[PATCH] D141173: AMDGPU: Use getConstantStringInfo for printf format strings

[Sameer Sahasrabuddhe via Phabricator via llvm-commits]

sameerds added a comment.

The change itself looks straightforward to me, but I am struggling to understand what is happening in the tests. @yaxunl , can you please comment?



================
Comment at: llvm/test/CodeGen/AMDGPU/opencl-printf.ll:549
+entry:
+  %call1 = call i32 (ptr addrspace(4), ...) @printf(ptr addrspace(4) @format.str.no.null.terminator, i32 %n)
+  ret void
----------------
I can't make out from the spec if OpenCL allows a format string with no terminator. Is this UB? Is it supposed to produce a diagnostic?


================
Comment at: llvm/test/CodeGen/AMDGPU/opencl-printf.ll:605
+entry:
+  %call1 = call i32 (ptr addrspace(4), ...) @printf(ptr addrspace(4) getelementptr ([11 x i8], ptr addrspace(4) @indexed.format.str, i64 0, i64 11), i32 %n)
+  ret void
----------------
Same question. UB? Diagnostic? 


================
Comment at: llvm/test/CodeGen/AMDGPU/opencl-printf.ll:625
 ; GCN-NEXT:    [[PRINTBUFFIDCAST:%.*]] = bitcast ptr addrspace(1) [[PRINTBUFFID]] to ptr addrspace(1)
-; GCN-NEXT:    store i32 12, ptr addrspace(1) [[PRINTBUFFIDCAST]], align 4
+; GCN-NEXT:    store i32 15, ptr addrspace(1) [[PRINTBUFFIDCAST]], align 4
 ; GCN-NEXT:    [[PRINTBUFFGEP:%.*]] = getelementptr i8, ptr addrspace(1) [[PRINTF_ALLOC_FN]], i32 4
----------------
Are the previous values plain wrong? Can this be reproduced in an actual OpenCL kernel? Will it crash the kernel?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D141173/new/

https://reviews.llvm.org/D141173