<html><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; "><div><div>Hi Zack,</div><div><br class="webkit-block-placeholder"></div><div>On Sep 27, 2007, at 09:54, Zack Rusin wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; "><span class="Apple-style-span" style="-webkit-text-stroke-width: -1; ">as some of you may know we're in process of experimenting with LLVM in Gallium3D (Mesa's new driver model), where LLVM would be used both in the software only (by just JIT executing shaders) and hardware (drivers will implement LLVM code-generators) cases.</span></div></blockquote><div><br class="webkit-block-placeholder"></div><div>Neat.</div><br><blockquote type="cite"><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">That is graphics hardware (basically every single programmable gpu) has<span class="Apple-converted-space"> <span class="Apple-style-span" style="-webkit-text-stroke-width: -1; ">instruction level support for vector swizzling and write masks.</span></span></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><br></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">For example the following represents a valid gpu shader instruction:</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">ADD dst.xyz <span class="Apple-converted-space"> </span>src1.yxzw<span class="Apple-converted-space"> </span>src2.zwxy</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">which performs an addition that stores the result to the dst operated (each<span class="Apple-converted-space"> <span class="Apple-style-span" style="-webkit-text-stroke-width: -1; ">operarand is a vector type of four data elements) The instruction uses source swizzle modifiers and destination mask modifier.</span></span></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><br></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">So if a language is capable of expressing such constructs (as GLSL, HLSL and<span class="Apple-converted-space"> <span class="Apple-style-span" style="-webkit-text-stroke-width: -1; ">few others are) I'd like to make sure that the code generator is actually capable of generating instructions with exactly those semantics. </span></span></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><br></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">Right now vector operations utilizing swizzling and write masks in LLVM IR<span class="Apple-converted-space"> <span class="Apple-style-span" style="-webkit-text-stroke-width: -1; ">have to expressed with series of load/extractelement/instertelement/store constructs. As in </span></span></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><br></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">vec2 = vec4.xy<span class="Apple-converted-space"> </span></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><br></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">would end up being:</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">%tmp = load <4 x float>* @vec4</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">%tmp1 = extractelement <4 x float> %tmp, i32 0</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">%tmp2 = insertelement <2 x float> undef, float %tmp1, i32 0</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">%tmp3 = extractelement <4 x float> %tmp, i32 1</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">%tmp4 = insertelement <2 x float> %tmp2, float %tmp3, i32 1<span class="Apple-tab-span" style="white-space:pre"> </span></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">store <2 x float> %tmp4, <2 x float>* @vec2</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">or the like.</div></blockquote><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><br></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; ">Loads and stores are always explicit; the code generator will fold them into the machine instructions if possible.</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><br class="webkit-block-placeholder"></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; ">You may be able to take advantage of the shufflevector instruction. Although its result will be a <4 x float> instead of a <2 x float>, same as the source vector. So you'll need to find a way to write "extract subvector" that codegens well. Perhaps this will work:</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><br class="webkit-block-placeholder"></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; ">%shufvec = shufflevector <4 x float> ...</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; ">%src1 = extractelement %shufvec, 0</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; ">%src2 = extractelement %shufvec, 1</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; ">%tmp = insertelement <2 x float> undef, %src1, 0</div>%res = insertelement %tmp, %src2, 1</div><div><br><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; ">If that's no good, then you might want to add intrinsics to do the job. You can then easily pattern match on (llvm.extractvector (shufflevector ...), which) where llvm.extractvector is your intrinsic.</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><br class="webkit-block-placeholder"></div><blockquote type="cite"><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">So I think my options come down to:</div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><br></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">1) figure out a way of having code generator be actually able to combine all<span class="Apple-converted-space"> </span></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">those IR instructions back into<span class="Apple-converted-space"> </span></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">OP dst.writemask src1.swizzle1 src2.swizzle2</div></blockquote><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><blockquote type="cite"><br></blockquote></div><blockquote type="cite"><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; ">2) have some kind of instruction level support for it in LLVM IR</div></blockquote><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; "><br></div><div style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; min-height: 14px; ">It's much easier to define intrinsic functions than instructions, so start there if you need to go that route.</div></div><div><span class="Apple-style-span" style="border-collapse: separate; border-spacing: 0px 0px; color: rgb(0, 0, 0); font-family: Trebuchet MS; font-size: 12px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; text-align: auto; -khtml-text-decorations-in-effect: none; text-indent: 0px; -apple-text-size-adjust: auto; text-transform: none; orphans: 2; white-space: normal; widows: 2; word-spacing: 0px; "><span class="Apple-style-span" style="border-collapse: separate; border-spacing: 0px 0px; color: rgb(0, 0, 0); font-family: Trebuchet MS; font-size: 12px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; text-align: auto; -khtml-text-decorations-in-effect: none; text-indent: 0px; -apple-text-size-adjust: auto; text-transform: none; orphans: 2; white-space: normal; widows: 2; word-spacing: 0px; "><span class="Apple-style-span" style="border-collapse: separate; border-spacing: 0px 0px; color: rgb(0, 0, 0); font-family: Trebuchet MS; font-size: 12px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; text-align: auto; -khtml-text-decorations-in-effect: none; text-indent: 0px; -apple-text-size-adjust: auto; text-transform: none; orphans: 2; white-space: normal; widows: 2; word-spacing: 0px; "><span class="Apple-style-span" style="border-collapse: separate; border-spacing: 0px 0px; color: rgb(0, 0, 0); font-family: Trebuchet MS; font-size: 12px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; text-align: auto; -khtml-text-decorations-in-effect: none; text-indent: 0px; -apple-text-size-adjust: auto; text-transform: none; orphans: 2; white-space: normal; widows: 2; word-spacing: 0px; "><span class="Apple-style-span" style="border-collapse: separate; border-spacing: 0px 0px; color: rgb(0, 0, 0); font-family: Trebuchet MS; font-size: 12px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; text-align: auto; -khtml-text-decorations-in-effect: none; text-indent: 0px; -apple-text-size-adjust: auto; text-transform: none; orphans: 2; white-space: normal; widows: 2; word-spacing: 0px; "><span class="Apple-style-span" style="border-collapse: separate; border-spacing: 0px 0px; color: rgb(0, 0, 0); font-family: Trebuchet MS; font-size: 12px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; text-align: auto; -khtml-text-decorations-in-effect: none; text-indent: 0px; -apple-text-size-adjust: auto; text-transform: none; orphans: 2; white-space: normal; widows: 2; word-spacing: 0px; "><div><br class="webkit-block-placeholder"></div>— Gordon<br class="Apple-interchange-newline"></span></span></span></span></span></span> </div><br></body></html>