For example, when loading 3.1415926535897931159979634685441851615905761718750, use lu12i.w $a0, 345154 ori $a0, $a0, 3352 lu32i.d $a0, -450053 lu52i.d $a0, $a0, 1024 movgr2fr.d $fa0, $a0 instead of pcalau12i $a0, .LCPI2_0 addi.d $a0, $a0, .LCPI2_0 fld.d $fa0, $a0, 0 Test this patch with 3A5000 on llvm13, the result shows that SPEC CPU2006 FP score increases 1% in average, 470.lbm score increases 8.9%. Thanks to @xry111 for the suggestion: https://reviews.llvm.org/D128898#3632140

Load FP immediates by movgr2fr from materialized integer if the bitcasted integer can be materialized within 2 instructions. For example, when loading double 1024.0, use lu52i.d $a0, $zero, 1033 movgr2fr.d $fa0, $a0 instead of pcalau12i $a0, .LCPI2_0 addi.d $a0, $a0, .LCPI2_0 fld.d $fa0, $a0, 0 Test this patch with 3A5000 on llvm13, the result shows that SPEC CPU2006 FP score increases 1.2% in average, 470.lbm score increases 11.9%. Thanks to @xry111 for the suggestion: https://reviews.llvm.org/D128898#3632140

For example, when loading 3.1415926535897931159979634685441851615905761718750, useLoad FP immediates by movgr2fr from materialized integer if the bitcasted integer lu12i.w $a0, 345154 ori $a0, $a0, 3352can be materialized within 2 instructions. lu32i.d $a0For example, when loading double 1024.0, -450053use lu52i.d $a0, $a0$zero, 10241033 movgr2fr.d $fa0, $a0 instead of pcalau12i $a0, .LCPI2_0 addi.d $a0, $a0, .LCPI2_0 fld.d $fa0, $a0, 0 Test this patch with 3A5000 on llvm13, the result shows that SPEC CPU2006 FP FP score increases 1.2% in average, 470.lbm score increases 811.9%. Thanks to @xry111 for the suggestion: https://reviews.llvm.org/D128898#3632140