- Jan 24, 2020
-
-
Michael Pratt authored
Add upstream note
-
Michael Pratt authored
This repository is now unused. Fixes #22
-
- May 27, 2017
-
-
Stefan O'Rear authored
Change-Id: I0641ccf864ac8562b220f75cd4808c21946bf469 Reviewed-on: https://review.gerrithub.io/352842 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
- May 25, 2017
-
-
Benjamin Barenblat authored
24a3215e (“cmd/internal/obj/riscv: Round int(float()) towards zero”) modified the float rounding mode for int-to-float conversions. Update the corresponding encoding tests to take the new encoding into account. Change-Id: Id0e43e414fde661584bd8cf811c3dad4b134d8ea Reviewed-on: https://review.gerrithub.io/362569 Reviewed-by:
Josh Bleecher Snyder <josharian@gmail.com>
-
- May 22, 2017
-
-
Benjamin Barenblat authored
065034e5 (“cmd/internal/obj/riscv: Fix immediate lowering 0x7ffff800 - 0x7fffffff”) modified the encoding for MOVing an immediate into a register. Update the corresponding encoding tests to take the new encoding into account. Change-Id: Ied626e939df5584874e26615183963bd38aa39a4 Reviewed-on: https://review.gerrithub.io/361807 Reviewed-by:
Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by:
Michael Pratt <michael@pratt.im>
-
- May 20, 2017
-
-
Benjamin Barenblat authored
ad5d4f9e (“cmd/internal/obj: large CALL relocations”) changed the expansion of CALL and JMP instructions. Update the corresponding encoding tests to take the new expansions into account. Fixes https://github.com/riscv/riscv-go/issues/17. Change-Id: I8b7e581f3ca1ff7262ac21e1025b404156ed321b Reviewed-on: https://review.gerrithub.io/361788 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Largely copied from MIPS, but RISC-V's arbitrary target register on JALR and the fused BLTU instruction allows us to shorten this a bit. Theoretical minimum is 3 instructions (LD, BLTU, JALR to a linker-generated trampoline that decrements T0 by 12 before acting on it). Change-Id: I4732edcd3560dfd44f283801db49b878673e3b7e Reviewed-on: https://review.gerrithub.io/352826 Reviewed-by:
Benjamin Barenblat <bbaren@google.com>
-
- Apr 20, 2017
-
-
Stefan O'Rear authored
We're now interpreting all FCVTLD instructions from the compiler as round to zero, which is less than ideal; can the SSA->Prog layer be made to set Scond directly? Change-Id: Idcc178e7a8f3b45ef662d1bdc9340c3e073f250c Reviewed-on: https://review.gerrithub.io/352823 Reviewed-by:
Josh Bleecher Snyder <josharian@gmail.com>
-
Stefan O'Rear authored
Traceback is called with pc = sp = -1 to indicate that the values should be fetched from the goroutine, however if a goroutine has no user frames on its stack traceback is retried with a different PC. Make sure that this does not break the -1 signal. Change-Id: Ibd284ad1250fcd80a0d5ecdaf30a67a25144ebb4 Reviewed-on: https://review.gerrithub.io/352798 Reviewed-by:
Josh Bleecher Snyder <josharian@gmail.com>
-
- Apr 19, 2017
-
-
Stefan O'Rear authored
Change-Id: Ia83b5d82330e20fdccf5fa9ee20c01039bf1966f Reviewed-on: https://review.gerrithub.io/352840 Reviewed-by:
Josh Bleecher Snyder <josharian@gmail.com>
-
Stefan O'Rear authored
If a function is longer than the range of JAL (1 MiB) this causes internal jumps to be rewritten to AUIPC + JALR. Change-Id: I0ce2c50f67a4d1daf26b9a2b36c3db0401a52ad0 Reviewed-on: https://review.gerrithub.io/352839 Reviewed-by:
Josh Bleecher Snyder <josharian@gmail.com>
-
Stefan O'Rear authored
Otherwise the SUB is being inferred as a boolean, which means that if the values being compared have the same least significant byte they may be erroneously seen as equal. Change-Id: I471f0048a787a07833206ed5e8e863a1dfcbd28a Reviewed-on: https://review.gerrithub.io/352838 Reviewed-by:
Josh Bleecher Snyder <josharian@gmail.com>
-
Stefan O'Rear authored
Was previously not zeroing the correct address range, leading to garbage in local variables, leading to crashes in select statement execution. Change-Id: I10e985b841d09a015a948ade204dcbca8965e800 Reviewed-on: https://review.gerrithub.io/352837 Reviewed-by:
Josh Bleecher Snyder <josharian@gmail.com>
-
Stefan O'Rear authored
live.go and fixedbugs/issue16249.go now pass Change-Id: I1e5bb9d3ff804de8d5e61af79a7cf29346d4c82c Reviewed-on: https://review.gerrithub.io/352836 Reviewed-by:
Josh Bleecher Snyder <josharian@gmail.com>
-
Stefan O'Rear authored
This removes a lot of temporaries allowing live2.go to pass. Change-Id: I39b8d60c9d0ed265c81665cfa19a0e1ea80f251d Reviewed-on: https://review.gerrithub.io/352834 Reviewed-by:
Josh Bleecher Snyder <josharian@gmail.com>
-
Stefan O'Rear authored
Change-Id: Iaf8dfb7d454f4b5f371652934f03b213374f14ff Reviewed-on: https://review.gerrithub.io/352833 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: Ieab240b9fa00931e09cbfb2d2e7e82ebd1bd6bbb Reviewed-on: https://review.gerrithub.io/352829 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: I9c2314650ce23365ec24d60d1d8f44ea07794e53 Reviewed-on: https://review.gerrithub.io/352828 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
(Should From3 be required in the preprocess stage? mips apparently doesn't.) This causes the compiler to generate a more helpful internal error instead of crashing if From3 is not set on a prog that reaches assemble. Change-Id: Ie501871bab1769da47624d04711712914f0bcd72 Reviewed-on: https://review.gerrithub.io/352827 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: I9e0cc4b5201c56e936f8caed6acedfe24f0bf58a Reviewed-on: https://review.gerrithub.io/352816 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
With this panic + recover is fully functional; remove the hack. Change-Id: I9dbba0cfddfed7ef085c51e7714ed2ca63e331e1 Reviewed-on: https://review.gerrithub.io/352815 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: I6ff63574eeedcc13b34340cfc35548a44cc1036e Reviewed-on: https://review.gerrithub.io/352814 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Improves compatiblity with other assemblers. Change-Id: I763a043dbb07d1cacc3aee470b40cbe735ef4cd1 Reviewed-on: https://review.gerrithub.io/352813 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: I4ca6228d2904369e2a1bbab03be9e2df5025b8e2 Reviewed-on: https://review.gerrithub.io/352812 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: I277a005b8f0b4e426243f1bd9ae7e037346e0cef Reviewed-on: https://review.gerrithub.io/352811 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
This prevents calls from being misidentified as dead code and deleted from a load with no natural exit. Fixes hang in test/235.go. Change-Id: I008d96dde915b5fdf41d30311dd9bab6511d502a Reviewed-on: https://review.gerrithub.io/352810 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: Ib143d79e55395455558914526d222328c6212d0a Reviewed-on: https://review.gerrithub.io/352809 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: Idef628a993ba728caeea656ef6a37f1832535964 Reviewed-on: https://review.gerrithub.io/352808 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
The code generation for LoweredMove and LoweredZero was not processing the last aligned unit, leading to crashes in select statments (go test context). Change-Id: I1120d1ef5ee709ed4efdaf0f9f50910c39d839f0 Reviewed-on: https://review.gerrithub.io/352807 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: Iaf3a2287f393fc21b8a7f79389225867e05ccddf Reviewed-on: https://review.gerrithub.io/352806 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Indirect and interface calls were not being recognized as requiring RA to be saved. Change-Id: I9499707f394ce809bc2b32a4a32b58d85c3fe2ae Reviewed-on: https://review.gerrithub.io/352805 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: Ie487904779dff19f7e8c97eaeaee5743f43d96fa Reviewed-on: https://review.gerrithub.io/352804 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: I62b1c1b6aca76d1daf63e4b8e0670c85c50955e9 Reviewed-on: https://review.gerrithub.io/352803 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
*This loses spike support* Change-Id: If29e76125814d51829fd49b0e621178541ec9d25 Reviewed-on: https://review.gerrithub.io/352802 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
This is the last piece needed to run Hello World and the riscvtests with the runtime enabled. Change-Id: Iec8766e3daec2f1f8dde65cdd03ea88ba4790521 Reviewed-on: https://review.gerrithub.io/352801 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
RISC-V is the first architecture supported by Go where the standard function calling sequence has variable length under usual conditions. For now I've set jmpdefer to assume a 12-byte sequence, but when cmd/link is regularly generating the 4/8-byte sequences this will need to be revisited. Change-Id: I4bf7f933664565d897d1a1286a0ee44f67ce24af Reviewed-on: https://review.gerrithub.io/352800 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Stack checks and reflectcall are not yet online. Change-Id: I2def5e4540330581e39ab2556a2c20da3f5c0b20 Reviewed-on: https://review.gerrithub.io/352799 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
Change-Id: I6288f1d9d19fd9b0a2b464000662b45c7a2dc4f1 Reviewed-on: https://review.gerrithub.io/352796 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
This gives a diagnostic if it turns out to be reachable and prevents a stack address from ever pointing one-past-the-end of a function. Change-Id: I9006b05a8593587821524071e28bb154644120c3 Reviewed-on: https://review.gerrithub.io/352794 Reviewed-by:
Michael Pratt <michael@pratt.im>
-
Stefan O'Rear authored
This is sufficient for scheduler boot but crashes during GC boot. Change-Id: Iae9c4cc46f8196107ac13d2ab74b1540e6d63da9 Reviewed-on: https://review.gerrithub.io/352792 Reviewed-by:
Michael Pratt <michael@pratt.im>
-