Compare changes
+12
−0
+42
−0
bin/run-tests.sh
0 → 100755
+50
−0
+8
−3
+6
−2
+6
−3
o Update fdlibm.h and setexception.c to support the inexact execption. o Use this in asinh. Tests pass.
o Add tests for this o Use setexception for inexact in e_asin.c.
o Add tests for this o Use setexception for inexact in e_exp.c.
Note that original code didn't actually signal inexact probably because the compiler constant-folded one - tiny to one.
This works like WITH-FLOAT-TRAPS-MASKED, except that the specified traps are enabled. Use this in fdlibm to enable the inexact trap.
Merge the body of both macros into one since they only differ in how the bits are merged with the actual mode bits.
WITH-FLOAT-TRAPS-ENABLED.
When enabling traps, need to take just the low 24 bits of the arg because (setf floating-point-modes) wants an (unsigned-byte 24) argument. The logorc2 makes the result negative when enabling traps.
Also replae WITH-INXACT-EXCEPTION-ENABLED with WITH-FLOAT-TRAPS-ENABLED. All tests still pass, as expected.
On sparc and ppc (setf vm:floating-point-modes) takes an (unsigned-byte 32) arg, so adjust the ldb byte appopriately. Clean up code by putting the docstring into the macro.
This makes it quite a bit easier to run the test suite instead of trying to remember exactly how to invoke it from the command line.
Printing of FP numbers can signal some traps like overflow, underflow, inexact, or denormalized-operand, so we need to disable these traps when printing out the error message for arithmetic errors.
* Add some comments. * Change x86 (setf floating-point-modes) to accept (unsigned-byte 32). * Remove unneeded x86 conditionalization on the byte size.
Fix issue #12 by replacing the code with an explicit call to set the inexact exception when needed.
To aid in debugging floating point modes, add two new functions: o %SET-FLOATING-POINT-MODES is like SET-FLOATING-POINT-MODES but applies the result to a specified mode value, returning the new mode value (as an integer). This is useful for investigating different mode values without modifying the actual hardware mode. o %GET-FLOATING-POINT-MODES is like GET-FLOATING-POINT-MODES but uses an integer argument instead of the actual floating-point mode. Useful when used with %SET-FLOATING-POINT-MODE or on its own.
Fix issue #14. WITH-FLOAT-TRAPS-ENABLED was leaving the accrued (and current) exceptions unchanged, but it should have cleared out any values there that matched the exceptions to be enabled. Without this, the next x87 operation would signal an exception if an accrued exception matched an enabled exception. This was the cause of issue #14. (Note that for x87, the accrued exception is the same as current exception.)
WITH-FLOAT-TRAPS-ENABLED was incorrectly setting accrued exceptions. Fix issue #14. WITH-FLOAT-TRAPS-ENABLED was leaving the accrued (and current) exceptions unchanged, but it should have cleared out any values there that matched the exceptions to be enabled. Without this, the next x87 operation would signal an exception if an accrued exception matched an enabled exception. This was the cause of issue #14. (Note that for x87, the accrued exception is the same as current exception.) See merge request !5
Allow search lists in pathname-match-p. For each arg, we enumerate the possible values of the search list and try to find a match between the path and the wild path. If there's a match, return true. Tests added for some cases of pathname-match-p with search lists. Fix issue #16.
We know the result of (expt 0 power) so return it immediately without first checking if the power exceeds the limit. Also took the opportunity to add a better message to the intexp-limit-error condition to make it more explicit what is being computed and why it's failing. Tests added too.
Handle search lists in pathname-match-p. Allow search lists in pathname-match-p. For each arg, we enumerate the possible values of the search list and try to find a match between the path and the wild path. If there's a match, return true. Tests added for some cases of pathname-match-p with search lists. Fix issue #16. See merge request !6
If the user specifies -dynamic-space-size 0, then use the platform-specific maximum heap size. Update the docstring for the switch too.
Forgot to put in an else clause if the specified size was 0.
Don't restore the fpu state from the context in the interrupt handler code (interrupt_handle_now) because that just makes the fpu state the same as the sigcontext that received the exception. I think this was done to reset the x87 fpu precision to 53-bit so that lisp can continue with default double-float rounding. This isn't needed anymore since lisp doesn't use x87 anymore. In SIGFPE-HANDLER, don't modify the current fp state based on the state from the sigcontext. These two items cause the arithmetic-error exception printer to trigger inexact when printing out the args. We add a few debugging prints for now so we can see what the FP state is in the handler. Remove them later.
In sigfpe-handler, don't modify the modes; just use whatever they are. (They should be the default values.) In with-float-traps-*, actually just restore the floating-point mode to the exact original mode instead of trying to mask things out.
We just want to return the original modes, so remove all the old commented out stuff that was modifying the original modes to some strange state. This makes a lot more sense to me if WITH-FLOAT-TRAPS-* actually restored the modes exactly as they were before running the body.
Put an unwind-protect around the error calls. The cleanup form restores the floating-point modes from the sigcontext so that the mode is restored. This is needed, I think, because we throw so that the signal handler doesn't return so the sigcontext isn't restored. If we don't restore the fpu state, it's set to the default processor state. We want the default state when calling error. In this way, things like (* 1d300 1d300) signals an overflow, and when we throw to top-level, the floating-point modes are restored to their original values they had before.
Handle FPU exceptions better Fix #15 Handle FPU exceptions better in sigfpe-handler. Linux now passes all of the tests, including the inexact tests. We can correctly print out the operands (if available) when we receive an inexact exception. The floating point modes are preserved as much as possible when we throw from an arithmetic-error. See merge request !7
o Need to define a new instruction format for these double shifts; they look like ext-reg-reg/mem but there's no width field. o Clean up emit-double-shift slightly. o Update double-shift-inst-printer-list to support both immediate shift or shifts in CL.
These are useful for multi-precision shifts. For x86, we can use the shld and shrd instructions. For others, we just use basic logical operations.
One temp variable for bignum-shld can be removed and the only temp variable for bignum-shld-c can be removed. Base this on the digit-ashr vops. This makes the vops simpler and faster.
Instead of storing the high and low words to memory and then loading it into the double-reg, use SSE2 instructions to directly create the double-float from the high and low words.
Remove the :load-if stuff. Don't think it's needed.
Mention that you need Motif to build motifd which is required for clm. Also say that the build error can be safely ignored if you do not need clm. Fix issue #20
According to the message from Chisheng Huang on cmucl-help, 2016-02-27, 32-bit Ubuntu 11.10 (in VirtualBox) cannot use the full heap space. There's something already allocated at address 0xb7b82000, so limit the max heap to address 0xb7b80000, for a total of 1530 MB instead of 1632 MB. (It would be nice to be able to detect this somehow.)