[AVX512][inline-asm] Fix AVX512 inline assembly instruction resolution when the size qualifier of a memory operand is not specified explicitly.

[coby]

Summary

This commit handles cases where the size qualifier of an indirect memory reference operand in Intel syntax is missing (e.g. "vaddps xmm1, xmm2, [a]").

GCC will deduce the size qualifier for AVX512 vector and broadcast memory operands based on the possible matches:
"vaddps xmm1, xmm2, [a]" matches only “XMMWORD PTR” qualifier.
"vaddps xmm1, xmm2, [a]{1to4}" matches only “DWORD PTR” qualifier.

This is different from the current behavior of LLVM, which deduces the size qualifier based on the size of the memory operand.
For "vaddps xmm1, xmm2, [a]"
"char a;" will imply "BYTE PTR" qualifier
"short a;" will imply "WORD PTR" qualifier.

This commit aligns LLVM to GCC’s behavior.

This is the clang part of the review.
The llvm part of the review: https://reviews.llvm.org/D26586

Differential Revision: https://reviews.llvm.org/D26587

[myatsina]

LGTM

Please emphasize in you commit message that this the test case for the llvm commit (and even better if you add the commit number of the llvm commit),

[coby]

superseded by rL302179

test/CodeGen/ms-inline-asm-avx512-memory-adjustments.c

+// RUN: %clang_cc1 %s -triple x86_64-unknown-linux-gnu -emit-llvm -target-cpu skylake-avx512 -fasm-blocks -o - | FileCheck %s
+
+// Case1: Check integrity of inspected patch upon broadcasting
+// Case2: Check integrity of inspected patch upon SIMD mem ref
+
+void F() {
+  char a;
+  // CHECK: vaddps xmm1, xmm2, dword ptr $0{1to4}
+  // CHECK: vaddps xmm1, xmm2, xmmword ptr $1
+  __asm vaddps xmm1, xmm2, [a]{1to4}
+  __asm vaddps xmm1, xmm2, [a]
+}