECL_FIELD_ABI_IGNORED
 	instead of cxx17_empty_base_field_p.
diff --git a/gcc/config/gcn/gcn.md b/gcc/config/gcn/gcn.md
index 8f5937781b2b..8cfb3a85d256 100644
--- a/gcc/config/gcn/gcn.md
+++ b/gcc/config/gcn/gcn.md
@@ -629,7 +629,7 @@
     rtx outhi = gen_highpart_mode (SImode, <MODE>mode, operands[0]);
 
     /* Ensure that overlapping registers aren't corrupted.  */
-    if (REGNO (outlo) == REGNO (inhi))
+    if (reg_overlap_mentioned_p (outlo, inhi))
       {
 	operands[0] = outhi;
 	operands[1] = inhi;
-- 
2.26.2