On gfx11, with a sequence such as
s_trap 2
s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
s_endpgm
the s_sendmsg does deallocate registers while the wave is supposed to be
stopped. As a result, the wave cannot do the expected context save
operations, and cannot context save.
To avoid this problem, park the wave in the trap handler for gfx11.
Note that gfx11 has implemented an instruction cache prefetch. When
parked, the prefetch tries to access memory past the end of trap handler
which causes memory violation exceptions to be reported. To avoid this,
we need to add padding at the end of the trap handler. The padding
consists of `s_code_end` instructions Given that the trap handler is
loaded at a 0x1000 aligned address the maximum prefetch amount (in
bytes) is given by `256 - (trap_handler_size % 64)`.
Change-Id: I5446da54a965a64f21cb0fd3ce3caa4b6137a933
gfx940 uses ttmp11 to hold the queue packet index so the first level
trap handler uses ttmp13 instead to save ib_sts.
Repurpose ttmp11[31] to mean that the ttmps are initialized. The issue
was that the debugger could not tell whether ttmp6 was written by the
trap handler when determining the stop reason.
If ttmp11[31]=0, then the trap handler has not been executed and ttmp6
should be assumed to be 0. If ttmp11[31]=1, then ttmp6 holds the
trap_id, if an s_trap instruction caused the exception.
Signed-off-by: Laurent Morichetti <laurent.morichetti@amd.com>
Signed-off-by: Lancelot Six <lancelot.six@amd.com>
Change-Id: I9af903abae044b9ec530306229caf3b883f3ee46
Includes some workarounds and HMM.
Conflicts:
opensrc/hsa-runtime/core/runtime/amd_topology.cpp
opensrc/hsa-runtime/core/util/flag.h
Change-Id: I22976f07964a43dbb228a6231777dbd599112b8d
Park the wave, if it is stopped, to avoid halting it at an s_endpgm
instruction if the architecture does not support it.
Free ttmp6 by converting the dispatch_ptr into a queue packet index
(25-bit) and storing it in ttmp7[24:0].
Save the exception PC in ttmp11[22:7] ttmp6[31:0].
Change-Id: Iaa3c5baf5b488c0b534044d338f12bffa63ddce2
Replace the stop reasons ttmp11.trap_raised and ttmp11.excp_raised
with ttmp11.wave_stopped which indicates that the trap handler has
halted the wave as the result of an event (trap, single-step or
exception).
If the wave is stopped because of a trap, also record the trap_id in
ttmp11.saved_trap_id[7:0].
Save status.halt in ttmp11.saved_status_halt, so that it can be
restored when resuming a wave (changing a wave's state from stopped to
running or single-stepping).
Change-Id: I7322f59b60e8cc1b92bf5f067dba606a3109ef49
This patch is to let ROCr recognize new gfx10.3.3 ISA.
Change-Id: Ied23eee2752e14c19c8c0a6d7789fded9940e31e
Signed-off-by: Huang Rui <ray.huang@amd.com>
To support single stepping the instruction preceding an s_endpgm,
unwind the PC by 8 bytes and set ttmp11[9] to notify the debugger
that the wave is halted with a modified PC.
Bump the debug r_version for this new trap handler ABI.
Change-Id: I55e4e0d65576f92da14a336266c31c513baab547
- Remove gfx800, gfx804 and gfx901 as they do not exist.
- Map the V2 note record of "AMD:AMDGPU:8:0:0" to gfx802 as they are
the same target just connected to a differnt motherboard.
- Correct typo for supporting gfx902:xnack+.
- Support agent names with a minor or stepping version greater than 9.
Change-Id: Ife933449f60ab4687e2aaab9baf4c9fc5b86339d
Add missing target names and make all parts consistent with which
targets are supported.
- Add gfx805 as a supported target.
- Add all ELF targets to genric code.
- Make offline loader match supported targets.
Change-Id: Idab4d69edc71645aecaa83aa55e29c1aeee4c1d6
Code object V2 had the ability to support the following queries:
- HSA_CODE_SYMBOL_INFO_KERNEL_KERNARG_SEGMENT_SIZE
- HSA_EXECUTABLE_SYMBOL_INFO_KERNEL_KERNARG_SEGMENT_SIZE
- HSA_CODE_SYMBOL_INFO_KERNEL_KERNARG_SEGMENT_ALIGNMENT
- HSA_EXECUTABLE_SYMBOL_INFO_KERNEL_KERNARG_SEGMENT_ALIGNMENT
However code object V3 onwards cannot support these as the kernel
descriptor changed. These queries need to be deprecated.
Until then return more reasonable values:
- For kernarg alignment return 16 which is the minimum alignment
required by the HSA standard.
- For kernarg size return the field from the kernel descriptor which
is a hint. If it is 0 then the compiler is not specifying the kernarg
size, or the kernel has no kernarg.
Change-Id: I19ce6cd0f3658a2bf62277492f39100ea5ab4256
Adds the following:
- New factory method to create a code object reader from
file with offset and size.
- A pair of queries on a loaded code object to get the URI name/length.
- A bump to the AMD vendor loader extension API and its associated table.
Change-Id: I17c83e9c2447d29a43c438459395365f786a3611
New trap handler ABI: Record in ttmp11[8:7] the event that caused the
trap handler to be entered. We currently record 2 events, trap_raised
if an s_trap instruction was executed, or excp_raised if an exception
(MEM_VIOL or ILLEGAL_INST) was raised.
Change-Id: Ie278c8277437b3b67c2737dcd1a12fe6511df428
Iterate the loaded shared objects to see if the given elf image binary
is part of a loaded segment.
Change-Id: I074cacd99eb5b59f883f4ce2bd901e0e35a660b8
Attribute optimize(0) doesn't appear to be helpful helpful. This
prevents optimization in the function but not at call sites to the
function. The function may still be inlined since it has no side
effect (in some cases that we currently don't support).
Having a side effect prevents a call site optimization that allows
removal of a noinline function call with no side effect. Call site
optimization should only happen (in GCC at least) when using whole
program optimization so this may be stronger than we strictly need.
Also added _amdgpu_r_debug to the exported symbol list (global) and
switched to the standard macro for an exported symbol (HSA_API).
Without being in the global list the debugger will not find this
symbol if the binary has been stripped.
Change-Id: Ieb00175ccc55fda4491deee44711cd55b3f24aeb
Allow user specified profiles if the HSAIL note is not found.
Konstantin reviewed and approved. HSAIL note is not generated by LLVM.
Change-Id: I40fbfbaedd6787b6a716507918f698d02007afe1
At the moment it is not possible to build ROCr with Clang. This is
a spurious limitation. The present PR addresses it by guarding GCC
only flags and by fixing some additional warnings that Clang triggers;
one of said warnings did outline a rather interesting issue with math
being done on void*s. - AlexVlx
Void ptr arithmetic had already been fixed in amd-master branch.
Change-Id: I5ee97e20b5c40b10dd73facecabe75f02ba46462
- Add support for R_AMDGPU_RELATIVE64 relocation record.
- Return status error if any unsupported relocation record encountered.
Change-Id: Icbb5dcb81109a70c1f2195412a0df58a11be9da1
Lancelot SIX
Laurent Morichetti
Konstantin Zhuravlyov
Jay Cornwall
Tony Tye
Sean Keely
Huang Rui
Chris Freehill
Ramesh Errabolu
Saleel Kudchadker
Qingchuan Shi
Scott Linder
Wilkin