yxsamliu wrote:
[AMD Official Use Only - General]
Thanks. I will fix it by memcpy instead of dereferencing int* casted from char*.
Sam
From: Joseph Huber ***@***.***>
Sent: Thursday, October 5, 2023 9:23 AM
To: llvm/llvm-project ***@***.***>
Cc: Liu, Yaxun (Sam) ***@***.***>; State change
jhuber6 wrote:
For reference, I've had similar problems in the past. Usually it happens
because the header needs to be aligned on a `uint64_t` to be read. The ELF
section has 8-byte alignment, but that can be incorrect when extracting from
something like an archive. I just make a copy in that
yxsamliu wrote:
[AMD Official Use Only - General]
I have reverted the commit. Will fix and reland. Thanks.
Sam
From: mikaelholmen ***@***.***>
Sent: Thursday, October 5, 2023 4:05 AM
To: llvm/llvm-project ***@***.***>
Cc: Liu, Yaxun (Sam) ***@***.***>; State change ***@***.***>
Subject: Re:
mikaelholmen wrote:
Hi,
I noticed that with UBSan this new testcase fails
```
09:59:08 Failed Tests (1):
09:59:08 Clang :: Driver/clang-offload-bundler-zlib.c
```
with
```
09:39:53 ../../clang/lib/Driver/OffloadBundler.cpp:1012:25: runtime error: load
of misaligned address 0x55ceca16e46c for
https://github.com/yxsamliu closed
https://github.com/llvm/llvm-project/pull/67162
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
https://github.com/jhuber6 approved this pull request.
I think in the long term we need to move away from the bundler, which would
suggest we probably want something similar in the "new driver". But overall the
changes here are pretty self contained and the concept is straightforward.
https://github.com/yxsamliu updated
https://github.com/llvm/llvm-project/pull/67162
>From c2d5f809ab0f1be50cee4baaffdf3994807e5ef2 Mon Sep 17 00:00:00 2001
From: "Yaxun (Sam) Liu"
Date: Thu, 21 Sep 2023 13:52:16 -0400
Subject: [PATCH] [HIP] Support compressing device binary
Add option
@@ -1183,6 +1183,10 @@ def fgpu_inline_threshold_EQ : Joined<["-"],
"fgpu-inline-threshold=">,
def fgpu_sanitize : Flag<["-"], "fgpu-sanitize">, Group,
HelpText<"Enable sanitizer for supported offloading devices">;
def fno_gpu_sanitize : Flag<["-"], "fno-gpu-sanitize">,
@@ -1183,6 +1183,10 @@ def fgpu_inline_threshold_EQ : Joined<["-"],
"fgpu-inline-threshold=">,
def fgpu_sanitize : Flag<["-"], "fgpu-sanitize">, Group,
HelpText<"Enable sanitizer for supported offloading devices">;
def fno_gpu_sanitize : Flag<["-"], "fno-gpu-sanitize">,
@@ -881,6 +900,184 @@ CreateFileHandler(MemoryBuffer ,
"'" + FilesType + "': invalid file type specified");
}
+OffloadBundlerConfig::OffloadBundlerConfig() {
+ auto IgnoreEnvVarOpt =
+
https://github.com/yxsamliu updated
https://github.com/llvm/llvm-project/pull/67162
>From b4b44809dcbc487239b4430751a232563c5ac590 Mon Sep 17 00:00:00 2001
From: "Yaxun (Sam) Liu"
Date: Thu, 21 Sep 2023 13:52:16 -0400
Subject: [PATCH] [HIP] Support compressing device binary
Add option
@@ -881,6 +900,184 @@ CreateFileHandler(MemoryBuffer ,
"'" + FilesType + "': invalid file type specified");
}
+OffloadBundlerConfig::OffloadBundlerConfig() {
+ auto IgnoreEnvVarOpt =
+
yxsamliu wrote:
> > We need to support compression/uncompression of standalone bundled bitcode,
> > and standalone bundled code objects that are not embedded in ELF or COFF
> > binaries. An OS-neutral binary format is preferred.
>
> I think my ideal solution would be LLVM supporting some
https://github.com/yxsamliu updated
https://github.com/llvm/llvm-project/pull/67162
>From b1c9550da1074414228a45594478fec7ecbf1779 Mon Sep 17 00:00:00 2001
From: "Yaxun (Sam) Liu"
Date: Thu, 21 Sep 2023 13:52:16 -0400
Subject: [PATCH] [HIP] Support compressing device binary
Add option
jhuber6 wrote:
>
> We need to support compression/uncompression of standalone bundled bitcode,
> and standalone bundled code objects that are not embedded in ELF or COFF
> binaries. An OS-neutral binary format is preferred.
I think my ideal solution would be LLVM supporting some module
@@ -1183,6 +1183,10 @@ def fgpu_inline_threshold_EQ : Joined<["-"],
"fgpu-inline-threshold=">,
def fgpu_sanitize : Flag<["-"], "fgpu-sanitize">, Group,
HelpText<"Enable sanitizer for supported offloading devices">;
def fno_gpu_sanitize : Flag<["-"], "fno-gpu-sanitize">,
https://github.com/yxsamliu updated
https://github.com/llvm/llvm-project/pull/67162
>From 71dc0a3afcc873d1afeb3ae265bc006860da1d82 Mon Sep 17 00:00:00 2001
From: "Yaxun (Sam) Liu"
Date: Thu, 21 Sep 2023 13:52:16 -0400
Subject: [PATCH] [HIP] Support compressing device binary
Add option
yxsamliu wrote:
> Unrelated, I need to hurry up and try to make the new driver the default for
> CUDA and HIP upstream at some point.
>
> I've had some thoughts about how to accomplish this in the past. For the new
> driver, `.llvm.offloading` is a single section that contains all of the
>
@@ -1183,6 +1183,10 @@ def fgpu_inline_threshold_EQ : Joined<["-"],
"fgpu-inline-threshold=">,
def fgpu_sanitize : Flag<["-"], "fgpu-sanitize">, Group,
HelpText<"Enable sanitizer for supported offloading devices">;
def fno_gpu_sanitize : Flag<["-"], "fno-gpu-sanitize">,
jhuber6 wrote:
Unrelated, I need to hurry up and try to make the new driver the default for
CUDA and HIP upstream at some point.
I've had some thoughts about how to accomplish this in the past. For the new
driver, `.llvm.offloading` is a single section that contains all of the
embedded
yxsamliu wrote:
ping
A little background: some HIP applications face size restrictions for their
packages. Compressing can significantly reduce the bundle size with acceptable
overhead of uncompressing. Therefore they need this feature.
https://github.com/llvm/llvm-project/pull/67162
https://github.com/yxsamliu updated
https://github.com/llvm/llvm-project/pull/67162
>From aa1aee8695a062c472dbb9efd1e87a2ba7cdc58b Mon Sep 17 00:00:00 2001
From: "Yaxun (Sam) Liu"
Date: Thu, 21 Sep 2023 13:52:16 -0400
Subject: [PATCH] [HIP] Support compressing device binary
Add option
llvmbot wrote:
@llvm/pr-subscribers-clang-driver
Changes
Add option -f[no-]offload-compress to clang to enable/disable compression of
device binary for HIP. By default it is disabled.
Add option -compress to clang-offload-bundler to enable compression of offload
bundle. By default it is
https://github.com/yxsamliu created
https://github.com/llvm/llvm-project/pull/67162
Add option -f[no-]offload-compress to clang to enable/disable compression of
device binary for HIP. By default it is disabled.
Add option -compress to clang-offload-bundler to enable compression of offload
24 matches
Mail list logo
