[gcc(refs/vendors/ibm/heads/gcc-17-future)] PR target/117251: Improve vector fusion #6
Michael Meissner
meissner@gcc.gnu.org
Thu Jun 25 09:10:15 GMT 2026
https://gcc.gnu.org/g:609a2dc198b9f0b6557818b3eee5c959d03a78a4
commit 609a2dc198b9f0b6557818b3eee5c959d03a78a4
Author: Michael Meissner <meissner@linux.ibm.com>
Date: Thu Jun 25 05:07:47 2026 -0400
PR target/117251: Improve vector fusion #6
See the following post for a complete explanation of what the patches
for PR target/117251:
* https://gcc.gnu.org/pipermail/gcc-patches/2025-June/686474.html
This is patch #2 of 45 to generate the 'XXEVAL' instruction on power10
and power11 instead of using the Altivec 'VANDC' instruction feeding
into 'VAND'. The 'XXEVAL' instruction can use all 64 vector registers,
instead of the 32 registers that traditional Altivec vector
instructions use. By allowing all of the vector registers to be used,
it reduces the amount of spilling that a large benchmark generated.
For vectors such as:
vector int a, b, c, d;
This patch will generate a call to xxeval if any of the registers are allocated
to traditional floating point registers:
Code: Old: New:
===== ==== ====
a = ~ ((c & ~ d) | b); vandc t,c,d xxeval a,b,c,d,208
vnor a,t,b
a = ~ ((c & ~ d) ^ b); vandc t,c,d xxeval a,b,c,d,210
veqv a,t,b
a = ~ ((c & d) | b); vand t,c,d xxeval a,b,c,d,224
vnor a,t,b
a = (~ (c & d)) ^ b; vnand t,c,d xxeval a,b,c,d,225
vxor a,t,b
a = (~ (c & d)) | b; vnand t,c,d xxeval a,b,c,d,239
vor a,t,b
a = ~ ((~ (c & d)) & b); vnand t,c,d xxeval a,b,c,d,241
vnand a,t,b
a = ~ ((c | ~ d) & b); vorc t,c,d xxeval a,b,c,d,244
vnand a,t,b
a = ~ ((~ (c ^ d)) & b); veqv t,c,d xxeval a,b,c,d,246
vnand a,t,b
Since fusion using 2 Altivec instructions is slightly faster than using
the 'XXEVAL' instruction we prefer to generate the Altivec instructions
if we can. In addition, because 'XXEVAL' is a prefixed instruction, it
possibly might generate an extra NOP instruction to align the 'XXEVAL'
instruction.
I have tested these patches on both big endian and little endian
PowerPC servers, with no regressions. Can I check these patchs into
the trunk?
2026-06-25 Michael Meissner <meissner@linux.ibm.com>
gcc/
PR target/117251
* config/rs6000/fusion.md: Regenerate.
* config/rs6000/genfusion.pl (gen_logical_addsubf): Add support
to generate vector/vector fusion if XXEVAL is supported.
Diff:
---
gcc/config/rs6000/fusion.md | 120 ++++++++++++++++++++++++-----------------
gcc/config/rs6000/genfusion.pl | 8 +++
2 files changed, 80 insertions(+), 48 deletions(-)
diff --git a/gcc/config/rs6000/fusion.md b/gcc/config/rs6000/fusion.md
index c700a623f811..2d355c435b29 100644
--- a/gcc/config/rs6000/fusion.md
+++ b/gcc/config/rs6000/fusion.md
@@ -2204,20 +2204,23 @@
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector vandc -> veqv
(define_insn "*fuse_vandc_veqv"
- [(set (match_operand:VM 3 "altivec_register_operand" "=&0,&1,&v,v")
- (not:VM (xor:VM (and:VM (not:VM (match_operand:VM 0 "altivec_register_operand" "v,v,v,v"))
- (match_operand:VM 1 "altivec_register_operand" "v,v,v,v"))
- (match_operand:VM 2 "altivec_register_operand" "v,v,v,v"))))
- (clobber (match_scratch:VM 4 "=X,X,X,&v"))]
+ [(set (match_operand:VM 3 "vector_fusion_operand" "=&0,&1,&v,wa,v")
+ (not:VM (xor:VM (and:VM (not:VM (match_operand:VM 0 "vector_fusion_operand" "v,v,v,wa,v"))
+ (match_operand:VM 1 "vector_fusion_operand" "v,v,v,wa,v"))
+ (match_operand:VM 2 "vector_fusion_operand" "v,v,v,wa,v"))))
+ (clobber (match_scratch:VM 4 "=X,X,X,X,&v"))]
"(TARGET_P10_FUSION)"
"@
vandc %3,%1,%0\;veqv %3,%3,%2
vandc %3,%1,%0\;veqv %3,%3,%2
vandc %3,%1,%0\;veqv %3,%3,%2
+ xxeval %x3,%x2,%x1,%x0,210
vandc %4,%1,%0\;veqv %3,%4,%2"
[(set_attr "type" "fused_vector")
(set_attr "cost" "6")
- (set_attr "length" "8")])
+ (set_attr "length" "8")
+ (set_attr "prefixed" "*,*,*,yes,*")
+ (set_attr "isa" "*,*,*,xxeval,*")])
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector veqv -> veqv
@@ -2369,38 +2372,44 @@
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector veqv -> vnand
(define_insn "*fuse_veqv_vnand"
- [(set (match_operand:VM 3 "altivec_register_operand" "=&0,&1,&v,v")
- (ior:VM (not:VM (not:VM (xor:VM (match_operand:VM 0 "altivec_register_operand" "v,v,v,v")
- (match_operand:VM 1 "altivec_register_operand" "v,v,v,v"))))
- (not:VM (match_operand:VM 2 "altivec_register_operand" "v,v,v,v"))))
- (clobber (match_scratch:VM 4 "=X,X,X,&v"))]
+ [(set (match_operand:VM 3 "vector_fusion_operand" "=&0,&1,&v,wa,v")
+ (ior:VM (not:VM (not:VM (xor:VM (match_operand:VM 0 "vector_fusion_operand" "v,v,v,wa,v")
+ (match_operand:VM 1 "vector_fusion_operand" "v,v,v,wa,v"))))
+ (not:VM (match_operand:VM 2 "vector_fusion_operand" "v,v,v,wa,v"))))
+ (clobber (match_scratch:VM 4 "=X,X,X,X,&v"))]
"(TARGET_P10_FUSION)"
"@
veqv %3,%1,%0\;vnand %3,%3,%2
veqv %3,%1,%0\;vnand %3,%3,%2
veqv %3,%1,%0\;vnand %3,%3,%2
+ xxeval %x3,%x2,%x1,%x0,246
veqv %4,%1,%0\;vnand %3,%4,%2"
[(set_attr "type" "fused_vector")
(set_attr "cost" "6")
- (set_attr "length" "8")])
+ (set_attr "length" "8")
+ (set_attr "prefixed" "*,*,*,yes,*")
+ (set_attr "isa" "*,*,*,xxeval,*")])
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector vnand -> vnand
(define_insn "*fuse_vnand_vnand"
- [(set (match_operand:VM 3 "altivec_register_operand" "=&0,&1,&v,v")
- (ior:VM (not:VM (ior:VM (not:VM (match_operand:VM 0 "altivec_register_operand" "v,v,v,v"))
- (not:VM (match_operand:VM 1 "altivec_register_operand" "v,v,v,v"))))
- (not:VM (match_operand:VM 2 "altivec_register_operand" "v,v,v,v"))))
- (clobber (match_scratch:VM 4 "=X,X,X,&v"))]
+ [(set (match_operand:VM 3 "vector_fusion_operand" "=&0,&1,&v,wa,v")
+ (ior:VM (not:VM (ior:VM (not:VM (match_operand:VM 0 "vector_fusion_operand" "v,v,v,wa,v"))
+ (not:VM (match_operand:VM 1 "vector_fusion_operand" "v,v,v,wa,v"))))
+ (not:VM (match_operand:VM 2 "vector_fusion_operand" "v,v,v,wa,v"))))
+ (clobber (match_scratch:VM 4 "=X,X,X,X,&v"))]
"(TARGET_P10_FUSION)"
"@
vnand %3,%1,%0\;vnand %3,%3,%2
vnand %3,%1,%0\;vnand %3,%3,%2
vnand %3,%1,%0\;vnand %3,%3,%2
+ xxeval %x3,%x2,%x1,%x0,241
vnand %4,%1,%0\;vnand %3,%4,%2"
[(set_attr "type" "fused_vector")
(set_attr "cost" "6")
- (set_attr "length" "8")])
+ (set_attr "length" "8")
+ (set_attr "prefixed" "*,*,*,yes,*")
+ (set_attr "isa" "*,*,*,xxeval,*")])
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector vnor -> vnand
@@ -2441,20 +2450,23 @@
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector vorc -> vnand
(define_insn "*fuse_vorc_vnand"
- [(set (match_operand:VM 3 "altivec_register_operand" "=&0,&1,&v,v")
- (ior:VM (not:VM (ior:VM (not:VM (match_operand:VM 0 "altivec_register_operand" "v,v,v,v"))
- (match_operand:VM 1 "altivec_register_operand" "v,v,v,v")))
- (not:VM (match_operand:VM 2 "altivec_register_operand" "v,v,v,v"))))
- (clobber (match_scratch:VM 4 "=X,X,X,&v"))]
+ [(set (match_operand:VM 3 "vector_fusion_operand" "=&0,&1,&v,wa,v")
+ (ior:VM (not:VM (ior:VM (not:VM (match_operand:VM 0 "vector_fusion_operand" "v,v,v,wa,v"))
+ (match_operand:VM 1 "vector_fusion_operand" "v,v,v,wa,v")))
+ (not:VM (match_operand:VM 2 "vector_fusion_operand" "v,v,v,wa,v"))))
+ (clobber (match_scratch:VM 4 "=X,X,X,X,&v"))]
"(TARGET_P10_FUSION)"
"@
vorc %3,%1,%0\;vnand %3,%3,%2
vorc %3,%1,%0\;vnand %3,%3,%2
vorc %3,%1,%0\;vnand %3,%3,%2
+ xxeval %x3,%x2,%x1,%x0,244
vorc %4,%1,%0\;vnand %3,%4,%2"
[(set_attr "type" "fused_vector")
(set_attr "cost" "6")
- (set_attr "length" "8")])
+ (set_attr "length" "8")
+ (set_attr "prefixed" "*,*,*,yes,*")
+ (set_attr "isa" "*,*,*,xxeval,*")])
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector vxor -> vnand
@@ -2477,38 +2489,44 @@
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector vand -> vnor
(define_insn "*fuse_vand_vnor"
- [(set (match_operand:VM 3 "altivec_register_operand" "=&0,&1,&v,v")
- (and:VM (not:VM (and:VM (match_operand:VM 0 "altivec_register_operand" "v,v,v,v")
- (match_operand:VM 1 "altivec_register_operand" "v,v,v,v")))
- (not:VM (match_operand:VM 2 "altivec_register_operand" "v,v,v,v"))))
- (clobber (match_scratch:VM 4 "=X,X,X,&v"))]
+ [(set (match_operand:VM 3 "vector_fusion_operand" "=&0,&1,&v,wa,v")
+ (and:VM (not:VM (and:VM (match_operand:VM 0 "vector_fusion_operand" "v,v,v,wa,v")
+ (match_operand:VM 1 "vector_fusion_operand" "v,v,v,wa,v")))
+ (not:VM (match_operand:VM 2 "vector_fusion_operand" "v,v,v,wa,v"))))
+ (clobber (match_scratch:VM 4 "=X,X,X,X,&v"))]
"(TARGET_P10_FUSION)"
"@
vand %3,%1,%0\;vnor %3,%3,%2
vand %3,%1,%0\;vnor %3,%3,%2
vand %3,%1,%0\;vnor %3,%3,%2
+ xxeval %x3,%x2,%x1,%x0,224
vand %4,%1,%0\;vnor %3,%4,%2"
[(set_attr "type" "fused_vector")
(set_attr "cost" "6")
- (set_attr "length" "8")])
+ (set_attr "length" "8")
+ (set_attr "prefixed" "*,*,*,yes,*")
+ (set_attr "isa" "*,*,*,xxeval,*")])
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector vandc -> vnor
(define_insn "*fuse_vandc_vnor"
- [(set (match_operand:VM 3 "altivec_register_operand" "=&0,&1,&v,v")
- (and:VM (not:VM (and:VM (not:VM (match_operand:VM 0 "altivec_register_operand" "v,v,v,v"))
- (match_operand:VM 1 "altivec_register_operand" "v,v,v,v")))
- (not:VM (match_operand:VM 2 "altivec_register_operand" "v,v,v,v"))))
- (clobber (match_scratch:VM 4 "=X,X,X,&v"))]
+ [(set (match_operand:VM 3 "vector_fusion_operand" "=&0,&1,&v,wa,v")
+ (and:VM (not:VM (and:VM (not:VM (match_operand:VM 0 "vector_fusion_operand" "v,v,v,wa,v"))
+ (match_operand:VM 1 "vector_fusion_operand" "v,v,v,wa,v")))
+ (not:VM (match_operand:VM 2 "vector_fusion_operand" "v,v,v,wa,v"))))
+ (clobber (match_scratch:VM 4 "=X,X,X,X,&v"))]
"(TARGET_P10_FUSION)"
"@
vandc %3,%1,%0\;vnor %3,%3,%2
vandc %3,%1,%0\;vnor %3,%3,%2
vandc %3,%1,%0\;vnor %3,%3,%2
+ xxeval %x3,%x2,%x1,%x0,208
vandc %4,%1,%0\;vnor %3,%4,%2"
[(set_attr "type" "fused_vector")
(set_attr "cost" "6")
- (set_attr "length" "8")])
+ (set_attr "length" "8")
+ (set_attr "prefixed" "*,*,*,yes,*")
+ (set_attr "isa" "*,*,*,xxeval,*")])
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector veqv -> vnor
@@ -2702,20 +2720,23 @@
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector vnand -> vor
(define_insn "*fuse_vnand_vor"
- [(set (match_operand:VM 3 "altivec_register_operand" "=&0,&1,&v,v")
- (ior:VM (ior:VM (not:VM (match_operand:VM 0 "altivec_register_operand" "v,v,v,v"))
- (not:VM (match_operand:VM 1 "altivec_register_operand" "v,v,v,v")))
- (match_operand:VM 2 "altivec_register_operand" "v,v,v,v")))
- (clobber (match_scratch:VM 4 "=X,X,X,&v"))]
+ [(set (match_operand:VM 3 "vector_fusion_operand" "=&0,&1,&v,wa,v")
+ (ior:VM (ior:VM (not:VM (match_operand:VM 0 "vector_fusion_operand" "v,v,v,wa,v"))
+ (not:VM (match_operand:VM 1 "vector_fusion_operand" "v,v,v,wa,v")))
+ (match_operand:VM 2 "vector_fusion_operand" "v,v,v,wa,v")))
+ (clobber (match_scratch:VM 4 "=X,X,X,X,&v"))]
"(TARGET_P10_FUSION)"
"@
vnand %3,%1,%0\;vor %3,%3,%2
vnand %3,%1,%0\;vor %3,%3,%2
vnand %3,%1,%0\;vor %3,%3,%2
+ xxeval %x3,%x2,%x1,%x0,239
vnand %4,%1,%0\;vor %3,%4,%2"
[(set_attr "type" "fused_vector")
(set_attr "cost" "6")
- (set_attr "length" "8")])
+ (set_attr "length" "8")
+ (set_attr "prefixed" "*,*,*,yes,*")
+ (set_attr "isa" "*,*,*,xxeval,*")])
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector vnor -> vor
@@ -3014,20 +3035,23 @@
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector vnand -> vxor
(define_insn "*fuse_vnand_vxor"
- [(set (match_operand:VM 3 "altivec_register_operand" "=&0,&1,&v,v")
- (xor:VM (ior:VM (not:VM (match_operand:VM 0 "altivec_register_operand" "v,v,v,v"))
- (not:VM (match_operand:VM 1 "altivec_register_operand" "v,v,v,v")))
- (match_operand:VM 2 "altivec_register_operand" "v,v,v,v")))
- (clobber (match_scratch:VM 4 "=X,X,X,&v"))]
+ [(set (match_operand:VM 3 "vector_fusion_operand" "=&0,&1,&v,wa,v")
+ (xor:VM (ior:VM (not:VM (match_operand:VM 0 "vector_fusion_operand" "v,v,v,wa,v"))
+ (not:VM (match_operand:VM 1 "vector_fusion_operand" "v,v,v,wa,v")))
+ (match_operand:VM 2 "vector_fusion_operand" "v,v,v,wa,v")))
+ (clobber (match_scratch:VM 4 "=X,X,X,X,&v"))]
"(TARGET_P10_FUSION)"
"@
vnand %3,%1,%0\;vxor %3,%3,%2
vnand %3,%1,%0\;vxor %3,%3,%2
vnand %3,%1,%0\;vxor %3,%3,%2
+ xxeval %x3,%x2,%x1,%x0,225
vnand %4,%1,%0\;vxor %3,%4,%2"
[(set_attr "type" "fused_vector")
(set_attr "cost" "6")
- (set_attr "length" "8")])
+ (set_attr "length" "8")
+ (set_attr "prefixed" "*,*,*,yes,*")
+ (set_attr "isa" "*,*,*,xxeval,*")])
;; logical-logical fusion pattern generated by gen_logical_addsubf
;; vector vnor -> vxor
diff --git a/gcc/config/rs6000/genfusion.pl b/gcc/config/rs6000/genfusion.pl
index 668bcb4eef29..9ba1ed618dee 100755
--- a/gcc/config/rs6000/genfusion.pl
+++ b/gcc/config/rs6000/genfusion.pl
@@ -246,6 +246,14 @@ sub gen_logical_addsubf
"veqv_vor" => 159,
"vorc_vxor" => 180,
"vorc_vor" => 191,
+ "vandc_vnor" => 208,
+ "vandc_veqv" => 210,
+ "vand_vnor" => 224,
+ "vnand_vxor" => 225,
+ "vnand_vor" => 239,
+ "vnand_vnand" => 241,
+ "vorc_vnand" => 244,
+ "veqv_vnand" => 246,
);
KIND: foreach $kind ('scalar','vector') {
More information about the Gcc-cvs
mailing list