Hi Richard,
Richard Biener wrote:
Looks good to me.
Thanks - I will commit it after lunch to see whether someone else has
additional comments.
+@item gfx1030
+Compile for RDNA2 gfx1030 devices (GFX10 series).
+
+@item gfx1100
+Compile for RDNA3 gfx1100 devices (GFX11 series).
Btw, "GFX10" series isn't precise as it's only the high-end parts
that are covered by gfx1030, there's gfx103[0-6] where hopefully
at least gfx1031, gfx1032 and gfx1034 (the dGPU variants) are
trivial to support as well(?).
Using gfx103x might be better
that way, OTOH if APU vs dGPU will make a compilation target
difference then gfx103d vs gfx103a maybe? "GFX10" series might
On the LLVM side (and also for the llvm-mc assembler), they distinguish
all of the gfx* for the -mcpu= argument. See
https://llvm.org/docs/AMDGPUUsage.html#id26 for that list.
Thus, I think it makes sense to do the same here. The last column on
that page lists the supported hardware but is it neither really up to
date nor complete.
Thus, I found it easier to just mention gfx1100 as that's unique. On the
ROCm side, AMD has:
https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/system-requirements.html
Tobias
