Now with patch ...
Tobias Burnus wrote:
Hi all, hi Richard & Andrew,
Am 24.01.24 um 17:01 schrieb Tobias Burnus:
This patch obviously depends on Andrew's; he wrote in the previous
email of this thread regarding his patch:
Andrew Stubbs wrote:
This is enough to get gfx1100 working for most purposes, on top of the
patch that Tobias committed a week or so ago; there are still some test
failures to investigate, and probably some tuning to do.
It might also get gfx1030 working too. @Richi, could you test it,
please?
If gfx1030 doesn't work, I would propose my patch previously in the
thread,
https://gcc.gnu.org/pipermail/gcc-patches/2024-January/643835.html
This patch assumes that both gfx1100 and gfx1030 are now working:
Okay to enable gfx1100 multilib building and to document gfx1100 in
the manual?
and, with this patch, additionally gfx1030?
OK for mainline, once Andrew's patch is in?
Tobias
amdgcn: config.gcc - enable gfx1030 and gfx1100 multilib; add them to the docs
gcc/ChangeLog:
* config.gcc (amdgcn-*-*): Add gfx1030 and gfx1100 to
TM_MULTILIB_CONFIG.
* doc/install.texi (Configuration amdgcn-*-*): Mention gfx1030/gfx1100.
* doc/invoke.texi (AMD GCN Options): Add gfx1030 and gfx1100 to
-march/-mtune.
libgomp/ChangeLog:
* testsuite/libgomp.c/declare-variant-4.h: Add variant functions
for gfx1030 and gfx1100.
* testsuite/libgomp.c/declare-variant-4-gfx1030.c: New test.
* testsuite/libgomp.c/declare-variant-4-gfx1100.c: New test.
Signed-off-by: Tobias Burnus <tbur...@baylibre.com>
gcc/config.gcc | 2 +-
gcc/doc/install.texi | 12 ++++++------
gcc/doc/invoke.texi | 6 ++++++
libgomp/testsuite/libgomp.c/declare-variant-4-gfx1030.c | 8 ++++++++
libgomp/testsuite/libgomp.c/declare-variant-4-gfx1100.c | 8 ++++++++
libgomp/testsuite/libgomp.c/declare-variant-4.h | 16 ++++++++++++++++
6 files changed, 45 insertions(+), 7 deletions(-)
diff --git a/gcc/config.gcc b/gcc/config.gcc
index b2d7d7dd475..a0f9c672308 100644
--- a/gcc/config.gcc
+++ b/gcc/config.gcc
@@ -4564,7 +4564,7 @@ case "${target}" in
TM_MULTILIB_CONFIG=
;;
xdefault | xyes)
- TM_MULTILIB_CONFIG=`echo "gfx900,gfx906,gfx908,gfx90a" | sed "s/${with_arch},\?//;s/,$//"`
+ TM_MULTILIB_CONFIG=`echo "gfx900,gfx906,gfx908,gfx90a,gfx1030,gfx1100" | sed "s/${with_arch},\?//;s/,$//"`
;;
*)
TM_MULTILIB_CONFIG="${with_multilib_list}"
diff --git a/gcc/doc/install.texi b/gcc/doc/install.texi
index 71593919389..5747b5a12fe 100644
--- a/gcc/doc/install.texi
+++ b/gcc/doc/install.texi
@@ -1258,12 +1258,12 @@ default set of libraries is selected based on the value of
@item amdgcn*-*-*
@var{list} is a comma separated list of ISA names (allowed values: @code{fiji},
-@code{gfx900}, @code{gfx906}, @code{gfx908}, @code{gfx90a}). It ought not
-include the name of the default ISA, specified via @option{--with-arch}. If
-@var{list} is empty, then there will be no multilibs and only the default
-run-time library will be built. If @var{list} is @code{default} or
-@option{--with-multilib-list=} is not specified, then the default set of
-libraries is selected.
+@code{gfx900}, @code{gfx906}, @code{gfx908}, @code{gfx90a}, @code{gfx1030},
+@code{gfx1100}). It ought not include the name of the default ISA, specified
+via @option{--with-arch}. If @var{list} is empty, then there will be no
+multilibs and only the default run-time library will be built. If @var{list}
+is @code{default} or @option{--with-multilib-list=} is not specified, then
+the default set of libraries is selected.
@item arm*-*-*
@var{list} is a comma separated list of @code{aprofile} and
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index 6ec56493e59..64c5ed2ffde 100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -21739,6 +21739,12 @@ Compile for CDNA1 Instinct MI100 series devices (gfx908).
@item gfx90a
Compile for CDNA2 Instinct MI200 series devices (gfx90a).
+@item gfx1030
+Compile for RDNA2 gfx1030 devices (GFX10 series).
+
+@item gfx1100
+Compile for RDNA3 gfx1100 devices (GFX11 series).
+
@end table
@opindex msram-ecc
diff --git a/libgomp/testsuite/libgomp.c/declare-variant-4-gfx1030.c b/libgomp/testsuite/libgomp.c/declare-variant-4-gfx1030.c
new file mode 100644
index 00000000000..d98d5ef54ec
--- /dev/null
+++ b/libgomp/testsuite/libgomp.c/declare-variant-4-gfx1030.c
@@ -0,0 +1,8 @@
+/* { dg-do link { target { offload_target_amdgcn } } } */
+/* { dg-additional-options -foffload=amdgcn-amdhsa } */
+/* { dg-additional-options -foffload=-march=gfx1030 } */
+/* { dg-additional-options "-foffload=-fdump-tree-optimized" } */
+
+#include "declare-variant-4.h"
+
+/* { dg-final { only_for_offload_target amdgcn-amdhsa scan-offload-tree-dump "= gfx1030 \\(\\);" "optimized" } } */
diff --git a/libgomp/testsuite/libgomp.c/declare-variant-4-gfx1100.c b/libgomp/testsuite/libgomp.c/declare-variant-4-gfx1100.c
new file mode 100644
index 00000000000..6ade35224cc
--- /dev/null
+++ b/libgomp/testsuite/libgomp.c/declare-variant-4-gfx1100.c
@@ -0,0 +1,8 @@
+/* { dg-do link { target { offload_target_amdgcn } } } */
+/* { dg-additional-options -foffload=amdgcn-amdhsa } */
+/* { dg-additional-options -foffload=-march=gfx1100 } */
+/* { dg-additional-options "-foffload=-fdump-tree-optimized" } */
+
+#include "declare-variant-4.h"
+
+/* { dg-final { only_for_offload_target amdgcn-amdhsa scan-offload-tree-dump "= gfx1100 \\(\\);" "optimized" } } */
diff --git a/libgomp/testsuite/libgomp.c/declare-variant-4.h b/libgomp/testsuite/libgomp.c/declare-variant-4.h
index a70352430c2..393a5e295cc 100644
--- a/libgomp/testsuite/libgomp.c/declare-variant-4.h
+++ b/libgomp/testsuite/libgomp.c/declare-variant-4.h
@@ -35,6 +35,20 @@ gfx90a (void)
return 0x90a;
}
+__attribute__ ((noipa))
+int
+gfx1030 (void)
+{
+ return 0x1030;
+}
+
+__attribute__ ((noipa))
+int
+gfx1100 (void)
+{
+ return 0x1100;
+}
+
#ifdef USE_FIJI_FOR_GFX803
#pragma omp declare variant(gfx803) match(device = {isa("fiji")})
#else
@@ -44,6 +58,8 @@ gfx90a (void)
#pragma omp declare variant(gfx906) match(device = {isa("gfx906")})
#pragma omp declare variant(gfx908) match(device = {isa("gfx908")})
#pragma omp declare variant(gfx90a) match(device = {isa("gfx90a")})
+#pragma omp declare variant(gfx90a) match(device = {isa("gfx1030")})
+#pragma omp declare variant(gfx90a) match(device = {isa("gfx1100")})
__attribute__ ((noipa))
int
f (void)
