incdir "tests"
include "dsp_base.inc"

input_ax:
; [0] - 0x0000'0000 - 0
CW 0
CW 0
; [1] - 0x0000'0001 - 1 in $ax0.l
CW 0
CW 1
; [2] - 0x0000'ffff - -1 in $ax0.l
CW 0
CW 0xffff
; [3] - 0x0001'0000 - 1 in $ax0.h
CW 1
CW 0
; [4] - 0x7fff'0000 - INT_MAX in $ax0.h
CW 0x7fff
CW 0
; [5] - 0x8000'0000 - INT_MIN in $ax0.h
CW 0x8000
CW 0
; [6] - 0xffff'0000 - -1 in $ax0.h
CW 0xffff
CW 0
input_ax_end:

input_acc:
; [0] - 0x00'0000'0000 - 0
CW 0
CW 0
CW 0
; [1] - 0x00'0000'0001 - 1 in $ac0.l
CW 0
CW 0
CW 1
; [2] - 0x00'0000'ffff - -1 in $ac0.l
CW 0
CW 0
CW 1
; [3] - 0x00'0001'0000 - 1 in $ac0.m
CW 0
CW 1
CW 0
; [4] - 0x00'7fff'0000 - INT_MAX in $ac0.m
CW 0
CW 0x7fff
CW 0
; [5] - 0x00'8000'0000 - INT_MIN in $ac0.m, but not sign extended
CW 0
CW 0x8000
CW 0
; [6] - 0x00'ffff'0000 - -1 in $ac0.m, but not sign extended
CW 0
CW 0xffff
CW 0
; [7] - 0x01'0000'0000 - 1 in $ac0.l
CW 1
CW 0
CW 0
; [8] - 0x7f'ffff'0000 - true INT_MAX
CW 0x7f
CW 0xffff
CW 0
; [9] - 0x80'0000'0000 - true INT_MIN
CW 0x80
CW 0
CW 0
; [10] - 0xff'8000'0000 - INT_MIN in $ac0.m, sign-extended
CW 0xff
CW 0x8000
CW 0
; [11] - 0xff'ffff'0000 - -1
CW 0xff
CW 0xffff
CW 0
input_acc_end:

/* Python script to generate the following result tables from a DSP dump:
import struct
def gen_tables(name, num_ax, num_acc):
	with open(name, "rb") as fin:
		data = fin.read()
	reg_values = list(struct.iter_unpack(">" + "H"*0x20, data))
	# Initial register values (there is no corresponding send_back call for these), then our two
	# default value checks, then the TSTAXH test, then the CMPAXH test, then the test results
	assert len(reg_values) == 1 + 2 + num_ax + num_ax * num_acc + 1
	print("result_table_tstaxh:")
	for ax in range(num_ax):
		# SR is register 0x13
		print("CW {:#04x}".format(reg_values[3 + ax][0x13]))
	print("result_table_tstaxh_end:")
	print()
	print("result_table_cmpaxh:")
	for ax in range(num_ax):
		print("; ax [{}]".format(ax))
		for acc in range(num_acc):
			print("CW {:#04x}".format(reg_values[3 + num_ax + ax * num_acc + acc][0x13]))
	print("result_table_cmpaxh_end:")

gen_tables("dsp_dump0.bin", 7, 12)
*/

result_table_tstaxh:
CW 0x22a4
CW 0x22a4
CW 0x22a4
CW 0x22a0
CW 0x2280
CW 0x2288
CW 0x22a8
result_table_tstaxh_end:

result_table_cmpaxh:
; ax [0]
CW 0x22a5
CW 0x22a1
CW 0x22a1
CW 0x22a1
CW 0x2281
CW 0x2291
CW 0x22b1
CW 0x22b1
CW 0x22b1
CW 0x22b9
CW 0x2289
CW 0x22a9
; ax [1]
CW 0x22a5
CW 0x22a1
CW 0x22a1
CW 0x22a1
CW 0x2281
CW 0x2291
CW 0x22b1
CW 0x22b1
CW 0x22b1
CW 0x22b9
CW 0x2289
CW 0x22a9
; ax [2]
CW 0x22a5
CW 0x22a1
CW 0x22a1
CW 0x22a1
CW 0x2281
CW 0x2291
CW 0x22b1
CW 0x22b1
CW 0x22b1
CW 0x22b9
CW 0x2289
CW 0x22a9
; ax [3]
CW 0x22a8
CW 0x22a8
CW 0x22a8
CW 0x22a5
CW 0x2281
CW 0x2281
CW 0x22b1
CW 0x22b1
CW 0x22b1
CW 0x22b3
CW 0x2299
CW 0x22a9
; ax [4]
CW 0x2288
CW 0x2288
CW 0x2288
CW 0x2288
CW 0x22a5
CW 0x22a1
CW 0x2291
CW 0x2291
CW 0x2291
CW 0x2293
CW 0x22b9
CW 0x2289
; ax [5]
CW 0x2290
CW 0x2290
CW 0x2290
CW 0x2290
CW 0x22b0
CW 0x22b0
CW 0x2290
CW 0x2290
CW 0x229a
CW 0x2298
CW 0x22a5
CW 0x2281
; ax [6]
CW 0x22a0
CW 0x22a0
CW 0x22a0
CW 0x22a0
CW 0x2290
CW 0x2290
CW 0x22b0
CW 0x22b0
CW 0x22ba
CW 0x22b8
CW 0x2288
CW 0x22a5
result_table_cmpaxh_end:

test_main:
	; Perform one test using the default values
	; ($acc0 is 14 0009 0007 and $ax0 is 8000 0003, but this can be changed in the DSPSpy UI)
	; Also, as a sanity check, record the computed sizes of the result tables
	LRI $ar0, #input_ax
	LRI $ix0, #(input_ax_end - input_ax)
	LRI $ar1, #input_acc
	LRI $ix1, #(input_acc_end - input_acc)
	LRI $ar2, #result_table_tstaxh
	LRI $ix2, #(input_ax_end - input_ax)/2
	LRI $ar3, #result_table_cmpaxh
	LRI $ix3, #((input_ax_end - input_ax)/2)*((input_acc_end - input_acc)/3)
	; Set the sticky overflow bit just so that we get consistent $sr values
	; before and after an overflow occurs
	SBSET #1
	CMPAXH $acc0, $ax0.h
	CALL send_back ; Expected $sr: 2290
	; $ar0 should match $ix0, etc
	ADDARN $ar0, $ix0
	LRI $ix0, #input_ax_end
	ADDARN $ar1, $ix1
	LRI $ix1, #input_acc_end
	ADDARN $ar2, $ix2
	LRI $ix2, #result_table_tstaxh_end
	ADDARN $ar3, $ix3
	LRI $ix3, #result_table_cmpaxh_end
	TSTAXH $ax0.h
	CALL send_back ; Expected $sr: 2288

	CLR $acc0
	CLR $acc1
	LRI $ax0.h, #0
	LRI $ax0.l, #0
	LRI $ax1.h, #0
	LRI $ax1.l, #0

	; Check TSTAXH...
	LRI $ar0, #input_ax
	LRI $ar2, #result_table_tstaxh

	; for (int ctr = input_ax.size(); ctr > 0; ctr--) {
	BLOOPI #(input_ax_end - input_ax)/2, check_tstaxh_last_ins
		; Note: if DSPSpy supported populating DMEM as well as IMEM, then there are several
		; instructions that could make this faster and cleaner... but it doesn't currently,
		; so we're stuck with ILRRI.

		; Load the test value into $ax0.h/$ax0.l via $ac0.m
		ILRRI $ac0.m, $ar0 ; $ac0.m = IMEM[$ar0++]
		MRR $ax0.h, $ac0.m
		ILRRI $ac0.m, $ar0
		MRR $ax0.l, $ac0.m
		; Load the expected value into $ac1.m
		ILRRI $ac1.m, $ar2 ; $ac1.m = IMEM[$ar2++]
		; Reduce noise in the results
		LRI $ac0.m, #0

		; Do the test
		TSTAXH $ax0.h
		CALL send_back

		; Check if $sr matches the value we expected.  If there is any difference,
		; note it via a nonzero $ax1.l.  (send_back saves the value of $sr)
		MRR $ac0.m, $sr
		CMP
		IFNZ
		LRIS $ax1.l, #1
check_tstaxh_last_ins:
	NOP
	; }

	; Check CMPAXH...
	CLR $acc0
	CLR $acc1
	LRI $ar0, #input_ax
	LRI $ar3, #result_table_cmpaxh

	; for (int ctr_ax = input_ax.size(); ctr_ax > 0; ctr_ax--) {
	BLOOPI #(input_ax_end - input_ax)/2, check_cmpaxh_last_ins_outer
		; Load the test value into $ax0.h/$ax0.l via $ac1.m
		ILRRI $ac1.m, $ar0
		MRR $ax0.h, $ac1.m
		ILRRI $ac1.m, $ar0
		MRR $ax0.l, $ac1.m

		LRI $ar1, #input_acc

		; for (int ctr_acc = input_acc.size(); ctr_acc > 0; ctr_acc--) {
		BLOOPI #(input_acc_end - input_acc)/3, check_cmpaxh_last_ins_inner

			; Load the test value into $ac0.h/$ac0.m/$ac0.l via $ac1.m
			ILRRI $ac1.m, $ar1
			MRR $ac0.h, $ac1.m
			ILRRI $ac0.m, $ar1 ; we can load it directly here
			ILRRI $ac1.m, $ar1
			MRR $ac0.l, $ac1.m

			; Load the expected value into $ac1.m
			ILRRI $ac1.m, $ar3

			; Do the test
			CMPAXH $acc0, $ax0.h
			CALL send_back

			; Check if $sr matches the value we expected.  If there is any difference,
			; note it via a nonzero $ax1.h.  (send_back saves the value of $sr)
			; We can overwrite $ac0.m here because we load it on the next iteration.
			MRR $ac0.m, $sr
			LRIS $ac0.l, #0
			LRI $ac0.h, #0
			CMP
			IFNZ
			LRIS $ax1.h, #1
check_cmpaxh_last_ins_inner:
		NOP
		; }
check_cmpaxh_last_ins_outer:
	NOP
	; }

	; We're done testing.  In the final send_back call, if $ax1.l or $ax1.h
	; is nonzero, the test failed.
	CALL send_back

	; We're done, DO NOT DELETE THIS LINE
	JMP end_of_test