f63b4f85aa
This commit works around a bug in xtensa-esp32-elf-gcc version esp-2020r3. The bug is in generation of loop constructs. The below code is generated by the xtensa-esp32 compiler. The first extract is the buggy machine code and the second extract is the corrected machine code. The test `basics/logic_constfolding.py` fails with the first code and succeeds with the second. Disassembly of section .text.push_result_rule: 00000000 <push_result_rule>: ... d6: 209770 or a9, a7, a7 d9: 178976 loop a9, f4 <push_result_rule+0xf4> d9: R_XTENSA_SLOT0_OP .text.push_result_rule+0xf4 dc: 030190 rsr.lend a9 df: 130090 wsr.lbeg a9 e2: a8c992 addi a9, a9, -88 e5: 06d992 addmi a9, a9, 0x600 e8: 130190 wsr.lend a9 eb: 002000 isync ee: 030290 rsr.lcount a9 f1: 01c992 addi a9, a9, 1 f4: 1494e7 bne a4, a14, 10c <push_result_rule+0x10c> f4: R_XTENSA_SLOT0_OP .text.push_result_rule+0x10c Disassembly of section .text.push_result_rule: 00000000 <push_result_rule>: ... d6: 209770 or a9, a7, a7 d9: 178976 loop a9, f4 <push_result_rule+0xf4> d9: R_XTENSA_SLOT0_OP .text.push_result_rule+0xf4 dc: 030190 rsr.lend a9 df: 130090 wsr.lbeg a9 e2: 000091 l32r a9, fffc00e4 <push_result_rule+0xfffc00e4> e2: R_XTENSA_SLOT0_OP .literal.push_result_rule+0x18 e5: 0020f0 nop e8: 130190 wsr.lend a9 eb: 002000 isync ee: 030290 rsr.lcount a9 f1: 01c992 addi a9, a9, 1 f4: 1494e7 bne a4, a14, 10c <push_result_rule+0x10c> f4: R_XTENSA_SLOT0_OP .text.push_result_rule+0x10c Work done in collaboration with @jimmo. Signed-off-by: Damien George <damien@micropython.org> |
||
---|---|---|
.. | ||
bare-arm | ||
cc3200 | ||
esp32 | ||
esp8266 | ||
javascript | ||
mimxrt | ||
minimal | ||
nrf | ||
pic16bit | ||
powerpc | ||
qemu-arm | ||
renesas-ra | ||
rp2 | ||
samd | ||
stm32 | ||
teensy | ||
unix | ||
windows | ||
zephyr |