On 11/7/24 18:02, Andrew Stubbs wrote:
On 07/11/2024 17:57, Robin Dapp wrote:
From: Robin Dapp <rd...@ventanamicro.com>
This patch adds an undefined else operand to the masked loads.
gcc/ChangeLog:
* config/gcn/predicates.md (maskload_else_operand): New
predicate.
* config/gcn/gcn-valu.md: Use new predicate.
---
gcc/config/gcn/gcn-valu.md | 23 +++++++++++------------
gcc/config/gcn/predicates.md | 2 ++
2 files changed, 13 insertions(+), 12 deletions(-)
diff --git a/gcc/config/gcn/gcn-valu.md b/gcc/config/gcn/gcn-valu.md
index cb2f4a78035..ce7a68f0e2d 100644
--- a/gcc/config/gcn/gcn-valu.md
+++ b/gcc/config/gcn/gcn-valu.md
@@ -3989,7 +3989,8 @@ (define_expand "while_ultsidi"
(define_expand "maskload<mode>di"
[(match_operand:V_MOV 0 "register_operand")
(match_operand:V_MOV 1 "memory_operand")
- (match_operand 2 "")]
+ (match_operand 2 "")
+ (match_operand:V_MOV 3 "maskload_else_operand")]
""
{
rtx exec = force_reg (DImode, operands[2]);
@@ -3998,11 +3999,8 @@ (define_expand "maskload<mode>di"
rtx as = gen_rtx_CONST_INT (VOIDmode, MEM_ADDR_SPACE
(operands[1]));
rtx v = gen_rtx_CONST_INT (VOIDmode, MEM_VOLATILE_P (operands[1]));
- /* Masked lanes are required to hold zero. */
- emit_move_insn (operands[0], gcn_vec_constant (<MODE>mode, 0));
-
emit_insn (gen_gather<mode>_expr_exec (operands[0], addr, as, v,
- operands[0], exec));
+ gcn_gen_undef (<MODE>mode), exec));
DONE;
})
@@ -4027,7 +4025,8 @@ (define_expand "mask_gather_load<mode><vnsi>"
(match_operand:<VnSI> 2 "register_operand")
(match_operand 3 "immediate_operand")
(match_operand:SI 4 "gcn_alu_operand")
- (match_operand:DI 5 "")]
+ (match_operand:DI 5 "")
+ (match_operand:V_MOV 6 "maskload_else_operand")]
""
{
rtx exec = force_reg (DImode, operands[5]);
@@ -4036,18 +4035,18 @@ (define_expand "mask_gather_load<mode><vnsi>"
operands[2], operands[4],
INTVAL (operands[3]), exec);
- /* Masked lanes are required to hold zero. */
- emit_move_insn (operands[0], gcn_vec_constant (<MODE>mode, 0));
-
if (GET_MODE (addr) == <VnDI>mode)
emit_insn (gen_gather<mode>_insn_1offset_exec (operands[0], addr,
const0_rtx, const0_rtx,
- const0_rtx, operands[0],
- exec));
+ gcn_gen_undef
+ (<MODE>mode),
+ operands[0], exec));
else
emit_insn (gen_gather<mode>_insn_2offsets_exec (operands[0],
operands[1],
addr, const0_rtx,
- const0_rtx, const0_rtx,
+ const0_rtx,
+ gcn_gen_undef
+ (<MODE>mode),
operands[0], exec));
I missed that this hunk replaces the wrong parameter, resulting in a
build failure in Newlib (PR117657).
I've committed the attached patch to fix it.
Andrew
DONE;
})
diff --git a/gcc/config/gcn/predicates.md b/gcc/config/gcn/predicates.md
index 3f59396a649..21beeb586a4 100644
--- a/gcc/config/gcn/predicates.md
+++ b/gcc/config/gcn/predicates.md
@@ -228,3 +228,5 @@ (define_predicate "ascending_zero_int_parallel"
return gcn_stepped_zero_int_parallel_p (op, 1);
})
+(define_predicate "maskload_else_operand"
+ (match_operand 0 "scratch_operand"))
LGTM.
Andrew
amdgcn: Fix build failure (PR117657)
The last patch did the right thing to the wrong parameter, which caused a build
failure in Newlib. This patch fixes it.
gcc/ChangeLog:
PR target/117657
* config/gcn/gcn-valu.md (mask_gather_load<mode><vnsi>): Fix bug in
maskload else patch.
diff --git a/gcc/config/gcn/gcn-valu.md b/gcc/config/gcn/gcn-valu.md
index ce7a68f0e2d..f7ed0b825a1 100644
--- a/gcc/config/gcn/gcn-valu.md
+++ b/gcc/config/gcn/gcn-valu.md
@@ -4038,16 +4038,17 @@ (define_expand "mask_gather_load<mode><vnsi>"
if (GET_MODE (addr) == <VnDI>mode)
emit_insn (gen_gather<mode>_insn_1offset_exec (operands[0], addr,
const0_rtx, const0_rtx,
+ const0_rtx,
gcn_gen_undef
(<MODE>mode),
- operands[0], exec));
+ exec));
else
emit_insn (gen_gather<mode>_insn_2offsets_exec (operands[0], operands[1],
addr, const0_rtx,
- const0_rtx,
+ const0_rtx, const0_rtx,
gcn_gen_undef
(<MODE>mode),
- operands[0], exec));
+ exec));
DONE;
})