<html><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; "><div>Hi,</div><div><br></div><div>The current definition of shuffle vector is</div><div> <span class="Apple-style-span" style="font-family: -webkit-monospace; font-size: 14px; white-space: pre; "><font class="Apple-style-span" size="1"><span class="Apple-style-span" style="font-size: 9px; "> <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> </span></font><i><font class="Apple-style-span" size="1"><span class="Apple-style-span" style="font-size: 9px; ">; yields <n x <ty>></span></font></i></span></div><div><div><br></div><div>The first two operands of a '<tt><font class="Apple-style-span" face="Helvetica">shufflevector</font></tt>' instruction are vectors with types that match each other and types that match the result of the instruction. The third argument is a shuffle mask, which has the same number of elements as the other vector type, but whose element type is always 'i32'. The shuffle mask operand is required to be a constant vector with either constant integer or undef values.</div><div><br></div><div><span class="Apple-style-span" style="font-family: -webkit-monospace; font-size: 14px; white-space: pre; "><font class="Apple-style-span" size="1"><font class="Apple-style-span" face="Helvetica" size="3"><span class="Apple-style-span" style="font-size: 12px; white-space: normal;">I am proposing to extend the shuffle vector definition to be</span></font><span class="Apple-style-span" style="font-size: 9px; "></span></font></span></div><div><span class="Apple-style-span" style="font-family: -webkit-monospace; font-size: 14px; white-space: pre; "><font class="Apple-style-span" size="1"><span class="Apple-style-span" style="font-size: 9px; "> <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> </span></font><i><font class="Apple-style-span" size="1"><span class="Apple-style-span" style="font-size: 9px; ">; yields <m x <ty>></span></font></i></span></div></div><div><br></div><div>The third argument is the shuffle mask and it is not required to have the same number of elements as the other vector type but the element type is still always 'i32'</div><div><br></div><div>The reason for this change is to allow us to handle vectors much more naturally, allows the generated IR to reflect the structure of the user code, and most importantly do a better job in code generation. Today, building a vector for different lengths requires a sequence of extracting each element in the vector and inserting each element into a new vector. With this new form, it is more straightforward to write and reason about</div><div><br></div><div><div>typedef __attribute__(( ext_vector_type(4) )) float float4;</div><div>typedef __attribute__(( ext_vector_type(8) )) float float8;</div><div>float8 f8;</div><div>float4 f4a, f4b, f4c;</div><div>f4a = f8.hi;</div><div>f8.hi = f4b; f8.lo = f4c;</div><div><br></div><div>where hi and lo represent the high half and low half of the vector. The outgoing IR is</div><div> %f4a = shufflevector <8xf32>%f8, undef, <4xi32> <0, 1, 2, 3></div><div><div> %f8 = shufflevector <4xf32>%f4b, <4xf32>%f4c, <8xi32> <0, 1, 2, 3, 4, 5, 6, 7></div><div><br></div><div>The problem with generating insert and extracts is that we can generate poor code</div><div><div> %tmp16 = extractelement <4 x float> %f4b, i32 0 </div><div> %f8a = insertelement <8 x float> %f8a, float %tmp16, i32 0 </div><div> %tmp18 = extractelement <4 x float> %f4b, i32 1 </div><div> %f8c = insertelement <8 x float> %f8b, float %tmp18, i32 1 </div><div> ... </div><div>For X86, legalize will convert each insertelement to become a vector shuffle. We are very careful in combining vector shuffles because we don't want to produce a vector shuffle whose mask is illegal or hard to code gen so we end up in this code to generate a sequence of unpcks and movhlps for this. With the new form, Legalize will divide the 8xf32 vector into two 4xf32 and since the two sides are the same, it will generate quad word moves to copy the values.</div><div><br></div><div>There are other cases when a user write vector code, the generation of extract element and insert elements will cause us to lose the structure of the original vector code and without this structure, the code generator will not generate code for it. </div><div><br></div><div>Please let me know if you have any comments, suggestions, or concerns,</div><div><br></div><div>Thanks,</div><div> -- Mon Ping</div></div></div><div><br></div><div><br></div></div><div><br></div></body></html>