.result; \
.popsection
+// TODO: assign a separate mem location for the comparison address?
+#define TEST_FP_OP_D32_INTERNAL( testnum, flags, result, val1, val2, val3, code... ) \
+test_ ## testnum: \
+ li TESTNUM, testnum; \
+ la a0, test_ ## testnum ## _data ;\
+ fld f0, 0(a0); \
+ fld f1, 8(a0); \
+ fld f2, 16(a0); \
+ lw a3, 24(a0); \
+ lw t1, 28(a0); \
+ code; \
+ fsflags a1, x0; \
+ li a2, flags; \
+ bne a0, a3, fail; \
+ bne t1, t2, fail; \
+ bne a1, a2, fail; \
+ .pushsection .data; \
+ .align 3; \
+ test_ ## testnum ## _data: \
+ .double val1; \
+ .double val2; \
+ .double val3; \
+ .result; \
+ .popsection
+
+#define TEST_FCVT_S_D32( testnum, result, val1 ) \
+ TEST_FP_OP_D32_INTERNAL( testnum, 0, double result, val1, 0.0, 0.0, \
+ fcvt.s.d f3, f0; fcvt.d.s f3, f3; fsd f3, 0(a0); lw t2, 4(a0); lw a0, 0(a0))
+
#define TEST_FCVT_S_D( testnum, result, val1 ) \
TEST_FP_OP_D_INTERNAL( testnum, 0, double result, val1, 0.0, 0.0, \
fcvt.s.d f3, f0; fcvt.d.s f3, f3; fmv.x.d a0, f3)
TEST_FP_OP_S_INTERNAL( testnum, flags, float result, val1, 0.0, 0.0, \
inst f3, f0; fmv.x.s a0, f3)
+#define TEST_FP_OP1_D32( testnum, inst, flags, result, val1 ) \
+ TEST_FP_OP_D32_INTERNAL( testnum, flags, double result, val1, 0.0, 0.0, \
+ inst f3, f0; fsd f3, 0(a0); lw t2, 4(a0); lw a0, 0(a0))
+// ^: store computation result in address from a0, load high-word into t2
+
#define TEST_FP_OP1_D( testnum, inst, flags, result, val1 ) \
TEST_FP_OP_D_INTERNAL( testnum, flags, double result, val1, 0.0, 0.0, \
inst f3, f0; fmv.x.d a0, f3)
TEST_FP_OP_S_INTERNAL( testnum, flags, dword result, val1, 0.0, 0.0, \
inst f3, f0; fmv.x.s a0, f3)
+#define TEST_FP_OP1_D32_DWORD_RESULT( testnum, inst, flags, result, val1 ) \
+ TEST_FP_OP_D32_INTERNAL( testnum, flags, dword result, val1, 0.0, 0.0, \
+ inst f3, f0; fsd f3, 0(a0); lw t2, 4(a0); lw a0, 0(a0))
+// ^: store computation result in address from a0, load high-word into t2
+
#define TEST_FP_OP1_D_DWORD_RESULT( testnum, inst, flags, result, val1 ) \
TEST_FP_OP_D_INTERNAL( testnum, flags, dword result, val1, 0.0, 0.0, \
inst f3, f0; fmv.x.d a0, f3)
TEST_FP_OP_S_INTERNAL( testnum, flags, float result, val1, val2, 0.0, \
inst f3, f0, f1; fmv.x.s a0, f3)
+#define TEST_FP_OP2_D32( testnum, inst, flags, result, val1, val2 ) \
+ TEST_FP_OP_D32_INTERNAL( testnum, flags, double result, val1, val2, 0.0, \
+ inst f3, f0, f1; fsd f3, 0(a0); lw t2, 4(a0); lw a0, 0(a0))
+// ^: store computation result in address from a0, load high-word into t2
+
#define TEST_FP_OP2_D( testnum, inst, flags, result, val1, val2 ) \
TEST_FP_OP_D_INTERNAL( testnum, flags, double result, val1, val2, 0.0, \
inst f3, f0, f1; fmv.x.d a0, f3)
TEST_FP_OP_S_INTERNAL( testnum, flags, float result, val1, val2, val3, \
inst f3, f0, f1, f2; fmv.x.s a0, f3)
+#define TEST_FP_OP3_D32( testnum, inst, flags, result, val1, val2, val3 ) \
+ TEST_FP_OP_D32_INTERNAL( testnum, flags, double result, val1, val2, val3, \
+ inst f3, f0, f1, f2; fsd f3, 0(a0); lw t2, 4(a0); lw a0, 0(a0))
+// ^: store computation result in address from a0, load high-word into t2
+
#define TEST_FP_OP3_D( testnum, inst, flags, result, val1, val2, val3 ) \
TEST_FP_OP_D_INTERNAL( testnum, flags, double result, val1, val2, val3, \
inst f3, f0, f1, f2; fmv.x.d a0, f3)
TEST_FP_OP_S_INTERNAL( testnum, flags, word result, val1, 0.0, 0.0, \
inst a0, f0, rm)
+#define TEST_FP_INT_OP_D32( testnum, inst, flags, result, val1, rm ) \
+ TEST_FP_OP_D32_INTERNAL( testnum, flags, dword result, val1, 0.0, 0.0, \
+ inst a0, f0, f1; li t2, 0)
+
#define TEST_FP_INT_OP_D( testnum, inst, flags, result, val1, rm ) \
TEST_FP_OP_D_INTERNAL( testnum, flags, dword result, val1, 0.0, 0.0, \
inst a0, f0, rm)
TEST_FP_OP_S_INTERNAL( testnum, flags, word result, val1, val2, 0.0, \
inst a0, f0, f1)
+#define TEST_FP_CMP_OP_D32( testnum, inst, flags, result, val1, val2 ) \
+ TEST_FP_OP_D32_INTERNAL( testnum, flags, dword result, val1, val2, 0.0, \
+ inst a0, f0, f1; li t2, 0)
+
#define TEST_FP_CMP_OP_D( testnum, inst, flags, result, val1, val2 ) \
TEST_FP_OP_D_INTERNAL( testnum, flags, dword result, val1, val2, 0.0, \
inst a0, f0, f1)
TEST_CASE(testnum, a0, correct, li a0, input; fmv.s.x fa0, a0; \
fclass.s a0, fa0)
+#define TEST_FCLASS_D32(testnum, correct, input) \
+ TEST_CASE(testnum, a0, correct, \
+ la a0, test_ ## testnum ## _data ;\
+ fld fa0, 0(a0); \
+ fclass.d a0, fa0) \
+ .pushsection .data; \
+ .align 3; \
+ test_ ## testnum ## _data: \
+ .dword input; \
+ .popsection
+
#define TEST_FCLASS_D(testnum, correct, input) \
TEST_CASE(testnum, a0, correct, li a0, input; fmv.d.x fa0, a0; \
fclass.d a0, fa0)
.float result; \
.popsection
+#define TEST_INT_FP_OP_D32( testnum, inst, result, val1 ) \
+test_ ## testnum: \
+ li TESTNUM, testnum; \
+ la a0, test_ ## testnum ## _data ;\
+ lw a3, 0(a0); \
+ lw a4, 4(a0); \
+ li a1, val1; \
+ inst f0, a1; \
+ \
+ fsd f0, 0(a0); \
+ lw a1, 4(a0); \
+ lw a0, 0(a0); \
+ \
+ fsflags x0; \
+ bne a0, a3, fail; \
+ bne a1, a4, fail; \
+ .pushsection .data; \
+ .align 3; \
+ test_ ## testnum ## _data: \
+ .double result; \
+ .popsection
+
#define TEST_INT_FP_OP_D( testnum, inst, result, val1 ) \
test_ ## testnum: \
li TESTNUM, testnum; \
.double result; \
.popsection
+// We need some special handling here to allow 64-bit comparison in 32-bit arch
+// TODO: find a better name and clean up when intended for general usage?
+#define TEST_CASE_D32( testnum, testreg1, testreg2, correctval, code... ) \
+test_ ## testnum: \
+ code; \
+ la x31, test_ ## testnum ## _data ; \
+ lw x29, 0(x31); \
+ lw x31, 4(x31); \
+ li TESTNUM, testnum; \
+ bne testreg1, x29, fail;\
+ bne testreg2, x31, fail;\
+ .pushsection .data; \
+ .align 3; \
+ test_ ## testnum ## _data: \
+ .dword correctval; \
+ .popsection
+
+// ^ x30 is used in some other macros, to avoid issues we use x31 for upper word
+
#-----------------------------------------------------------------------
# Pass and fail code (assumes test num is in TESTNUM)
#-----------------------------------------------------------------------
projects / riscv-tests.git / blobdiff