9475f6af81
Previous commit message: >Previous commit message: > >> Original commit message: >> >>>When users explicitly specify a PTX version via -mattr=+ptxNN that's insufficient for their target SM, we now emit a fatal error. Previously, we silently upgraded the PTX version to the minimum required for the target SM. >>> >>>When no SM or PTX version is specified, we now use PTX 3.2 (the minimum for the default SM 3.0) instead of PTX 6.0. >> >>The following commits should fix the failures that arose when I previously tried to land this commit: >> >> >>9fc5fd0ad6should address the llvm-nvptx*-nvidia-* build failures: https://github.com/llvm/llvm-project/pull/174834#issuecomment-3742242651 >> >> >>600514a637should address the MLIR failures > >The previous commit was reverted withd23cb79ba4because the [mlir-nvidia](https://lab.llvm.org/buildbot/#/builders/138/builds/24797) and [mlir-nvidia-gcc7](https://lab.llvm.org/buildbot/#/builders/116/builds/23929) Buildbots were failing. > >Those tests failed because MLIR's default SM was 5.0, which caused NVPTX to target PTX ISA v4.0, which did not support the intrinsics used in the failing tests. > >243f011577should address this by bumping MLIR's default SM to 7.5. Now, using MLIR's new default SM, NVPTX targets the PTX ISA v6.3, which supports the intrinsics used in the failing tests. --- The previous commit was reverted with e9b578a4d77025e18318efedd0f3f3764338d859 [because](https://github.com/llvm/llvm-project/pull/179304#issuecomment-3856301333) the clang driver set the default PTX ISA version to v4.2 when no CUDA installation is found. However, given our patch, we should not set a default; instead, let the LLVM backend select the appropriate PTX ISA version for the target SM.
52 lines
2.4 KiB
LLVM
52 lines
2.4 KiB
LLVM
; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_103a -mattr=+ptx90 2>&1 | FileCheck %s --check-prefix=CHECK-SM103A-HIGH
|
|
; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_103a 2>&1 | FileCheck %s --check-prefix=CHECK-SM103A
|
|
; RUN: not llc < %s -mtriple=nvptx64 -mcpu=sm_103a -mattr=+ptx87 2>&1 | FileCheck %s --check-prefix=CHECK-SM103A-LOW
|
|
; RUN: %if ptxas-sm_103a && ptxas-isa-9.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_103a -mattr=+ptx90 | %ptxas-verify -arch=sm_103a %}
|
|
; RUN: %if ptxas-sm_103a %{ llc < %s -mtriple=nvptx64 -mcpu=sm_103a | %ptxas-verify -arch=sm_103a %}
|
|
|
|
; Test that sm_120a defaults/requires PTX 8.7
|
|
; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_120a 2>&1 | FileCheck %s --check-prefix=CHECK-SM120A
|
|
; RUN: not llc < %s -mtriple=nvptx64 -mcpu=sm_120a -mattr=+ptx86 2>&1 | FileCheck %s --check-prefix=CHECK-SM120A-LOW
|
|
; RUN: %if ptxas-sm_120a %{ llc < %s -mtriple=nvptx64 -mcpu=sm_120a | %ptxas-verify -arch=sm_120a %}
|
|
|
|
; Test that sm_90a defaults/requires PTX 8.0
|
|
; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_90a 2>&1 | FileCheck %s --check-prefix=CHECK-SM90A
|
|
; RUN: not llc < %s -mtriple=nvptx64 -mcpu=sm_90a -mattr=+ptx78 2>&1 | FileCheck %s --check-prefix=CHECK-SM90A-LOW
|
|
; RUN: %if ptxas-sm_90a %{ llc < %s -mtriple=nvptx64 -mcpu=sm_90a | %ptxas-verify -arch=sm_90a %}
|
|
|
|
; Test older SM targets
|
|
; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_80 2>&1 | FileCheck %s --check-prefix=CHECK-SM80
|
|
; RUN: not llc < %s -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx63 2>&1 | FileCheck %s --check-prefix=CHECK-SM80-LOW
|
|
; RUN: %if ptxas-sm_80 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_80 | %ptxas-verify -arch=sm_80 %}
|
|
|
|
; CHECK-SM103A-HIGH: .version 9.0
|
|
; CHECK-SM103A-HIGH: .target sm_103a
|
|
|
|
; CHECK-SM103A: .version 8.8
|
|
; CHECK-SM103A: .target sm_103a
|
|
|
|
; CHECK-SM103A-LOW: LLVM ERROR: PTX version 8.7 does not support target 'sm_103a'.
|
|
; CHECK-SM103A-LOW: Minimum required PTX version is 8.8.
|
|
|
|
; CHECK-SM120A: .version 8.7
|
|
; CHECK-SM120A: .target sm_120a
|
|
|
|
; CHECK-SM120A-LOW: LLVM ERROR: PTX version 8.6 does not support target 'sm_120a'.
|
|
; CHECK-SM120A-LOW: Minimum required PTX version is 8.7.
|
|
|
|
; CHECK-SM90A: .version 8.0
|
|
; CHECK-SM90A: .target sm_90a
|
|
|
|
; CHECK-SM90A-LOW: LLVM ERROR: PTX version 7.8 does not support target 'sm_90a'.
|
|
; CHECK-SM90A-LOW: Minimum required PTX version is 8.0.
|
|
|
|
; CHECK-SM80: .version 7.0
|
|
; CHECK-SM80: .target sm_80
|
|
|
|
; CHECK-SM80-LOW: LLVM ERROR: PTX version 6.3 does not support target 'sm_80'.
|
|
; CHECK-SM80-LOW: Minimum required PTX version is 7.0.
|
|
|
|
define void @foo() {
|
|
ret void
|
|
}
|