marvin.weiler
/
musl-cross-make
mirror of https://github.com/richfelker/musl-cross-make.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
musl-cross-make/patches/gcc-9.2.0/0018-m68k-sqrt.diff

19 lines
695 B
Diff
Raw Permalink Normal View History

add patch fixing broken m68k sqrt gcc pr 95921. the m68k sqrt rtl handles excess precision wrong. I couldn't figure out how to make gcc generate the insn only for long double, but since long double usage is rare anyway, just remove it. this can be replaced with a proper fix later if there's ever one upstream.
2020-06-30 22:35:55 +02:00
--- gcc-9.2.0.orig/gcc/config/m68k/m68k.md 2019-01-01 07:31:55.000000000 -0500
improve m68k sqrt patch not to suppress insn on archs where it works the 60840 fpu and later, and coldfire fpus, have precision-specific sqrt instructions that get used and do the right thing. only when targeting baseline (or near-baseline) ISA does the ld80-only instruction get used, and that's all that needs to be suppressed.
2020-07-02 17:50:26 +02:00
+++ gcc-9.2.0/gcc/config/m68k/m68k.md 2020-07-01 15:57:20.528408009 -0400
@@ -4126,13 +4126,13 @@
(define_expand "sqrt<mode>2"
[(set (match_operand:FP 0 "nonimmediate_operand" "")
(sqrt:FP (match_operand:FP 1 "general_operand" "")))]
add patch fixing broken m68k sqrt gcc pr 95921. the m68k sqrt rtl handles excess precision wrong. I couldn't figure out how to make gcc generate the insn only for long double, but since long double usage is rare anyway, just remove it. this can be replaced with a proper fix later if there's ever one upstream.
2020-06-30 22:35:55 +02:00
- "TARGET_HARD_FLOAT"
improve m68k sqrt patch not to suppress insn on archs where it works the 60840 fpu and later, and coldfire fpus, have precision-specific sqrt instructions that get used and do the right thing. only when targeting baseline (or near-baseline) ISA does the ld80-only instruction get used, and that's all that needs to be suppressed.
2020-07-02 17:50:26 +02:00
+ "(TARGET_68881 && TARGET_68040) || TARGET_COLDFIRE_FPU"
"")
add patch fixing broken m68k sqrt gcc pr 95921. the m68k sqrt rtl handles excess precision wrong. I couldn't figure out how to make gcc generate the insn only for long double, but since long double usage is rare anyway, just remove it. this can be replaced with a proper fix later if there's ever one upstream.
2020-06-30 22:35:55 +02:00
improve m68k sqrt patch not to suppress insn on archs where it works the 60840 fpu and later, and coldfire fpus, have precision-specific sqrt instructions that get used and do the right thing. only when targeting baseline (or near-baseline) ISA does the ld80-only instruction get used, and that's all that needs to be suppressed.
2020-07-02 17:50:26 +02:00
(define_insn "sqrt<mode>2_68881"
[(set (match_operand:FP 0 "nonimmediate_operand" "=f")
(sqrt:FP (match_operand:FP 1 "general_operand" "f<FP:dreg>m")))]
- "TARGET_68881"
+ "TARGET_68881 && TARGET_68040"
{
if (FP_REG_P (operands[1]))
return "f<FP:round>sqrt%.x %1,%0";