<html><head><meta http-equiv="Content-Type" content="text/html charset=us-ascii"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;">I have a generic question about these tables. If the cost is impacted by how the target / sdisel lowers the node, then the entries can be outdated if lowering changes. It would be nice if we have a mechanism to detect that.<div><br></div><div>Evan</div><div><br><div><div>On Apr 1, 2013, at 3:23 AM, Benjamin Kramer <<a href="mailto:benny.kra@googlemail.com">benny.kra@googlemail.com</a>> wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><div style="letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px;">Author: d0k<br>Date: Mon Apr 1 05:23:49 2013<br>New Revision: 178459<br><br>URL:<span class="Apple-converted-space"> </span><a href="http://llvm.org/viewvc/llvm-project?rev=178459&view=rev">http://llvm.org/viewvc/llvm-project?rev=178459&view=rev</a><br>Log:<br>X86TTI: Add accurate costs for itofp operations, based on the actual instruction counts.<br><br>Modified:<br> llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp<br> llvm/trunk/test/Analysis/CostModel/X86/cast.ll<br> llvm/trunk/test/Transforms/LoopVectorize/X86/conversion-cost.ll<br><br>Modified: llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp<br>URL:<span class="Apple-converted-space"> </span><a href="http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp?rev=178459&r1=178458&r2=178459&view=diff">http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp?rev=178459&r1=178458&r2=178459&view=diff</a><br>==============================================================================<br>--- llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp (original)<br>+++ llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp Mon Apr 1 05:23:49 2013<br>@@ -271,10 +271,33 @@ unsigned X86TTI::getCastInstrCost(unsign<br> { ISD::ZERO_EXTEND, MVT::v4i64, MVT::v4i32, 1 },<br> { ISD::TRUNCATE, MVT::v4i32, MVT::v4i64, 1 },<br> { ISD::TRUNCATE, MVT::v8i16, MVT::v8i32, 1 },<br>- { ISD::SINT_TO_FP, MVT::v8f32, MVT::v8i8, 1 },<br>- { ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i8, 1 },<br>- { ISD::UINT_TO_FP, MVT::v8f32, MVT::v8i8, 1 },<br>- { ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i8, 1 },<br>+<br>+ { ISD::SINT_TO_FP, MVT::v8f32, MVT::v8i1, 8 },<br>+ { ISD::SINT_TO_FP, MVT::v8f32, MVT::v8i8, 8 },<br>+ { ISD::SINT_TO_FP, MVT::v8f32, MVT::v8i16, 5 },<br>+ { ISD::SINT_TO_FP, MVT::v8f32, MVT::v8i32, 1 },<br>+ { ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i1, 3 },<br>+ { ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i8, 3 },<br>+ { ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i16, 3 },<br>+ { ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i32, 1 },<br>+ { ISD::SINT_TO_FP, MVT::v4f64, MVT::v4i1, 3 },<br>+ { ISD::SINT_TO_FP, MVT::v4f64, MVT::v4i8, 3 },<br>+ { ISD::SINT_TO_FP, MVT::v4f64, MVT::v4i16, 3 },<br>+ { ISD::SINT_TO_FP, MVT::v4f64, MVT::v4i32, 1 },<br>+<br>+ { ISD::UINT_TO_FP, MVT::v8f32, MVT::v8i1, 6 },<br>+ { ISD::UINT_TO_FP, MVT::v8f32, MVT::v8i8, 5 },<br>+ { ISD::UINT_TO_FP, MVT::v8f32, MVT::v8i16, 5 },<br>+ { ISD::UINT_TO_FP, MVT::v8f32, MVT::v8i32, 9 },<br>+ { ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i1, 7 },<br>+ { ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i8, 2 },<br>+ { ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i16, 2 },<br>+ { ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i32, 6 },<br>+ { ISD::UINT_TO_FP, MVT::v4f64, MVT::v4i1, 7 },<br>+ { ISD::UINT_TO_FP, MVT::v4f64, MVT::v4i8, 2 },<br>+ { ISD::UINT_TO_FP, MVT::v4f64, MVT::v4i16, 2 },<br>+ { ISD::UINT_TO_FP, MVT::v4f64, MVT::v4i32, 6 },<br>+<br> { ISD::FP_TO_SINT, MVT::v8i8, MVT::v8f32, 1 },<br> { ISD::FP_TO_SINT, MVT::v4i8, MVT::v4f32, 1 },<br> { ISD::ZERO_EXTEND, MVT::v8i32, MVT::v8i1, 6 },<br><br>Modified: llvm/trunk/test/Analysis/CostModel/X86/cast.ll<br>URL:<span class="Apple-converted-space"> </span><a href="http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/CostModel/X86/cast.ll?rev=178459&r1=178458&r2=178459&view=diff">http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/CostModel/X86/cast.ll?rev=178459&r1=178458&r2=178459&view=diff</a><br>==============================================================================<br>--- llvm/trunk/test/Analysis/CostModel/X86/cast.ll (original)<br>+++ llvm/trunk/test/Analysis/CostModel/X86/cast.ll Mon Apr 1 05:23:49 2013<br>@@ -77,3 +77,78 @@ define i32 @masks4(<4 x i1> %in) {<br> ret i32 undef<br>}<br><br>+define void @sitofp4(<4 x i1> %a, <4 x i8> %b, <4 x i16> %c, <4 x i32> %d) {<br>+ ; CHECK: cost of 3 {{.*}} sitofp<br>+ %A1 = sitofp <4 x i1> %a to <4 x float><br>+ ; CHECK: cost of 3 {{.*}} sitofp<br>+ %A2 = sitofp <4 x i1> %a to <4 x double><br>+<br>+ ; CHECK: cost of 3 {{.*}} sitofp<br>+ %B1 = sitofp <4 x i8> %b to <4 x float><br>+ ; CHECK: cost of 3 {{.*}} sitofp<br>+ %B2 = sitofp <4 x i8> %b to <4 x double><br>+<br>+ ; CHECK: cost of 3 {{.*}} sitofp<br>+ %C1 = sitofp <4 x i16> %c to <4 x float><br>+ ; CHECK: cost of 3 {{.*}} sitofp<br>+ %C2 = sitofp <4 x i16> %c to <4 x double><br>+<br>+ ; CHECK: cost of 1 {{.*}} sitofp<br>+ %D1 = sitofp <4 x i32> %d to <4 x float><br>+ ; CHECK: cost of 1 {{.*}} sitofp<br>+ %D2 = sitofp <4 x i32> %d to <4 x double><br>+ ret void<br>+}<br>+<br>+define void @sitofp8(<8 x i1> %a, <8 x i8> %b, <8 x i16> %c, <8 x i32> %d) {<br>+ ; CHECK: cost of 8 {{.*}} sitofp<br>+ %A1 = sitofp <8 x i1> %a to <8 x float><br>+<br>+ ; CHECK: cost of 8 {{.*}} sitofp<br>+ %B1 = sitofp <8 x i8> %b to <8 x float><br>+<br>+ ; CHECK: cost of 5 {{.*}} sitofp<br>+ %C1 = sitofp <8 x i16> %c to <8 x float><br>+<br>+ ; CHECK: cost of 1 {{.*}} sitofp<br>+ %D1 = sitofp <8 x i32> %d to <8 x float><br>+ ret void<br>+}<br>+<br>+define void @uitofp4(<4 x i1> %a, <4 x i8> %b, <4 x i16> %c, <4 x i32> %d) {<br>+ ; CHECK: cost of 7 {{.*}} uitofp<br>+ %A1 = uitofp <4 x i1> %a to <4 x float><br>+ ; CHECK: cost of 7 {{.*}} uitofp<br>+ %A2 = uitofp <4 x i1> %a to <4 x double><br>+<br>+ ; CHECK: cost of 2 {{.*}} uitofp<br>+ %B1 = uitofp <4 x i8> %b to <4 x float><br>+ ; CHECK: cost of 2 {{.*}} uitofp<br>+ %B2 = uitofp <4 x i8> %b to <4 x double><br>+<br>+ ; CHECK: cost of 2 {{.*}} uitofp<br>+ %C1 = uitofp <4 x i16> %c to <4 x float><br>+ ; CHECK: cost of 2 {{.*}} uitofp<br>+ %C2 = uitofp <4 x i16> %c to <4 x double><br>+<br>+ ; CHECK: cost of 6 {{.*}} uitofp<br>+ %D1 = uitofp <4 x i32> %d to <4 x float><br>+ ; CHECK: cost of 6 {{.*}} uitofp<br>+ %D2 = uitofp <4 x i32> %d to <4 x double><br>+ ret void<br>+}<br>+<br>+define void @uitofp8(<8 x i1> %a, <8 x i8> %b, <8 x i16> %c, <8 x i32> %d) {<br>+ ; CHECK: cost of 6 {{.*}} uitofp<br>+ %A1 = uitofp <8 x i1> %a to <8 x float><br>+<br>+ ; CHECK: cost of 5 {{.*}} uitofp<br>+ %B1 = uitofp <8 x i8> %b to <8 x float><br>+<br>+ ; CHECK: cost of 5 {{.*}} uitofp<br>+ %C1 = uitofp <8 x i16> %c to <8 x float><br>+<br>+ ; CHECK: cost of 9 {{.*}} uitofp<br>+ %D1 = uitofp <8 x i32> %d to <8 x float><br>+ ret void<br>+}<br><br>Modified: llvm/trunk/test/Transforms/LoopVectorize/X86/conversion-cost.ll<br>URL:<span class="Apple-converted-space"> </span><a href="http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopVectorize/X86/conversion-cost.ll?rev=178459&r1=178458&r2=178459&view=diff">http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopVectorize/X86/conversion-cost.ll?rev=178459&r1=178458&r2=178459&view=diff</a><br>==============================================================================<br>--- llvm/trunk/test/Transforms/LoopVectorize/X86/conversion-cost.ll (original)<br>+++ llvm/trunk/test/Transforms/LoopVectorize/X86/conversion-cost.ll Mon Apr 1 05:23:49 2013<br>@@ -33,11 +33,10 @@ define i32 @conversion_cost2(i32 %n, i8*<br><br>.lr.ph: ; preds = %0, %.lr.ph<br> %indvars.iv = phi i64 [ %indvars.iv.next, %.lr.ph ], [ 9, %0 ]<br>- %2 = add nsw i64 %indvars.iv, 3<br>- %3 = trunc i64 %2 to i32<br>- %4 = sitofp i32 %3 to float<br>- %5 = getelementptr inbounds float* %B, i64 %indvars.iv<br>- store float %4, float* %5, align 4<br>+ %add = add nsw i64 %indvars.iv, 3<br>+ %tofp = sitofp i64 %add to float<br>+ %gep = getelementptr inbounds float* %B, i64 %indvars.iv<br>+ store float %tofp, float* %gep, align 4<br> %indvars.iv.next = add i64 %indvars.iv, 1<br> %lftr.wideiv = trunc i64 %indvars.iv.next to i32<br> %exitcond = icmp eq i32 %lftr.wideiv, %n<br><br><br>_______________________________________________<br>llvm-commits mailing list<br><a href="mailto:llvm-commits@cs.uiuc.edu">llvm-commits@cs.uiuc.edu</a><br><a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</a></div></blockquote></div><br></div></body></html>