<html> <head> <meta http-equiv="Content-Type" content="text/html; charset=utf-8"> </head> <body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""> <div> <blockquote type="cite" class=""> <div class="">On Jan 18, 2018, at 1:10 PM, Roland Scheidegger <<a href="mailto:sroland@vmware.com" class="">sroland@vmware.com</a>> wrote:</div> <br class="Apple-interchange-newline"> <div class=""><span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">Am 17.01.2018 um 23:33 schrieb George Kyriazis:</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <blockquote type="cite" style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-size-adjust: auto; -webkit-text-stroke-width: 0px;" class=""> The texture swizzle was not doing the right thing for avx512-style<br class=""> 16-wide loads.<br class=""> <br class=""> Special-case the post-load swizzle operations for avx512 so that we move<br class=""> the xyzw components correctly to the outputs.<br class=""> <br class=""> cc: Jose Fonseca <<a href="mailto:jfonseca@vmware.com" class="">jfonseca@vmware.com</a>><br class=""> ---<br class=""> src/gallium/auxiliary/gallivm/lp_bld_pack.c | 40 +++++++++++++++++++++++++++--<br class=""> 1 file changed, 38 insertions(+), 2 deletions(-)<br class=""> <br class=""> diff --git a/src/gallium/auxiliary/gallivm/lp_bld_pack.c b/src/gallium/auxiliary/gallivm/lp_bld_pack.c<br class=""> index e8d4fcd..7879826 100644<br class=""> --- a/src/gallium/auxiliary/gallivm/lp_bld_pack.c<br class=""> +++ b/src/gallium/auxiliary/gallivm/lp_bld_pack.c<br class=""> @@ -129,6 +129,31 @@ lp_build_const_unpack_shuffle_half(struct gallivm_state *gallivm,<br class=""> }<br class=""> <br class=""> /**<br class=""> + * Similar to lp_build_const_unpack_shuffle_half, but for AVX512<br class=""> + * See comment above lp_build_interleave2_half for more details.<br class=""> + */<br class=""> +static LLVMValueRef<br class=""> +lp_build_const_unpack_shuffle_16wide(struct gallivm_state *gallivm,<br class=""> + unsigned lo_hi)<br class=""> +{<br class=""> + LLVMValueRef elems[LP_MAX_VECTOR_LENGTH];<br class=""> + unsigned i, j;<br class=""> +<br class=""> + assert(lo_hi < 2);<br class=""> +<br class=""> + // for the following lo_hi setting, convert 0 -> f to:<br class=""> + // 0: 0 16 4 20 8 24 12 28 1 17 5 21 9 25 13 29<br class=""> + // 1: 2 18 6 22 10 26 14 30 3 19 7 23 11 27 15 31<br class=""> + for (i = 0; i < 16; i++) {<br class=""> + j = ((i&0x06)<<1) + ((i&1)<<4) + (i>>3) + (lo_hi<<1);<br class=""> +<br class=""> + elems[i] = lp_build_const_int32(gallivm, j);<br class=""> + }<br class=""> +<br class=""> + return LLVMConstVector(elems, 16);<br class=""> +}<br class=""> +<br class=""> +/**<br class=""> * Build shuffle vectors that match PACKxx (SSE) instructions or<br class=""> * VPERM (Altivec).<br class=""> */<br class=""> @@ -325,8 +350,8 @@ lp_build_interleave2(struct gallivm_state *gallivm,<br class=""> }<br class=""> <br class=""> /**<br class=""> - * Interleave vector elements but with 256 bit,<br class=""> - * treats it as interleave with 2 concatenated 128 bit vectors.<br class=""> + * Interleave vector elements but with 256 (or 512) bit,<br class=""> + * treats it as interleave with 2 concatenated 128 (or 256) bit vectors.<br class=""> *<br class=""> * This differs to lp_build_interleave2 as that function would do the following (for lo):<br class=""> * a0 b0 a1 b1 a2 b2 a3 b3, and this does not compile into an AVX unpack instruction.<br class=""> @@ -343,6 +368,14 @@ lp_build_interleave2(struct gallivm_state *gallivm,<br class=""> *<br class=""> * And interleave-hi would result in:<br class=""> * a2 b2 a3 b3 a6 b6 a7 b7<br class=""> + *<br class=""> + * For 512 bits, the following are true:<br class=""> + *<br class=""> + * Interleave-lo would result in (capital letters denote hex indices):<br class=""> + * a0 b0 a1 b1 a4 b4 a5 b5 a8 b8 a9 b9 aC bC aD bD<br class=""> + *<br class=""> + * Interleave-hi would result in:<br class=""> + * a2 b2 a3 b3 a6 b6 a7 b7 aA bA aB bB aE bE aF bF<br class=""> */<br class=""> LLVMValueRef<br class=""> lp_build_interleave2_half(struct gallivm_state *gallivm,<br class=""> @@ -354,6 +387,9 @@ lp_build_interleave2_half(struct gallivm_state *gallivm,<br class=""> if (type.length * type.width == 256) {<br class=""> LLVMValueRef shuffle = lp_build_const_unpack_shuffle_half(gallivm, type.length, lo_hi);<br class=""> return LLVMBuildShuffleVector(gallivm->builder, a, b, shuffle, "");<br class=""> + } else if ((type.length == 16) && (type.width == 32)) {<br class=""> + LLVMValueRef shuffle = lp_build_const_unpack_shuffle_16wide(gallivm, lo_hi);<br class=""> + return LLVMBuildShuffleVector(gallivm->builder, a, b, shuffle, "");<br class=""> </blockquote> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">This is not really "interleave_half", more like "interleave_quarter"...</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">That said, avx512 certainly follows the same rules as avx256, so 128bit</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">pieces are treated independently. So maybe this should be renamed like</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">"interleave_native" or something like that.</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">Also, I believe it is definitely a mistake to restrict this to dword</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">interleaves here. You should handle all type widths, just like the</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">256bit case can handle all widths.</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">And I'm not sure through which paths you reach this, but I'm not sure</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">why you don't need the corresponding unpack2_native and pack2_native</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">adjustments - it should not really be a special case, avx512 should</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">generally handle things like this (if you'd want to extend the gallivm</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">code to use avx512...). For that matter, the commit log and shortlog is</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">confusing, because this isn't directly related to texture fetching.</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> </div> </blockquote> <div><br class=""> </div> Thanks Roland, <div class=""><br class=""> </div> <div class="">This check-in is a prerequisite for some avx-512 changes I am about to check in to the swr driver. I’ve hit this for piglit test vs-texelfetch-isampler1d and similar ones that try to fetch from a texture from a VS. You are right, there are probably other cases where this gets hit, but that was the place where I’ve hit it first. If you could recommend some tests that hit the other width cases, it would be great to fix them too; since didn’t know where those cases would be hit, I decided to make this “surgical” change. <div> <blockquote type="cite" class=""></blockquote> </div> </div> <div class=""><br class=""> </div> <div class="">My follow-on changes depend on this change for correctness, and I’d like to get the the other changes in before the mesa 18.0 date (tomorrow). AVX512 support is an ongoing effort on our side, so they will definitely be additional check-ins along the same lines.</div> <div class=""><br class=""> </div> <div class="">Thoughts on checking in as-is for 18.0, with upcoming changes slowly creeping in to fix additional issues, and make things more general?</div> <div class=""><br class=""> </div> <div class="">Thanks,</div> <div class=""><br class=""> </div> <div class="">George</div> <div class=""><br class=""> </div> <div class=""><br class=""> </div> <blockquote type="cite" class=""> <div class=""><span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">Roland</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <blockquote type="cite" style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-size-adjust: auto; -webkit-text-stroke-width: 0px;" class=""> } else {<br class=""> return lp_build_interleave2(gallivm, type, a, b, lo_hi);<br class=""> }<br class=""> <br class=""> </blockquote> <br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">_______________________________________________</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <span style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; float: none; display: inline !important;" class="">mesa-dev mailing list</span><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <a href="mailto:mesa-dev@lists.freedesktop.org" style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-size-adjust: auto; -webkit-text-stroke-width: 0px;" class="">mesa-dev@lists.freedesktop.org</a><br style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px;" class=""> <a href="https://lists.freedesktop.org/mailman/listinfo/mesa-dev" style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: normal; letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-size-adjust: auto; -webkit-text-stroke-width: 0px;" class="">https://lists.freedesktop.org/mailman/listinfo/mesa-dev</a></div> </blockquote> </div> <br class=""> </body> </html>