[llvm] [NVPTX] Add support for f16 fabs (PR #116107)

[Alex MacLean via llvm-commits]

================
@@ -0,0 +1,99 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; ## Some FP16 support but not for abs
+; RUN: llc < %s -mtriple=nvptx64-nvidia-cuda -mcpu=sm_53  \
+; RUN:          -O0 -disable-post-ra -frame-pointer=all -verify-machineinstrs \
+; RUN: | FileCheck -allow-deprecated-dag-overlap -check-prefixes CHECK-NOF16 %s
+; RUN: %if ptxas %{                                                           \
+; RUN:   llc < %s -mtriple=nvptx64-nvidia-cuda -mcpu=sm_53  \
+; RUN:          -O0 -disable-post-ra -frame-pointer=all -verify-machineinstrs \
+; RUN:   | %ptxas-verify -arch=sm_53                                          \
+; RUN: %}
+
+; ## FP16 support explicitly disabled.
+; RUN: llc < %s -mtriple=nvptx64-nvidia-cuda -mcpu=sm_53  \
+; RUN:          -O0 -disable-post-ra -frame-pointer=all --nvptx-no-f16-math \
+; RUN:           -verify-machineinstrs \
+; RUN: | FileCheck -allow-deprecated-dag-overlap -check-prefixes CHECK-NOF16 %s
+; RUN: %if ptxas %{                                                           \
+; RUN:   llc < %s -mtriple=nvptx64-nvidia-cuda -mcpu=sm_53  \
+; RUN:          -O0 -disable-post-ra -frame-pointer=all --nvptx-no-f16-math   \
+; RUN:           -verify-machineinstrs                                        \
+; RUN:   | %ptxas-verify -arch=sm_53                                          \
+; RUN: %}
+
+; ## FP16 is not supported by hardware.
+; RUN: llc < %s -O0 -mtriple=nvptx64-nvidia-cuda -mcpu=sm_52  \
+; RUN:          -disable-post-ra -frame-pointer=all -verify-machineinstrs \
+; RUN: | FileCheck -allow-deprecated-dag-overlap -check-prefixes CHECK-NOF16 %s
+; RUN: %if ptxas %{                                                               \
+; RUN:   llc < %s -O0 -mtriple=nvptx64-nvidia-cuda -mcpu=sm_52  \
+; RUN:          -disable-post-ra -frame-pointer=all -verify-machineinstrs         \
+; RUN:   | %ptxas-verify -arch=sm_52                                              \
+; RUN: %}
+
+; ## Full FP16 support.
+; RUN: llc < %s -O0 -mtriple=nvptx64-nvidia-cuda -mcpu=sm_53 -mattr=+ptx70 \
----------------
AlexMaclean wrote:

This test only covers f16 abs. bf16 abs was already supported and has coverage in the bf16 tests. Should I add bf16 cases to this test as well?

https://github.com/llvm/llvm-project/pull/116107